Preclinical Milestones in MECP2 Gene Transfer for Treating Rett Syndrome Open Access

Rett syndrome (RTT) is an X-chromosome-linked neurodevelopmental disorder caused by inactivating mutations in the transcription regulator methyl-CpG-binding protein 2 (MeCP2) [1]. Symptoms of RTT include intellectual disability, speech abnormalities, seizures, microcephaly, stereotypy, gait abnormalities, and respiratory abnormalities [1, 2]. In mice and humans, symptoms of RTT partially overlap with those of a severe reciprocal disorder known as MECP2 duplication syndrome (Fig. 1) [1, 2]. These genetically opposite yet debilitating disorders underscore the central question of gene therapy for RTT: Can MECP2 gene transfer be both effective and safe?

Gene therapy for RTT is conceptually challenging because the gene dose sensitivity of MECP2 is complicated by the mosaic expression of endogenous MeCP2 in female patients (there are far fewer documented cases of male RTT patients) [3, 5, 6]. Due to random X chromosome inactivation, approximately half of the cells in a female RTT patient express mutated MeCP2 (skewed inactivation can also happen [1, 7]); the remaining cells express WT MeCP2 (Fig. 1) [1, 3]. Even if an unregulated gene therapy were used to treat defective cells successfully in mosaic RTT patients, the expression of exogenous MeCP2 in transduced WT cells might be toxic.

This review summarizes how researchers have approached the goal of achieving safe and effective gene therapy in mice modeling RTT. Tables 1 and 2 list outcomes for gene transfer studies in Mecp2^−/y knockout (KO) mice treated during adolescence and do not list details for vectors administered only to mouse neonates (e.g., the first vector design published by Gadalla et al. [9]). Note that neonatal administration in rodents translates approximately to the second to third trimester in humans (preterm infancy), a treatment age that is not yet under consideration for MECP2 gene therapy clinical trials (NCT05606614, NCT06152237, and NCT05898620 are evaluating patients aged 12 years and older, children aged 5–8 years, and children aged 4–10 years, respectively) [20, 21, 25‒28]. Preterm infancy in humans precedes the median diagnosis age (which occurs after symptom onset) in patients by approximately 3–4 years (classical and atypical RTT, respectively) [29]. For simplicity, this review names most MECP2 gene therapy paradigms (route, capsid, and viral genome combinations) by numbers (#i–xiii) assigned in consistent register in Tables 1 and 2. Other experimental considerations (e.g., choice of single-stranded or self-complementary inverted terminal repeats, injection route, and treatment age and rationale for self-complementary adeno-associated viral vector serotype 9 [scAAV9]) are beyond the scope of this review and have been described previously [2, 10, 15].

There are over two dozen mouse models of RTT, but the Mecp2^−/y KO model is the most commonly used model among MECP2 gene therapy publications [1, 4, 30, 31]. In contrast to the female Mecp2^+/− mouse model, the male KO model’s severity of symptoms and the age of symptom onset align more closely with those observed in female RTT patients [32].

Results for the earliest or “first-generation” MECP2 gene therapies in KO mice were promising. Studies by Garg et al. [8] and Gadalla et al. [9] formed the foundation of this field (Table 2). Together, they demonstrate efficacy at molecular, compartmental, and organismal levels, as well as across male and female mouse models and across treatment ages [8, 9]. After adolescent administration in KO mice, MECP2 vectors improved somal size, GABA expression, and/or survival (#i–ii; Table 1) [8, 9]. After neonatal administration in KO mice, a MECP2 gene therapy with a nonphysiological chicken β-actin (CBA) promoter also improved survival, treadmill performance, nuclear size, and aggregate severity scores (referring to a standardized gross health assessment biased toward limb and motor phenotypes [33]) [9]. The goal of the field thereafter has been to develop a gene therapy that overcomes toxicity while meeting the examples of Garg et al. [8] and Gadalla et al. [9] on diligent efficacy assessments across readouts, models, and/or treatment ages.

To be clear, this review summarizes the earliest signs of toxicity, not for the sake of being critical but to set the stage for subsequent work in which neurological and hepatic toxicity were diligently elucidated [2, 10, 11, 22]. To begin, the absence of vector-treated WT control mice for #i (MeP730 promoter; in Tables 1, 2) precluded a safety assessment [8]. Limited toxicity was documented for the first MECP2 gene therapy that featured a more physiologically relevant promoter variant (MeP229 promoter; #ii in Table 2) [9]. This toxicity included quantitative liver dysfunction after adolescent administration (WT mice) and anecdotal overexpression-related toxicity after neonatal administration in KO mice. Surprisingly, in contrast to the neonatally administered MeP229 vector (published anecdotal information), the AAV9/CBA-MECP2-myc-SV40 pA vector did not cause neurological toxicity across a broad range of behavioral tests after neonatal administration [9]. Neonatal administration of these two vector designs (those featuring MeP229 [#ii in Table 1] and CBA promoters) could not be compared because the dose was not disclosed for the MeP229 vector [9]. Finally, dose-dependent toxicity could not be assessed in the earliest studies because many (not all) of the readouts evaluated only one dose per treatment paradigm [8, 9].

The first quantification of dose-dependent neurological side effects in WT mice after unregulated MECP2 gene transfer marked a pivotal moment in the field (#ii in Table 2) [10]. In parallel, other researchers have noted qualitative liver toxicity with the same vector [11]. New 5′ and 3′ regulatory elements mitigated side effects across treatment routes, but behavioral toxicity persisted for the highest intracisternal (intracisterna magna [ICM]) dose (#iii in Table 2) [10, 11]. These new elements included a longer truncated variant of the physiological MECP2 promoter and a synthetic 3′ UTR modeled after the longer of the two endogenous MECP2 3′ UTRs [10, 11]. Notably, viral genome #iii was the first MECP2 gene therapy to feature literature-based miRNA binding sites in its 3′-UTR [10, 11]. As new regulatory elements were introduced, the increasing viral genome size posed a growing risk of inconsistent packaging integrity in scAAV. Fortunately, Tillotson et al. [30] reported their human-derived miniMECP2 gene in the same year (Table 3). The space-efficient pairing of miniMECP2 with new viral genome regulatory elements in scAAV paved the way for the additional innovation described in the next section [2].

In summary, this second round of preclinical gene therapies advanced RTT gene therapists closer to the clinic. New AAV9 vectors improved body weight, respiratory measures, and/or survival after adolescent administration in KO mice (#iii–viii in Table 1) [10‒12] and improved gross behavior and/or survival after neonatal administration in KO mice (#iii, xi; in Table 1) [11, 30].

Conceptually, multiple technical parameters may impact the cellular levels of the total MeCP2 protein after gene therapy. First, a host’s mosaic genotype poses the clearest contributing risk factor for toxic overexpression. Second, some administration routes (e.g., ICM) may lend themselves to high focal transduction near the injection site. Therefore, tissues proximal to the injection site may be more likely to express supraphysiological expression levels of exogenous MeCP2 [2]. Third, treatment age may impact transduction efficiency. This, in turn, may affect the number of viral genome copies in a cell and, therefore, the cellular MeCP2 expression levels [15]. Although a second-generation MECP2 gene therapy featured an miRNA target panel for regulating exogenous MeCP2 expression, it is important to note the limitations of the literature-based design for this target panel (#iii; Table 1) [10, 11]. To our knowledge, miRNA-focused RTT literature at the time did not address gene therapy applications [36‒38]. In other words, the published approaches identifying MeCP2-driven miRNAs did not account for the totality of host genotype, gene therapy dose, exogenous regulatory elements, tissue type, age, injection route, proximity to the injection site, and circulation patterns of the cerebrospinal fluid. Importantly, some of these conditions cannot be replicated in mouse or human cell lines under conventional in vitro culture conditions [36].

To reverse-engineer an miRNA target panel tailored for this totality, researchers profiled endogenous brain and spinal cord miRNAs that were upregulated after an in vivo overdose in adolescent mice [2]. This panel was named “miRARE” (miRNA-responsive autoregulatory element) and inserted into the miniMECP2-myc vector (#xi) to create #xii in Tables 1 and 2. This regulated AAV9/miniMECP2-myc-miRARE vector outperformed other designs in a head-to-head comparison after intrathecal (IT) administration in KO mice (#iii, xi–xii in Tables 1, 2) [2]. Notably, AAV9/miniMECP2-myc-miRARE permitted genotype-dependent miniMECP2-myc protein expression in the pons and midbrain (Table 3) [2]. However, miniMeCP2 (without myc) was nearly undetectable in the midbrain of KO mice treated with a human-ready version of the vector after IT administration (#xiii in Tables 1, 2) [14]. Together, these publications warrant further mechanistic characterization of miRARE [2, 14]. Nevertheless, AAV9/miniMECP2-miRARE (#xiii in Table 1) was shown to be effective across multiple readouts (primarily survival, body weight, and respiratory efficacy) [14]. Sinnett et al. [2] published a conceptual diagram of the intended regulation strategy.

Due diligence warrants thoughtful acknowledgment of key survival-related considerations recently identified by gene therapists. For example, although all MECP2 gene therapy publications have used KO mice to assess survival (Table 1), older KO mice (i.e., not neonates [39]) may experience immune responses to exogenous MeCP2 [13]. These immune responses can manifest as tail lesions [2, 13]. Daily immunosuppression has been shown to unconfound (improve and extend) median survival in MECP2 vector-treated KO mice [13].

To be clear, our intention in this section is to help readers interpret meaningful consistencies and comparisons within and across MECP2 gene therapy publications. We also seek to guide younger researchers toward consistent, transparent methods that foster reproducibility. These lessons are applicable across disease models. Indeed, Scott et al. [40] presented a careful assessment of technical considerations (e.g., censoring) that affect survival analyses in mouse models of amyotrophic lateral sclerosis. Interestingly, the meaning of the word “censor” itself may vary across sources as it may mean “exclude” in one publication [40] but not in another [41, 42]. Because all MECP2 gene therapy publications evaluating KO mice have identified survival efficacy, readers can be confident in the therapeutic potential of MECP2 gene transfer [2, 8‒14, 18, 30]. The research community can carry that confidence forward while opening the door for broader discussions on reproducibility, interpretation, and rigor.

To begin, the extensions listed in Table 1 should be interpreted thoughtfully for three reasons: First, survival may be confounded by immune responses that may or may not be clearly identified in cited publications (Sinnett et al. [2] and Luoni et al. [13], e.g., acknowledge these lesions). Second, the choice to euthanize mice for tail lesions may vary across veterinarians and study sites. Third, the brevity of published methods may mask differences, if any, in how investigators account for veterinarian-requested euthanasias (see related discussion points published by Sadhu et al. [14]). One approach is to list veterinarian-requested euthanasias as deaths in statistical software and graphically flag the relevant data points, as was done by Sinnett et al. in 2021 [2, 14]. Alternatively, some investigators may enter veterinarian-requested euthanasias as study dropouts (not deaths; censored and included) for Kaplan-Meier analyses [14, 42]. The rationale for this latter approach would be to mathematically account for partial survival without entering a potentially subjective or artifactual date of death. Neither approach is ideal as conceptually, the former may sometimes have a higher risk of generating false-negative conclusions; the latter may sometimes have a higher risk of generating false-positive conclusions [14]. Ultimately, the choice of analytical methods may be secondary to a more competitive solution: to assess efficacy across multiple mouse models of RTT [8, 11, 13, 18, 22].

Perhaps prudent advice for MECP2 gene therapists is to (1) clearly describe analytical and graphing methods for survival studies; (2) clearly describe veterinary adverse events; and/or (3) present multiple efficacy readouts for a well-rounded interpretation [14]. To generate data in the presence of an attenuated immune response, researchers may use mosaic and missense mutation mouse models of RTT. Alternatively, researchers may maintain a long-term immunosuppression regimen for virus-treated KO mice, as previously published [13, 14].

The Hippocratic theme unifies MECP2 gene therapy publications, which consistently advocate for careful safety and efficacy testing [2, 9, 11‒13, 18, 22, 30]. For each vector design, do the initial safety conclusions stand the test of time across preclinical publications? The AAV9/miniMECP2-myc vector appeared to be safe and effective after neonatal administration in WT and KO mice, respectively [30]. Specifically, the AAV9/miniMECP2-myc vector significantly extended survival and qualitatively delayed the onset of severe aggregate severity scores in KO mice (statistics were not described) [30]. Survival and aggregate severity data indicated that the same vector appeared to be functionally equivalent to the vehicle in WT mice after neonatal administration [30] (note that the “3′ UTR” was not named in the publication by Tillotson et al. [30], but the authors do cite a prequel study naming RDH1pA [11, 30]). However, AAV9/miniMECP2-myc was harmful and ineffective after IT administration during adolescence in WT and KO mice, respectively (#xi in Tables 1, 2) [2]. A safer miRARE-regulated version of this vector proceeded to the clinical trial (#xiii in Tables 1 and 2; clinical trial numbers NCT05606614 and NCT06152237 [20, 21]).

The published narrative of AAV9/miniMECP2-myc suggests two key lessons. First, safety and efficacy are evolving narratives that can span multiple publications. Vector designs that appear to be safe and effective after neonatal intracerebroventricular (ICV) administration should be tested across different paradigms, such as clinically relevant treatment ages. For instance, AAV9/miniMECP2-miRARE (#xiii; Table 1) was shown to be effective after treatment at postnatal day 2 (ICV) and P7-P28 (IT) in male KO mice and safe after administration in postnatal day 2 (ICV) male WT mice and adult female RTT and WT mice (IT) [14]. These data for the human-ready vector (#xiii in Tables 1, 2) helped support the initiation of a clinical trial (NCT05606614) [14, 20]. Second, miRNA-binding regulatory elements (e.g., those in investigational gene therapy candidates for NCT05606614, NCT06152237, and NCT05898620 [20, 21, 26]) may not necessarily be required to achieve safety after neonatal administration in WT mice. The reassuring safety data for WT mice treated with AAV9/miniMECP2-myc – without miRARE – is a prime example [30]. In addition, ssAAV9/CBA-MECP2-myc and scAAV9/P546-MECP2 achieved rigorous safety and efficacy goals after neonatal administration without having any regulatory miRNA-binding sites in their viral genomes (survival extensions for scAAV9/P546-MECP2-treated KO mice revealed a U-shaped dose response, but behavioral tests revealed consistent results across doses) [9, 18]. The former design was abandoned; the latter design has not yet been assessed in older mice to our knowledge. Certainly, the viral genome designs for ssAAV9/CBA-MECP2-myc and scAAV9/P546-MECP2 are not directly comparable to published miRARE designs [2, 9, 14, 18]. However, it is fair to acknowledge that Gadalla et al. [9] and Powers et al. [18] presented diligent safety studies for vectors featuring conventional regulatory elements. Together, these studies after neonatal vector administration in WT mice spanned assessments for survival, aggregate severity scores, body weight, treadmill, open field, respiration, and nuclear volume (which was significantly increased [9]) [9, 18].

The published narratives for neonatal vector administration invite the following question: If miRARE does not yet appear essential for behavioral well-being after neonatal administration, why does it appear to be essential after IT and ICM administration in older mice [2]? The prequel (“Recent endeavors in MECP2 gene transfer for gene therapy of Rett syndrome” by Sinnett and Gray) to this review may guide readers toward one of multiple hypotheses [15]. Because apparent transduction efficiencies for gene therapies may decline with increasing treatment age, gene therapists may be forced to increase the dose to compensate and maintain efficacy. An increased dose of MECP2 gene therapy may, in turn, increase the risk of supraphysiological MeCP2 expression in transduced cells. Thus, improved regulation may be warranted for higher doses of gene therapies administered at older treatment ages.

Lastly, it would be wise to address conventional wisdom: the convention in the MECP2 gene therapy field has been to evaluate safety in WT mice because they are better suited for modeling the symptoms of MECP2 duplication syndrome after gene transfer. In addition, overexpression-related side effects are more easily detectable in mice that do not have MeCP2 loss-of-function phenotypes, some of which overlap with duplication phenotypes [10]. Importantly, however, Matagne et al. [22] demonstrated liver damage and lethality in female Mecp2^−/+ mice after intravenous administration of their codon-optimized gene therapy. This vector design was safer in WT mice and effective in KO mice (#iv in Table 1) [12, 22]. This observed lethality in female Mecp2^−/+ mice incentivizes safety assessments in female RTT mice. Recently, Sadhu et al. reported efficacy in KO mice without compromising survival safety in WT and female Mecp2^−/+ mice after administration of a human-ready miniMECP2-miRARE gene therapy (#xiii in Table 1) [14]. Moving forward, preclinical researchers should continue to evaluate the safety of MECP2 gene therapies across genotypes and genders.

The first- and second-generation MECP2 vectors demonstrated positive effects in attenuating phenotypic symptoms in KO mice (Table 1). However, these beneficial effects were overridden by liver and behavioral toxicity (Table 2). The dose-dependent toxicity of MECP2 gene transfer warranted tight regulation of cellular total MeCP2 protein levels [2]. The recently developed miRARE panel represents one strategy to circumvent adverse events without compromising efficacy [2, 14]. Furthermore, the human-ready (AAV9/miniMECP2-miRARE) vector has demonstrated efficacy across multiple treatment ages and doses in KO mice, providing supportive evidence for the ongoing clinical trial NCT05606614 [14, 20]. Although the clinical trials NCT05606614, NCT06152237, and NCT05898620 have already begun, additional peer-reviewed protein expression studies will continue to inform the research community and may be useful for dose-sensitive gene therapy applications outside RTT [14, 20, 21]. Importantly, these protein expression studies should be conducted using in vivo animal models (in addition to or in lieu of cell lines) [14]. In vivo models will elucidate how miRARE functions within the context of host miRNAs, whose baseline expression levels can sometimes vary across brain regions, developmental stages, and sex [14, 43]. Collaborations were essential for the 10-year journey from the first MECP2 gene therapy publications to the first patient injections [20, 26]. The first clinical injections are a shared milestone.

The authors thank the former lab members Emily Boyle and Christopher Lyons for their technical support and dedication.

S.E.S. has received royalties and/or other income for miRARE technology licensed to Taysha Gene Therapies and Abeona Therapeutics Inc.

This work was supported by the independent laboratory funds for S.E.S. These funds were generated from inventors’ revenue.

I.J., J.O., and S.E.S. participated in the writing process. S.E.S. prepared tables and figures. I.J. and J.O. completed the due diligence fact-checking and assisted with figure planning. All the authors have reviewed and agreed to the published version of the manuscript.