Development of an adeno-associated virus vector for gene replacement therapy of NF1-related tumors

Abstract

Neurofibromatosis type 1 (NF1) is a tumor predisposition syndrome caused by alterations in NF1 gene that lead to tumor growth throughout the nervous system, which can cause morbidity and mortality, and transform to malignancy. NF1 gene replacement therapy, though promising, is hindered by NF1 gene's large size and delivery challenges. We introduced a membrane-targeted, truncated neurofibromin comprising the GAP-related domain (GRD) fused to the KRAS4B C-terminal domain, which effectively inhibits the RAS signaling pathway and restores Schwann cell differentiation in an NF1 iPSC-derived model. For systemic application, we engineered an adeno-associated virus (AAV) vector using in vivo capsid evolution through sequential DNA shuffling and peptide library screening in a NF1 xenograft mouse model. This tailored vector, AAV-NF, exhibits greatly reduced liver uptake, enhanced tumor targeting across various NF1-related MPNST, neurofibromas and glioma models, and therapeutic efficacy in xenografts of MPNST. This study not only advances a viable AAV vector for NF1 treatment but also outlines a replicable strategy for vector and payload development in other monogenic and tumor-associated disease manifestations.

Fig. 1. Developing a transgene for NF1 gene replacement therapy.

A Different RAS hypervariable region (HVR) sequences were attached to NF1 GRD for optimization. B NF1 null cell lines ST88-14 and ipNF9511bc, and immortalized normal human Schwann cell line ipn02.3λ, were used for the following experiments. Anti-NF1 western blotting showed the status of neurofibromin expression. C Inhibition of NF1 cells by GRD fused with various RAS HVR sequences. ipNF9511bc cells were transfected with indicated GRD constructs packaged in AAV-DJ at different multiplicity of infection (MOI). After 3 days, viable cells were measured via WST-8 at Abs 450 nm, with AAV-DJ-EGFP as control (100%). GRD-KRAS4B-C24 showed the most effective suppression of NF1 null cells. Three biological replicates were used. Data are presented as mean values with Standard Deviation (SD) and analyzed by two-tailed t-test. D Lower impact of GRDC24 construct in normal human Schwann cells. GRDC24 construct package in AAV was transfected in ipn02.3λ cells and compared with AAV-DJ-EGFP as control. Three biological replicates were used. Data are presented as mean values with SD and analyzed by two-tailed t-test. E Optimization of NF1 GRD domain. Different lengths of GRD sequences surrounding the 230 AA core GRD sequence were tested in ipNF9511bc cells and the expression as well as the inhibition of pERK1/2 were revealed by western blotting. The 333 AA version (NF1 AA 1200-1532) showed the highest expression and most potent inhibition of pERK1/2. F Membrane localization of GRDC24 in NF1 cells. ipNF9511bc cells were infected with AAV carrying 2HA-GRD or 2HA-GRDC24 (KRAS4B) transgene and anti-HA immunofluorescent staining showed increased localization of GRDC24 in the plasma membrane (white arrow heads), in comparison to the diffused expression pattern of GRD alone. Scale bar = 20 µm. G Suppression of phosphorylated ERK pathway by GRDC24 in NF1 cells. ipNF9511bc cells were transfected with AAV-GRDC24 and harvested after 36 hr. GRDC24 inhibited pERK1/2 more potently than GRD alone. H GRDC24 induced apoptosis in NF1 cells. ipNF9511bc cells were transfected with AAV-GRDC24 and after 24 h cells were stained with annexin V.

Fig. 2. Expression of GRDC24 rescued the NF1^−/− neural crest (NC) cells’ differentiation to Schwann cells.

A Western blotting of NF1 protein in iPSC derived from normal human fibroblast (FiPS) and the isogenic FiPS cells with NF1 knockout (NF1^−/−, D12). B FiPS and NF1^−/− iPSC cells were differentiated to NC cells. Flow cytometry of NC marker NGFR confirmed the NC differentiation (blue) of FiPS and NF1^−/− cells. C NF1^−/− NC cells were transfected with AAV-DJ-GRDC24 for 48 h, which led to suppression of pERK1/2. D Inhibition of NF1^−/− NC cells by GRDC24. FiPS and NF1^−/− NC cells were transfected with GRDC24 constructs packaged in AAV-DJ at different multiplicity of infection (MOI). After 3 days, viable cells were measured, with AAV-DJ-EGFP as control (100%). GRDC24 showed the most effective suppression of NF1^−/− NC cells. Three biological replicates were used. Data are presented as mean values with SD and analyzed by two-tailed t-test. E GRDC24 restored the differentiation of NF1^−/− SC. FiPS and NF1^−/− NCs were transfected by AAV-DJ-GRDC24 at indicated time before or shortly after SC differentiation and co-cultured with rat dorsal root ganglia (DRG) neuron in SC differentiation media for 20 days as depicted in the left diagram. Immunofluorescence staining of neuron marker βIII-tubulin and myelination marker MPZ showed myelinated SC along the neuron (right panel). Scale bar = 20 µm.

Fig. 3. Capsid DNA shuffling and selection in NF1 xenograft mice created AAV vectors for NF1 tumors.

A Example of an orthotopic xenograft ST88-14 tumor grown in the mouse sciatic nerve. ST88-14 cells were injected in the sciatic nerve of NSG mice and tumor was formed in the nerve. B Capsid 557-2 is the top candidate selected from a capsid DNA shuffling library. A DNA shuffling library was created from 12 natural serotypes of AAV capsids and packaged in 293T cells. Selection in ST88-14 tumor-bearing mice resulted in 557-2, whose resemblance to the template capsids was shown in the graph generated by Xover 3.0. The solid black lines represent the parts of the 557-2 sequence found in the capsids of natural serotypes. C Improved transduction by 557-2-GFP in ST88-14 tumor compared to AAV9-GFP. AAVs were injected IV to the ST88-14 tumor-bearing mice at the dose of 5 × 10¹¹ vg (viral genome)/mouse and after 2 weeks tumors were harvested for anit-GFP IHC. Scale bar = 10 µm. D AAV-557-2 showed a significantly reduced liver retention with more tumor distribution in a panel of three orthotopic NF1 xenografts. Three xenografts of NF1-related tumor cells, ST88-14, ipNF03.3 and RHT92, were injected in the sciatic nerve of NSG mice. AAV9-GFP or 557-2-GFP were injected IV to the mice at a dose of 5 × 10¹¹ vg. Tissues were harvested after two weeks and quantitative PCR was used to determine titers of AAV genome in different tissues with the linearized AAV plasmid as standard. Three mice were used in each data point. Data are presented as mean values with SD and analyzed by two-tailed t-test. E AAV-557-2-GRDC24 significantly slowed down the ST88-14 tumor growth in mice. ST88-14 cells were implanted in the sciatic nerve in the NSG mice and after 14 days, tumor sizes were assessed by IVIS imaging and AAV 557-2-GRDC24 was injected IV to at a dose of 1 × 10¹² vg per mouse. Two weeks later, the tumor growth was imaged and compared to the control. Data are presented as mean values with SD and analyzed by one-tailed t-test. Con: n = 14 mice; 557-2: n = 13 mice.

Fig. 4. Improving capsid 557-2 via selection of random peptide library created AAV-K55, a vector preferentially delivers in NF1-related tumors.

A A predicted structure of 557-2 generated by AlphaFold 3. The position between residue N588 and T589 in the VR-VIII loop were identified as the insertion site of a 7mer random peptide library. B A 7mer random peptide library was selected in xenograft mice carrying ST88-14 tumor with Cre recombinase expression. ST88-14-luc-cre cells were implanted in the sciatic nerve of NSG mice and AAV 557-2 library was injected IV. After 2 weeks, tumors were harvested and the capsid sequences were recovered by a Cre-dependent PCR. The recovered capsid DNA was cloned into library vector, which was used for the next round of selection. After two rounds of selection, NGS was done to determine the enrichment of mutant capsids. C A predicted structure of the top candidate K55 generated by AlphaFold 3. The inserted 7mer peptide is displayed in the VR-VIII loop. D TEM pictures of AAV vectors K55-GFP and K55-GRDC24. Scale bar: 100 nm. E Markedly improved transduction of K55-GFP in ST88-14 and RHT92 xenograft tumors. A dose of 1 × 10¹² AAV9-GFP or K55-GFP was injected IV in the mice bearing tumors in the sciatic nerve. After two weeks, tumors were harvested and the expression of GFP was evaluated by anti-GFP IHC. RHT92 is a patient-derived NF1-related MPNST cell line. Scale bar: 50 µm. F AAV-K55-GFP showed significantly reduced liver retention and substantially improved tumor transduction compared to the benchmark AAV9-GFP, with another candidate AAV-K57-GFP as comparison, in a panel of two ST8814, one ipNF03.3 and one RHT92 tumors. AAVs were injected IV at the 10¹² vg dose and viral genome was quantified in tissues using qPCR. Four mice were used and data are presented as mean values with SD and analyzed by two-tailed t-test. G. Distribution of AAV-K55 in NF1 xenografts and PDXs, including LN229 NF1^−/− glioma, and a panel of solid human xenograft tumors. AAV-K55-GFP was injected IV at the dose of 10¹² vg and tumors were harvested and GFP-positive tumor cells were quantified by flow cytometry. Data are presented as mean values with SD. S462 and JH-0-031: n = 6 xenograft tumors; RHT92, JH-2-002 and LN229: n = 4 xenograft tumors; all others: n = 3 xenograft tumors, as biological replicates.

Fig. 5. AAV-K55 transduced S100B + pNF xenograft tissues and showed a seroprevalence profile similar to that of AAV9.

A A pNF iPSC-derived tumor (3MM, 3PNF_SiPSsv_MM_11, NF1^−/−) implanted together with pNF patient’s fibroblasts (FB, NF^+/-), was shown to be substantially transduced by AAV-K55-GFP via IV administration. Xenograft was harvested two weeks after the AAV treatment. Right image: H&E staining. Right image: anti-GFP IHC. Scale bar = 50 µm. B AAV-K55-GFP transduced mostly the S100B-positive population (tumor) in 3MM-FB neurofibroma xenograft. A section was stained with anti-GFP (green) and anti-S100B (red) antibodies. Scale bar = 50 µm. C Circulation of AAV-K55-GFP in mouse plasma. 10¹² vg of AAV were injected IV in NSG mice and viral genome in the plasma was quantified via serial blood draw and qPCR (n = 3 mice). Data are presented as mean values with SD. D A direct comparison of the AAV9-binding antibody (B-Ab) profile and AAV-K55-binding antibody profile in a panel of heathy human serum samples. E Profiles of AAV-K55-binding antibody (B-Ab) and neutralizing antibody (N-Ab) in healthy human serum samples. AAV K55-luc was used for the assays.

Fig. 6. AAV-K55-GRDC24 showed anti-tumor efficacies in 2 NF1 xenograft models.

A Efficacies of AAV K55-GRDC24 in ST88-14 xenograft tumor. NSG mice were implanted with ST88-14 tumor cells in the sciatic nerve and after 2 weeks, one dose of 10¹² AAV was injected IV. Tumor growth was monitored via IVIS imaging and showed significant inhibition by AAV-K55-GRDC24 compared to the untreated control. However, after four weeks, the effects of the treatment became insignificant (n = 5 mice). Data are presented as mean values with SD and analyzed by one-tailed t-test. B In ST88-14 xenograft tumor model, two doses of AAV K55-GRDC24 enabled a more sustained response. In a procedure similar to A, a second dose of 10¹² AAV was injected IV one week after the first dose (n = 4 mice, Con: untreated control). Data are presented as mean values with SD and analyzed by one-tailed t-test. C Comparing the payload expression of one dose vs two doses AAV-K55-GFP. AAV-K55-GFP was injected IV in the ST88-14 tumor-bearing mice at a dose of 10¹² vg. Another dose was given in the double dose cohort 5 days later. All tumors were harvested 14 days after the initial dose and Western blotting indicated the enhanced expression of GFP in the tumors treated with two doses. D Efficacies of AAV K55-GRDC24 in RHT92 xenograft tumor. One dose of 10¹² AAV was injected IV one month after the tumor implantation in the sciatic nerve (n = 5 mice, Con: untreated control). Data are presented as mean values with SD and analyzed by one-tailed t-test. E In RHT92 xenograft tumor model, two doses of AAV K55-GRDC24 resulted in a substantial response. In a procedure similar to C, a second dose of 10¹² AAV was injected IV 5 days after the first dose (n = 4 mice, Con: untreated control). Data are presented as mean values with SD and analyzed by one-tailed t-test.