AAV Gene Replacement Therapy for NF1-related Tumors

Introduction

Neurofibromatosis type 1 (NF1) is a common autosomal dominant genetic disorder (1 in 2,500) caused by loss of function in the NF1 gene, encoding the tumor suppressor neurofibromin, a Ras-GTPase–activating protein. Loss of NF1 leads to hyperactive Ras signaling, driving benign and malignant tumor development (e.g., plexiform neurofibromas, malignant peripheral nerve sheath tumors (MPNST), low-grade gliomas). Current therapies like MEK inhibitors (selumetinib, mirdametinib) can control some tumor growth but require continuous dosing, cause toxicities, and don’t address the genetic root cause.

Adeno-associated virus (AAV) vectors are clinically validated for gene replacement in several monogenic diseases, but NF1 gene therapy faces two major hurdles:

• The NF1 cDNA is ~8.4 kb — too large for AAV packaging.
• Natural AAV serotypes have poor tropism for NF1 tumor cells.

This study develops a truncated NF1 therapeutic payload and engineers a novel AAV vector with high NF1 tumor selectivity and reduced liver transduction, aiming to establish a viable systemic gene therapy for NF1-related tumors.

Methods

The authors combined payload engineering and vector evolution:

1. Payload Optimization:

The researchers optimized the NF1 payload by testing truncated GRD versions (230-367 amino acids) fused to hypervariable regions (HVRs) from RAS isoforms (HRAS, NRAS, KRAS4A, KRAS4B) of varying lengths, selecting the most effective based on expression, RAS pathway inhibition (via Western blotting for pERK1/2), and cytotoxicity in NF1-null cell lines (e.g., MPNST ST88-14, pNF ipNF9511bc) versus normal Schwann cells (ipn02.3λ). Payloads were packaged in self-complementary AAV vectors with a CBh promoter for in vitro assays like cell viability (WST-8), apoptosis (Annexin V flow cytometry), and immunofluorescence (anti-HA for localization).

2. Capsid Engineering:

For vector engineering, a capsid DNA shuffling library was created from 12 natural AAV serotypes (1-11, 32/33), with diversity ~3.3 × 10^6, and selected over two rounds in orthotopic ST88-14 xenograft mice (implanted in sciatic nerve of NSG mice). Top candidates (e.g., 557-2) were refined by inserting a randomized heptamer peptide library into the VR-VIII loop (N588-T589) using a Cre-dependent method, followed by two rounds of in vivo biopanning in ST88-14-Cre tumors. Enrichment was analyzed via NGS sequencing, and candidates validated by GFP transduction (IHC, flow cytometry) and biodistribution (qPCR for viral genomes).

3. Validation:

In vivo efficacy was tested in xenograft models (ST88-14, RHT92 MPNST; ipNF03.3 pNF; LN229 glioma; NF1-/- iPSC-derived neurofibromas) with IV AAV injections (10^11-10^12 vg/mouse), monitoring tumor growth via IVIS luciferase imaging. Schwann cell differentiation was assessed in NF1-/- iPSCs (D12 line) co-cultured with rat neurons, using markers like S100B (IF) and MPZ (myelination). Seroprevalence was evaluated in human sera (n=40-50) via ELISA for binding antibodies and a luciferase reporter for neutralizing antibodies. Statistical analyses used two-tailed t-tests, with biological replicates (n=3-6 per group).

Key Results

A. Optimizing the GRD Payload

The research team first had to design a therapeutic gene payload that was both effective and small enough to fit within an AAV. Since the full-length NF1 gene was too large, they focused on its critical functional domain, the GAP-Related Domain (GRD). They systematically tested GRD constructs of varying lengths, each fused to membrane-targeting sequences from different RAS proteins to enhance its localization. Figure 1 illustrates the outcome of this optimization: the fusion of a 333-amino-acid GRD with the C-terminal domain of KRAS4B (dubbed GRDC24) emerged as the superior candidate. GRDC24 not only showed the strongest localization to the plasma membrane (Fig. 1F), where active RAS resides, but also potently suppressed RAS pathway signaling (evidenced by reduced phosphorylated ERK, Fig. 1G) and induced apoptosis specifically in NF1-deficient cells (Fig. 1H), while having a minimal impact on normal Schwann cells (Fig. 1D). This established GRDC24 as a potent and selective “miniNF1” therapeutic.

Figure 1. Optimization of NF1 GRD as payload for rAAV vectors

B. Schwann Cell Differentiation Rescue

The next set of experiments assessed whether GRDC24 could restore Schwann cell differentiation, a key pathogenic defect in NF1. Using an iPSC-based differentiation model, NF1−/− iPSCs (D12) were directed to form neural crest cells and then Schwann cells. Normally, NF1−/− cells exhibit excessive ERK activation and fail to differentiate properly. AAV-mediated expression of GRDC24 corrected these defects: treated neural crest cells showed reduced pERK and lower cell viability in NF1−/− but not wild-type cells (Fig. 2C–D). When induced to differentiate, NF1−/− cells expressing GRDC24 acquired normal Schwann cell morphology and S100B marker expression, and were able to myelinate dorsal root ganglion neurons, mimicking wild-type behavior (Fig. 2E). This provided a critical functional validation that the engineered mini-gene can compensate for NF1 loss in relevant cell lineages.

Figure 2. Expression of GRDC24 in NF1−/− cells rescued Schwann cell differentiation

C. Engineering the AAV-NF (K55) Vector

With the payload established, the authors turned to the vector delivery challenge. Natural AAV serotypes show poor tropism for NF1 tumors and high liver uptake, so they applied capsid DNA shuffling and in vivo selection in human ST88-14 xenograft mouse models to evolve tumor-targeted vectors. After two rounds of selection, clone 557-2 was enriched and validated (Fig. 3B). When injected systemically, AAV-557-2-GFP achieved stronger tumor transduction compared to AAV9, and importantly, showed markedly reduced liver biodistribution while enhancing delivery to multiple NF1 tumor xenografts (ST88-14, ipNF03.3, RHT92) (Fig. 3C–D). When packaging GRDC24, 557-2 slowed tumor growth significantly, though effects waned over time, likely due to rapid tumor proliferation (Fig. 3E). This step demonstrated that capsid engineering can yield vectors with selective NF1 tumor targeting and reduced off-target exposure.

Figure 3. Capsid DNA shuffling and selection in NF1 xenograft mice

To further improve tropism, the team inserted random heptapeptides into the VR-VIII loop of 557-2 and performed iterative in vivo selection. Among enriched candidates, K55 was identified as the top-performing mutant. Structural modeling showed the SKVPLPN peptide exposed in VR-VIII, and transmission electron microscopy confirmed normal capsid morphology (Fig. 4C–D). K55 demonstrated superior transduction efficiency in NF1 xenograft models, including MPNST (ST88-14, RHT92), patient-derived xenografts, and NF1-deficient glioma (LN229), achieving 30–60 % tumor cell transduction in many models (Fig. 4F–G). At the same time, K55 maintained low liver uptake and showed negligible transduction in non-NF1 tumors, confirming a selective targeting profile.

Figure 4. Improving capsid 557-2 via selection of random peptide library

D. Therapeutic Efficacy in Animal Models

Finally, the study evaluated therapeutic efficacy in vivo. In ST88-14 xenografts, a single systemic injection of AAV-K55-GRDC24 led to a significant but transient reduction in tumor growth, whereas two injections given one week apart achieved more durable suppression and higher transgene expression within the tumor (Fig. 5A–C). Similar efficacy was observed in the RHT92 xenograft model (Fig. 5D–E). In contrast, AAV9-GRDC24 showed no benefit and caused systemic toxicity, likely due to high liver transduction. Combining K55-GRDC24 with selumetinib enhanced therapeutic impact, indicating potential for synergistic regimens. Collectively, these results demonstrate that systemic delivery of a rationally engineered AAV vector carrying an optimized mini-NF1 gene can achieve meaningful tumor control in aggressive NF1 xenograft models.

Figure 5. AAV-K55-GRDC24 showed anti-tumor efficacies in 2 NF1 xenograft models

Conclusion

The study successfully developed a miniaturized NF1 transgene (GRDC24) that effectively suppresses RAS signaling and restores Schwann cell differentiation. Using directed evolution through capsid DNA shuffling and peptide library screening in a human xenograft mouse model, the authors engineered an AAV vector (K55) with selective and efficient tropism for NF1 tumors and reduced off-target liver delivery. Systemic intravenous administration of AAV-K55-GRDC24 significantly inhibited NF1 tumor growth in multiple xenograft models with a favorable safety profile. The findings show a viable gene replacement strategy for NF1 tumors, overcoming major barriers like payload size and tumor targeting.

This study marks a critical advance in gene therapy for NF1-related tumors, a genetically defined cancer with limited treatment options. The strategy and vectors described here may serve as a blueprint for developing gene therapies in other monogenic tumor syndromes and complex cancers with limited current therapeutic options. It also highlights the necessity of vector engineering for tumor specificity to enhance safety and efficacy in systemic gene therapies for cancer. This comprehensive paper establishes a strong foundation for future clinical translation of AAV-based gene replacement therapies in NF1 and potentially other oncology indications driven by tumor suppressor gene loss.

References:

https://pmc.ncbi.nlm.nih.gov/articles/PMC12480499/