Improving Gene Therapy for Angelman Syndrome with Secreted Human UBE3A

Abstract

The rare genetic neurodevelopmental disease Angelman syndrome (AS) is caused by the loss of function of UBE3A, a ubiquitin ligase. The disease results in a lifetime of severe symptoms, including intellectual disability and motor impairments for which there are no effective treatments. One avenue of treatment for AS is the use of gene therapy to reintroduce a functional copy of the UBE3A gene. Our group had previously shown that recombinant adeno-associated virus (rAAV) expressing mouse Ube3a could rescue deficits in a mouse model of AS. Here, we expand on this work and show that this approach could be successfully replicated in a second AS model using the human UBE3A gene. Furthermore, we address the challenge of limited vector distribution in the brain by developing a novel modified form of UBE3A. This modified protein, termed STUB, was designed with a secretion signal and a cell-penetrating peptide. This allowed transduced cells to act as factories for the production of UBE3A protein that could be taken up by neighboring non-transduced cells, thus increasing the number of neurons receiving the therapeutic protein. Combining this construct with intracerebroventricular injections to maximize rAAV distribution within the brain, we demonstrate that this novel approach improves the recovery of behavioral and electrophysiological deficits in the AS rat model. More importantly, a comparison of rAAV-STUB to a rAAV expressing the normal human UBE3A gene showed that STUB was a more effective therapeutic. These data suggest that rAAV-STUB is a new potential approach for the treatment of AS.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13311-022-01239-2.

Introduction

The genetic neurodevelopmental disorder Angelman syndrome (AS) is a severe disease with symptoms including, but not limited to, severe intellectual disability, speech impairment or inability to produce speech, ataxia, the occurrence of seizures, and an unusual personality characterized by unprovoked laughter and a fascination with water [1]. Currently, no disease-modifying therapeutics are available. The majority of available treatments target an individual disease symptom, with most having minimal efficacy. Studies have identified UBE3A, encoding an E3 ubiquitin ligase, as the gene responsible for the development of AS. This gene is deleted in most cases of AS, with the remaining cases arising due to mutations in the UBE3A gene, methylation defects, or paternal uniparental disomy [1–5].

For treatment purposes, we can consider AS a monogenic disorder, which is evidenced by AS patients with disease-causing point mutations in UBE3A. This suggests gene replacement therapy as a promising avenue of treatment for this monogenic disorder. Investigations into gene therapy-based treatments for neurological disorders have been increasing for several years, especially for the treatment of monogenic disorders. The recent FDA approval of gene therapy approaches for RPE65 mutation-associated retinal dystrophy and type I spinal muscular atrophy supports the progression of other monogenic disorders like AS. This makes UBE3A a clear target for a disease-modifying treatment [6, 7]. Our group has previously shown recovery of deficits in the mouse model of AS using recombinant adeno-associated viral vectors (rAAV) expressing mouse Ube3a [8]. While there is a clear potential for rAAV-mediated gene therapy, the primary challenge is getting rAAV to transduce a majority of UBE3A-deficient neurons, while minimizing high titer and the number of injections. Therefore, this work aimed to address rAAV vector diffusion and transduction limitations in order to develop effective rAAV-based central nervous system (CNS) therapies. To address this, we developed a novel secreted form of UBE3A which we combined with intracerebroventricular (ICV) injections in a rat model of AS.

We recently reported on the development of a new AS rat model which was generated using CRISPR technology resulting in a full deletion of the rat Ube3a gene. This rat model displays the typical loss of Ube3a expression within neurons, disruption in general motor coordination, gait alterations, increased cortical and hippocampal delta EEG power, and sociability and memory deficits [9–11]. These studies were the first to identify the presence of extracellular UBE3A protein in the rat. While the protein is present in the brain of wild type (WT) animals, we found that extracellular UBE3A in the AS rat was absent [12]. We also demonstrated that the levels of extracellular UBE3A appeared to be regulated in an activity-dependent manner, with increases in UBE3A in the fear conditioning task [12]. These data strongly suggested to us that UBE3A may have an important role extracellularly for learning and memory. This led to our hypothesis that the release of UBE3A protein from transduced cells may be advantageous in a gene therapy approach.

Here, we explore a combination of two approaches to facilitate the distribution of UBE3A throughout the CNS. We combined ICV injections, which can give increased vector distribution, with a novel secreted human UBE3A construct we termed STUB (secreted TAT UBE3A). STUB was designed with a secretion signal along with a cell-penetrating peptide sequence. This allows the vector to create cellular “protein factories” from transduced cells in which the protein will be created and secreted. The cell-penetrating peptide sequence allows neighboring non-transduced cells to take up the protein. In addition to demonstrating the effective use of the human UBE3A gene in a rAAV as an AS treatment, our data also show that STUB, via ICV injection, has a greater effect compared to the native non-secreted UBE3A vector in rescuing behavioral and electrophysiological deficits in the Ube3a deletion rat model of AS [11].

Methods

Animals

Ube3a deletion rats were bred and maintained as described previously [11]. 129-Ube3a^tm1Alb/J mice [13] were bred and maintained as described previously [8]. AS mice were used for the initial characterization of the electrophysiological testing before moving to the rat model for more in-depth testing. Male and female animals were housed in a standard 12-h light/dark cycle and supplied with food and water ad libitum at the University of South Florida and were group-housed. The investigators were blind to treatments for all behavior testing. All procedures were conducted in compliance with the NIH Guidelines for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee of USF (approval number A4100-01).

Construct Generation and Vector Production

The human UBE3A gene variant 1 (GenBank: BC002582.2) was cloned into the rAAV expression vector pTR12.1-MCSW. This vector has AAV2 terminal repeats, CBA/CAG promoter, WPRE, and BGH poly signal (Fig. 1a). Starting with our hUBE3A construct, a secretion signal —(ATGAAGTTATGGGATGTCGTGGCTGTCTGCCTGGTGCTGCTCCACACCGCGTCCGCC) and a cell-penetrating peptide sequence — (TAT, TACGGCAGAAAGAAGAGGAGGCAGAGAAGGAGA) were added in frame on the 5’ end of hUBE3A (Fig. 1a). For the in vitro experiment shown in Fig. 1b, an HA tag was added to the vector in order to detect the protein easily. For the remaining work, the HA-tag free vector was used. In vitro testing was performed by transfecting HEK293 cells with the STUB construct using lipofectamine and incubating for 72 h before media was collected for further testing via western blot. Viral particles were generated as previously described [8, 14]. Briefly, HEK293 cells were co-transfected with either the pTR-hUBE3A plasmid or the pTR-hSTUB plasmid, pXX6 helper plasmid, and pAAV9 serotype plasmid using polyethyleneimine transfection. The virus was harvested from cells and purified using an iodixanol gradient and concentrated using centrifugal filtration. Viral titer was determined with a dot blot assay as described previously [14]. Vectors are described as vector genomes (vg)/mL.

Fig. 1.

Generation of a Novel Secreted Form of UBE3A. a Diagram of AAV vector constructs. Using the human isoform 1 UBE3A coding sequence STUB was created by attaching a GDNF signal for secretion and a TAT sequence for reuptake to the N-terminus. Both hUBE3A and STUB were cloned into pTR expression vectors with the CBA (or CAG) promoter. b Western blot using anti-HA antibody of cell culture media collected from HEK293 cells transfected with pTR-hUBE3A-HA or pTR-STUB-HA showed expression and secretion of the STUB-HA protein into the cell culture media

Western Blotting

Samples were heated at 95 °C for 5 min before running on an SDS-PAGE gel with Laemmli sample buffer containing β-mercaptoethanol. Protein was transferred to a nitrocellulose membrane using the Trans-Blot Turbo system (Bio-Rad) and REVERT stain (Li-Cor Biosciences) was used to quantify total protein before blocking with Intercept T20 blocking buffer (Li-Cor Biosciences). Blots were incubated overnight at 4 °C with gentle rocking. For UBE3A detection, we used anti-UBE3A primary antibody solution (Sigma-Aldrich, SAB1404508) at 1:1000 dilution. For anti-HA detection, we used the Roche antibody clone 3F10 at 1:1000 dilution. Blots were incubated in secondary antibody (Li-Cor Biosciences, IgG-800CW) at 1:10,000 for 2 h at room temperature. Detection was performed using the Odyssey CLx scanner (Li-Cor Biosciences). Analysis was performed using LI-COR Image Studio. Data are shown as the ratio of UBE3A band intensity/total protein intensity (REVERT).

Immunohistochemical (IHC)

To test for expression of the newly generated STUB construct, in vivo immunohistochemical (IHC) staining was performed. Animals were anesthetized with a lethal dose of ketamine/xylazine and then perfused with PBS before removal of the brain and submersion of the brain in cold 4% PFA. The issue was fixed overnight before being incubated in 30% sucrose for 3 days. Sections were obtained with a microtome. Tissue sections were processed using the free-floating IHC procedure with nickel and diaminobenzidine. Endogenous peroxidases were quenched with 0.3% hydrogen peroxide for 30 min. Blocking solution (5% horse serum, 0.25% Triton-X100 in PBS) was applied for 1 h. Tissue sections were then incubated overnight with anti-UBE3A antibody at a 1:3000 dilution (Sigma-Aldrich, SAB1404508). After washing with PBS, secondary antibody (horse anti-mouse IgG(H + L), peroxidase, Vector Labs) was applied for 1 h at room temperature. Color was developed using diaminobenzidine and nickel chloride with hydrogen peroxide. Tissue was mounted on glass slides before dehydration and coverslipping using DPX. Slides were scanned with the Zeiss Axioscan and analyzed using NearCyte. Data is shown as percent area positive DAB staining of the hippocampus normalized to the AS control group. One-way ANOVA was used to analyze data, followed by Tukey’s post hoc test.

Animal Injection

Animals were anesthetized using isoflurane and positioned in a World Precision Instruments stereotactic surgery apparatus. An incision was made on the sagittal surface of the skull. AS mice were injected into the hippocampus (HPC) (X ± 2.7, Y-2.7, Z-3.0 from bregma) with 2 μL of either rAAV9-GFP (n = 4: M = 1, F = 3) or rAAV9-STUB (n = 7: M = 3, F = 4) at 1.0 × 10¹³ vg/ml. Wild type littermate controls were injected with 2 μL per site with sterile PBS (n = 3 M). For rats, the coordinates (X =  ± 1.5, Y =  −0.5, Z =  −4.3 from bregma) were used for bilateral ICV injections and (X = 4.6, Y =  −6.0, Z =  −5.0 from bregma) for unilateral intrahippocampal (IHPC) injections into the right hippocampus. A 10-μL volume for ICV and 2-μL volume for IHPC injections (per injection site) were applied at the coordinates above using convection-enhanced delivery method [15, 16]. Vectors were injected into AS rats at the following concentrations: hippocampal STUB, 3.410¹³ vg/mL (n = 8: M = 4, F = 4); hippocampal UBE3A, 2 × 10¹³ (SD + / −0.56 × 10¹³) vg/mL (n = 8: M = 4, F = 4); ICV STUB, 3.0 × 10¹³ (SD + / −0.58 × 10¹³) vg/ml (n = 17: M = 12, F = 5); ICV UBE3A, 2 × 10¹³ vg/mL (n = 15: M = 10, F = 5).

Motor Deficit Testing

Motor deficits were tested in rats using the hind limb clasping test (AS STUB n = 13: M = 10, F = 3; AS UBE3A n = 15: M = 10, F = 5; WT = 11: M = 6, F = 5; AS PBS n = 14, M = 9, F = 5) along with the accelerating rotarod (AS STUB n = 17: M = 12, F = 5; AS UBE3A n = 15: M = 10, F = 5; WT = 14; M = 8, F = 6; AS PBS n = 14, M = 9, F = 5). Testing was performed as previously reported [11]. Hind limb clasping was performed by suspending the animal by the tail for 30 s and observing the hind limbs. The animal behavior was recorded and later scored by a researcher blind to animal treatment. The clasping phenotype was measured according to the commonly used 0–3, scale with 0 corresponding with no clasping, 1 with one withdrawn limb, 2 with both limbs withdrawn, and 3 with clasping together of the hind paws. Rotarod was performed by placing the animals onto an accelerating rotating cylinder (4–40 rpm, Ugo Basile) for up to 5 min, and the latency to fall was recorded. Testing was performed for 4 trials each day over two consecutive days with 30 min between each trial.

Testing of Associative Learning

Associative learning was tested using contextual fear conditioning as previously reported [11] (AS STUB n = 11: M = 8, F = 3; AS UBE3A n = 12: M = 8, F = 4; WT = 13: M = 8, F = 5; AS PBS n = 13, M = 8, F = 5). For conditioning, animals were placed in a sound attenuation chamber and allowed to acclimate for 2 min. The conditioned stimulus tone (95 dB, 5000 Hz) played for 30 s with the animals receiving a 1 mA foot shock during the last 2 s. Animals then remained in the chamber for 3 min while freezing behavior (lack of movement for two consecutive seconds) was recorded. For the contextual testing, animals were placed back into the chamber 72 h after conditioning and freezing behavior was recorded. Freezing behavior was analyzed through tracking software (Ethovision XT).

Electrophysiological Testing

Mouse hippocampal long-term potentiation (LTP) was performed as described previously [8] (AS STUB injected n = 20 slices, 7 animals; AS GFP injected n = 14 slices, 3 animals; WT n = 6 slices, 3 animals). Rat hippocampal LTP testing was performed on both IHPC and ICV cohorts (IHPC: AS STUB injected hemisphere n = 5 slices, 3 animals: M = 2, F = 1; AS STUB uninjected hemisphere n = 6 slices, 3 animals: M = 2, F = 1; AS UBE3A injected hemisphere n = 5 slices, 3 animals: M = 2, F = 1; AS UBE3A uninjected hemisphere n = 7 slices, 4 animals: M = 2, F = 2; AS PBS n = 7 slices 4 animals: M = 2, F = 2. ICV: AS STUB n = 11 slices, 5 animals: M = 3, F = 2; AS UBE3A n = 11 slices, 5 animals: M = 4, F = 1; WT n = 8 slices, 4 animals: M = 2, F = 2; AS PBS n = 11 slices 4 animals: M = 3, F = 2). LTP was tested as previously reported with minor modifications [17]. Briefly, the brain was quickly removed and immediately submerged in 95:5 O₂: CO₂ saturated ice-cold cutting solution. Sections were obtained using a vibrating microtome. Hippocampi were dissected and allowed to rest in a solution of 50:50 cutting solution: artificial cerebrospinal fluid (ACSF). Slices were transferred to the recording chamber and allowed to recover for 1 h 30 m. Stimulation was applied to Schaffer collaterals using a concentric bipolar electrode (WPI), and extracellular field recordings were obtained from neurons of the CA1. To induce LTP, theta burst stimulation, consisting of five trains of four pulses at 100 Hz with an inter-burst interval of 20 s, was applied at an intensity that resulted in ~ 50% maximal response as obtained from an input/output curve. Data were analyzed using ClampFit 10.7 software. For analysis, six sweeps over 2 min were averaged, and data were normalized to the averaged value of the initial slope of the fEPSP from the 20-min baseline recording. Data are presented as 2-min average timepoints. All data are represented as the mean ± SEM.

Statistical Analysis

Graphs were created using GraphPad Prism with data presented as the mean with standard error of the mean. Comparison between two groups was done by t-test. For multiple groups, data was analyzed by one-way ANOVA followed by Tukey’s multiple comparison test or mixed-effects ANOVA and Fisher’s LSD. p value of 0.05 was chosen as the cutoff for significance. Significance was represented as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Results

Generation of the Novel STUB Construct

To optimize the efficiency of rAAV-based gene therapy for the CNS, our team designed a secreted form of hUBE3A we have termed STUB (secreted TAT UBE3A) (Fig. 1a). This construct was created by cloning a secretion signal and a cell-penetrating peptide sequence onto the N-terminus of hUBE3A into an rAAV plasmid. The STUB construct was first tested for expression in mammalian cells via cell culture using HEK293 cells and compared to unmodified human UBE3A. Due to low sensitivity and nonspecific bands when using an anti-UBE3A antibody with the culture media, we added a HA tag to the C-terminus of each construct for this experiment only. We designated the STUB construct with HA as STUB-HA. We observed secretion of STUB-HA into the cell culture media (with no detection seen by the unmodified UBE3A-HA) as expected, indicating that the secretion signal peptide was functional (Fig. 1b). The expression of STUB protein generated by rAAV was validated using the AS rat model and IHC staining (Fig. 2). UBE3A staining was found in the injected hemisphere of the HPC but not the contralateral hemisphere. We noted that the AAV-UBE3A staining appeared stronger than the AAV-STUB (Fig. 2a, c) even though the titer for the UBE3A was slightly lower. We believe that the most likely explanation is due to secretion. From our experience with other proteins that have been secreted, we see what appears to be a lower expression but is likely just due to a decrease in staining because of a more diffuse presence of the protein in the extracellular space [18]. However, we cannot exclude the possibility that the additional sequences in STUB are reducing the expression of the protein.

Fig. 2.

Hippocampal expression of UBE3A and STUB from unilateral hippocampal-injected AS rats verified by immunohistochemistry. To examine the ability of various rAAVs to express UBE3A in vivo, AS rats were unilaterally injected into the hippocampus with vectors containing STUB or UBE3A. After allowing 6 weeks for protein expression, tissue was collected and probed for UBE3A. Representative images are shown. a UBE3A had a strong signal in the injected ipsilateral hemisphere. b There is no staining in the uninjected contralateral hemisphere of UBE3A animals. c STUB had a diffuse expression in the injected hemisphere. d The uninjected hemisphere from STUB animals had no staining. e Wild type hippocampus stained for UBE3A. Scale bar = 200 μm. f Quantification of staining represented as the percentage area of positive staining. WT=wild type, Co=contralateral, Ip=ipsilateral

Functional Activity of STUB

After confirmation of vector expression, we wanted to ensure that STUB would generate active protein after the addition of the modifications. First, STUB was tested for its ability to recover the hippocampal LTP deficit present in AS mice. AS mice were injected bilaterally into the hippocampus with either rAAV-STUB or a rAAV-GFP control vector, and hippocampal electrophysiology was performed as described previously [8]. We observed a complete recovery of the hippocampal LTP deficit in the AS mice injected with rAAV STUB compared to control injected animals (Fig. 3). This rescue was similar to that observed previously with the mouse Ube3a gene in AAV [8]. Furthermore, it demonstrates that the modifications to hUBE3A to generate STUB did not alter its functional ability to rescue this deficit in animal models. We next wanted to confirm potential secretion and further characterize the functional activity of STUB in vivo. Given that hippocampal structures are connected across hemispheres by the hippocampal commissural fibers [19], we hypothesized that delivery of STUB to the ipsilateral hippocampus could have potential benefits on the contralateral side due to secretion of STUB. This effect should not be observed with gene delivery of the native hUBE3A because we would expect to have a limited release of protein into the contralateral hippocampus. Ube3a^m−/p+ rats were injected unilaterally into the hippocampus with either PBS, rAAV-hUBE3A, or rAAV-STUB. Rats recovered for at least 8 weeks to allow for gene expression, and then LTP was induced in both the ipsilateral and contralateral hippocampus. We demonstrate that the AS rats showed a significant LTP deficit and that expression of either hUBE3A or STUB was able to recover this deficit in the ipsilateral hippocampus (Fig. 4). More importantly, we showed recovery of LTP deficits in the contralateral hippocampus for rAAV-STUB injected animals, whereas non-secreted hUBE3A injections failed to recover deficits in the contralateral hippocampus (Fig. 4). These data suggest that the STUB construct produced a functional UBE3A protein, and based on the ability of rAAV vectors to take advantage of axonal transportation, this protein was likely being secreted from cells in the injected hippocampus into the contralateral hippocampus [19–23].

Fig. 3.

STUB retains function in treated AS mice. a Hippocampal LTP testing demonstrates that STUB recovers the LTP deficit present in AS mice. Data shown as mean and standard error of the mean. b An average of the final 10 min of electrophysiological testing shows a recovery of LTP to wildtype levels. ANOVA results indicate a significant difference between groups (F (2, 15) = 267.7). Tukey’s multiple comparison test found that the average of EPSP slopes was significantly different between the GFP group and STUB group and the GFP group and the wildtype group. No difference was found between the STUB and wild-type groups. Data shown as mean and standard error of the mean. ****p < 0.0001

Fig. 4.

Hippocampal long-term potentiation in unilaterally injected AS rats is rescued in both contralateral and ipsilateral hemispheres by STUB treatment. a HPC LTP testing in animals treated by unilateral HPC injection of STUB (STUB uninjected n = 6 slices, 3 animals; STUB injected n = 5 slices, 3 animals; PBS n = 7 slices, 3 animals). c HPC LTP testing in animals treated by unilateral HPC injection of UBE3A (UBE3A uninjected n = 7 slices, 4 animals; UBE3A injected n = 5 slices, 3 animals; PBS n = 7 slices, 3 animals). b and d Quantification of the final 10 min of each experiment. ANOVA results: F (2, 12) = 9.442, p = 0.0034. Statistical significance as tested by Tukey’s test shown as asterisks

Effects of ICV Gene Replacement Therapy Injections on Hippocampal LTP

We next explored the potential recovery of the LTP deficits in AS rats using ICV injections of rAAV-STUB and rAAV-hUBE3A. We observed a significant improvement in hippocampal LTP in the rAAV-hUBE3A injected group compared to the control injected AS animals (Fig. 5). This result showed that the human UBE3A gene could recover LTP in the AS rat in a similar manner to what we observed previously using rAAV with the mouse Ube3a gene in the AS mouse [8], although it did not completely achieve the level of potentiation of wild type animals (Fig. 5). Interestingly, the administration of rAAV-STUB was able to recover LTP to a level indistinguishable from WT animals (Fig. 5), suggesting that the STUB construct is more effective for LTP recovery when delivered ICV than the hUBE3A vector.

Fig. 5.

Electrophysiological deficits are recovered in ICV STUB injected AS rats. a Hippocampal LTP deficit shows recovery after treatment with STUB via ICV injection, with a partial recovery shown in animals treated with an identical injection of UBE3A. b Quantification of the last 10 min of electrophysiology recordings shows a recovery with the STUB treatment and a partial recovery with UBE3A treatment (wild -type n = 10 slices, 4 animals; STUB injected n = 11 slices, 4 animals; UBE3A injected n = 11 slices, 4 animals, UBE3A injected n = 11 slices, 5 animals) ANOVA results: F (3, 16) = 88.96, p < 0.0001. Statistical significance as tested by Tukey’s test, shown as asterisks

Behavioral Testing of Animals Treated with ICV STUB

To determine if the effects seen in the recovery of electrophysiological deficits translated to functional recovery, behavioral tests were performed on Ube3a^m−/p+ rats ICV injected with rAAV-STUB, rAAV-hUBE3A, or PBS. After 8 weeks, to allow for gene expression, animals were tested on the accelerating rotating cylinder (rotarod) test to determine effects on locomotor abilities. By the end of the second day of testing, STUB-treated animals showed a latency to fall comparable to WT levels, whereas hUBE3A-treated animals were indistinguishable from the PBS-treated control group (Fig. 6a). Animals were also examined for hind limb clasping. Although there was a trend for recovery, the rAAV-hUBE3A group was not significantly different from the PBS group (Fig. 6b). On the other hand, the rAAV-STUB group showed significant improvement and a recovery to WT levels. Both tasks suggest improved motor function with the rAAV-STUB treatment but not with the rAAV-hUBE3A treatment. To examine the effects of ICV treatment on associative learning and memory deficit in the Ube3a^m−/p+ animals, fear conditioning testing was performed. Results showed a partial recovery in the freezing behavior of the STUB and UBE3A groups above the PBS control group but not completely to wildtype levels (Fig. 6c), suggesting an improvement in learning and memory in the treated animals.

Fig. 6.

STUB improves the behavioral deficits present in the AS rat. ICV treatment of AS rats with STUB shows an improvement in 3 different behavioral tests. a ICV UBE3A treatment shows no improvement in motor function as measured by Rotarod. STUB treatments result in an improvement in motor learning on day two to wild type levels. The UBE3A and PBS groups demonstrate a lack of learning on day 2. A reduced latency to fall was found between UBE3A and PBS groups in trials 7 and 8. Repeated measures ANOVA performed (F (21, 434) = 1.679 p = 0.031). b The hind limb clasping phenotype is recovered with ICV STUB treatment but not UBE3A treatment. ANOVA results indicate a difference between groups (F (3, 52) = 7.242). Tukey’s multiple comparison test found a difference between wildtype and both UBE3A (p = 0.017) and PBS (p = 0.002). The STUB group had a score equivalent to wildtype. c ANOVA results indicate a difference between groups (F (3, 47) = 3.121). STUB and UBE3A both result in a partial improvement of associated learning as tested by contextual fear conditioning; however, the magnitude of improvement is greater with STUB treatment. Tukey’s multiple comparison test found a difference between PBS and wild type groups (p = 0.0222) but no difference between other groups. *p < 0.05; **p < 0.01 (hind limb clasp: wildtype n = 14, STUB n = 13, UBE3A n = 15, PBS n = 14; Rotarod: wildtype n = 13, STUB n = 18, UBE3A n = 17, PBS n = 18; fear conditioning: wildtype n = 13, STUB n = 13, UBE3A n = 12, PBS n = 13)

Discussion

In this study, we have demonstrated an improved AS gene therapy using two rodent models of AS through our innovative secreted UBE3A construct delivered via rAAV. This novel secreted form of human UBE3A variant 1 showed expression in cell culture and in vivo before functional testing was performed. After adding modifications allowing for the secretion and reuptake of the protein, we first sought to verify that these changes did not affect its treatment potential in the AS mouse. Our initial functional examination in vivo showed that the STUB construct did not lose the ability to recover the LTP deficit in the AS mouse model, indicating that it was suitable for further study. More importantly, when we performed a unilateral hippocampal injection of STUB, we were able to improve LTP in both the ipsilateral and contralateral hemispheres. This supports the notion that STUB was able to secrete to the contralateral hemisphere and is thus acting in a manner consistent with its design. We do not believe that the improvement seen on the contralateral side is due to anterograde or retrograde transport of the rAAV, which has previously been reported [20–23], because the non-secreted hUBE3A failed to show improvements in the contralateral hippocampus. A second interesting aspect of this study is the level of expression on the contralateral side. By immunohistochemistry, there was little increase in the levels of detectable UBE3A protein in the contralateral hippocampus, which might suggest that we do not need a detectable increase in the level of UBE3A to achieve a therapeutic effect. This may suggest that a low level of transduction could be sufficient to have significant effects in vivo. Further studies exploring the levels of secreted hUBE3A protein using microdialysis could help determine the minimally required levels of UBE3A protein needed for therapeutic improvements. It is important to point out that the recovery seen could be due to an undetermined mechanism such as simply increasing overall synaptic function, which results in an increase in LTP.

One major roadblock to translating gene therapy from animal models into the patient is addressing the significantly larger volume of the human brain alongside restrictions on the amount of vector that can be introduced to patients. In addition to the engineering of STUB, we increased distribution through ICV injections. The brain’s ability to transport molecules through cerebrospinal fluid throughout the brain makes targeting the ventricles an obvious strategy for gene replacement therapy. Indeed, several groups have had success in using ICV injections to enhance rAAV-based treatments [24–26]. CSF administration of rAAV vectors is also currently being explored in clinical trials for other neurological disorders (e.g., Study NCT03634007). We show here a similar increase in distribution using ICV injections with the use of rAAV-STUB and rAAV-hUBE3A. Both STUB and hUBE3A vectors showed significant improvements in hippocampal LTP, suggesting that this delivery method could be advantageous for a therapeutic delivery route. The STUB vector also showed additional improvements over the hUBE3A vector, with improved recovery of behavioral deficits in the Ube3a^m−/p+ rat model. This rescue suggests that secretion of hUBE3A adds an additional enhancement to the vector design above native hUBE3A.

One potential explanation for the improved function of STUB is an increase in extracellular UBE3A protein levels. Our recent discovery of extracellular UBE3A protein [12] hinted at the possibility of an important extracellular function of UBE3A protein and suggested to us that the release of UBE3A protein from cells could have potential effects on neighboring cells. If this is the case, it might be unnecessary to transduce every cell to recover UBE3A function in the CNS. In fact, our results show that by treating one hemisphere with STUB, recovery of hippocampal LTP is found both ipsilaterally and contralaterally. We hypothesize two mutually non-exclusive explanations for the results demonstrating STUB as superior to non-secreted UBE3A in ICV injections: (1) STUB increases the spread of the UBE3A protein and with the TAT sequence, allows uptake by neighboring cells resulting in an increase in the number of cells affected by the therapy; and (2) STUB increases the extracellular pool of UBE3A which is important for improved synaptic function. These potential mechanisms would require further studies to elucidate. It is likely that UBE3A has important functions in many compartments of the neurons and that the cell-penetrating peptide is required for UBE3A’s intracellular functions. However, if UBE3A has a critical extracellular function, then STUB has a greater potential to increase the levels of extracellular UBE3A protein. Additionally, this would also benefit neighboring non-transduced cells. Further studies exploring a purely secreted UBE3A may help reveal if it is primarily one phenomenon over the other or a synergetic effect of the two.

It is important to emphasize that adult rats were used in this study. The animals were injected at 3 months of age, an age representing young adulthood [27]. Previous work with a Cre-dependent inducible Ube3a AS mouse suggested that there are distinct developmental windows in which Ube3a restoration can restore AS deficits [28]. In their study, they reinstated Ube3a at 4 distinct ages, embryonic, juvenile (3 weeks), adolescent (6 weeks), and adult (14 weeks), and then tested the animals at 10, 16, 22, and 28 weeks of age, respectively. LTP was rescued at all ages. Motor deficits were rescued in embryonic and juvenile-treated animals with some improvement in adolescent animals but no improvements in adult animals. Interestingly, anxiety, repetitive behavior, and seizures were only rescued when Ube3a was reinstated during early development. This suggested that there may be a time window for which an effective therapy could be given to patients and that adult AS patients may have limited recovery of most aspects associated with this disease. This may explain the improvement in learning in memory but not complete recovery in fear conditioning task in this study. Earlier intervention with these vectors could give a much greater rescue of the AS phenotype and should be explored in future studies.

We do not see the rescue of the motor deficits in the adult AS rat with the injections of rAAV-hUBE3A, a finding that is consistent with the conditional mouse study as well as our previous rAAV delivery of mouse Ube3a. However, we surprisingly report a rescue of the motor deficits in adult rats with the injection of rAAV-STUB. This is a highly significant finding as no other therapies have shown motor rescue in adult animals, with the exception of the recent exciting work by Adhikari et al. (2021) demonstrating a rescue using a lentivirally transduced stem cell transplant [29]. Whether this motor rescue in the STUB but not the UBE3A-injected animals is due to a significant improvement in the number of cells receiving the therapeutic or due to an increase in the extracellular UBE3A remains to be seen. Understanding the molecular mechanism for this rescue would certainly advance our understanding of UBE3A biology and its function in neurons.

The work presented here in the AS rat (summarized in Table 1) builds upon our previous work in the AS mouse using a gene replacement approach and our recent observations of extracellular UBE3A [8, 30]. Through the addition of secretion and cell-penetrating peptides, this novel STUB construct can increase the effectiveness of treatment beyond locally transduced neurons by supplying the replacement of UBE3A to cells that were not transduced with the viral administration. Behavioral rescue with STUB to a greater extent than standard AAV-mediated gene replacement raises an interesting premise for the treatment of specific learning and memory phenotypes observed in the Ube3a deficiency rat model. Either the greater distribution offered with STUB or the presence of extracellular UBE3A protein is responsible for this observation. The rescue of contralateral LTP following STUB treatment would suggest the latter explanation is the more likely. These results introduce a new consideration when designing a gene therapy treatment for human AS and shed much-needed light on the yet unknown UBE3A mechanism of action underlying the severe effects of UBE3A deficiency.

Table 1.

Comparison of AAV9-UBE3A and AAV9-STUB phenotypic rescue in the AS rat

Test/outcome measure		UBE3A	STUB
HPC injection	• Ipsilateral HPC LTP • Contralateral HPC LTP	Improved	Improved
		No improvement	Improved
ICV injection, HPC LTP		Partial rescue	Rescue
Rotarod		No rescue	Rescue
Hindlimb clasp		No rescue	Rescue
Contextual fear conditioning		Partial rescue	Partial rescue

Supplementary Information

Below is the link to the electronic supplementary material.