Visual Recovery Reflects Cortical MeCP2 Sensitivity in Rett Syndrome

Authors: Alex Joseph Simon, Nathalie Picard, Valeria d'Andrea, Enchi Chang, Joseph Leffler, Eleonora Centofante, Matthew Taylor, Francesca Bardi, Francesca Cavicchiolo, Takao K. Hensch, Stefano Panzeri, Chinfei Chen, Michela Fagiolini

Abstract

Objective

Rett syndrome (RTT) is a devastating neurodevelopmental disorder with developmental regression affecting motor, sensory, and cognitive functions. Sensory disruptions contribute to the complex behavioral and cognitive difficulties and represent an important target for therapeutic interventions. Although genetic medicine‐based therapies targeting MeCP2 have successfully restored motor and respiratory functions in animal models, their ability to reverse sensory deficits across levels of the visual pathway remains largely unexplored.

Methods

Using genetically reversible mouse models of MeCP2 deficiency (Mecp2^ stop/y ^ and Mecp2^ stop/x ^), we applied advanced electrophysiological, anatomical, and behavioral techniques to evaluate visual function, a critical sensory domain impaired in both animal models and RTT patients.

Results

In Mecp2 ^ stop/y ^ mice, initiating MeCP2 expression after postnatal day 35 (P35) reversed progressive cortical dysfunction, prevented thalamic circuit disorganization, and restored visual function, despite some remaining cortical anatomical abnormalities. Even in fully regressed adult Mecp2 ^ stop/x ^ heterozygous female mice, MeCP2 reactivation was sufficient to reduce the symptoms.

Interpretation

These findings highlight the remarkable sensitivity of cortical circuits to MeCP2 expression in both developing and mature brain. Importantly, restoring just 60%–70% of MeCP2 protein levels was sufficient to rescue sensory functions, even after the onset of regression. This underscores the transformative potential of genetic medicine‐based therapies in RTT, suggesting that even partial restoration of MeCP2 can meaningfully improve sensory processing and quality of life for patients.

Introduction

Loss or partial reduction in the expression of the X‐linked gene, methyl‐CpG‐binding protein 2 (MECP2) causes Rett Syndrome (RTT), a rare neurodevelopmental disorder affecting mainly girls [1, 2]. Groundbreaking studies in animal models have demonstrated that restoring MeCP2 expression after the onset of RTT phenotype reverses some symptoms, including motor deficits [3, 4, 5]. These pivotal results have launched therapeutic efforts to deliver healthy copies of the MECP2 gene to compensate for their loss in individuals with RTT [6, 7, 8, 9, 10], and clinical trials are now underway in multiple countries (NCT05898620, NCT06152237, NCT05606614). Yet, whether genetic intervention can rescue impairments in sensory modalities, cognition, communication, and social interaction remains unknown. RTT is often diagnosed after neurobehavioral regression begins (median age 2.7 years‐old) [11, 12], long after the onset of neuronal network maturation, which is even accelerated in the absence of MeCP2 [13]. This timing raises crucial questions about the potential effectiveness of MeCP2 expression in adulthood, as the brain's ability to rewire declines with age [14, 15].

Here, we explored whether the functional and structural visual system deficits caused by the absence of MeCP2 throughout postnatal maturation could be rescued once plasticity had diminished in adulthood. Individuals with Rett syndrome show impaired sensory modalities that are crucial for accurately perceiving and interacting with their environment, including a decline in both auditory and visually evoked potential (VEP) amplitudes and reduced visual acuity, compared to typically developing individuals [16, 17, 18, 19, 20]. Mouse models of RTT faithfully recapitulate the overall disorder phenotype, including the abnormal sensory function seen in patients [17, 21, 22, 23, 24]. Mecp2 null males exhibit RTT phenotypes that are robust, highly reproducible, and closely mimic the onset and progression of the human condition, and therefore are considered a valid model for studying RTT. Their visual acuity develops to typically mature levels by postnatal day P30, but then rapidly regresses after P35‐40 as cortical circuits fall silent in adulthood [21, 25]. In the adult primary visual cortex (V1), VEP waveforms are abnormal, pyramidal cell activity in response to visual stimulation is sharply reduced, and visual acuity is strongly impaired [17, 21, 25]. This reflects in part the accelerated maturation of fast‐spiking, large‐basket, parvalbumin‐positive (PV) inhibitory cells that are hyper‐connected onto pyramidal cells [13, 21, 25, 26], weakly connected to each other [13, 27], and receive increased intracortical and thalamocortical excitatory input [13, 21, 27, 28]. The timing and progression of dysfunction in the thalamus and visual cortex remain poorly understood, limiting insight into how MeCP2 deficiency affects sensory circuits and their potential for recovery. To address this gap, we first mapped visual circuit abnormalities in the thalamus and visual cortex of male Mecp2‐deficient mice. We then investigated in both male and female MeCP2 mouse models whether ubiquitously restoring gene expression in adulthood (using the tamoxifen‐inducible Cre‐ER system [4]) would alleviate symptoms, reverse cortical abnormalities, and rescue sensory modality (vision) later in life (after the peak of the critical period of plasticity [13, 29]). We found progressive activation of MeCP2 expression in young adult males rescued cortical physiology, prevented subcortical deficits, and restored functional vision. Furthermore, activation of MeCP2 expression in fully regressed, adult MeCP2 heterozygote female mice reduced the RTT phenotype in a dose‐dependent manner. Overall, our findings support the potential for genetic re‐expression as a therapeutic approach to mitigate disease symptoms and enhance sensory functions, which could ultimately improve the quality of life for patients.

Methods

All procedures were approved by the Boston Children's Hospital Institutional Animal Care and Use Committee. Mice were housed on a 12‐h light/dark cycle with ad libitum food and water. Experimenters were masked to genotype and treatment. Mecp2‐null and Mecp2 ^stop/y^ and Mecp2 ^stop/+^ mice were originally generated by A. Bird and colleagues [4, 30]. The Cre‐ER mouse (S. Hayashi and A.P. McMahon [31]) allows Cre activation by tamoxifen (20 mg/mL (males) or 6 mg/mL (females) in corn oil, 100 mg/kg body weight, intraperitoneally (i.p.) as previously shown by Guy et al. [4] and Lang et al. [5]). Unless stated otherwise, wildtype (WT), Cre‐ER, WT+TMX, and Cre‐ER+ TMX age‐matched littermates were pooled together as WT. Per protocol [3, 4], IACUC recommendations, and due to the constraint of the animals' health, only TMX‐injected animals were studied to account for TMX toxicity and isolate the Mecp2 expression effect.

Phenotypic Scoring

Presence of RTT‐like symptoms was assessed with the 12‐point phenotypic scoring as previously described [4].

Western Blot

Protein lysates were processed for Western Blot analysis as previouslydescribed [25]. For primary antibodies, we used a rabbit polyclonal antibody directed to the C‐terminal of MeCP2 (EMD Millipore 07‐013, 500) and a mouse antibody to GAPDH (Abcam ab8245, 20,000). For secondary antibodies, we used goat anti‐mouse IgG (HRP) and goat anti‐rabbit IgG (HRP) (Abcam). Fluorescent signal was detected by photographic film and analyzed by densitometry with Fiji.

Immunofluorescence

Immunofluorescence experiments were performed as described in our previous works [21, 26]. Primary antibody rabbit anti‐MeCP2 (c/o M. Greenberg Lab, 2000 dilution), guinea pig anti‐parvalbumin (Swant, 1000 dilution), mouse anti‐GAD65 (Developmental Studies Hybridoma Bank, 1000), and NeuroTrace Nissl (Invitrogen, 500) were prepared in 3% normal goat serum, 0.5% Triton X‐100. Secondary antibodies were goat anti‐rabbit Alexa 647 (Invitrogen, 800 dilution) and goat anti‐guinea pig Alexa 594 (Invitrogen, 500 dilution) in PBS.

Confocal Microscopy

Brain sections were analyzed using laser scanning confocal microscope (Olympus FluoView FV1000) in multi‐channel mode. Analysis was masked and performed on 3–5 mice per genotype. Mean pixel intensity of PV or MeCP2 in nuclei was measured in primary visual cortex images (1024 × 1014) at 20×. For MeCP2, 100 nuclei per slice were analyzed. Parvalbumin cell area and presynaptic structures were assessed using a 100× oil‐immersion objective. PV puncta were quantified from parvalbumin, GAD65, and DAPI‐stained images. PV puncta and NeuN‐positive cell density were analyzed using ImageJ's “analyze particles” (0.05–10 mm^2^) and Volocity software.

Single‐Cell Electrophysiology In Vivo and Visual Evoked Potentials (VEP)

Electrophysiological recordings in vivo were performed as previously described by the lab [21, 25] under Nembutal (50 mg/kg, i.p.) anesthesia and chlorprothixene (0.2 mg, i.m.). Cortical activity in V1b and dLGN was recorded using multichannel probes (A1x16‐3 mm 50‐177; Neuronexus Technologies, Ann Arbor, Michigan), and VEP was measured with a tungsten electrode (1.5 MΩ, FST) placed in layer 4 of V1b. Local Field Potential (LFP) analysis was performed between 5 and 100 Hz across cortical layers.

Electrophysiology In Vitro

Whole‐cell voltage‐clamp recordings of relay neurons were performed in dLGN slices as previously described [32, 33]. The optic tract was stimulated with paired pulses (50 ms interval, 20 s inter‐trial interval), and synaptic responses were measured at −70 mV (AMPA) and +40 mV (NMDA). If a response occurred at one potential but not the other, the trial was repeated. Failures at both potentials led to incremental stimulus intensity increases (0.25 μA) until the minimal response for both AMPAR and NMDAR was reliably evoked. Retinal ganglion cell innervation was estimated using the fiber fraction method [33].

Eye‐Specific Segregation

Mice anesthetized with 2% isoflurane were injected intraorbitally with 2–3 μL of a 2% solution of cholera toxin β subunit conjugated with Alexa 488 (Green) or 594 (Red) (Invitrogen, Carlsbad, CA). After 2–4 days, mice were deeply anesthetized with Nembutal (50 mg/kg i.p.) and transcardially perfused. Brains were cut (1–3 slices/animal) and analyzed using previously described threshold‐independent quantitative measures of eye‐specific layer segregation [34].

Optomotor Visual Acuity Task

Behavioral visual acuity was assessed using the optomotor task (CerebralMechanics Inc.) as described previously [35]. Mice were placed on an elevated platform, and head tracking of the rotating grating indicated visual detection. The highest tracked spatial frequency was recorded as the acuity threshold.

Computation of the E/I Ratio

E/I was defined as the ratio of the median of firing rates of regular spiking (RS) cells (fr ~ RS ) to the median of firing rates of FS cells (fr ~ FS ). We used a non‐parametric permutation to test the difference between E/I ratios of two groups. We generated a null hypothesis distribution of E/I values by randomly permuting the group label for each cell and computing the E/I ratios distribution for 1000 different random permutations of each group. The one‐tailed p‐value was computed comparing the E/I value of real, unpermuted data with those of the null hypothesis permuted distribution. Error bars and distributions of values of frRS and frFS, and E/I ratio values were obtained by bootstrapping (N = 1000 bootstraps) [36]. Bootstrap values were only used for data visualization of error bars but were not used for the computation of significance of difference in E/I across groups, which was based on the permutation test described above.

Spectral Analysis of LFPs

LFP analysis was performed between 5 and 100 Hz across cortical layers. For each trial, the post‐stimulus period was divided into three non‐overlapping segments of 1 s length. We excluded the first post‐stimulus 1 s segment because of the non‐stationary signal. We checked that the spectral estimates of the other two segments were very similar, so we pooled data from both the second and third window in all spectral analyses. We estimated the power spectral density (psd) in each segment by using the multi‐taper technique [37]. We used n = 1024, the number of points of the Fourier transform, and multitaper parameter LW = 2. Psd of each trial was normalized to the total power, and then we averaged psd across trials independently for each channel. Psd plots report median and 68% bootstrap confidence intervals. To evaluate total power in specific bands, we summed psd values in selected frequency bands.

Results

Since RTT onset occurs early in postnatal life, we focused the initial phase of our study on the Mecp2 male model, which exhibits disease progression like humans but faster than MeCP2 heterozygous female mice [4]. We further leveraged the extensive analysis of the visual pathway in the Mecp2 null model [21, 24, 25, 26, 28, 35, 36, 37]. We specifically used the Mecp2 ^stop/y^mouse previously described and characterized as hypomorphic [3, 4, 5] (Figure 1A). They exhibited a milder phenotype than Mecp2 null mice, a prolonged progression of the disorder, as shown by their survival curve shifted toward older ages (Figure 1B; median survival of 94.5 vs. 53 days in Mecp2 ^ stop/y ^ and null, respectively) [21, 24] and delayed RTT onset (> P45 as compared to P30 in null, Figure 1C). Mecp2 ^stop/y^ mice exhibited an increase in all six symptoms used to determine the overall RTT severity score, ultimately resulting in a severe phenotype by P110 (Figure 1D) [3, 4, 5].

Cortical Abnormalities Precede the Onset of RTT Symptoms While dLGN Remains Intact

While the visual pathway is well studied in Mecp2 null mice [13, 21, 24, 25, 26, 35, 37, 38], the precise timing of these deficits and the sequence of disruptions between the dLGN and V1 remain unclear. Understanding these aspects is crucial for mapping the trajectory of visual pathway disturbances in relation to disease progression and identifying potential therapeutic targets. Therefore, we first examined the dLGN at P30 before the onset of RTT symptoms. We recorded excitatory postsynaptic currents (EPSCs) from relay neurons in dLGN in vitro from WT and Mecp2 ^stop/y^ mice by gradually increasing the intensity of optic tract stimulation as described previously [39] (Figure 1E). Single fiber synaptic strength (quantified by the peak single‐fiber AMPA and NMDA receptor EPSC amplitudes to minimal stimulation) did not significantly differ between Mecp2 ^stop/y^ mice and their WT controls at P30 (Figure 1F,G). Maximal AMPA and NMDA currents were also not impaired at this age in the Mecp2 ^stop/y^ mice (Figure 1H). To address whether synapse pruning during the strengthening of retinogeniculate synapses was affected, we compared fiber fraction ratios, which estimate the relative degree of retinal ganglion cell convergence [40]. In contrast to our previous findings in Mecp2 null mice [35], synaptic pruning was intact in male Mecp2 ^stop/y^ mice at P30 (Figure 1I).

We also conducted extracellular recordings of thalamic relay cells in anesthetized mice to investigate potential functional abnormalities in vivo. Using multi‐channel probes, we quantified both evoked and spontaneous responses to moving gratings at low spatial frequency (Figure 1J). At P30, we observed no differences in the spontaneous activity of dLGN neurons in Mecp2 ^stop/y^ mice (Figure 1K). Additionally, the subset of Mecp2 ^stop/y^ relay cells responsive to high‐contrast gratings displayed spontaneous and maximal evoked activities comparable to those of WT mice (Figure 1L).

We then recorded in vivo spontaneous and evoked activities from putative excitatory regular spiking (RS) pyramidal neurons across V1 cortical layers and the reliability of the visually evoked responses to repeated presentations both in Mecp2 ^ stop/y ^ mice before the onset of RTT phenotype and in Mecp2‐null mice at the onset of symptoms (P30) under anesthesia (Figure 2A). Both models exhibited a decrease in spontaneous and evoked spiking activity of regular spiking putative pyramidal (RS) cells compared to WT (Figure 2B,C). Additionally, response reliability to the preferred orientation was impaired, as reflected in an increased coefficient of variation (Figure 2D).

At the anatomical level, cortical thickness was diminished in Mecp2 ^stop/y^ compared to WT mice (Figure 2E) [41]. We also observed an increase in PV intensity and a higher number of PV puncta onto the somata of pyramidal cells in agreement with the PV hyper‐connectivity described previously in Mecp2 null mice (Figure 2F) [13, 21, 25, 26]. Together these results indicate that cortical circuits are already disrupted before the onset of symptoms while the dLGN circuits remain intact early in the disorder. This suggests that the cortex is extremely sensitive to MeCP2 expression and raises the question of whether late gene activation (after the peak of plasticity in visual cortex [13, 29]) would be sufficient to prevent the onset of subcortical deficits, stop further regression, or even reverse the cortical impairments.

Late Expression of MeCP2 Reverses RTT Phenotype and Cortical Dysfunction in Mature Mecp2

^ stop/y ^ Males

To activate MeCP2 expression, we crossed Mecp2 ^stop/+^ heterozygous female mice with transgenic mice expressing an estrogen receptor/cre‐recombinase transgene (CreER) and administered tamoxifen to induce MeCP2 expression in adulthood (Figure 3A) [4]. Starting at P35, tamoxifen was given over 25 days, allowing an increase in MeCP2 protein levels to 60%–70% of WT level in both cortical and subcortical circuits by P75 (Figure 3C). MeCP2 expression was expressed in ~60% of neurons by P105 (Figure 3D). Mecp2 expression activation extended survival (Figure 4A), halted RTT progression (Figure 4B), and prevented the emergence of new symptoms, except for hindlimb clasping (Figure 4B bottom panels). These mice are referred to as the LateMecp2 group.

Remarkably, following MeCP2 expression, spontaneous and evoked activities of pyramidal RS neurons were higher than in Mecp2 ^ stop/y ^ mice and comparable to WT activity levels at P110 (Figure 5A,B, left). The reliability of neuronal responses was also restored, with a decrease in the coefficient of variation returning to WT values (Figure 5C, left). A direct comparison of the cortical activity of the Mecp2 ^ stop/y ^ group at P30 versus P110 showed that without treatment, spontaneous, evoked activities, and the reliability of pyramidal neurons further declined with progression of the disorder (Figure 5A–C, right). In LateMecp2 mice, the evoked activity and reliability were improved compared to P30 values in Mecp2 ^ stop/y ^, revealing that MeCP2 expression not only stopped the regression but also reversed such deficits (Figure 5A–C, right).

Late Expression of MeCP2 Ameliorates Vision Deficits

To directly evaluate the impact on vision, we quantified VEP amplitude in a separate cohort of adult LateMecp2 mice viewing alternating black and white sinusoidal bars of increasing spatial frequency. We then determined spatial resolution (acuity), a robust translatable measure in human subjects that is impaired in RTT subjects [16, 20]. The VEP negative peak amplitude (N1) and acuity were reduced in P110 Mecp2 ^stop/y^ mice, similar to Mecp2 Het females (Figure 6A–C) [16, 20]. LateMecp2 mice displayed a greater N1 amplitude associated with a higher VEP acuity compared to Mecp2 ^stop/y^ mice; their amplitude and acuity were no different from WT (Figure 6B,C).

Using the optomotor task (OPT) [42] (Figure 6D, left), we behaviorally assessed vision (visual acuity), which is known to decline after P30 in Mecp2 null mice [21]. OPT assesses subcortical visual processing by engaging brainstem and pretectal circuits that mediate reflexive head and eye movements in response to large‐field visual motion. It provides a sensitive, non‐invasive measure for tracking visual performance and overall visual system integrity over time [38]. In Mecp2 ^ stop/y ^ mice, OPT acuity was decreased at P45 and was severely impaired by P110 (Figure 6D, right). Late expression of MeCP2 was sufficient to slow this decline (Figure 6D). We also examined whether Late expression could prevent eye‐specific de‐segregation in the thalamus, which begins at P75 in Mecp2 ^stop/y^ mice (vs. P46–51 in the Mecp2‐null [35]) and reaches statistical significance by P100–110, when Mecp2 ^stop/y^ mice were behaviorally fully symptomatic (Figure 6E,F). Our results showed no dLGN impairment in the LateMecp2 group at P75 and P100 (Figure 6E,F). Instead, LateMecp2 expression failed to rescue cortical thickness (Figure 6G).

Abnormalities in Brain Oscillations Partially Rescued by Late Expression of MeCP2

To assess the degree of functional recovery in visual processing at the systems level beyond the single‐neuron scale, we recorded Local Field Potentials (LFPs), a key indicator of network activity dynamics and the synchronization of oscillatory structures. We analyzed LFP power across a broad frequency range, including delta/theta (1–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and low‐ (30–55 Hz) and high‐gamma (65–100 Hz) bands, within the superficial layers of V1 [43]. At P30, spontaneous oscillations were already disrupted in mutant mice and shifted toward higher frequencies compared to WT (Figure 7A). Specifically, the beta, low‐, and high‐gamma bands showed an increase in power while the lower frequency delta/theta band was reduced. No changes were detected in the alpha range at this age. By P110, Mecp2 ^stop/y^ mice still exhibited an abnormal spectral profile, now also displaying increased alpha power (Figure 7B). Interestingly, LateMecp2 mice exhibited an intermediate spectrum that was different from both WT and mutant beta power recovered to WT values, the delta/theta range was intermediate, and alpha, low‐, and high‐gamma remained indistinguishable from those in the Mecp2 ^stop/y^ mice (Figure 7B).

Partial Rescue of E/I Balance by Late Expression of MeCP2

The balance between recurrent excitation (E) and inhibition (I) is crucial for regulating network dynamics, generating oscillations, and processing information. To investigate this, we calculated the E/I firing rate ratio for each group. Bootstrap analysis of fast spiking (FS) inhibitory cells revealed an increase in evoked firing rate between P30 and P100 in WT (Figure 7C) while pyramidal cell activity did not change over time (2.27 ± 01.03 spikes/s and 2.52 ± 0.84 spikes/s; n = 72 and 203 cells; 6 and 4 mice, respectively; Mann–Whitney test, ns). This resulted in a net decrease in E/I ratio between P30 and P100 due to the increase in inhibition (Figure 7D). In Mecp2^stop/y^, the analysis revealed an increase in FS firing rate at P30 but a decrease at P100 when compared to WT (Figure 7C). However, due to the robust reduction of excitatory activity at both ages, the E/I ratio was lower in MeCP2 ^ stop/y ^ compared to WT at both ages, suggesting a shift in E/I balance toward inhibition (Figure 7D). In LateMecp2 mice, the activity of FS cells decreased compared to MeCP2^stop/y^, driving the E/I ratio to an intermediate state between WT and MeCP2 ^ stop/y ^ mice levels (Figure 7A–D). Anatomically, PV intensity in large‐basket cells was renormalized, while their connectivity onto pyramidal cells remained higher than in WT even after MeCP2 activation (Figure 7E).

Beyond preventing behavioral regression (Figure 4), our data show that even a late and partial activation of MeCP2 expression starting well after cortical deficits arise was still sufficient to reverse V1 functional abnormalities and improve sensory processing in Mecp2‐deficient mice.

Late‐Stage MeCP2 Expression Reduces Rett Syndrome Phenotype in Female Mecp2 Lox‐Stop Mice

Lastly, we assessed whether late‐stage genetic intervention in fully symptomatic females could result in functional recovery. Mecp2 ^ stop/+ ^; Cre‐ER positive and Mecp2 ^ stop/+ ^; Cre‐ER negative females were followed from 3 mos to 11 mos old to track the progression of the disorder (Figure 8A, n = 6 CreER‐ and 5 CreER+). In Mecp2 ^ stop/+ ^ females, the RTT score increased over time and reached higher values compared to WT (pooled littermates and others at 8 to 11 mos old) by the time of treatment initiation (Figure 8B, n = 11 Mecp2^ stop/x ^ females and 8 WT). Before treatment, Mecp2 ^ stop/+ ^ mice also exhibited lower visual acuity compared to WT females (mean acuity = 0.21 ± 0.04 vs. 0.33 ± 0.02 cpd, p = 0.03 in n = 11 Mecp2 ^ stop/+ ^ and 8 WT mice, respectively). These findings are consistent with observations in previous mouse models and RTT patients [17].

Mecp2 ^ stop/+ ^; CreER‐negative mice, and five Mecp2 ^ stop/+ ^; CreER‐positive females were then treated with tamoxifen for 1 week, following the protocol described by Lang et al. [5] and monitored for 6 weeks (Figure 8C). At the endpoint, brain tissues were collected, and Western blot analysis confirmed that MECP2 expression in CreER‐positive mice was, on average, close to WT levels, but showed inter‐animal variability (range: 0.55–1.35, normalized to WT). Such variability in reactivation efficiency, including occasional overexpression, is an expected characteristic of tamoxifen‐induced CreER systems and is consistent with previous reports [3, 4]. In contrast, Mecp2 ^ stop/+ ^; CreER‐negative mice (n = 5) exhibited only 50% of WT expression (Figure 8D).

Behaviorally, RTT score progressively decreased over time in Mecp2 ^ stop/+ ^; CreER‐positive mice and was lower compared to Mecp2 ^ stop/+ ^; CreER‐negative mice at both 3‐ and 4‐weeks post‐treatment, as determined by the Kruskal‐Wallis test (Figure 8E). Analysis of ∆RTT (endpoint—baseline) score using bootstrap analysis revealed a decrease in RTT values in Mecp2 ^ stop/+ ^; Cre‐positive compared to Cre‐negative females (∆RTT = −1.44 [−2.5, −0.25] vs. +0.57 [−0.20, +1.45], respectively, mean [95% CI] bootstrap analysis, Figure 8F). Furthermore, a strong correlation between ∆RTT and Mecp2 expression was observed in Mecp2 ^ stop/+ ^; CreER‐positive females (Figure 8G, linear regression, p = 0.023).

Notably, Cre‐positive females demonstrated a trend toward improved visual acuity, as measured by the OPT task, at 6 weeks post‐treatment (0.26 ± 0.04 vs. 0.19 ± 0.05 in Cre‐positive vs. Cre‐negative mice, ns, data not shown). Together, these findings highlight the efficacy of Mecp2 re‐expression in reducing core RTT phenotypes, even at advanced stages of disease progression.

Discussion

Animal studies have shown that LateMecp2 expression can reverse RTT phenotypic score, improve respiratory and motor deficits, and ameliorate cortical epileptiform activity [3, 4, 5, 7, 8]. However, it remained unclear whether deficits in sensory processing—previously reported in Mecp2‐deficient mice and individuals with RTT [17, 20, 21, 26, 39, 41]—could be reversed, and to what degree. Our study addressed two (1) how the loss of MeCP2 impacts the postnatal development of visual pathways in a RTT animal model; and (2) whether the associated functional and structural deficits could be reversed once the natural window of plasticity had diminished. We found (1) reduced Mecp2 expression disrupted visual cortex function even before the onset of RTT behavioral phenotype; (2) impairments in the first stage retino‐thalamic circuits also followed later in adulthood; (3) restoring 60%–70% of MECP2 levels was sufficient to reverse cortical dysfunction, improve visual function, and prevent subsequent regression of subcortical circuits; and (4) Mecp2 re‐expression was effective in rescuing core RTT symptoms in females, even at advanced stages of disease progression.

Our data brings us closer to unlocking the mechanisms underlying the progression of the disorder and the degree of efficacy of possible genetic medicine‐based therapies. The onset of symptoms in children with de novo mutations in Mecp2 mainly appears between 6 and 18 months of age, with the disorder being diagnosed around 2–3 years after birth (median age 2.7 years‐old) [15, 17, 20, 40, 42]. During these early years, the brain undergoes experience‐dependent maturation and refinement, followed by the consolidation of neuronal circuits. While female mouse models share with RTT girls the mosaicism of MECP2 expression (only one copy is affected, generating a mix of healthy and unhealthy cells), they exhibit RTT phenotypes with a degree of variability higher than in mutant males and develop symptoms later than humans (at 8–12 postnatal weeks in Mecp2 het and at 11–17 postnatal weeks in Mecp2 ^ stop/+ ^ female mice, equivalent to ~20 and ~26 years of age in humans). Therefore, mutant male models are crucial to address questions related to the early development of the disease.

Compared to Mecp2 null mice, Mecp2 ^ stop/y ^ mice still express minimal levels of MeCP2 (Figures 1 and 3) [3, 4, 5], resulting in a delayed/slower progression of the disorder. This prolonged regression provided us with a unique opportunity to elucidate the sequence of circuit dysfunction. Deeper insight into how abnormalities emerge across circuits and structures helps to determine whether impairments arise independently within microcircuits or propagate sequentially in a macroscopic feedforward or top‐down manner across circuits while also identifying potential differences in sensitivity to MeCP2 loss.

Prior to RTT symptom onset, P30 in Mecp2 ^ stop/y ^ mice, retinogeniculate synapses appeared largely intact, suggesting that subcortical circuits do not require high levels of MeCP2 for their initial wiring. In contrast, cortical abnormalities were already pronounced, with a dramatic reduction in both spontaneous and evoked activity of pyramidal neurons (Figure 2). We also identified aberrant spectral power (Figure 7A), with suppressed rhythms in the thalamocortical loop (< 8 Hz) [43, 44, 45, 46] and enhanced local intracortical E/I loops (> 12 Hz) [47, 48, 49] but not long cortico–cortical networks (8–12 Hz) [50]. Altogether, these results point to a higher sensitivity of cortical circuits to MeCP2 levels at P30.

Only when full regression was reached (P110) did thalamic circuits desegregate retinal axon topography alongside further cortical decline (Figures 5 and 6E,F) when all networks, including cortico‐cortical loops, showed a disruption (Figure 7B). This sequence of impairments aligns with previous studies in Mecp2 null mice where PV circuit hyperconnectivity and altered cortical network function are already present after eye opening (P15) [13, 21], preceding thalamic synaptic circuit alterations and later eye‐specific layer disruptions in the dLGN [32]. This “top‐down” sequence of circuit dysfunction is also consistent with findings that disruption of cortical activity during development can alter connectivity of thalamic circuits [51].

Activating MeCP2 expression was sufficient to slow down the progression of the disorder (Figure 2A) and extend lifespan (Figure 2B). It also restored cortical pyramidal cell activity and response reliability in adulthood (Figure 5), improved visual acuity, prevented further impairment in the dLGN (Figure 6), and partially reversed brain oscillations with a full return of the 12–30 Hz band to WT levels and an intermediate state for low frequencies and high gamma rhythms (Figure 7B). At the anatomical level, cortical thickness remained reduced (Figure 6G), and PV circuits were still hyper‐connected even at P110 (Figure 7E). Experimental and theoretical studies have shown PV networks play a prominent role in synchronizing network activity in the gamma band (30–100 Hz) [46, 47]. Thus, abnormal PV connectivity could explain the incomplete rescue of gamma power at P110 (Figure 7). Interestingly, we found that older LateMecp2 mice (P160) exhibited a reduced and renormalized gamma power (2 ± 0.4 vs. 8 ± 2 at P160 vs. P110 LateMecp2, respectively; p = 0.01; WT: 4 ± 0.4, n = 3 at P160, 6 at P110, and 4 WT), suggesting that PV circuits may also be subject to rewiring but just at a slower time scale than pyramidal cells.

Our findings in Mecp2 ^stop/+^ female mice demonstrate that late‐stage MeCP2 re‐expression can partially reverse core RTT phenotypes, even in fully symptomatic animals. Despite the progressive nature of RTT, CreER‐positive females showed improvements in RTT scores and visual acuity following tamoxifen‐induced MeCP2 restoration (Figure 8). Recent studies in CDKL5 deficiency disorder [52] suggest that late‐stage gene re‐expression may offer only limited therapeutic benefit. On the contrary, our results demonstrate that MeCP2 reactivation can still restore a significant amelioration of cortical function and behavioral phenotypes in late‐stage RTT. This implies that RTT‐associated neural circuits may retain some degree of plasticity, enabling at least partial recovery even after symptom onset. However, it remains uncertain whether further improvements will be achieved as circuits may still need some time to rearrange after P110 to lead to long‐term effects. Importantly, the correlation between MeCP2 expression levels and functional recovery suggests a dose‐dependent effect essential for optimizing maximal therapeutic benefit.

Overall, our study supports the potential of MeCP2 re‐expression as a viable therapeutic strategy for male and female RTT patients and supports the notion that even partial restoration of Mecp2 can lead to measurable improvements in visual function. Functional recovery in both behavioral and sensory domains, both early and late in the disorder, emphasizes the plasticity of MeCP2‐deficient circuits across stages, leaving open the possibility for even greater recovery starting prior to symptom onset. Further research is essential to refine genetic medicine‐based therapies, evaluating the optimal level and timing to maximize therapeutic benefits for individuals with RTT.

Author Contributions

A.J.S.: conception and design of the study, acquisition, and analysis of in vivo electrophysiology, immune and behavioral data, drafting initial manuscript and figures. N.P.: acquisition and analysis of single unit and LFP data, preparing the figures, writing the manuscript and its revision. V.d.A.: analysis and computation of single unit and LFP data. E.C.: acquisition of behavioral and immunohistochemistry data. J.L.: Tracer eye injections and eye‐specific segregation data analysis. E.C.: VEP data acquisition and analysis. M.T.: Tracer eye injections and eye‐specific segregation data analysis. F.B. and F.C.: acquisition of behavioral data and Western Blot analysis in Mecp2 heterozygous females. T.K.H.: drafting of the manuscript and revisions. S.P.: analysis and computation of single unit and LFP data and drafting of manuscript. C.C.: conception and design of the study, in vitro data acquisition and analysis, and drafting of manuscript. M.F.: conception and design of the study, drafting of figures, writing of the manuscript, and revisions.

Conflicts of Interest

The authors declare no conflicts of interest.