Microglia are dispensable for experience-dependent refinement of mouse visual circuitry

Abstract

To test the hypothesized crucial role of microglia in the developmental refinement of neural circuitry, we depleted microglia from mice of both sexes with PLX5622 and examined the experience-dependent maturation of visual circuitry and function. We assessed retinal function, receptive field tuning of visual cortex neurons, acuity and experience-dependent plasticity. None of these measurements detectibly differed in the absence of microglia, challenging the role of microglia in sculpting neural circuits.

Microglia are proposed to be critical mediators of neural circuit refinement¹. However, data demonstrating that microglia alter neuronal function or corresponding neural circuits relevant to behavior remain limited. We reasoned that, if microglia perform functions essential for the maturation of neural circuits, then eliminating microglia should disrupt the tuning properties of neurons and their plasticity, which are refined during the experience-dependent phase of visual development (approximately postnatal days 14 to 45 (~P14–P45+)). Here, we depleted microglia throughout the period of experience-dependent refinement from soon after eye opening (~P14) and measured the functional consequences for visual circuitry.

Microglia require signaling through the colony-stimulating factor 1 receptor (CSF1R) for survival². To deplete microglia from the developing brain, we fed dams and pups chow containing the highly selective CSF1R inhibitor PLX5622 (1,200 mg kg⁻¹, PLX chow)³. Treatment was initiated at P14 for all experiments; mice can be maintained on PLX for durations lasting months^3,4.

First, we examined retinal function because any alteration in the retina would necessarily impact all downstream visual pathways. Heterozygous Cx3cr1^GFP mice express green fluorescent protein (eGFP) in microglia⁵. Treatment with PLX from P14 to P50 depleted microglia throughout the retina, consistent with published studies⁶ (Fig. 1a,b). To test for functional changes at the level of the photoreceptors and ON bipolar cells, we compared electroretinograms (ERGs) of wild-type (WT) mice treated with PLX and controls. The scotopic a-wave and b-wave, which represent the respective rod photoreceptor and rod bipolar cell response, were not different between groups (Fig. 1c). The photopic a-wave and b-wave, which represent the cone photoreceptor and cone ON bipolar cell response, and the flicker ERG were also indistinguishable between groups (Fig. 1c). Calcium imaging of retinal ganglion cells (RGCs) in a whole-mount preparation from WT mice carrying the Thy1-GCaMP6f transgene displayed no significant differences in response amplitudes (normalized change in fluorescence, ΔF/F) of ON-type, OFF-type and ON–OFF-type responses between the PLX and control groups (Fig. 1d–f)⁷. The frequency distribution of response types was also similar. We conclude that loss of microglia does not cause evident changes in visual signaling in the retina. This finding is consistent with a previous study that reported no short-term effect of microglial ablation on retinal structure and visual function but did report a decrease in amplitude of the scotopic rod response⁸.

Fig. 1 |. Visual signaling in retina of mice without microglia.

a, Schematic distribution of microglia (green) within the retinal anatomy (gray). b, Confocal fluorescence images of the outer plexiform layer (OPL) of control (C, top) and PLX-treated (PLX, bottom) retinas of CX3CR1-EGFP heterozygous mice (green, left), with immunostaining for the microglia marker Iba1 (magenta, center); right, the merged image (scale bar, 50 μm). c, Dark-adapted and light-adapted full-field ERGs with light flashes at three intensities (C, black, n = 10 mice, 5 male (m) and 5 female (f) mice; PLX-treated, red, n = 10, 6m and 4f mice; gray and pink lines, ±1.0 s.e.m.). The scotopic response amplitude (0.9 log cd s m⁻², flash) did not differ between groups (mean ± s.e.m.: C, black, 120 ± 16; PLX, red, 95 ± 10; P = 0.21, determined by Welch’s two-sided t-test (W2t-test)) nor did the photopic response amplitude (3.2 log cd s m⁻², flash; C, 54 ± 7; PLX, 53 ± 6; P = 0.88, determined by W2t-test) and the s.d. of the photopic flicker response (C, 5.7 ± 0.8; PLX, 5.2 ± 1.1; P = 0.67, determined by W2t-test). d, Two-photon fluorescence image of GCaMP6f-expressing ganglion cells. Scale bar, 10 μm. Bottom, ROI masks (blue) used to extract each cell’s calcium response. e, Fluorescence responses of the cells shown in d. Gray bars, dark phase of the 1-Hz, contrast-reversing spot stimulus. f, Response amplitudes for all cells recorded in C mice (gray; black, mean ± 1.0 s.e.m.; n = 6, 3m and 3f mice) and PLX-treated mice (pink; red, mean ± 1.0 s.e.m.; n = 7, 4m and 3f mice). Left, cells grouped by primary response waveform: nonmodulating (f0; C, n = 189; PLX, n = 140); modulating at the stimulus frequency (f1; C, n = 565; PLX, n = 582); modulating at twice the stimulus frequency (f2; ON–OFF response; C, n = 35; PLX, n = 38). Right, response amplitudes of OFF-type and ON-type ganglion cells with peak response at the stimulus frequency (1 Hz; f1 population of left panel: OFF, C, n = 234; PLX, n = 309; ON, C, n = 331; PLX, n = 273). P values were determined by one-way analysis of variance (ANOVA) and Holm–Šidák’s multiple-comparison test.

Next, we examined central visual circuitry. In the mouse, retinogeniculate afferents from the contralateral eye and ipsilateral eye during development are initially intermingled upon their arrival at the dorsal lateral geniculate nucleus (dLGN) but then separate into distinct termination zones with minimal overlap by P14 (refs. 9,10). To avoid the potential impact of depleting microglia on retinogeniculate segregation^11,12, we initiated treatment with PLX beginning at P14, after retinogeniculate segregation is complete. Eradication of microglia was extensive after 4 days of treatment. In the primary visual cortex (V1), only a handful of microglia were evident at P18 by staining against ionized calcium-binding adaptor molecule 1 (Iba1) (Fig. 2a and Extended Data Fig. 1). The brain was almost completely devoid of microglia by P21 and depletion persisted for the duration of treatment, consistent with other studies^3,4.

Fig. 2 |. Retinogeniculate segregation in LGN, neuronal tuning properties in V1 and acuity of mice without microglia.

a, Images of V1 from C and PLX-treated mice immunostained for Iba1. Scale bar, 125 μm. Inset, ×2.5 magnification. b, Number of microglia per V1 imaging field. Symbols represent mice: circles, m mice; triangles, f mice (P18: C, n = 3, 2m and 1f mice; PLX, n = 4, 2m and 2f mice; P21: C, n = 4, 1m and 3f mice; PLX, n = 3, 2m and 1f mice; P28: C, n = 3, 2m and 1f mice; PLX, n = 3, 3m and 0f mice; P55: C, n = 3, 1m and 2f mice; PLX, n = 3, 1m and 2f mice). c, Schematic of retinogeniculate tracing; axons from ipsilateral (green) and contralateral (magenta) eye projecting to LGN. d, Example images of retinogeniculate overlap for C and PLX-treated mice. Scale bar, 100 μm. e, Percentage overlap of retinogeniculate axons in C (black, n = 8, 5m and 3f mice) and PLX-treated (red, n = 6, 2m and 4f mice) mice (mean ± s.d.: C, 1.4% ± 0.4%; PLX, 1.5% ± 0.6%; P = 0.67, determined by W2t-test). f–h, Excitatory L2 and L3 neurons in V1 for neurons responsive to the contralateral eye for C (black, n = 7, 3m and 4f mice) and PLX-treated (red, n = 7, 4m and 3f mice) mice. Gray and pink circles represent neurons (C, n = 598; PLX, n = 750). Lines indicate the mean for mice. f, Median preferred SF (mean ± s.d.: C, 0.06 ± 0.02; PLX, 0.06 ± 0.02; P = 0.76, determined by W2t-test). g, The percentage of visually responsive neurons per mouse with significant responses to each SF (P = 0.44, determined by two-way ANOVA F(12,84) = 1.01). h, OSI (mean ± s.d.: C, 0.51 ± 0.06; PLX, 0.48 ± 0.05; P = 0.26, determined by W2t-test). i, The percentage of binocular neurons per mouse (mean ± s.d.: C, 29% ± 10%; PLX, 24% ± 10%; P = 0.40, determined by W2t-test). j, The difference in preferred orientation for binocular neurons (C, n = 203; PLX, n = 161; P = 0.16, determined by Kolmogorov–Smirnov test). k, Acuity for C (n = 9, 4m and 5f mice) and PLX-treated (n = 9, 5m and 4f mice) mice (mean ± s.d.: C, 0.45 ± 0.05 cpd; PLX 0.44 ± 0.04 cpd; P = 0.56, determined by W2t-test). l, The number of blocks of ten trials to reach the acuity testing criteria for each mouse (mean ± s.d.: C, 4.2 ± 2.1 blocks; PLX, 3.4 ± 1.1 blocks; P = 0.35, determined by W2t-test).

We mapped the distribution of inputs from the two eyes with dual-color anterograde tracing with cholera toxin subunit B (CTB) for mice raised from P14 on either PLX or control chow until at least P24 (Fig. 2c). Inputs to each hemisphere were dominated by the contralateral eye and surrounded a smaller patch represented by the ipsilateral eye, matching the known normal anatomy¹³ (Fig. 2d). Consistent with previous studies, the overlap of territory occupied by inputs from both eyes was small in control mice (1–2%)¹⁰ (Fig. 2e). Depleting microglia did not alter the overlap between to the two eyes. Thus, the absence of microglia after eye opening does not degrade the established segregation of these inputs at the first stage of central visual processing.

Then, we measured the response properties of neurons in layers 2 and 3 (L2 and L3) of V1 with two-photon calcium imaging at cellular resolution in alert WT mice that expressed the genetically encoded calcium sensor GCaMP6s in excitatory neurons^14–16 (Supplementary Videos 1 and 2). In mouse, vision drives the refinement of receptive field tuning in V1 during the first weeks after eye opening^17–20. The preferred spatial frequency (SF) and orientation selectivity of excitatory neurons in L2 and L3 significantly increase between P18 and P36, as does the fraction of binocular neurons¹⁸. The matching of orientation preference for binocular neurons also improves during this interval^15,18,21. Mice were treated with PLX from P14 until imaging at P28–P32. The median preferred SF, fraction of neurons with significant responses at each SF tested, orientation tuning, percentage of binocular neurons and binocular matching orientation preference were all similar between groups (Fig. 2e–j). Overall, there was no significant difference in the tuning properties of mice depleted of microglia.

The maturation of neuronal tuning properties is concomitant with improved vision. In mice, visual acuity measured with a behavioral assay nearly doubles in the 3 weeks after weaning (~P21) to attain adult levels around P40 (refs. 22,23). Visual experience is required to improve acuity²⁴. The acuity of mice treated with PLX from P14 to the completion of testing (~P45) was indistinguishable from controls (Fig. 2k). We conclude that acuity is normal in mice lacking microglia from soon after eye opening through the period of experience-dependent maturation of acuity.

Given the lack of a detectable effect of depletion of microglia on the experience-dependent refinement of visual circuitry with normal vision, we examined whether microglia contribute to plasticity associated with abnormal vision. Brief monocular deprivation (MD) during the developmental critical period alters ocular dominance (OD)²⁵. Nondeprived mice exhibited a pronounced bias toward responsiveness to visual stimuli presented to the contralateral eye, resulting in contralateral bias index (CBI) scores near 0.7 (Fig. 3a). Furthermore, 4 days of MD initiated during the zenith of the critical period (P26–P28) reduced the CBI scores to near 0.5 as the OD shifted significantly away from the contralateral (deprived) eye (Fig. 3a). Nondeprived mice treated with PLX from P14 to P32 had normal CBI scores, whereas mice receiving 4 days of MD at P28 displayed typical OD plasticity as measured with multiunit electrophysiologic recordings (Fig. 3a).

Fig. 3 |. OD plasticity of mice without microglia.

a, CBI scores from multiunit electrophysiologic recordings. Circles, m mice; triangles, f mice. Nondeprived C (black, n = 10, 8m and 2f mice) and PLX-treated (red, n = 9, 5m and 4f mice) mice, as well as both groups receiving 4 days of MD (n = 9, 4m and 5f mice; n = 9, 3m and 6f mice). Nondeprived C mice display higher CBI scores than mice following MD (mean ± s.d.: C, 0.67 ± 0.03; 4 days MD, 0.50 ± 0.04; P < 0.0001), as do PLX mice (mean ± s.d.: C, 0.67 ± 0.03; 4 days MD, 0.47 ± 0.04; P < 0.0001), while MD mice have similar CBI scores between groups (P = 0.47). P values were determined by a Welsh ANOVA with Dunnett’s multiple-comparison test. b, ODI scores from calcium imaging for nondeprived C (black, n = 7, 3m and 4f mice) and PLX-treated (red, n = 7, 4m and 3f mice) mice, as well as mice receiving 4 days of MD (n = 6, 3m and 3f mice; n = 7, 4m and 3f mice). Nondeprived C mice display higher ODI scores than mice following MD (mean ± s.d.: C, 0.43 ± 0.16; 4 days MD, −0.12 ± 0.14; P = 0.0001), as do PLX mice (mean ± s.d.: C, 0.44 ± 0.11; 4 days MD, 0.00 ± 0.15; P = 0.0002), while MD mice have similar CBI scores between groups (P = 0.42, determined by a Welsh ANOVA with Dunnett’s multiple-comparison two-sided test). c, Histogram of ODI scores for neurons corresponding to the nondeprived and MD C mice in b (C, n = 622; MD, n = 737). d, Histogram of ODI scores for neurons corresponding to the nondeprived and 4 days MD PLX mice in b (C, n = 733; MD, n = 931).

We also measured OD plasticity with two-photon calcium imaging at neuronal resolution (Fig. 3b–d). OD plasticity during the critical period involves a complex change in the tuning properties of excitatory cortical neurons. In L2 and L3, MD reduces the number of neurons responsive only to the contralateral (deprived) eye and increases the number responsive only to the ipsilateral (nondeprived) eye¹⁶. MD yielded similar OD shifts in L2 and L3 of both PLX-treated mice and controls that were consistent with our results with electrophysiology (Fig. 3a). MD decreased the fraction of neurons that were predominantly monocular and responsive to the contralateral (deprived) eye and increased the fraction of neurons that were predominantly monocular and responsive to the ipsilateral (nondeprived) eye, with only modest effects on binocular neurons (Fig. 3c,d). Thus, microglia do not appear to contribute to OD plasticity as measured with electrophysiology and calcium imaging.

These results may appear surprising, considering that preceding studies concluded that microglia have a critical role in the development and plasticity of visual circuitry, although evidence of changes in neuronal function is largely absent^{1,11,12,26–28}. Mice lacking the gene for either C1q or C3 are reported to display impaired retinogeniculate segregation^11,12. However, the magnitude of this deficit may be much smaller than first reported and no subsequent studies identified any deficits in visual processing or vision by C1q-mutant or C3-mutant mice^29,30. C1q-mutant and Cx3cr1-mutant mice also possess normal binocularity and OD plasticity^31–33. If microglia have a role in the refinement and plasticity of visual circuitry, whether by synaptic pruning or another mechanism, it may be restricted to the thalamus during early developmental ages and its functional importance is unclear.

Deletion of the purinergic receptor P2Y12 gene, a microglial gene, prevents OD plasticity as measured with optical imaging of intrinsic signals²⁸. This technique measures changes in reflectance from brain tissue to represent neural activity by the hemodynamic response. It remains unclear how disrupting P2Y12 signaling by microglia would disrupt OD plasticity when eradicating microglia altogether does not. However, a recent study reported that disrupting microglial P2Y12 signaling impairs neurovascular coupling, which may have influenced measurements of binocularity inferred from optical imaging of intrinsic signals³⁴.

In addition, previous studies examined the visual circuitry after treating mice with PLX^4,35. One reported a decrease in orientation tuning and predicted that mice lacking microglia have impaired visual acuity⁴. In contrast, we did not detect any difference in orientation tuning of neurons in V1 despite presenting visual stimuli with identical resolution for orientation (30° spacing), better resolution for SF (octave versus half-octave) and superior sampling (20 trials versus 40 trials on average). We also measured acuity and determined it was normal in mice depleted of microglia. A separate study reported the absence of OD plasticity in mice depleted of microglia³⁵. Unfortunately, this study was unable to detect the established magnitude of OD plasticity in control mice and whether these recordings were sufficient for interpretation is in question. By comparison, we performed electrophysiologic recordings as reported previously, as well as calcium imaging at neuronal resolution^{16,22,36–39}. According to our results using these two techniques, the OD plasticity in mice depleted of microglia is normal.

In conclusion, we observe that removing microglia and, thus, the functional roles attributed to microglia, including spine pruning, are without evident functional consequences for the experience-dependent phase of visual development. We propose that microglia are dispensable for experience-dependent refinement and plasticity of visual circuitry.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements ofdata and code availability are available at https://doi.org/10.1038/s41593-024-01706-3.

Experimental model and subjects

All procedures were approved by the University of Louisville Institutional Animal Care and Use Committee and were in accord with guidelines set by the US National Institutes of Health (NIH). Mice were anesthetized by isoflurane inhalation and killed by carbon dioxide asphyxiation or cervical dislocation following deep anesthesia in accordance with approved protocols. Mice were housed in groups of five or fewer per cage in a 12-h light and 12-h dark cycle. Animals were naive subjects with no prior history of participation in research studies.

Mice

Retinal experiments were performed on CX3CR1-EGFP (The Jackson Laboratory; strain 005582) and Thy1-GCaMP6f mice (The Jackson Laboratory; strain 025393) on mice age P50–P60. Calcium imaging in visual cortex was performed on mice expressing GCaMP6s in excitatory neurons in forebrain at P28–P32. The CaMKII-tTA (stock no. 007004) and TRE-GCaMP6s (stock no. 024742) transgenic mouse lines were obtained from The Jackson Laboratory^14,40. All other experiments were performed on mice from crosses of CaMKII-tTA and TRE-GCaMP6s that lacked one of the two transgenes. Unless stated otherwise, mice were P50–P60 at the time of experiment. Mice were genotyped with primer sets suggested by The Jackson Laboratory.

Treatment with PLX5622

PLX5622 (Chemgood) was mixed into OpenStandard diet with 15 kcal% fat (Research Diets) at 1,200 mg kg⁻¹ and stored at 4 °C until use. Cages of mice were provided chow containing PLX or the chow alone (controls) beginning at P14 until the completion of experiments.

Immunohistology of retina

We tested for the presence of microglia in PLX-treated and control retinas using immunohistochemistry against microglia marker Iba1. Eyes were enucleated and hemisected and the posterior eyecup including the retina was placed in 4% PFA (30 min at 4 °C) followed by a wash in 0.1 M PBS. The fixed retina was removed from the eyecup and incubated with blocking buffer (24 h at 4 °C) followed by incubation with a primary antibody to Iba1 (rabbit polyclonal; 1:1,000; 3–5 days at 4 °C) and incubation with a fluorescence-tagged secondary antibody (goat anti-rabbit; 1:1,000, 24 h at 4 °C). Retinas were radially incised for flat-mounting on a glass microscope slide in Vectashield mounting medium (Vector Laboratories). Images were acquired on an Olympus Fluoview 1200 confocal fluorescence microscope.

Retinal experiments

Mice receiving PLX chow were examined 35 days after beginning treatment. Gross visual function was assessed by ERG as in our previously published experiments⁴¹. Briefly, mice were dark-adapted overnight and anesthetized with a ketamine + xylazine solution (127 + 12 mg kg⁻¹, respectively) diluted in normal mouse Ringer’s solution and prepared for ERG recordings under dim red light. Pupils were dilated and accommodation-relaxed with topical applications of phenylephrine hydrochloride and tropicamide. The corneal surface anesthetized using proparacaine HCl. Body temperature was maintained by an electric heating pad (TC1000 Temperature Controller, CWE). A clear acrylic contact lens with a gold electrode (LKC Technologies) was placed on the cornea and wet with artificial tears (Tears Again, OCuSOFT). Ground and reference needle electrodes were placed in the tail and on the midline of the forehead, respectively. For scotopic responses, stimuli were presented following >2 h dark adaptation. For photopic responses, the animals were light-adapted for 15 min and test flashes were presented on a rod-saturating background.

Neuronal responses to visual stimulation were assessed using two-photon fluorescence calcium imaging of RGCs in Thy1-GCaMP6f mice as in our previously published experiments⁴². Briefly, mice were dark-adapted for ∼30 min, anesthetized with isoflurane and killed by cervical dislocation under dim red illumination. Eyes were enucleated and hemisected in oxygenated Ames medium (95% O₂, 5% CO₂; Sigma-Aldrich) under infrared illumination using night-vision scopes (OWL Night Vision Scopes; B. E. Meyers) mounted on a dissecting microscope (Olympus SZ61). The retina was radially incised for flattening, separated from the eyecup at the retinal pigment epithelium–photoreceptor layer junction and mounted ganglion cell side up on a nitrocellulose filter paper disk (Millipore Sigma); holes in the filter paper (1.2 mm in diameter) enabled visual stimulation of the photoreceptors through the condenser light path of the microscope. The paper disk with retina preparation was placed in a custom-designed, three-dimensionally printed recording chamber and mounted onto the stage of a custom-built two-photon fluorescence microscope. The retina preparation was continuously perfused with nonrecycled oxygenated Ames medium at physiological temperature (∼6 ml min⁻¹; 33–35 °C) for the duration of the experiment—approximately 3 h per retina. Two-photon fluorescence imaging was performed with a modified Olympus microscope controlled with ScanImage 3.8 software and an Olympus ×60, 1.0 numerical aperture (NA) LUMPlanFL/IR objective. The scan laser (Chameleon Ultra II; Coherent) was tuned to 910 nm. Fluorescence responses were recorded with X–Y frame scans at 15 fps.

Immunohistology of visual cortex

Sectioning, immunostaining and imaging were performed as previously described but with an anti-Iba1 antibody diluted 1:200 (5 μg ml⁻¹) and fluorescence-tagged secondary antibody (goat anti-rabbit; 1:200)^22,37.

Composite coronal sections of each hemisphere were assembled from overlapping image fields in Adobe Photoshop using the ‘auto-align layers’ function. All images were collected in the same session using identical parameters.

Tracing of retinogeniculate projections and analysis of overlap in dLGN

Mice were perfused transcardially with 4% PFA in PBS and the brains were sliced coronally into 70-μm sections with a vibratome. The sections spanning the dLGN were mounted serially. The middle three sections of the dLGN for each hemisphere were imaged and analyzed. Epifluorescence microscopy was performed with a BX-51 upright microscope (Olympus) and Retiga EX B 12-bit monochrome camera. Images were through a ×10, 0.25 NA PLAN objective. Images were downsampled to 8 bit with Adobe Photoshop. The brightness of the image was adjusted to fill the full dynamic range. The background was then determined by measuring the mean pixel intensity in a region outside of the dLGN. This value was subtracted for each pixel in the image. A mask was drawn at the circumference of the dLGN. The images were then imported into ImageJ (NIH) and binarized. Each channel was then passed through a threshold set at 0.1 and the images were multiplied to yield the pixels overlapping between the two images. The number of overlapping pixels was measured as the numerator for the percentage of overlap. The number of pixels in the unmasked area was measured as the denominator for determining the percentage of overlap.

Calcium imaging to measure neuronal tuning properties

Imaging and analysis were performed blind to treatment. Methods were identical to our previously published experiments¹⁶. In brief, mice were administered carprofen (5 mg kg⁻¹) and buprenorphrine (0.1 mg kg⁻¹) for analgesia and anesthetized with isoflurane (4% induction, 1–2% maintenance). The hair on the scalp was clipped and mice were mounted on a stereotaxic frame with a palate bar; their body temperature was maintained at 37 °C with a heat pad controlled by feedback from a rectal thermometer (TCAT-2LV, Physitemp). The scalp was resected, the connective tissue was removed from the skull and an aluminum headbar was affixed with C&B metabond (Parkell). A circular region of bone 3 mm in diameter centered over the left visual cortex was removed using a high-speed drill (Foredom). Care was taken to not perturb the dura. A sterile 3-mm circular glass coverslip was sealed to the surrounding skull with cyanoacrylate (Pacer technology) and dental acrylic (ortho-jet, Lang Dental). The remaining exposed skull was likewise sealed with cyanoacrylate and dental acrylic. Mice recovered on a heating pad. Mice were left to recover for at least 2 days before imaging.

After implantation of the cranial window, the binocular zone of visual cortex was identified with wide-field calcium imaging similar to our method for optical imaging of intrinsic signals³⁶. In brief, mice were anesthetized with isoflurane (4% induction), provided a low dose of the sedative chlorprothixene (0.5 mg kg⁻¹ intraperitoneally; C1761, Sigma) and secured by the aluminum headbar. The eyes were lubricated with a thin layer of ophthalmic ointment (Puralube, Dechra Pharmaceuticals). Body temperature was maintained at 37 °C with a heating pad regulated by a rectal thermometer (TCAT-2LV, Physitemp). Visual stimulus was provided through custom-written software (MATLAB, Mathworks). A monitor was placed 25 cm directly in front of the animal and subtended +40° to −40° of visual space in the vertical axis. A horizonal white bar (2° high and 20° wide) centered on the 0° azimuth drifted from the top to bottom of the monitor with a period of 8 s. The stimulus was repeated 60 times. Cortex was illuminated with blue light (475 ± 30 nm) (475/35, Semrock) from a stable light source (Intralux dc-1100, Volpi). Fluorescence was captured using a green filter (HQ535/50) attached to a tandem lens (50-mm lens, Computar) and camera (Manta G-1236B, Allied Vision). The imaging plane was defocused to approximately 200 μm below the pia. Images were captured at 10 Hz as images of 1,024 × 1,024 pixels and 12-bit depth. Images were binned spatially 4 × 4 before the magnitude of the response at the stimulus frequency (0.125 Hz) was measured by Fourier analysis.

For imaging at cellular resolution, mice were mounted by the headplate atop a spherical treadmill. The monitor was centered on the 0° azimuth and elevation at a distance 35 cm away from the mouse and subtended 45° (vertical) by 80° (horizontal) of visual space. A battery of static sinusoidal gratings were generated in real time with custom software (Processing, MATLAB) as described¹⁵. Stimulus presentation was synchronized to the imaging data by time stamping the presentation of each visual stimulus to the image acquisition frame number, with a transistor-to-transistor logic (TTL) pulse generated with an Arduino at each stimulus transition. Orientation was sampled at equal intervals of 30° from 0° to 150° (six orientations). SF was sampled in eight steps on a logarithmic scale at half-octaves from 0.028 to 0.48 cycles per degree. An isoluminant gray screen was included (blank), provided as a ninth step in the SF sampling as a control. The spatial phase was equally sampled at 45° intervals from 0° to 315° for each combination of orientation and SF. Gratings with random combinations of orientation, SF and spatial phase were presented at a rate of 4 Hz on a monitor with a refresh rate of 60 Hz. Imaging sessions were 10 min (2,400 presentations in total). Consequently, each combination of orientation and SF was presented 40 times on average (range, 29–56).

Imaging was performed with a resonant scanning two-photon microscope controlled by Scanbox image acquisition and analysis software (Neurolabware). The objective lens was fixed at vertical for all experiments. Fluorescence excitation was provided by a tunable wavelength infrared laser (Ultra II, Coherent) at 920 nm. Images were collected through a ×16 water-immersion objective (Nikon, 0.8 NA). Images (512 × 796 pixels, 520 × 740 μm) were captured at 15.5 Hz at depths between 150 and 400 μm. Eye movements and changes in pupil size were recorded using a Dalsa Genie M1280 camera (Teledyne Dalsa) fitted with a 50-mm 1.8 lens (Computar) and an 800-nm long-pass filter (Edmunds Optics). Imaging was performed on alert mice positioned on a spherical treadmill by the aluminum headbar affixed to the skull. The visual stimulus was presented to each eye separately by covering the fellow eye with a small custom occluder.

The imaging series for each eye were motion-corrected with the SbxAlign tool. Regions of interest (ROIs) corresponding to excitatory neurons were selected manually with the SbxSegment tool following computation of pixel-wise correlation of fluorescence changes over time from 350 evenly spaced frames (~4%). ROIs for each experiment were determined by correlated pixels the size similar to that of a neuronal soma. The fluorescence signal for each ROI and the surrounding neuropil were extracted from this segmentation map.

The fluorescence signal for each neuron was extracted by computing the mean of the calcium fluorescence within each ROI and subtracting the median fluorescence from the surrounding perimeter of neuropil^15,43. An inferred spike rate (ISR) was estimated from adjusted fluorescence signal with the Vanilla algorithm⁴⁴. A reverse correlation of the ISR to stimulus onset was used to calculate the preferred stimuli^15,43,45,46. Neurons that satisfied three criteria were categorized as visually responsive: (1) the ISR was highest with the optimal delay of 4–9 frames following stimulus onset. This delay was determined empirically for this transgenic GCaMP6s mouse¹⁵; (2) the SNR was at least one s.d. greater than spontaneously active neurons. The signal is the mean of the spiking s.d. at the optical delay between frames 4 and 9 after stimulus onset and the noise of this value at frames −2 to 0 before the stimulus onset or frames 15–18 after it^15,45; and (3) the percentage of responses to the preferred stimulus was at least one s.d. greater than spontaneously active neurons. Visual responsiveness for every neuron was determined independently for each eye. The visual stimulus capturing the preferred orientation and SF was the determined from the matrix of all orientations and SFs presented as the combination with highest average ISR.

The preferred orientation for each neuron was calculated as follows:

$$\text{Orientation}=\frac{\arctan \left({\displaystyle {\sum }_n{O}_n}\times e^{i\times 2\times \pi \times {\theta }_n/180}\right)}{2}$$

The orientation selectivity index (OSI) was calculated as follows:

$$\text{OSI}=\sqrt{{\left({\displaystyle \sum _n{O}_n}\times \sin \left(2{\theta }_n\right)\right)}^2+{\left({\displaystyle \sum _n{O}_n}\times \cos \left(2{\theta }_n\right)\right)}^2/{\displaystyle \sum _n{O}_n}}$$

$O_n$ is a 1 × 6 array of the mean z scores associated with the calculation of the ISR at orientations $O_n$ (0° to 150°, spaced every 30°). The orientation calculated with this formula is in radians and was converted to degrees.

The preferred SF for each neuron was calculated as follows:

$$\text{SF}={10}^{\frac{{\displaystyle {\sum }_k{\text{Sf}}_k}\times {\log }_{10}{\omega }_k}{{\displaystyle {\sum }_k{\text{Sf}}_k}}}$$

${\text{Sf}}_{\kappa }$ is a 1 × 8 array of the mean z scores at SFs $w_{\kappa }$ (eight equal steps on a logarithmic scale from 0.028 to 0.481 cycles per degree). Tails of the distribution were clipped at 25% of the peak response. The tuning width was the full width at half-maximum of the preferred SF in octaves. The percentage of visually responsive neurons with significant responses at each SF was determined by comparing the distribution of ISR values at each SF versus the stimulus blank with a Kruskal–Wallis test with Dunn’s correction for eight comparisons. Neurons with P < 0.01 for a given SF were considered significant responses at that SF⁴⁷.

Binocular matching was measured as the absolute difference in the preferred orientation calculated for visual stimuli presented to the contralateral eye and ipsilateral eye along the 180° cycle^19,21,48.

Neuronal ODI was calculated as (C − I)/(C + I), where C and I are the mean normalized change in fluorescence $\left(\Delta F/F\right)$ for the preferred visual stimulus for the contralateral eye and ipsilateral eye, respectively. In cases where neurons displayed no significant response to visual stimuli provided to one eye, they were considered monocular for the other eye and assigned ODI values of 1 (contralateral) and −1 (ipsilateral)⁴⁷. Summed ODI was calculated by summing the $\Delta F/F$ for the preferred visual stimulus for the all neurons visually responsive to the contralateral eye and ipsilateral eye for each mouse, respectively. The summed ODI per mouse was then calculated as (C − I)/(C + I) for each mouse.

MD

One eye lid was sutured shut on P26–P28 with 6–0 polypropylene monofilament (Prolene 8709H; Ethicon) under brief isoflurane anesthesia for 4 days. The knot was sealed with cyanoacrylate glue. Upon removing the suture, the eye was examined under a stereomicroscope and mice with scarring of the cornea were eliminated from the study.

Multiunit electrophysiologic recordings

Recordings and analysis were performed blind to treatment. Methods were adapted from previously published methods^38,39. In brief, mice were anesthetized with isoflurane (4% induction, 1–2% maintenance in O₂ during surgery). The mouse was placed in a stereotaxic frame and body temperature was maintained at 37 °C by a homeostatically regulated heat pad (TCAT-2LV, Physitemp). Dexamethasone (4 mg kg⁻¹ subcutaneously; American Reagent) was administered to reduce cerebral edema. The eyes were flushed with saline and the corneas were protected thereafter by covering the eyes throughout the surgical procedure with silicone oil (10838, Millipore Sigma). A craniotomy was made over the visual cortex in the left hemisphere and a custom-designed aluminum headbar was attached with Metabond over the right hemisphere to immobilize the animal during recording. Before transfer to the recording setup, a dose of chlorprothixene (0.5 mg kg⁻¹ intraperitoneally; C1761, Sigma) was administered to decrease the level of isoflurane required to maintain anesthesia to 0.8%.

Recordings were made with Epoxylite-coated tungsten microelectrodes with tip resistances of 10 MΩ (FHC). The signal was amplified (model 3600; A-M Systems), low-pass filtered at 3,000 Hz, high-pass filtered at 300 Hz and digitized (micro1401; Cambridge Electronic Design). Multiunit activity was recorded from four to six locations separated by >90 μm in depth for each electrode penetration. In each mouse, there were four to six penetrations separated by at least 200 μm across the binocular region of primary visual cortex, defined by a receptive field azimuth < 25°. Responses were driven by drifting sinusoidal gratings (0.1 cycles per degree (cpd), 95% contrast), presented in six orientations separated by 30° (custom software, MATLAB). The gratings were presented for 2 s of each 4-s trial. The grating was presented in each orientation in a pseudorandom order at least four times, interleaved randomly by a blank, which preceded each orientation once. Action potentials (APs) were identified in recorded traces with Spike2 (Cambridge Electronic Design). Only waveforms extending beyond four s.d. above the average noise were included in subsequent analysis. For each unit, the number of APs in response to the grating stimuli was summed and averaged over the number of presentations. If the average number of APs for the grating stimuli was not greater than 50% above the blank, the unit was discarded.

The OD index (ODI) was calculated as (C − I)/(C + I), where C and I are the average number of APs elicited in a given unit when showing the same visual stimulus to each eye independently. Units were then assigned to one of seven OD categories (1–7) where units assigned to category 1 are largely dominated by input from the contralateral eye and units assigned to category 7 are largely dominated by input.

Behavioral measure of visual acuity

The visual water task was used to measure acuity²³. Training and testing were performed as previously described^22,38. In brief, two monitors were positioned at the wide end of a trapezoidal tank behind clear plexiglass. One monitor displayed a sinusoidal SF grating at 95% contrast, while the other displayed an isoluminant gray screen. The luminance of the two monitors was matched and gamma-corrected with computer software (Eye-One Match 3). Inside the tank, the monitors were separated by a 46-cm divider. The SF was determined relative to the length of this divider. The tank was filled with water and a hidden platform was submerged below the surface of the water in front of the monitor displaying the grating.

Using a low SF (0.1 cpd), mice were trained to swim toward the monitor displaying the grating and hidden platform after a molding phase during which mice gradually learned to swim from a release chute at the back of the tank toward the monitors. During the training phase, when a mouse chose incorrectly, it repeated the trial on the same side until it chose correctly and was then returned to its home cage. For both the training and the subsequent testing phase, mice swam blocks of ten interleaved trials in groups of five for a maximum of four blocks of trials per day.

During the testing phase, the SF was increased in small, sequential increments until an animal consistently fell to 70% accuracy. Starting at 0.1 cpd, mice had to succeed at three consecutive trials before proceeding to the next special frequency, which presented one more complete cycle of the sinusoidal grating. Following the first failure, mice were required to achieve five correct trials in a row or eight out of ten correct trials at each SF before proceeding to the next higher frequency. Once a mouse failed to complete eight out of ten correct trials at a given SF, it was briefly retrained at half that SF to eliminate any potential ‘side bias’. Then, testing resumed at the SF below the original failure. The threshold for visual acuity was established once a mouse exhibited a consistent pattern of performance. Acuity thresholds were estimated as the SF average from three or more failures at adjacent SFs. Throughout the testing phase, any mouse that failed to find the hidden platform on the first try repeated the trial one more time before it was returned to its home cage, whether or not it chose correctly the second time.

Statistics and reproducibility

No statistical methods were used to predetermine sample sizes but our sample sizes are similar to those reported in previous publications^{16,38,39,41,42}. Pre-established exclusion criteria for calcium imaging were any mice for which the quality of the cranial window was compromised such that fewer than ~40 visually evoked neurons could be examined; for multiunit recording, this was extended to any mouse for which at least two full penetrations could not be completed. Litters of mice were chosen at random for each treatment condition. The investigators were not blinded to allocation during data collection because investigator performing data collection also performed procedures to prepare animals for data collection including removal of sutures for MD. Investigators were blinded to allocation during data analysis. All statistical analyses were performed using Prism 8 software (GraphPad Software). All datasets were examined for normality and two-tailed t-tests were only used for data consistent with normal distributions. The nonparametric Kolmogorov–Smirnov test of cumulative distribution and Kruskal–Wallis test with Dunn’s correction for multiple comparisons were used for data that did not conform to a normal distribution. Each mouse was considered an independent experiment except in Fig. 1f where neurons were pooled for comparison.

Extended Data

Extended Data Figure 1. Timeline of PLX treatment and representative coronal images of control and PLX-treated brains.

(a) The PLX treatment was initiated at P14. Experiments to examine the refinement of visual circuitry and plasticity began after P24 as indicated. (b) Representative images of coronal sections of brains from control mice and PLX-treated mice at P18, P21, P28 stained with antibodies directed at Iba-1.

Supplementary Material

Ext Data Movie 1

Supplemental movies 1 and 2. Calcium imaging of in V1 at neuronal resolution. Images were collected at 15.5 frames per second. Each video presents images at 100 frames per second (~7X speed). Movie 1 is from a control mouse. Movie 2 is from a PLX-treated mouse. Videos were down sampled to 480p.

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41593-024-01706-3.

Data availability

The deposited data listed in the Methods table were deposited to Mendeley data at https://data.mendeley.com/datasets/4x96trx9tv/1.

Methods

Reagent or resource	Source	Identifier
Deposited data
Retinal ERG composite data		NA
Calculated response amplitudes from RGCs		NA
Images of Iba1 staining of visual cortex		NA
Images of retinogeniculate tracing		NA
Calculated tuning properties for all neurons from calcium imaging		NA
Calculated OD for all units from multiunit electrophysiologic recordings		NA
Experimental models: organisms and strains
Mouse: B6;DBA-Tg(tetO-GCaMP6s)2Niell/j	The Jackson Laboratory	RRID: ISMR_JAX: 024742
Mouse: B6;Cg-Tg(Camk2a-tTA)1Mmay/DboJ	The Jackson Laboratory	RRID: ISMR_JAX: 007004
Mouse: C57BL/6J-Tg(Thy1-GCaMP6f)GP5.17Dkim/J	The Jackson Laboratory	RRID: IMSR_JAX:025393
Mouse: B6.129P2(Cg)-Cx3cr1tm1Litt/J	The Jackson Laboratory	RRID: IMSR_JAX:005582
Software and algorithms
MATLAB (version 2017b)	Mathworks	https://www.mathworks.com/
Processing2	Processing	https://processing.org/
ImageJ (version 1.54)	NIH	RRID: SCR_003070
QCapture (version 2.9.10)	Quantitative Imaging Corporation
Adobe Photoshop (version 22.0.0)	Adobe	RRID: SCR_014199
Reagents
Anti-Iba1 antibody (Fig. 1), cat. no. 019-19741	Wako Chemicals	RRID: AB_839504
Anti-Iba1 antibody (Fig. 2), cat. no. GTX635363	Genetex	RRID: AB_2888516
goat anti-rabbit conjugated to Alexa Fluor 594, cat. no. 111-585-003	Jackson Immunoresearch	RRID: AB_2338059
goat anti-rabbit conjugated to Alexa Fluor 488, cat. no. 111-545-003	Jackson Immunoresearch	RRID: AB_2338046
CTB 488 conjugate, cat. no. C22841	Invitrogen
CTB 594 conjugate, cat. no. C34777	Invitrogen

NA, not applicable; RRID, research resource identifier.