Gene therapy for Lafora disease in the Epm2a^−/− mouse model

Abstract

Lafora disease is a rare and fatal form of progressive myoclonic epilepsy typically occurring early in adolescence. The disease results from mutations in the EPM2A gene, encoding laforin, or the EPM2B gene, encoding malin. Laforin and malin work together in a complex to control glycogen synthesis and prevent the toxicity produced by misfolded proteins via the ubiquitin-proteasome system. Disruptions in either protein cause alterations in this complex, leading to the formation of Lafora bodies containing abnormal, insoluble, and hyperphosphorylated forms of glycogen. We used the Epm2a^−/− knockout mouse model of Lafora disease to apply gene therapy by administering intracerebroventricular injections of a recombinant adeno-associated virus carrying the human EPM2A gene. We evaluated the effects of this treatment through neuropathological studies, behavioral tests, video-electroencephalography, electrophysiological recordings, and proteomic/phosphoproteomic analysis. Gene therapy ameliorated neurological and histopathological alterations, reduced epileptic activity and neuronal hyperexcitability, and decreased the formation of Lafora bodies. Moreover, differential quantitative proteomics and phosphoproteomics revealed beneficial changes in various molecular pathways altered in Lafora disease. Our results represent proof of principle for gene therapy with the coding region of the human EPM2A gene as a treatment for EPM2A-related Lafora disease.

Graphical abstract

José Serratosa, Marina Sánchez, and colleagues successfully restored the EPM2A gene in the brain of Epm2a^−/− mice using an rAAV vector, improving neurological alterations and epilepsy. They identified potential mechanisms induced by laforin, providing a new avenue for treating Lafora disease, a rare, fatal, and untreatable form of childhood epilepsy.

Introduction

Lafora disease (OMIM: 254780. ORPHA: 501) is a rare and fatal form of neurodegenerative disease that appears in early adolescence with seizures and a progressive neurological decline with dementia resulting in death within 5–15 years.^1,2,3 Unfortunately, there is no specific therapy, and patients can only receive antiseizure medications to temporarily manage seizures.⁴ The disease is caused by autosomal recessive inherited mutations in either the EPM2A (OMIM: 607566) gene, encoding the dual-specificity phosphatase laforin,^5,6,7,8 or the EPM2B (OMIM: 608072) gene, encoding the E3-ubiquitin ligase malin.^9,10 Laforin is a glucan phosphatase acting as an adapter protein of enzymes involved in glycogen synthesis, an adapter protein in endoplasmic reticulum (ER) stress and protein clearance, and a tumor suppressor protein.¹¹ The laforin-malin complex regulates glycogen metabolism by inducing proteasome-dependent degradation of muscle glycogen synthase (GS), glycogen debranching enzyme (GDE), and protein targeting to glycogen (PTG).^11,12 Additionally, the laforin-malin complex helps mitigate the toxicity produced by misfolded proteins through the ubiquitin-proteasome system (UPS).¹³ Disruptions in laforin or malin lead to the formation of aggregates of abnormal, insoluble, poorly branched, and hyperphosphorylated forms of glycogen, known as Lafora bodies (LBs).^14,15,16,17 Alterations in oxidative stress, protein misfolding, and proteasomal dysfunction also contribute to the pathophysiology of the disease.^18,19,20

Different murine models of Lafora disease have been generated, including the Epm2a^−/− and Epm2b^−/− knockout mice.^21,22 These models replicate, although with a milder phenotype, most of the neurological alterations seen in patients, such as neuroinflammation, neurodegeneration, LB formation, neuronal hyperexcitability, cognitive deficits, and motor impairment.²³ Epm2a^−/− and Epm2b^−/− knockout mouse models have been used to assay different putative treatments to cure or ameliorate the symptoms of the disease. Thus, we showed in these models that metformin improved many neurological alterations.^24,25 Subsequently, we showed that, in early-stage patients, metformin slows the progression of symptoms and the decline in the performance of activities of daily living²⁴ compared to treatments in patients in more advanced stages of the disease.²⁶ Alternative therapeutic strategies, such as sodium selenate,²⁷ VAL-0417,²⁸ the antisense oligonucleotide Gys1-ASO that targets the mRNA of the brain-expressed GS 1,²⁹ and various modulators of neuroinflammation, have also been assessed in animal models.³⁰ Downregulation of the Gys1 gene through AAV9-delivered Gys1-directed CRISPR-Cas9³¹ or microRNA (miRNA)³² also resulted in a remarkable reduction in polyglucosan body formation and a slight reduction of neuroinflammatory markers.

Recombinant adeno-associated viruses (rAAVs) are widely used vectors for gene therapy. They are non-pathogenic, produce low immunological responses, rarely integrate into the host genome, have a broad tropism, and allow long-term transgene expression. These viruses are small, non-enveloped, single-stranded (ss) DNA particles belonging to the Parvoviridae family, genus Dependoviridae. To date, more than 10 different adeno-associated virus (AAV) serotypes and more than 100 variants have been isolated from adenovirus stocks and human/non-human primate tissues. Different serotypes have different tropisms and recent progress in the field has led to precise targeting toward specific tissues of interest.^{33,34,35,36,37,38,39,40,41}

Here, we show that the rAAV2/9-CAG-hEPM2A (rAAV-hEPM2A) vector, containing the coding region of the human EPM2A (hEPM2A) gene, significantly diminishes neurological and histopathological alterations, reduces epileptic activity, and rescues electrophysiological deficits in the Epm2a^−/− mouse model of Lafora disease. We also show, through differential quantitative proteomic and phosphoproteomic analysis, that human laforin produces beneficial changes in certain molecular pathways altered in Lafora disease. This molecular analysis is crucial for comprehending gene therapy mechanisms of action and understanding specific processes that are corrected by replacing EPM2A.

Results

The rAAV-hEPM2A vector efficiently transduces central nervous system cells in Epm2a^−/− mice, leading to transcription and translation of the hEPM2A transgene

Three months after a single intracerebroventricular (i.c.v.) injection of rAAV-GFP or rAAV-hEPM2A in 3-month-old Epm2a^−/− mice, we verified the successful transduction of the vectors in central nervous system (CNS) cells. First, we confirmed hEPM2A transgene transcription through RT-PCR (Figure S1A). Next, using an antibody against GFP, we observed expression of the GFP transgene in the hippocampus, cortex, septal nuclei, and fimbria (Figure S1B).

Transcription of the hEPM2A transgene was quantified by RT-qPCR in RNA samples from the total brain of Epm2a^−/− mice 3 and 9 months after a single i.c.v. injection of rAAV-hEPM2A vector (Figures S2A and S2B). Gene expression levels were similar to those of endogenous Epm2a in wild-type (WT) animals and did not change over time, although we did observe inter-animal variability (Figures S2A and S2B). Additionally, RNA translation was analyzed with an antibody against laforin, demonstrating effective protein expression in the hippocampus, cortex, the region of the fornix and the septofimbrial and lateral septal nuclei, both 3 and 9 months after rAAV-hEPM2A injection, without showing differences in expression over time (Figure S2C). To identify the cell type where laforin was expressed, the colocalization of laforin and neuronal nuclei (NeuNs), and laforin and glial fibrillary acidic protein (GFAP), was assessed (Figures 1 and S3). Laforin preferentially labeled NeuN-positive pyramidal neurons of cortical layer V and the CA1 and CA2–CA3 fields of the hippocampus (Figures 1A and S3A). A slight level of laforin expression was also found in GFAP-positive cells in the same regions (Figures 1B and S3B).

Figure 1.

Expression of the transgene in neurons and astrocytes

(A) Double IF-P staining was performed to identify cells expressing laforin (green) and NeuN (red), confirming that laforin is correctly expressed in neurons (yellow) of the deep layers of the cortex and in those of the CA1 and CA2 regions of the hippocampus, following i.c.v. injection of rAAV-hEPM2A. (B) Similarly, laforin (green) and GFAP (red) immunostaining showed that laforin is expressed in astrocytes (yellow) of the cortex and hippocampus, although at lower levels compared to neuronal transduction. Scale bar, 50 μm. See also Figure S3.

i.c.v. injection of rAAV-hEPM2A prevents LB formation in the brain of Epm2a^−/− mice

We analyzed and quantified the number of LBs in the CA1 (Figures 2A and 2C) and CA2–CA3 (Figures 2B and 2D) regions of the hippocampus and in layers IV–V of the sensorimotor cortex (Figures 2E and 2F) 3 and 9 months after injections of rAAV-hEPM2A and rAAV-Null. We also analyzed the progression of LB formation at baseline and 3 and 9 months after i.c.v. injections (Figure 2G). Treated mice showed a significant decrease in the number of LBs in the CA1 and CA2–CA3 regions of the hippocampus (Figures 2C and 2D) 9 months after treatment administration, as well as a slower progression of the formation of these aberrant molecules compared to animals injected with rAAV-Null (Figure 2G). When analyzing the number of LBs in layers IV–V of the sensorimotor cortex (magnified box in Figure 2E), we did not observe significant differences between treated and control animals (Figure 2F); however, we found that the progression of LB formation was slower in mice treated with rAAV-hEPM2A (Figure 2G).

Figure 2.

Quantification of LB formation in the CA1 and CA2–CA3 regions of the hippocampus and in layers IV and V of the sensorimotor cortex of Epm2a^−/− mice treated with rAAV-hEPM2A

(A, B, and E) Periodic acid–Schiff (PAS)-Diastase (PAS-D) staining in the CA1 (A) and CA2–CA3 (B) regions of the hippocampus and in layers IV and V of the sensorimotor cortex (E) of Epm2a^−/− mice 3 and 9 months after i.c.v. administration of rAAV-hEPM2A or rAAV-Null. (C, D, and F) Quantitative comparison of LB number in the different regions. In the sensorimotor cortex, it the region zoomed in was quantified (width, 747 px; height, 1,550 px), corresponding to layers IV and V. Results are expressed as the median of independent samples. The bars in the box plots show the minimum and maximum values. Values were normalized using values from Epm2a^−/− mice injected with rAAV-Null. A non-parametric Mann-Whitney test was performed. Nine months after injection, CA1 region, Epm2a^−/− + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A p = 0.0043; 9 months after injection, CA2–CA3 region, Epm2a^−/− + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A p = 0.0173). (G) Comparative progression of LB formation over time in different brain regions of Epm2a^−/− mice treated with rAAV-hEPM2A or injected with rAAV-Null. A two-way ANOVA with Tukey’s multiple comparisons test was performed. Data are shown as mean (SD). Asterisks (∗) indicate differences in the progression of LB formation over time in Epm2a^−/− mice injected with rAAV-Null; hash symbols (#) indicate differences in the progression of LB formation over time in Epm2a^−/− mice treated with rAAV-hEPM2A. CA1region: Epm2a^−/− mice injected with rAAV-Null from 3 to 6 months, p = <.001; Epm2a^−/− mice injected with rAAV-Null from 3 to 12 months, p = <.001; Epm2a^−/− mice injected with rAAV-Null from 6 to 12 months, p = 0.0067; Epm2a^−/− mice treated with rAAV- hEPM2A from 3 to 6 months, p = 0.0019; Epm2a^−/− mice treated with rAAV- hEPM2A from 3 to 12 months p = <.001. CA2–CA3 region: Epm2a^−/− mice injected with rAAV-Null from 3 to 6 months, p = <.001; Epm2a^−/− mice injected with rAAV-Null from 3 to 12 months, p = <.001; Epm2a^−/− mice treated with rAAV- hEPM2A from 3 to 6 months, p = 0.0175. Layers IV-V of the sensorimotor cortex: Epm2a^−/− mice injected with rAAV-Null from 3 to 12 months, p = <.001; Epm2a^−/− mice injected with rAAV-Null from 6 to 12 months, p = <.001; Epm2a^−/− mice treated with rAAV- hEPM2A from 3 to 12 months, p = 0.0017; Epm2a^−/− mice treated with rAAV- hEPM2A from 6 to 12 months, p = 0.0129. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; N.S., not significant. n = 4–6 mice per group. Scale bar, 25 μm (A and B) and 50 μm (E).

Treatment with rAAV-hEPM2A in Epm2a^−/− mice decreases astrogliosis, microgliosis, and cell death

The effect of rAAV-hEPM2A on neuroinflammation in Epm2a^−/− mice was assessed 3 and 9 months after i.c.v. injections using GFAP (Figures 3 and S4A–S4C) and Iba1 (Figures S4D–S4F) antibodies to stain sections from different regions of the hippocampus and cerebral cortex. We analyzed the CA1 field with GFAP and observed a reduction in reactive astrocytes 3 and 9 months after injection (Figures 3A–3C). In the sensorimotor cortex, a decrease in the number of GFAP-positive cells was also found 3 months post rAAV-hEPM2A injection, although 9 months later the results were not statistically significant (Figures 3D–3F). We observed a significant decrease in the number of reactive astrocytes in the CA2–CA3 region of the hippocampus 3 months after rAAV-hEPM2A injection in Epm2a^−/− mice (Figures S4A and S4B). Although there was a trend toward decrease in the number of reactive astrocytes 9 months after injection, the results did not reach statistical significance (Figures S4A and S4C). When analyzing microglia with the Iba-1 antibody in the CA2–CA3 fields of the hippocampus of Epm2a^−/− mice injected with the rAAV-Null vector, we found a significantly higher number of Iba-1-positive cells compared to WT mice 9 months after injection (Figures S4D and S4F). Treatment with rAAV-hEPM2A normalized the number of microglial cells in Epm2a^−/− mice compared to WT mice (Figures S4D and S4F).

Figure 3.

Astrogliosis in the CA1 region of the hippocampus and in layers IV and V of the sensorimotor cortex in Epm2a^−/− mice

(A and D) IHC with anti-GFAP antibody in the CA1 region of the hippocampus (A) and in layers IV and V of the sensorimotor cortex (D) of WT mice injected with rAAV-Null and Epm2a^−/− mice treated with rAAV-hEPM2A or injected with rAAV-Null 3 and 9 months after i.c.v. injection. (B and C) Quantification of reactive astrocytes in the CA1 region of the hippocampus of WT mice injected with rAAV-Null and Epm2a^−/− mice treated with rAAV-hEPM2A or injected with rAAV-Null 3 months (Epm2a^−/− + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A, p = 0.0374) (B) and 9 months (Epm2a^−/− + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A p = 0.0325) (C) after injection. (E and F) Quantification of reactive astrocytes in layers IV and V of the sensorimotor cortex in WT mice injected with rAAV-Null and Epm2a^−/− mice treated with rAAV-hEPM2A or injected with rAAV-Null 3 months (WT + rAAV-Null vs. Epm2a^−/− + rAAV-Null p = 0.0045) (E) and 9 months (F) after injection. The quantification was carried out in the region zoomed in (width, 747 px; height, 1550 px), corresponding to layers IV and V. Results are expressed as the median of independent samples. The bars in the box plots show the minimum and maximum values. Values were normalized using the values of WT mice injected with rAAV-Null. A non-parametric Kruskal-Wallis test followed by Dunn’s multiple comparisons test was performed. ∗p < 0.05, ∗∗p < 0.01. n = 4–6 mice per group. Scale bar, 50 μm.

The effect of rAAV-hEPM2A on neurodegeneration was examined using the NeuN antibody to analyze different regions of the hippocampus and the sensorimotor cortex (Figures S6 and S7). A tendency toward reducing neuronal loss in the CA1 region was observed in Epm2a^−/− mice 3 months after the treatment administration; however, no statistically significant changes in neurodegeneration were observed in any brain region (Figures S6A–S6C and S7).

Given the clear tendency to recover neuronal loss in Epm2a^−/− mice treated with rAAV-hEPM2A at 3 months post injection, we analyzed the levels of cleaved caspase-3 in the CA1 and CA2–CA3 regions of the hippocampus and in the layers IV–V of the sensorimotor cortex. We observed a significant decrease in cleaved caspase 3-positive cells in the CA1 and CA2–CA3 fields of the hippocampus of treated mice (Figure S5), but not in the sensorimotor cortex (data not shown). These data suggest that treatment with rAAV-hEPM2A reduces cell death in the hippocampus.

Gene therapy with rAAV-hEPM2A i.c.v. injection in Epm2a^−/− mice delays the onset of memory decline and diminishes motor impairments

We analyzed the effects of laforin expression on episodic memory, motor coordination, and spontaneous locomotor activity. Memory performance was significantly improved in Epm2a^−/−-treated mice 3 months after injection, as indicated by their higher discrimination index (Figure 4A). This effect, however, was not seen in Epm2a^−/−-treated mice 9 months post injection (Figure 4B) since memory failures were absent at this time even in control Epm2a^−/− mice (injected with rAAV-Null).

Figure 4.

Behavioral studies in Epm2a^−/− mice treated with rAAV-hEPM2A or rAAV-Null 3 and 9 months after injection

(A and B) Memory assessment based on DI evaluation in the ORT, 3 months (WT vs. Epm2a^−/− + rAAV-Null, p = 0.0332; Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null, p < 0.001) (A) and 9 months (B) after injection. (C and D) Evaluation of motor coordination based on the latency time to fall from the cylinder of Epm2a^−/− mice 3 months (WT vs. Epm2a^−/− + rAAV-Null third attempt, p = 0.0102; WT vs. Epm2a^−/− + rAAV-Null fourth attempt, p = 0.0083; no significant differences were found in WT vs. Epm2a^−/− + rAAV-hEPM2A) 3 months (C) and 9 months (D) after injection. (E and F) Analysis of spontaneous accumulated, stereotyped, and rearing movements of Epm2a^−/− mice 3 months after injection (E) (accumulated movement counts: Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null 30′, p = 0.0417, 45′ and 60′, p < 0.001; Epm2a^−/− + rAAV-hEPM2A vs. WT + rAAV-Null 45′, p = 0.0414; Epm2a^−/− + rAAV-hEPM2A vs. WT + rAAV-Null 60′, p = 0.0239. Stereotyped movement counts: Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null 45′, p = 0.0019, 60′, p < 0.001; Epm2a^−/− + rAAV-hEPM2A vs. WT + rAAV-Null 30′, 45′ and 60′, p < 0.001. Rearing movement counts: Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null 60′, p = 0.0168) and 9 months after injection (F) (stereotyped movement counts: Epm2a^−/− + rAAV-Null vs. WT + rAAV-Null 45′, p = 0.0214; Epm2a^−/− + rAAV-Null vs. WT + rAAV-Null 60′, p < 0.001; Epm2a^−/− + rAAV-hEPM2A vs. WT + rAAV-Null 60′, p = 0.0040. Rearing movement counts: Epm2a^−/− + rAAV-Null vs. WT + rAAV-Null 45′, p = 0.0497; Epm2a^−/− + rAAV-Null vs. WT + rAAV-Null 60′, p = 0.0043). Data are shown as mean (SD). One-way and two-way ANOVA tests with Tukey’s multiple comparisons were performed between the experimental groups. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, n.s., not significant. Asterisk indicates Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null; hash symbol (#) indicates WT + rAAV-Null vs. Epm2a^−/− + rAAV-Null; dollar sign ($) indicates WT + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A. n = 15–25 mice per group and experiment.

Motor coordination was also improved with treatment, as shown by increased latency to fall from the rod on the rotarod 3 and 9 months post injection (Figures 4C and 4D). Three months after i.c.v. injection, Epm2a^−/− control mice (Epm2a^−/− + rAAV-Null) showed a significantly shorter cylinder drop latency time in the rotarod test than the healthy WT control (Figure 4C), showing the alterations in motor coordination present in the model. However, the Epm2a^−/− mice that received the treatment lasted longer in the cylinder than the Epm2a^−/− mice injected with rAAV-Null and experienced no significant differences compared to the WT injected with rAAV-Null (Figure 4C). Nine months after the i.c.v. injection, this difference became more noticeable, completely normalizing the cylinder drop latency times between treated Epm2a^−/− and WT mice (Figure 4D).

Analysis of spontaneous motor activity in the actimeter revealed improved performance of Epm2a^−/−-treated mice 3 months after injection. However, the treatment resulted in a greater spontaneous motor activity compared to WT mice, denoting certain hyperactivity (Figure 4E). Nine months after treatment with rAAV-hEPM2A, Epm2a^−/− mice did not exhibit an increased spontaneous locomotor activity compared to Epm2a^−/− injected with rAAV-Null, although treated mice presented a partially reduced hyperactivity (Figure 4F).

Treatment with rAAV-hEPM2A in Epm2a^−/− mice reduces EEG power, frequency of interictal epileptiform discharges, and their heightened sensitivity to pentylenetetrazole

Video-electroencephalography (EEG) recordings were performed in 12-month-old mice (9 months after injection) to examine epileptic-like activity. Epm2a^−/−-treated mice exhibited normalized basal activity rhythms, showing lower beta and gamma wave power spectra compared to Epm2a^−/− mice injected with rAAV-Null (Figure 5A). Following the administration of a subconvulsive dose of pentylenetetrazole (PTZ) (30 mg/kg), power spectra decreased in all groups. However, the decrease was more significant in Epm2a^−/− mice injected with rAAV-Null compared to WT or Epm2a^−/−-treated mice (Figure 5B). Furthermore, analysis of spontaneous (Figure 5C) and 30-mg/kg PTZ-induced (Figure 5D) interictal epileptiform discharges (IEDs) indicated that Epm2a^−/−-treated mice experienced fewer spontaneous IEDs than those injected with rAAV-Null (Figure 5C).

Figure 5.

Analysis of EEG and susceptibility to PTZ in WT and Epm2a^−/− mice 3 and 9 months after injection of rAAV-hEPM2A or rAAV-Null vectors

(A) Representative normalized power spectra were obtained from baseline EEG recordings (48 h). (B) Representative normalized power spectra obtained from EEG recordings after i.p. injection of PTZ at subconvulsive doses (30 mg/kg). Data are shown as mean ± SEM. A one-way ANOVA test with Tukey’s multiple comparisons was performed between the AUC obtained after plotting the EEG powers of all experimental groups. Spontaneous: Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null delta and alpha frequency waves p < 0.001, theta frequency waves p = 0.0104, beta frequency waves p = 0.0026, and gamma frequency waves p = 0.0067; WT + rAAV-Null vs. Epm2a^−/− + rAAV-Null delta frequency waves p = 0.0014, alpha frequency waves p = 0.0040, and beta and gamma frequency waves p < 0.001; WT + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A gamma frequency waves p = 0.0041. PTZ injection: Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null delta, alpha, and gamma frequency waves p < 0.001, theta frequency waves p = 0.0073; WT + rAAV-Null vs. Epm2a^−/− + rAAV-Null delta, theta, alpha, beta, and gamma frequency waves p < 0.001; WT + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A delta and theta frequency waves p < 0.001, and alpha frequency waves p = 0.0088. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.000.1. Asterisk indicates Epm2a^−/− mice treated with rAAV-hEPM2A vs. Epm2a^−/− mice injected with rAAV-Null; hash symbol (#) indicates WT injected with rAAV-Null vs. Epm2a^−/− mice injected with rAAV-Null; plus sign (+) indicates WT injected with rAAV-Null vs. Epm2a^−/− mice treated with rAAV-hEPM2A. n = 3–4 mice per group. (C and D) Analysis of spontaneous (C) and PTZ-induced (D) IEDs. A non-parametric Kruskal-Wallis test was performed followed by Dunn’s multiple comparisons. Data are shown as normalized mean ± SD. Values were normalized using WT mice injected with rAAV-Null values. Spontaneous: WT + rAAV-Null vs. Epm2a^−/− + rAAV-Null, p = 0.0447. n = 3–4 mice per group. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.000.1. (E and F) Percentage of animals with myoclonic jerks after i.p. injection of 30 mg/kg PTZ in WT and Epm2a^−/− mice 3 months (Epm2a^−/− + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A, p = 0.0255) (E) and 9 months (F) after injection. (G) Percentage of animals with GTC seizures, and (H) percentage of lethality after i.p. injection of 50 mg/kg PTZ in WT and Epm2a^−/− mice 9 months after i.c.v. administration of rAAV-hEPM2A or rAAV-Null. WT + rAAV-Null vs. Epm2a^−/− + rAAV-Null, p = 0.0070; Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null, p = 0.0468. Data are shown as percentages. A Fisher’s exact test was performed between the experimental groups. ∗p < 0.05, ∗∗p < 0.01. n = 15–25 mice per group and experiment.

Additionally, Epm2a^−/−-treated mice displayed a reduced incidence of myoclonic jerks in response to intraperitoneal (i.p.) injection of PTZ (30 mg/kg), compared to Epm2a^−/− mice injected with rAAV-Null, 3 months after injection (Figure 5E). We found that treatment produced a tendency toward fewer myoclonic jerks 9 months post injection (Epm2a mice with myoclonic jerks: 60% Epm2a^−/− +rAAV-Null vs. 33.3% Epm2a^−/− + rAAV-hEPM2A) (Figure 5F), although differences were not statistically significant. Nine months after an i.p. injection of PTZ (50 mg/kg), Epm2a^−/−-treated mice exhibited a significant lower mortality (Figure 5H) and fewer generalized tonic-clonic (GTC) seizures (Figure 5G).

The rAAV-hEPM2A vector restores physiological firing properties in the CA1 region of Epm2a^−/− mice

Electrophysiological characterization of the hippocampal circuit of Epm2a^−/− mice showed unaltered membrane properties of CA1 pyramidal neurons. Indeed, membrane capacitance, voltage-current relationship, and resting membrane potential, albeit with a trend toward greater depolarization for Epm2a^−/− cells, showed no significant differences between experimental groups (Figures S8A and S8B). As shown in the histograms, gene therapy treatment did not alter the physiological membrane properties of pyramidal neurons. In contrast, the analysis of the number of action potentials (APs) evoked by the injection of depolarizing steps of current showed a significant increase in the number of APs in the first step (50 pA) of injected current. This demonstrates an initial intrinsic hyperexcitability; however, this is not maintained due to a probable failure of the repolarization mechanisms of CA1 pyramidal neurons in Epm2a^−/− animals injected with rAAV-Null (Figure 6A). In parallel, we found a trend toward reduced rheobase current and a significant depolarized threshold potential in Epm2a^−/− animals injected with rAAV-Null (Figure 6B). Treatment with the rAAV-hEPM2A vector completely restored the firing properties of CA1 pyramidal neurons and the threshold potential 3 months after injection (Figures 6A and 6B). Analysis of the spontaneous excitatory postsynaptic currents (sEPSCs) and long-term potentiation (LTP) of CA1 pyramidal neurons revealed no differences between WT and Epm2a^−/− animals (Figures S8C and S8D).

Figure 6.

Action potential discharge of CA1 pyramidal neurons from WT and Epm2a^−/− mice after injection of rAAV-hEPM2A or rAAV-Null vectors

(A and B) (A) AP firing patterns in response to 50-pA stepped depolarizing current injections (duration 1,200 ms) from CA1 pyramidal neurons in WT (green trace) and Epm2a^−/− mice injected with either rAAV-Null (pink trace) or rAAV-hEPM2A (red trace) vectors. The plot shows mean (±SEM) number of APs at the different input currents (n = 10 neurons for WT, n = 9 for Epm2a^−/− + rAAV-Null, n = 10 for Epm2a^−/− + rAAV-hEPM2A; 50 pA, WT vs. Epm2a^−/− + rAAV-Null p = 0.0207; 250 pA, WT vs. Epm2a^−/− + rAAV-Null, p = 0.0176; 300 pA, WT vs. Epm2a^−/− + rAAV-Null, p = 0.0019). A student’s t test was performed. ∗p < 0.05, ∗∗p < 0.01. (B) Example current-clamp recordings (5-pA stepped depolarizing current injections; 50 ms), scaled to show the AP threshold, in CA1 pyramidal neurons of WT (green), Epm2a^−/− + rAAV-Null (pink), and Epm2a^−/− + rAAV-hEPM2A (red) mice. The plots show similar rheobase current (left) for the three groups (WT, 45.71 ± 5.6 pA, n = 14; Epm2a^−/− + rAAV-Null, 36.07 ± 5.8 pA, n = 14; Epm2a^−/− + rAAV-hEPM2A, 59.08 ± 8.4 pA, n = 13) while AP threshold (right) of Epm2a^−/− + rAAV-Null is significantly depolarized with respect to WT (WT, −41.01 ± 1.7 mV, n = 13; Epm2a^−/− + rAAV-Null, −31.87 ± 2 mV, n = 13; WT vs. Epm2a^−/− + rAAV-Null, p = 0.0020). A student’s t test was performed. ∗∗p < 0.01. Note that injection of rAAV-hEPM2A vector rescues the AP threshold of Epm2a^−/− mice to control values (Epm2a^−/− + rAAV-Null, −31.87 ± 2 mV, n = 13; Epm2a^−/− + rAAV-hEPM2A, −39.74 ± 2.1 mV, n = 13; Epm2a^−/− + rAAV-Null vs. Epm2a^−/− + rAAV-hEPM2A, p = 0.0131). A student’s t test was performed. ∗p < 0.05. Data are reported as means ± SEM.

i.c.v. administration of rAAV-hEPM2A vector reduces epileptic-like activity and restores physiological LTP in dentate gyrus granule cells

Analysis of granule cell membrane properties and current-voltage relationships showed no differences between WT and Epm2a^−/− animals (Figures S9A–S9D). Treatment with the rAAV-hEPM2A vector did not alter the physiological properties of the granule cell membrane properties (Figures S9A–S9D). Analysis of the sEPSCs of dentate gyrus (DG) granule cells showed a trend toward a higher frequency of spontaneous events, normalized by the treatment (Figure S9E). In contrast, no differences were found between WT and Epm2a^−/− animals in the amplitude of spontaneous events (Figure S9E). Using the electrophysiological technique of field potentials, we performed the analysis of elicited epileptic-like activity as described previously⁴² on DG granule cells. The analysis showed increased epileptic-like activity in DG slices from rAAV-Null-injected Epm2a^−/− mice compared with WT mice, as revealed by increased PS amplitude (Figure 7A). The rAAV-hEPM2A vector was able to completely reduce epileptic-like activity to the levels of WT animals (Figure 7A). Similarly, we found a complete inability of Epm2a^−/− granule cells to perform LTP, as shown in Figure 7B. Hence, the rAAV-hEPM2A vector proved to be an excellent strategy to fully restore LTP as well (Figure 7B).

Figure 7.

Epileptic-like activity and LTP in DG slices of WT and Epm2a^−/− mice after injection of rAAV-hEPM2A or rAAV-Null vectors

(A and B) Representative traces of field potentials (FPs) and time course graph of population spikes (PS) amplitude recorded in the DG of WT (green), Epm2a^−/− + rAAV-Null (pink), and Epm2a^−/− + rAAV-hEPM2A (red) mice in a magnesium-free artificial cerebrospinal fluid (ACSF) in the presence of 0.1 μM bicuculline, showing a time-dependent increase of the epileptic-like activity in Epm2a^−/− + rAAV-Null mice compared to WT (WT, 170.1% ± 39.9%, n = 9; Epm2a^−/− + rAAV-Null, 380.3% ± 42.5%, n = 10; WT vs. Epm2a^−/− + rAAV-Null, p = 0.0085). A two-way ANOVA test was performed. ∗∗p < 0.01. Note that injection of rAAV-hEPM2A vector rescues the epileptic-like activity to control values (WT, 170.1% ± 39.9%, n = 9; Epm2a^−/− + rAAV-hEPM2A, 153.6% ± 40.6%, n = 8). (B) Representative PS traces recorded before (left) and 40 min after (right) the HFS protocol in DG slices of WT (green), Epm2a^−/− + rAAV-Null (pink), and Epm2a^−/− + rAAV-hEPM2A (red) mice. The time-course plot of PS amplitudes recorded in the DG before and after HFS protocol shows a significative impairment of LTP in Epm2a^−/− + rAAV-Null mice compared to WT (WT, 246.4% ± 40.3%, n = 7; Epm2a^−/− + rAAV-Null, 155.0% ± 11.3%, n = 6; WT vs. Epm2a^−/− + rAAV-Null, p = 0.0350). A two-way ANOVA test was performed. ∗p < 0.05. Note that injection of rAAV-hEPM2A vector rescues LTP to physiological levels (WT, 246.4% ± 40.3%, n = 7; Epm2a^−/− + rAAV-hEPM2A, 278.2% ± 34.7%, n = 5). Data are reported as means ± SEM.

Expression of human laforin in the brain of Epm2a^−/− mice led to significant changes in critical molecular pathways, as revealed by proteomic and phosphoproteomic analyses of the hippocampus

We performed a proteomic analysis, 3 months after injection, on tissue samples from the hippocampus of Epm2a^−/−-treated mice compared to Epm2a^−/− and WT mice injected with the rAAV-Null vector. Given the pathophysiological characteristics of Lafora disease, our attention was focused on proteins involved in glycogen metabolism, oxidative stress and regulation of misfolded proteins, protein degradation via UPS, and neuronal hyperexcitability. Proteins that exhibited a Zq comparison with a p value less than 0.05 (p < 0.05) were classified as differentially abundant proteins (DAPs) in each respective comparison.

Epm2a^−/− mice injected with the rAAV-Null vector exhibited an increased abundance of enzymes involved in glycogen metabolism, including GS, glycogenin (GYG), glycogen phosphorylase (PYGB/GP), and glycogen debranching (AGL/GDE), compared to the WT group (Figure 8A). Expression of human laforin reduced these levels in the hippocampus of Epm2a^−/−-treated mice (Figures 8A and 8B).

Figure 8.

Variations in the abundance of proteins in the hippocampus of WT and Epm2a^−/− mice 3 months after i.c.v. injection of rAAV-hEPM2A or rAAV-Null vectors

(A) Heatmap representing differences in the abundance of proteins (ΔZq) involved in glycogen metabolism between WT + rAAV-Null, Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null. (B) Box plot showing the differences in the abundance of proteins involved in glycogen metabolism between Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null. (C) Heatmap representing variations in the protein abundance related to proteostasis and reactive oxygen species (ROS) regulation between WT + rAAV-Null, Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null. (D) Box plots showing variations in the protein abundance related to proteostasis and ROS regulation between Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null. Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null: HSPA5 p = 0.0013, TRX1 p = 0.0385, ERp29 p = 0.0130, SELENO W p = 0.0143, and SELENO T p = 0.0021. (E) Heatmap representing differences in the abundance of proteins involved in protein degradation through UPS between WT + rAAV-Null, Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null. (F) Box plots showing differences in the abundance of proteins involved in protein degradation through UPS between Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null.. Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null: NEDD4-2 p = 0.0406, TRIM33 p = 0.0058, ZNRF2 p = 0.0026, XIAP p = 0.0485, and HECW1 p = 0.0269. (G) Heatmap representing variations in the abundance of proteins associated with neuronal excitability between WT + rAAV-Null, Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null. (H) Box plots showing variations in the abundance of proteins associated with neuronal excitability between Epm2a^−/− + rAAV-hEPM2A and Epm2a^−/− + rAAV-Null. Epm2a^−/− + rAAV-hEPM2A vs. Epm2a^−/− + rAAV-Null: GLUA4 p = 0.0432, TARP γ-8 p = 0.0139, GAT2 p = 0.0015, and KIR4.1 p = 0.0351. Data are shown as means of a standardized log2 ratio at the protein level (Zq). Whiskers in box plots show the minimum and maximum values. A t test was carried out between the experimental groups. ∗p < 0.05, ∗∗p < 0.01. n = 4 mice per group.

In Epm2a^−/− mice injected with the rAAV-Null vector, a reduced abundance of proteins involved in protein folding and oxidative stress was observed, compared to WT mice. These proteins included binding immunoglobulin protein chaperone (HSPA5/binding immunoglobulin protein [BiP]), thioredoxin (TRX1), endoplasmic reticulum protein 29 (ERp29), selenoprotein W (SELENOW), and selenoprotein T (SELENOT) (Figure 8C). Remarkably, the expression of human laforin resulted in increased levels of these proteins (Figures 8C and 8D). Next, we explored whether the increased abundance of HSPA5/BiP, a participant in the unfolded protein response (UPR), was a consequence of the abundant release of exogenous laforin or a beneficial effect of laforin expression. To address this, we evaluated the interaction between this chaperone and the transmembrane stress sensor proteins inositol-requiring kinase 1 (IRE1) and PKR-related ER kinase (PERK). Coordinated protein response analysis revealed no significant changes in these proteins in Epm2a^−/−-treated mice compared to Epm2a^−/− mice injected with rAAV-Null vector, 3 months after injection (Figures S11A and S11B), suggesting that the increase in proteins related to protein misfolding in the ER is likely a positive response to laforin expression.

Epm2a^−/− mice injected with rAAV-Null also exhibited a decreased abundance of several E3 ubiquitin ligases, including tripartite motif-containing 33 (TRIM33); zinc and ring finger 2 (ZNRF2); and HECT, C2, and WW-domain containing E3 ubiquitin protein ligase 1 (HECW1), compared to WT mice (Figure 8E). Treatment with rAAV-hEPM2A led to an increased abundance of these proteins in the hippocampus of mice (Figures 8E and 8F). Notably, expression of human laforin increased the abundance of X-linked inhibitor of apoptosis (XIAP), an E3 ubiquitin ligase with anti-apoptotic function, and decreased the abundance of NEDD4-like E3 ubiquitin protein ligase (NEDD4.2), which has a role in the ubiquitination and endocytosis of GLT-1 in Lafora disease models (Figure 8F).

In the hippocampus of 6-month-old Epm2a^−/− mice injected with rAAV-Null, there was an increased abundance of proteins related to neuronal hyperexcitability (Figure 8G), including the glutamate ionotropic receptor α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) type subunit 4 (GLUA4/GRIA4) and calcium voltage-gate channel auxiliary subunit gamma 8 (TARP γ-8). Conversely, there was a reduction in proteins such as glial γ-aminobutyric acid (GABA) transporter-2 (GAT-2) and potassium inwardly rectifying channel, subfamily J, member 10 (KIR4.1) compared to WT mice (Figure 8G). Expression of human laforin restored protein levels to those observed in WT mice (Figures 8G and 8H).

To analyze coordinated protein responses, a standardized log2 ratio, Zc, was calculated for some functional protein categories. In Epm2a^−/−-treated mice, we observed changes in relevant molecular pathways, which were not altered in the Epm2a^−/− mouse model injected with rAAV-Null (Figures S10A and S10B). These signaling pathways include mammalian target of rapamycin (mTOR), RAP1 and RAS signaling cascades, epidermal growth factor (EGF)-induced KSR1-MEK-BRAF-ERK signaling, and phosphatidylinositol signaling system, which are downregulated (Figure S10A). Additionally, we found upregulated pathways related to the immune response and complement activation (Figure S10B).

Finally, we analyzed the levels of peptide phosphorylation in the hippocampus of 6-month-old Epm2a^−/−-treated mice compared to Epm2a^−/− and WT mice injected with the rAAV-Null vector. The abundance of phosphorylated/dephosphorylated peptides was assessed using the standardized log2 ratio, Zp, for several proteins relevant to multiple molecular pathways in Lafora disease (Table S1). These included enzymes related to glycogen metabolism, such as GS, 1,4-alpha-glucan branching enzyme (GBE1), and PYGB/GP, and many kinases and phosphatases, such as RAC-beta serine/threonine-protein kinase (AKT2), AKT3, various protein kinase C (PKC) isoforms, phosphatidylinositol 3,4,5-trisphosphate 3-phosphatase and dual-specificity protein phosphatase (PTEN), kinase suppressor of RAS 1 (KSR1), and mitogen-activated protein kinase 7 (MAPK7). Differential phosphorylation was also observed in neurotransmitter receptors and channels, such as ionotropic glutamate receptor, NMDA2A, glutamate receptor (GRIA1), isoform 4 of glutamate receptor 2 (GRIA2), potassium voltage-gated channel subfamily KQT member 2 (KCNQ2), KCNQ3, and disc large homolog 1 or synapse-associated protein 97 (DLG1/SAP97) (Table S1), all of which may be involved in the molecular mechanisms underlying Lafora disease.

Expression of human laforin in the hippocampus of Epm2a^−/− mice led to a decrease in GS and an increase in HSPA5/BiP expression, validating the proteomic results

To validate some of the most relevant findings observed by proteomic analysis, we examined the levels of GS and BiP expression by immunohistochemistry (IHC) in the hippocampus of Epm2a^−/− mice 3 months post injection (Figure S12). Consistent with the findings of the proteomic analysis, we found that the hippocampus of treated Epm2a^−/− mice exhibited lower levels of GS expression compared to the levels present in the hippocampus of Epm2a^−/− mice injected with the rAAV-Null vector (Figures S12A and S12B). Furthermore, also in agreement with the findings revealed by proteomic analysis, we observed that the levels of the chaperone HSPA5/BiP in the CA2 region of the hippocampus of Epm2a^−/− mice treated with rAAV-hEPM2A were significantly higher than that in Epm2a^−/− mice injected with the rAAV-Null vector (Figures S12C and S12D).

Discussion

This study demonstrates the significant therapeutic impact of a gene therapy in a mouse model of Lafora disease. This is the first-ever example of a successful gene therapy with enduring benefits in a preclinical mouse model of Lafora disease. We developed a gene therapy with an rAAV2/9 vector to restore the expression of the EPM2A gene and recover the function of laforin in the Epm2a^−/− mouse model of Lafora disease. Evaluation of the effects of this form of therapy involved testing memory abilities, motor coordination, spontaneous locomotor activity, epileptic activity, and neuronal hyperexcitability, along with neuropathological and molecular changes. Treatment with the rAAV-hEPM2A vector improved many neurological outcomes 3 and 9 months after a single i.c.v. injection. Proteomic and phosphoproteomic analysis revealed changes in important molecular signaling pathways associated with the expression of human laforin, suggesting possible molecular pathways through which human laforin could produce the observed beneficial effects.

Disruption of either laforin or malin results in the formation of aggregates of polyglucosans.^14,15,16,17 In our study, we observed a significant reduction in LB formation in Epm2a^−/−-treated mice compared to mice injected with rAAV-Null. The laforin-malin complex controls glycogen synthesis by causing the proteasome-dependent degradation of GS, GDE, and PTG.^11,12 Proteomic analysis revealed that untreated Epm2a^−/− mice exhibited an abnormal abundance of these enzymes involved in glycogen metabolism in the hippocampus compared to WT mice. Gene therapy reversed this enzymatic imbalance, restoring enzyme levels to those observed in WT mice, thus correcting abnormal glycogen accumulation and LB formation.

Furthermore, besides the formation of LBs, which sequester a variety of proteins involved in proteostasis and autophagy,^22,43,44 disruptions in the laforin-malin complex lead to the generation of misfolded proteins, coupled with dysfunctions in the UPS and autophagy pathways. These pathophysiological conditions result in heightened ER stress.^{13,19,22,45,46,47,48} These anomalies are further exacerbated by issues related to mitochondrial functionality, following reactive oxygen species (ROS) production, and challenges in counteracting these reactive species due to disruptions in antioxidant mechanisms.^18,49,50,51 Cumulatively, along with the diminished glutamate reuptake capacity of astrocytes and its accumulation within the synaptic cleft described in murine models of Lafora disease,^52,53,54 this results in a notable surge in astrogliosis, neuroinflammation, and neuronal hyperexcitability,^{52,53,55,56,57,58,59} ultimately leading to apoptosis and neurodegeneration.²⁰ Our investigation showed a marked reduction in neuroinflammation following i.c.v. injection of the rAAV-hEPM2A vector, observed both 3 and 9 months post injection across various hippocampal regions and in layers IV–V of the sensorimotor cortex in Epm2a^−/− mice. Furthermore, a significant decline in active caspase-3 levels and a modest increase in neuron count were noted in the CA1 region of the hippocampus in these mice 3 months post-i.c.v. administration.

Chaperones,^25,60 antioxidants,^25,27 and autophagy stimulators^24,25,61 are examples of compounds that have shown promise in lowering or rescuing neuroinflammation in Lafora disease mouse models. In this regard, our proteomic analysis revealed a substantial increase in the abundance of proteins associated with protein folding, UPS, and the regulation of ROS in Epm2a^−/−-treated mice. Thus, the abundances of TRX1, an antioxidant and proteasome-related protein that is decreased in fibroblasts from Lafora disease patients,⁶² and BiP chaperone¹⁹ are increased in Epm2a^−/−-treated mice. The ERp29 chaperone and the antioxidant selenoproteins T and W are also more abundant in Epm2a^−/−-treated mice. Furthermore, rAAV-hEPM2A increased the abundance of several E3 ubiquitin ligases. However, NEDD4-2 ubiquitin ligase, which has previously been reported to induce the ubiquitination and endocytosis of GLT-1 in Lafora disease models,⁵⁴ exhibits lower abundance. The reduction in the abundance of this protein could produce decreased ubiquitination of GLT-1, increasing the levels of the glutamate transporter at the plasma membrane and the glutamate uptake capacity.^54,63 These data strongly suggest that the expression of laforin in treated mice improves neurological functions by reducing misfolded proteins, stimulating UPS and glutamate reuptake, and alleviating alterations in ROS. Consequently, this leads to a decrease in neuroinflammation and cell death, ultimately correcting neurological alterations in memory abilities, spontaneous motor activity, and motor coordination. Strangely, for unknown reasons, although our group of Epm2a^−/− animals injected with rAAV-Null vector showed memory deficits 3 months after injection of the rAAV-Null vector and these were reduced with rAAV-hEPM2A treatment, their memory failures were absent 9 months after injection. Therefore, we could not evaluate the treatment with rAAV-hEPM2A vector in memory performance 9 months after treatment administration.

Mice treated with rAAV-hEPM2A showed fewer myoclonic jerks and GTC seizures and a lower mortality after i.p. administration of both subconvulsive and convulsive PTZ doses. Furthermore, video-EEG analysis revealed that Epm2a^−/−-treated mice exhibited a restoration of normal power spectra, particularly in the gamma range, which is known to be associated with epileptic activity,^64,65,66 as well as a reduced number of IEDs. Epm2a^−/−-treated mice displayed less decrease in EEG power after PTZ administration across high-frequency waves, indicating decreased PTZ susceptibility.⁶⁷

Our electrophysiological findings reveal region-specific impairments within the hippocampus of Epm2a^−/− mice, particularly highlighting intrinsic hyperexcitability in CA1 pyramidal neurons. Indeed, our analysis indicates that Epm2a^−/− mice injected with rAAV-Null exhibit an increased mean number of APs elicited in pyramidal neurons in response to small depolarizing current steps compared to their WT and Epm2a^−/− mice treated with rAAV-hEPM2A counterparts. Furthermore, CA1 neurons from Epm2a^−/− mice injected with rAAV-Null showed a reduced ability to maintain firing rates when faced with increasing current injections, unlike their WT and Epm2a^−/− mice treated with rAAV-hEPM2A counterparts. This phenomenon of rapidly declining firing performance has been similarly observed in models of aging and various neurodegenerative disorders, as reported previously.⁶⁸ These results highlight that treatment with rAAV-hEPM2A reverts the intrinsic hyperexcitability in CA1 pyramidal neurons. Moreover, we have observed that treatment with rAAV-hEPM2A reverses the enhanced epileptic-like activity to WT conditions and restores detrimental LTP in the DG of Epm2a^−/− mice injected with rAAV-Null.

Given that the rAAV-hEPM2A vector preferably transduces CA1 neurons, improvements in the neuronal function of the DG could be achieved through functional enhancements in the CA1-DG axis, related to memory consolidation.^69,70 These electrophysiological changes corroborate the behavioral and epileptic analyses, which showed that treatment reduces sensitivity to PTZ, normalizes EEG profiles, and corrects cognitive deficits in these mice.

Alterations in ionotropic receptor expression, trafficking, subunit composition, and biophysical properties might disrupt information transmission, contributing to both hyperexcitability and synaptic plasticity issues.⁷¹ Other changes consistent with EEG, PTZ sensitivity, and electrophysiological alterations were noted in proteomic and phosphoproteomic studies. A higher abundance of the GABA transporter GAT-2 was observed in the hippocampus of mice treated with the therapeutic vector. Reduced GAT-2 levels have been associated with temporal lobe epilepsy.⁷² This GABA transporter may contribute to reducing neuronal hyperexcitability in treated mice through glutamate-induced GABA release.^73,74,75 Phosphorylation of N-methyl-d-aspartate (NMDA) modulatory subunits (NR1, NR2A, and NR2B) and AMPA receptors regulates neuronal excitability, among other neuronal pathways. In the hippocampus of mice treated with rAAV-hEPM2A, we observed significant dephosphorylation of NR2A and AMPAR subunits GLUA1 and GLUA2, along with a notable decrease in GLUA4 abundance. These findings suggest that gene therapy with rAAV-hEPM2A may decrease neuronal hyperexcitability by reducing glutamatergic transmission.

In addition, we observed a reduction in neuronal hyperexcitability resulting from molecular pathways beyond the glutamatergic and GABAergic systems. Potassium channels are crucial for maintaining the balance of brain excitability. Thus, treated mice exhibited restoration of normal KIR4.1 channel abundance. Additionally, mutations in KCNQ2 and KCNQ3 are linked to benign familial neonatal convulsions (BFNCs), and phosphorylation of these channels by cyclic AMP (cAMP)-dependent protein kinase (PKA) stimulates their activity, reducing epileptic activity.^76,77 Phosphorylation patterns suggest increased PKA activity in mice treated with therapeutic vectors, with higher phosphorylation of KCNQ2 and KCNQ3 in the hippocampus. These findings may collectively explain the decrease in epileptic activity in Epm2a^−/−-treated mice.

Proteomic studies revealed changes in different pathways not previously associated with Lafora disease, including mTOR, RAP1, RAS, KSR1-MEK-BRAF-ERK, and phosphatidylinositol signaling pathways. Interestingly, the phosphatidylinositol 3-kinase (PI3K)/AKT/mTOR signaling pathway has been observed to be elevated in the brains of patients with epilepsy in various genetic and acquired epilepsy models in vivo and in laforin-deficient cells.^11,78 This pathway plays a role in controlling processes related to seizures, synaptic plasticity, cell growth, and ion channel protein expression, which potentially contribute to epileptogenesis.^11,79 Although not previously described in Lafora disease, human laforin may reduce epileptogenic activity in Epm2a^−/− mice by regulating proteins within the PI3K/AKT/mTOR pathway through its protein phosphatase activity. Furthermore, pathways associated with immune response and complement activation may experience an increase in activity due to rAAV-hEPM2A-induced laforin expression in Epm2a^−/− mice.

In conclusion, our study demonstrates that gene therapy inducing laforin expression with the coding region of the human EPM2A gene results in significant neurological improvements in a mouse model of Lafora disease. Treatment at an early symptomatic stage markedly reduced neuroinflammation and LB formation, delayed memory and motor alterations, improved motor coordination, and reduced epileptic activity and electrophysiological impairments. Furthermore, proteomic and phosphoproteomic analyses shed light on various mechanisms through which laforin may be inducing these improvements, indicating potential new targets in Lafora disease. Our results with gene therapy open a new avenue for treating this devastating disease.

Materials and methods

Experimental animals

We used the Epm2a^−/− mouse model of Lafora disease, generated following previously described methods,²¹ and age-matched WT mice (C57BL6). Since no-gender-related phenotype differences have been described in mice⁸⁰ or patients with Lafora disease,^26,81 we analyzed data from male and female mice indistinctively. The mouse colonies were bred in the Animal Facility Service of the Instituto de Investigación Sanitaria-Fundación Jiménez Díaz and were housed in isolated cages in ventilated racks with a 12:12 light/dark cycle at a constant temperature of 23°C, with free access to food and water. All experiments were conducted with the utmost care to use and sacrifice the minimum number of animals while minimizing their suffering. The experimental procedures adhered to the "Principles of Laboratory Animal Care" (NIH publication no. 86–23, revised 1985), as well as the European Communities Council Directive (2010/63/EU) and were approved by the Ethical Review Board of the Instituto de Investigación Sanitaria-Fundación Jiménez Díaz and by the Animal Care and Use Committee at the University of Perugia (authorization no. 2B818.N.9JX).

Production of rAAV2/9-CAG-hEPM2A, rAAV2/9-CAG-GFP, and rAAV2/9-CAG-Null vectors

The rAAV2/9-CAG-hEPM2A (rAAV-hEPM2A) vector, containing the cDNA of the EPM2A gene transcript variant 1 (hEPM2A) (GeneBank: NM_005670.4); the rAAV2/9-CAG-Null (rAAV-Null) vector, containing a non-coding DNA; and the rAAV2/9-CAG-GFP (rAAV-GFP) vector, containing the green fluorescent protein (GFP) gene, were generated in the Unitat de Producció de Vectors (UPV; www.viralvector.eu). The production of those vectors was performed following the triple transfection system: (1) the ITR-containing plasmid, (2) the plasmid encoding AAV capsid (VP1, VP2, and VP3 proteins) and replicate genes, and (3) the adenoviral helper plasmid. To remove empty capsids, AAV vectors were purified by iodixanol-based ultracentrifugation.⁸²

Stereotaxic intracerebroventricular injections

Three-month-old Epm2a^−/− mice received a single i.c.v. injection of rAAV-hEPM2A, rAAV-GFP, or rAAV-Null vectors. Mice were anesthetized in an induction chamber filled with 4% isoflurane and 2% O₂ and maintained with 2% isoflurane and 1.5% O₂. Mice were fixed in a stereotaxic frame (Stoelting, Illinois, USA) and body temperature was maintained using a heating pad at 37°C. Hydration was controlled with a subcutaneous saline injection (1 mL), and ophthalmic gel was applied to prevent dry eyes. The total number of mice subjected to rAAV-hEPM2A and rAAV-Null i.c.v. injections was 25–35 per group and condition.

The incision site was sterilized with 70% ethanol and was made in the midline, starting behind the eyes. Bregma and lambda were identified with H₂O₂ on the skull. A small burr hole was drilled according to stereotaxic coordinates of the right cerebral lateral ventricle relative to bregma (anterioposterior −0.3 and mediolateral −0.9). A Hamilton syringe (Thermo Fisher Scientific, MA, USA, catalog #10664301) was introduced at −2.5 cm in the dorsoventral axis to deliver 3 μL of viral suspension with a titer of 1.26 – 10¹² vg/mL at a rate of 1 μL/min. The incision was sutured, and meloxicam (Boehringer Ingelheim, Georgia, USA) (5 mg/kg) was administered as an analgesic.

RNA extraction and RT-qPCR

RNA was extracted from brain samples previously homogenized on ice with TRIzol Reagent (Thermo Fisher Scientific, MA, US). The RNA pellets were washed, dried, resuspended, -treated with DNase Enzyme (Thermo Fisher Scientific, MA, US), and quantified with a NanoDrop ND-1000 spectrophotometer (Thermo Fisher, MA, US). RT-PCR experiments were carried out using the High-Capacity cDNA Reverse Transcription Kit with RNase inhibitor (Thermo Fisher Scientific, MA, US) with 1 μg of RNA per reaction. The RT-PCR conditions were 25°C for 10 min, 37°C for 120 min, and 85°C for 5 min. For RT-qPCR, cDNA from the hEPM2A transcript variant 1 was used as a template. The reaction was performed with TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific, MA, USA), with Epm2a, EPM2A, and Gapdh probes (Thermo Fisher Scientific, MA, USA). The qPCR conditions were 50°C for 2 min, 95°C for 2 min, 95°C for 1 s, and 60°C for 20 s (40 cycles). Analysis was performed using the 2^−ΔΔCT method.

Immunofluorescence, IHC, and PAS-D staining

Mice were anesthetized and transcardially perfused with 4% phosphate-buffered paraformaldehyde, and their brains were removed, dehydrated, and paraffin embedded. Blocks were then sectioned into 5-μm-thick consecutive sections. For PAS-D staining, sections were rehydrated using decreasing graded alcohols, treated with porcine pancreas α-amylase (5 mg/mL in dH₂O) (Merck, Darmstadt, Germany), and processed with the PAS Kit (Merck, Darmstadt, Germany). Following this, they were counterstained with Gill no. 3 hematoxylin (Merck, Darmstadt, Germany). For paraffin IHC and immunofluorescence (IF-P), rehydrated sections underwent boiling in 0.1 M sodium citrate buffer, pH 6.0. Samples were then incubated in blocking buffer (1% bovine serum albumin, 5% fetal bovine serum, 2% Triton X-100, diluted in PBS) and then with primary antibodies diluted in blocking buffer. For IHC, the primary antibodies used were GFP (1:100 dilution; Abcam, Cambridge, UK; catalog #ab183734), ionized calcium-binding adapter molecule 1 antibody (Iba1) (1:100 dilution; Thermo Fisher Scientific, MA, USA; #MA536257), NeuN (1:100 dilution; Millipore, Temecula, CA, USA; catalog # MAB377), GFAP (1:1,000 dilution; Millipore, Temecula, CA, USA; catalog #MAB360), laforin (3 μg/mL; Lifespan Biosciences, WA, US, catalog #LS B6474), GS (1:100 dilution; Abcam, Cambridge, U.K.; catalog #EP817Y), BiP chaperone (1:1,000 dilution; Millipore, Temecula, CA, USA.; catalog # MABC675), and cleaved caspase-3 (1:400 dilution; Cell Signaling Technology, Danvers, MA, USA.; catalog #9661). Sections were stained with the Vectastain ABC kit (Vector Laboratories, Burlingame, CA, USA). Immunoreactivity was revealed using diaminobenzidine (Dako Cytomation, CA, USA) and H₂O₂ and counterstained with Carazzi hematoxylin (Panreac Quimica, Barcelona, Spain). For IF-P, the primary antibody was GFAP (1:1,000 dilution; Millipore, Temecula, CA, USA; catalog #MAB360), NeuN (1:100 dilution; Millipore, Temecula, CA, USA; catalog # MAB377), and laforin (3 μg/mL; Lifespan Biosciences, WA, US, catalog #LS B6474). Secondary antibodies were conjugated to Alexa Fluor 594 (donkey anti-mouse, 1:400 dilution; Abcam, Cambridge, UK, catalog #ab150108). Samples from four to six mice per group were used, and two consecutive sections per animal were stained and analyzed. Images from the different areas of the hippocampus were acquired using a Leica DMLB 2 microscope (Leica, Wetzlar, Germany) connected to a Leica DFC320 FireWire digital microscope camera (Leica, Wetzlar, Germany) for IHC-processed sections, and with a Zeiss Axioscope 5 (Zeiss, Jena, Germany) connected to an Axiocam 208 color camera (Zeiss, Jena, Germany) for sections processed via IF-P. Subsequently, LBs and NeuN-, Iba1-, and GFAP- positive cells were quantified by two researchers using ImageJ software (NIH, Bethesda, MD, USA). GS expression levels were quantified by measuring the percentage of stained area with ImageJ and Fiji (2.15.0). Reported values represent the average of these quantifications.

Object recognition task

The object recognition task (ORT) was used to assess episodic memory retention. Mice were individually familiarized with a dark box in an open field for 10 min. Two hours later, two identical objects (A and B) were placed in the center of the box. Each mouse had 10 min to freely explore the objects with exploration times (tA and tB) recorded. After 2 h, a new object (C) replaced object B, and the exploration times (tA and tC) were measured. Virtual timers generated by the XNote Stopwatch software were used to measure exploration times when mice examined objects from 2 cm or less. A discrimination index (DI) was calculated using the following equation: DI = (tC − tA)/(tC + tA).

Motor coordination

Motor coordination and balance were assessed using the rotarod test (Harvard Apparatus, Holliston, MA, USA). The mice underwent 2 days of training. On the first day, mice were placed on the rotarod at a constant speed of 4 rpm for 60 s. On the second day, they were trained with speed increasing from 4 to 8 rpm. Only mice able to stay on the rod for 60 s were included in the tests to minimize learning-related variations. In subsequent test sessions over 2 days, the latency time to fall from the cylinder was recorded during two sessions each day, with speed increasing from 4 to 40 rpm, and a maximum time limit of 5 min.

Spontaneous locomotor activity

Spontaneous movements were monitored with a computerized actimeter (Harvard Apparatus, Holliston, MA, USA) that recorded the number of times each mouse crossed the open field through infrared light beam breaks. The SEDACOM 1.4 software (Harvard Apparatus, Holliston, MA, USA) was utilized to analyze the spontaneous, rearing, and stereotyped movements at intervals of 5, 10, 15, 30, 45, and 60 min.

Video-EEG analysis

A plastic pedestal (Plastics1, Virginia, USA) with trimmed electrodes was surgically implanted and secured with acrylic resin onto the skull. Post-surgical pain was managed with meloxicam (5 mg/kg) (Boehringer Ingelheim, GA, USA). Animals were allowed 1 week for recovery before testing. Video-EEG recordings were obtained using a wireless transmitter (Epoch, CA, USA) attached to the pedestal, and the data were digitally recorded on a computer under free-motion conditions. Mice were observed in their home cages under basal conditions and, after 48 h, recorded for 30 min following PTZ injections. The sampling rate was 250 Hz and a 50-Hz notch filter was applied. Mouse behavior was captured using digital video cameras. EEG data were analyzed automatically and manually using the Acknowledge 5.0 software (Epoch, CA, USA) excluding periods with signal loss or artifacts. After applying the Comb Band Stop filter and a Blackman window, power spectra were calculated using a spectral estimator based on autoregressive processes ensuring normalized amplitudes for consistent peak-to-peak ranges. An automated seizure analysis was also conducted, closely monitoring and thoroughly analyzing both seizures and IEDs in the video-EEG recordings.

Sensitivity to PTZ

To analyze neuronal hyperexcitability, PTZ was administered via intraperitoneal injection, using two doses: 30 mg/kg (subconvulsive dose, rarely causing GTC seizures in WT animals) and 50 mg/kg (convulsive dose). The subconvulsive dose was used to determine the percentage of mice displaying myoclonic jerks. The convulsive dose was administered to evaluate the percentage of animals experiencing GTC seizures, their duration, the time to the first myoclonic or GTC seizure, and the lethality. Each animal was observed for 45 min by two researchers.

Electrophysiology

Mice were sacrificed by cervical dislocation. Brain collection and slice preparation were performed as previously reported.⁸³ For whole-cell patch-clamp recordings, pyramidal neurons of the CA1 and granule cells of the DG were visualized using infrared differential interference contrast (Olympus). Thin-wall borosilicate glass electrodes (3–6 MΩ) were filled with (in mM) K-gluconate 120, KCl 20, MgCl₂ 2, EGTA 0.02, HEPES 10, Mg-ATP 2, and Na-GTP 0.3 (pH 7.4). Access resistance was monitored online throughout each experiment and recordings were discarded if either access resistance or holding current increased by more than 25% during the experiment. No liquid junction potential correction was implemented. Membrane capacitance and resistance were taken online from the membrane seal test function of pClamp 10.7 (−5 mV step, 10 ms). Injection of depolarizing and hyperpolarizing current steps (1,200 ms, 50-pA increments) were used to obtain current-voltage curves and AP numbers at suprathreshold responses. Depolarizing current steps of increasing amplitudes (5-pA increments) were used to determine the AP threshold and rheobase currents. When recording spontaneous excitatory postsynaptic currents (sEPSCs), picrotoxin (50 ΩM) was added to the ACSF to block GABAA currents. Neurons were clamped at the holding potential (Vh) of −60 mV (CA1) and −70 mV (DG). Data were acquired with pClamp 10.7 (Molecular Devices) and currents were filtered at 0.1 kHz, digitized at 200 μs using Clampex 10.7, analyzed offline using the automatic detection, and subsequently checked manually for accuracy.

For extracellular recordings, the Axoclamp 2B amplifier (Molecular Devices, MA, USA) was used. The stimulating electrode was inserted into the Schaffer collaterals or in the perforant path fibers and the recording electrode, made of borosilicate glass capillaries filled with 2 M NaCl (resistance 10–15 MΩ), was placed into the stratum radiatum of the cornu ammonis 1 (CA1) or in the DG close to the granular layer, respectively. Epileptic-like activity in hippocampal DG was induced as previously reported.^42,83 LTP was induced by high-frequency stimulation (HFS) at 100 Hz (1 s), consisting of one train of stimuli for the CA1 area and three trains (5-min intervals) for the DG area, after recording a stable baseline for 10 min. Traces were filtered at 3 kHz, digitized at 10 kHz, and stored. The initial slope (for the CA1) and PS amplitude (for DG) of extracellular fEPSPs were used to assess alterations in synaptic strength.

Differential quantitative proteomics by isobaric labeling: TMT11plex

For differential quantitation in the proteomic and phosphoproteomic analysis, peptides were labeled with the tandem mass tag 11 plex (TMT11-Plex) technique. Identification and quantification of proteins were performed using liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS).^84,85

Data were analyzed using a logarithmic statistical model (log2), enabling the estimation of peptide and protein abundances (Zq). For the identification of alterations in functional biological processes beyond individual protein responses, each category was assigned a category Z value (Zc). This facilitated the analysis of coordinated protein changes within categories. Finally, for the quantification of changes in peptide phosphorylation, the abundance of phospho-peptides was assessed using the corresponding standardized log2 ratio (Zp).^84,85

Statistical analysis

Values are presented as means ± SEM or standard deviation (SD) or as percentages. To analyze the differences between experimental groups, we employed one- or two-way ANOVA, Fisher’s exact test, Student’s t test, and non-parametric Kruskal-Wallis or Mann-Whitney test, as indicated in each specific case. For EEG analysis, the area under the curve (AUC) was obtained to compare the differences in the power spectra between groups. For proteomics and phosphoproteomics, changes in protein abundance, categories, and abundance of phospho-peptides were determined by comparing the means of Zq, Zc, and Zp, respectively, between groups and selecting those with a p value of 0.05 or lower. For electrophysiological recordings, data analysis was performed offline using Clampfit 10.7 (Molecular Devices). Statistical analyses were conducted using GraphPad Prism 8.0 or GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA). Statistical tests were two tailed and the statistical significance thresholds were denoted as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001.

Data and code availability

Data supporting the findings of this study are available from the corresponding author.

Supplemental information

Table S1. Phosphorylation/dephosphorylation profile

Phosphorylation/dephosphorylation profile of peptides related to Lafora disease in Epm2a-/- mice treated with rAAV-hEPM2A compared to Epm2a-/- mice injected with rAAV-Null. The p-value indicates the significant differences between the means of Zp (ΔZp).