Novel Cross-Correction–Enabled Gene Therapy for CDKL5-Deficiency Disorder

Neurological disorders (ND) are the leading cause of disability and the second leading cause of mortality worldwide, with a significant increase in their incidence predicted in a growing and aging global population [1]. Consequently, the need for innovative paradigm-shifting strategies for advancements in treatment modalities remains urgent. To date, approximately 1700 identified monogenic neurological disorders (MND) afflict millions of people, encompass a broad range of brain symptoms, and present a substantial health threat and care burden [1]. Monogenic neurological disorders (MND) are attributable to inherited genetic or de novo mutations in a single gene that disrupt nervous system development, impair neuronal function, or lead to neurodegeneration.

Current treatment modalities for monogenic diseases have largely focused on complementation of the affected cells, organs, or entire organisms with a functional (1) gene copy (e.g., gene replacement therapy (GRT)) or (2) protein (e.g., enzyme replacement therapy (ERT)). However, limitations associated with each of these therapeutic approaches restrict their use in treating diseases of the central nervous system (CNS). Gene replacement therapy (GRT) utilizes viral vectors to induce long-term production of functional proteins in infected cells. Major hurdles associated with viral GRTs in the treatment of CNS disorders include (1) efficient targeting of the vector to the CNS without induction of genotoxicity and systemic toxicity, (2) insufficient transduction efficiency resulting in limited numbers of brain cells that express the therapeutic protein, and (3) the generation of neutralizing antibodies in response to previous infection by the viral vector, which can inhibit efficient gene transduction [2]. Classic ERT is largely used for treating lysosomal storage diseases, but enzymes do not cross the blood–brain-barrier sufficiently to render a therapeutic benefit in the CNS [3]. For neurological disorders, direct enzyme administration to the CNS (e.g., intracisternal or intrathecal) must be continuous or repeated, which presents significant practical difficulties and risks to patients [4, 5].

Notably, in this issue of Neurotherapeutics, Medici et al. [6] combine technical aspects of GRT and ERT to deliver to the CNS cyclin-dependent kinase-like 5 (CDKL5), a protein with cytosolic and nuclear localization. The virally delivered transgene encodes the therapeutic protein, which was engineered to be secreted and have cell-penetrating properties. Using this approach, virally transduced cells constitutively produce and secrete the therapeutic protein, which can additionally traffic through the extracellular space making it available for uptake by non-transduced cells (cross-correction). By means of this novel approach, Medici et al. [6] demonstrate in vivo efficacy of this cross-correction strategy and its therapeutic superiority over conventional gene delivery, where the cytosolic/nuclear therapeutic protein is exclusively expressed and restricted to virally transduced cells.

Cross-correction has previously been observed in experiments using viral delivery of genes encoding lysosomal proteins. Massaro et al. suggest that cross-correction is responsible for widespread correction of cells in the CNS in several pre-clinical lysosomal GRT studies where the number of virally transduced cells was limited [7]. The phenomenon of lysosomal cross-correction is a direct consequence of endogenous cellular mechanisms that enable sorting of lysosomal proteins. Soluble lysosome-targeted proteins carry several signals that facilitate their trafficking from the endoplasmic reticulum, via the trans-Golgi Network, to lysosomes. The post-translational modification of proteins by mannose 6-phosphate (M6P) residues is certainly the best characterized sorting signal for lysosome-targeted enzymes. The M6P receptor (MPR), located in the membrane of late Golgi compartments and endosomes, binds M6P-tagged lysosome-targeted proteins and, through fusion with lysosomes, introduces its protein cargo into the lysosomal compartment [8]. However, it has been demonstrated that up to 10% of the MPR escape their endosomal fate and become integrated into the cell membrane [9, 10]. Additionally, 5–20% of lysosome-targeted proteins escape their lysosomal fate and are secreted. Binding of secreted lysosome-targeted proteins to the MPR on the cell surface leads to (1) internalization of the receptor and their M6P-tagged cargo in a clathrin-dependent manner, and (2) integration of the formed vesicles into the endosomal compartment, which then allows trafficking of the internalized proteins to lysosomes [11, 12]. Thus, secreted lysosomal proteins/enzymes are recaptured by MPRs on the plasma membrane and are delivered to lysosomes through MPR-mediated endocytosis. This mechanism of lysosomal protein uptake from the extracellular space has been therapeutically harnessed by ERT for lysosomal storage disorders [13, 14].

ERT delivery of exclusively lysosomal proteins can theoretically be surmounted through use of short (usually under 30 amino acids) cell-penetrating peptides (CPPs). When fused to therapeutic ‘cargo’, the CPP-cargo fusion protein can translocate across the cell membrane [15]. CPPs can be synthetic peptides or be derived from natural proteins, such as the HIV trans-activator of transcription protein TAT or penetratin (a peptide derived from the drosophila homeobox protein, antennapedia, also referred to as Antp) [16]. While the mechanism for CPP-mediated protein internalization has not yet been deciphered, two general pathways are proposed (1) energy-independent translocation through the plasma membrane to the cytosolic compartment, and (2) endocytosis. For various CPPs, several forms of endocytosis have been demonstrated [17–19]. Using the endocytic pathway, the therapeutic potential of the CPP-cargo fusion is dependent on its ability to escape the endosomal compartment, to elude lysosomal degradation, and to reach the desired subcellular compartment (e.g., cytosol). Several recent studies show linkage of several transactivator of transcription (TAT) peptides to one another, or to other CPPs, improves cytosolic delivery [20–25].

CPPs that facilitate translocation of CPP-fusion protein into the cytosol open the door for their application in cross-correction-enabled gene therapy of cytosolic disorders. In combination with an N-terminal secretion signal, cross-correction is enabled for non-lysosomal GRT via (1) secretion of the therapeutic CPP-fusion protein from virally transduced cells, and (2) cytosolic uptake of the fusion protein by both virally transduced and non-transduced cells. Thus, theoretically, the limitation of CNS-directed gene therapy to reach only a limited number of brain cells can be overcome. While CPPs have been employed in many studies involving protein therapy, such as ERT, limited studies exist in which CPPs have been encoded by a transgene. Previously, Koutsokeras et al. [26] demonstrated cross-correction of a NF-kB inhibitor fused with the cell-penetrating peptide TAT and a secretory signal. In this study, investigators engineered conditionally immortalized DBA/1 embryonic fibroblasts to produce a fusion protein, which was shown to penetrate co-cultured 57A HeLa reporter cells and have an anti-inflammatory effect in mouse models of NF-kB-mediated paw inflammation. Similarly, Ma et al. [27] fused brain-derived neurotrophic factor (BDNF) to the CPPs TAT and HA2, derived from a subunit of influenza hemagglutinin, showing improved CNS spread of the naturally secreted BDNF after nasal virus delivery. Medici et al. [6] virally delivered the first cross-correction-enabled transgene for treatment of a neurological disorder caused by the deficiency of a protein (i.e., CDKL5) with cytosolic and nuclear localization.

Cyclin-dependent kinase-like 5 (CDKL5) disorder is a rare X-linked neurodevelopmental disease caused by mutations in the CDKL5 gene that lead to a deficiency in CDKL5 protein. Clinically, CDKL5 encephalopathy involves severe neurodevelopmental impairment, intellectual disability, motor impairment and potential respiratory dysregulation. The disorder is characterized by early-onset epileptic seizures, hypotonia, abnormal eye tracking, and severe visual impairment [28–31]. Since CDKL5 disorder is a monogenic disease, it is a prime candidate for ERT or GRT. In fact, Trazzi et al. recently developed a therapy using systemically injected recombinant TATk-CDKL5 protein. Using a CDKL5-deficient mouse model, the authors demonstrated blood–brain-barrier passage of the TAT fusion protein. This treatment was shown to rescue various neuroanatomical and behavioral defects, including breathing pattern and visual responses [32]. Compelled by these encouraging findings, the group developed a secreted version of TATk-CDKL5, which was vectorized for delivery by an engineered version of adeno-associated virus (AAVPHP.B).

In the novel study by Medici et al., two AAVPHP.B viruses were generated; each encoding human CDKL5 fused to an HA-tag to facilitate immuno-detection of the therapeutic protein. In one AAVPHP.B virus, additional N-terminal elements were encoded to enable cross-correction (cross-correcting GRT). These elements include (1) the secretion signal of an immune-globulin kappa chain to facilitate secretion, followed by (2) TATk, a mutated form of TAT, which has previously been shown to improve secretion of TAT-fusion proteins [33]. The other virus lacked these cross-correction-enabling elements (conventional GRT). Comparing the effects of infection with these two viruses directly identified the impact of the cross-correction-enabling elements.

First, the authors show successful transgene expression following transduction of mouse primary hippocampal neurons by both virus preparations. Additionally, they present evidence of ex vivo cross-correction by TATk-CDKL5, which was secreted from HEK293T cells and translocated into the cytosol of a few co-cultured primary neurons. In in vivo studies, evidence of cross-correction was demonstrated following intracerebroventricular virus injections of neonatal CDKL5-deficient mice. Imaging of cells positive for CDKL5 mRNA (in situ hybridization) and protein (immunohistochemistry for HA-tag) revealed the presence of cells containing CDKL5 protein but not the respective mRNA in various brain regions; implying that non-infected cells acquired the protein through cross-correction. Compared to conventional GRT, the authors estimate that cross-correcting GRT increased the number of cells containing CDKL5 protein by approximately 6%. Notably, the authors demonstrate viral transduction of 30% of brain cells. Together these data imply that by means of cross-correction, the total number of corrected brain cells was increased by ~ 2%. While this level of enhancement could be viewed as modest, a pronounced therapeutic benefit in cross-correction enabled gene therapy was evident in vivo compared to conventional GRT.

To investigate the therapeutic benefit, the authors treated 3–4-month-old CDKL5-deficient mice to compare the effects of both conventional and cross-correcting GRT 60 days post intracarotid virus infusion. In most assays, assessing autism-like behavior, stereotypic movements, and open-field motor function, the cross-correcting GRT significantly improved the CDKL5-deficiency phenotype, whereas conventional GRT had no significant therapeutic effect. However, one observed exception was in learning behavior, where both cross-correcting and conventional GRT yielded similar results that were significant but partial in rescue. Whole-body plethysmography, used to assess breathing patterns during REM and non-REM sleep phases, showed apneas in REM sleep phases, present in CDKL5-deficient mice, were significantly reduced to WT levels in both cross-correcting and conventional GRT. By contrast, cortical visual responses measured using non-invasive optical signal imaging were normalized to WT levels only in mice that received cross-correcting GRT. In regard to CDKL5-related brain pathology, dendritic arborization, spine structure and stabilization as well as excitatory synaptic contacts and hippocampal neuron survival were examined since all were previously reported to be significantly reduced in the well-established CDKL5 deficiency mouse model [34–40]. Apical and basal dendritic length was determined to be significantly improved to WT levels, and the number of apical and basal branches showed a trend toward improvement in animals treated with cross-correcting GRT. Dendritic spine density, percentage of mature spines, number of excitatory synaptic contacts, together with the number of hippocampal neurons (in the CA1 region), were normalized by cross-correcting GRT, demonstrating a significant benefit over conventional GRT. Cross-correction GRT also simultaneously led to a significant and WT-like reduction of microglia activation compared to both untreated CDKL5-deficient mice and to conventional GRT.

The multitude of data from behavioral and functional assays, as well as from examinations of brain pathology, demonstrate the clear benefit of cross-correcting GRT over conventional GRT in the study of Medici et al. [6]. However, correction of cells was only estimated to be increased by approximately 6% in cross-correcting GRT compared to conventional GRT. This raises the question as to whether it is indeed the cross-correction-enabling strategy that led to the observed benefits or whether other factors, such as differences in viral transduction efficiency, played a role. To answer this question, the authors carried out thorough analyses of viral transduction efficiency for both cross-correcting and conventional GRTs. To this end, the authors compare CDKL5 mRNA levels in various brain regions of treated CDKL5-deficient mice and in WT mice. Virally induced CDKL5 mRNA expression differed between brain regions and to respective WT levels. However, the authors found no significant difference in virally induced CDKL5 mRNA when comparing conventional to cross-correcting GRT in each brain region. This indicates that viral transduction, though distinct in brain regions, was similar for both cross-correcting and conventional GRT, suggesting that the remarkable benefit of the cross-correcting GRT cannot be explained solely by higher infection rates or by differences in the quality of the two virus preparations.

Notably, the authors provide only a quantitative estimate of detected cross-corrected cells for the hindbrain, where viral transduction efficiency was comparatively high. It is likely that in brain regions with lower viral transduction efficiency, such as hippocampus, the number of non-infected cells that receive a detectable amount of CDKL5 protein through cross-correction is significantly lower. Nonetheless, the authors do show numerous phenotypic improvements involving the hippocampus that are clearly enhanced in the cross-correcting GRT approach. It is, therefore, likely that the methodology used to assess cross-correction underestimates the number of cells that receive a therapeutic level of protein. Specifically, non-transduced cells that received a low amount of CDKL5 through cross-correction may have remained undetectable using the immunofluorescence assays employed. Thus, based on the improved phenotypic effects of the cross-correcting GRT, it is likely that these undetectable protein quantities have a therapeutic effect.

The work of Medici et al. [6] demonstrates that enabling cross-correction via the combination of secretion and cell-penetration is an effective strategy to overcome current hurdles of viral gene therapy for diseases of the CNS. This novel approach is theoretically suitable for many CNS disorders caused by the deficiency of proteins with cytosolic and/or nuclear localization. However, the work also serves to highlight the need for continued research. Advancements in the sensitivity of methods used for the detection of therapeutic proteins will improve their quantification and provide a more accurate measure of cross-correction modalities. Additionally, safety of cross-correcting GRT, in general, and CDKL5-related GRT specifically require standardized methodologies for evaluation. In the study reviewed herein, the authors report no effects of their treatment regimen on animal well-being. The authors base their conclusion on the absence of detectable differences in mouse body weight and sleep pattern, as well as microglial cell number and cell survival between treated CDKL5-deficient mice and untreated WT mice. However, a comprehensive safety study with GRT in WT mice was not conducted.

In regard to the development of a clinical treatment for CDKL5 encephalopathy, the innovative work by Medici et al. represents a major leap forward. Nevertheless, it remains to be determined how the efficacy demonstrated in the mouse model will translate to larger animals (e.g., non-human primates) and humans. Further optimization of the therapeutic approach is warranted for the following reasons. First, the authors used an engineered AAV serotype (PHP.B) with a greatly enhanced transduction profile in particular mouse backgrounds including the one used in their study. While this certainly benefited this proof-of-concept study, it has been shown that other serotypes are more suitable for the treatment of non-human primates and likely humans [41]. Secondly, even under optimal conditions for murine gene therapy realized by Medici et al. [6], many of the phenotypic parameters of the CDKL5 mouse model were only partially rescued. Since the physical distance to be overcome by cross-correcting proteins presents a greater challenge in larger brains, the elements of cross-correction (secretion signal and cell-penetrating peptide) will certainly require optimization to achieve a significant clinical benefit in humans. Fortunately, biosynthetic research over the last 50 years has provided a multitude of options in regard to secretion signals and cell-penetrating peptides/proteins. The latter include effective receptor-mediated protein translocation systems, such as non-toxic variants of diphtheria toxin [42, 43] and antibody fragments [44, 45].

Taken together, the promising findings of Medici et. al. [6] indicate that cross-correcting fusion proteins can enhance the efficacy of clinical GRT for CDKL5 encephalopathy and strongly suggest their potential use for other CNS-related diseases caused by deficiencies of proteins with cytosolic and/or nuclear localization. Thus, the novel work of Medici et al. [6] serves as an example of innovative paradigm-shifting advancements in treatment modalities required to counter the global burden of neurological disorders.

Supplementary Information

Below is the link to the electronic supplementary material.