A Tripartite AAV System with Engineered Lox Sites Enables Efficient Delivery of the EYS Gene for Retinal Gene Therapy

Plasmid construction

A full-length human EYS cDNA clone (pcDNA3.1-hEYS; NM_001142800.2) was obtained from GenScript and used as a template for PCR amplification. The EYS coding sequence (CDS; 9,435 bp) was divided into three segments (1,644, 3,903, and 3,888 bp), PCR-amplified using Q5 High-Fidelity DNA polymerase (New England Biolabs), and cloned into AAV shuttle plasmids (pAAV-GRK1p-15-inCREv1 [with T2A], pAAV-GRK1p-15-IRES-inCRE [with internal ribosomal entry site (IRES)], pFBAAV-17-mid-1522, and pFBAAV2-1722-pA), ¹⁷ using a GenBuilder Cloning kit (GenScript). The resulting constructs were designated pAAV-GRK1p-EYS-E1-15-inCRE, pAAV-GRK1p-EYS-E1-15-IRES-inCRE, pFBAAV-17-EYS-E2-1522, and pFBAAV2-1722-EYS-E3-pA, respectively. Primer sequences are available upon request.

Cell culture and transfection

HEK293T/17 cells (ATCC #CRL-11268) were maintained and transfected as previously described. ¹⁷

Animals and subretinal injection

Wild-type C57BL/6J mice (strain #: 000664) were acquired from the Jackson Laboratory. All procedures were approved by the Institutional Animal Care and Use Committee of the University of Iowa and conducted in accordance with the recommendations outlined in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Mice were kept on a 12-h light/dark cycle with ad libitum access to standard mouse chow.

For subretinal injection, AAV-EYS vectors (serotype: AAV8) were produced by the University of Iowa Viral Vector Core. On the day of injection, AAV vectors were thawed on ice and mixed at a 1:1:1 ratio (3 × 10⁹ genome copies [GC]/µL per vector). Both male and female mice at 1 month of age were anesthetized with a ketamine/xylazine mixture (87.5 mg/kg ketamine, 12.5 mg/kg xylazine), and 10% povidone-iodine and 1% tropicamide solutions were applied. Under a Zeiss SteREO Discovery.V8 dissecting microscope, eyeballs were slightly pulled out using forceps, and a small hole was made using a 30-gauge needle near the limbus. Limbal-approach trans-retinal subretinal injections were performed using a NanoFil syringe attached to a blunt-end 35-gauge needle (World Precision Instrument) as previously described, ¹⁸ delivering 1 µL of the vector solution. After injections, an antibiotic/steroid ophthalmic ointment (neomycin and dexamethasone) was applied. Antisedan (Zoetis) was administered right after injection to facilitate recovery. Animals were excluded from follow-up analysis if no obvious blebs were observed or if significant hemorrhage occurred at the time of injection.

Viral DNA and RNA extraction from mouse retinas, reverse transcription, and PCR

Mice were euthanized by CO₂ asphyxiation followed by cervical dislocation, and eyes were enucleated. Anterior segments were removed with microdissecting scissors, and posterior segments were homogenized in 1 mL of TRIzol Reagent (Invitrogen) using a Polytron PT 1200E homogenizer (Kinematica). Episomal viral DNA and total RNA were extracted following the manufacturer’s RNA extraction protocol. To assess EYS expression cassette reconstitution, extracted nucleic acids were directly used as templates for PCR with GoTaq G2 Flexi DNA polymerase (Promega).

For reverse transcription (RT), samples were treated with DNase I (New England Biolabs) according to the manufacturer’s instructions. After DNase I inactivation by phenol-chloroform extraction, 1 µg of RNA was reverse-transcribed using SuperScript IV reverse transcriptase (Invitrogen) and random hexamers (Invitrogen). Following heat inactivation, cDNAs were used for PCR with GoTaq G2 Flexi DNA polymerase. The following junction-specific primers were used to detect recombination between vector segments: for the 5′-middle junction, F1 (5′-CCAAGATAAAGGTCCTGCTCAA-3′) and R1 (5′-CAGAGGCCATGCACTGATATAC-3′); and for the middle-3′ junction, F2 (5′-CGTCTTCCTCCATGTCTGTAAT-3′) and R2 (5′-CAGAAGTCCATAGGAGCTGAAG-3′). The F3 primer sequence was 5′-GCCAGATTGGGCTGGAAATAC-3′. Primer sequences for mouse Gapdh RT-PCR were 5′-ACCACAGTCCATGCCATCAC-3′ (forward) and 5′-TCCACCACCCTGTTGCTGTA-3′ (reverse).

Protein extraction, SDS-PAGE, immunoblotting, and immunohistochemistry

Proteins were extracted from HEK293T/17 cells and mouse retinas as previously described. ¹⁷ Extracted proteins were loaded on 3–8% NuPAGE Tris-Acetate gels (Invitrogen), transferred onto nitrocellulose membranes (BioRad), and immunoblotted following standard protocols. Antibodies used were rabbit anti-EYS (Invitrogen, PA5-55507), rabbit anti-Cre (Cell Signaling, 5036), mouse anti-β-actin (Sigma, A1978), horseradish peroxidase (HRP)-conjugated antimouse IgG (Cell Signaling, 7076), and HRP-conjugated antirabbit IgG (Cell Signaling, 7074). Detection was performed using SuperSignal West Dura Extended Duration Substrate (Thermo Scientific) or SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific) and imaged with a ChemiDoc Imaging system (Bio-Rad). Immunohistochemistry on mouse retinal sections was performed as previously described, ¹⁷ using rabbit anti-EYS (Invitrogen, PA5-55507) and mouse anti-ABCA4 (Sigma, MABN2439) antibodies.

Design of a tripartite AAV vector system for full-length EYS gene delivery

To enable AAV-mediated delivery of the full-length EYS CDS (9,435 bp, NM_001142800.2, isoform 1), we engineered a tripartite vector system comprising three AAV constructs, each carrying a distinct segment of the EYS CDS (Fig. 1A). These segments were designed to be seamlessly reconstituted in target cells via Cre-mediated, unidirectional DNA recombination.

OPEN IN VIEWER

Figure 1.

Strategy for reconstituting the full-length EYS gene using a triple AAV vector system. (A) Overview of the Cre-mediated recombination system. The EYS coding sequence (CDS) was split into three segments (CDS1, CDS2, and CDS3) and incorporated into three separate AAV vectors: 5′, middle, and 3′. In the reconstituted cassette, splice donor (SD) and splice acceptor (SA) sites flanking the CDS fragments form functional introns. Black arrowheads (F1, R1, F2, and R2) on the reconstituted therapeutic cassette denote the binding sites of PCR primers used to amplify the reconstituted junctional regions. The location of the EYS antibody immunogen (residues C795-R861) is indicated by a black bar. Lox site variants were labeled as follows: 15:P (loxJT15 with a loxP spacer), 15:2272 (loxJT15 with a lox2272 spacer), 17:P (loxJTZ17 with a loxP spacer), 17:2272 (loxJTZ17 with a lox2272 spacer), 15/17:P (loxJT15/JTZ17 LE/RE double mutant with a loxP spacer), and 15/17:2272 (loxJT15/JTZ17 LE/RE double mutant with a lox2272 spacer). (B) Lox sites for recombination between the 5′ and middle vectors. This recombination is mediated by loxJT15 and loxJTZ17 sites, which are loxP variants containing mutations (underlined) in either the left element (LE) or the right element (RE). Cre-mediated recombination between these two sites produces one LE/RE double mutant (lox15/17) and one canonical loxP site. The double mutant lox15/17 site binds Cre with very low affinity, preventing the reverse reaction. (C) Lox sites for recombination between the middle and 3′ vectors. Lox15:2272 and lox17:2272 are LE and RE mutant lox sites, respectively, with a lox2272 spacer (GGATACTT) instead of a loxP spacer (GCATACAT). Mutations in the lox2272 spacer are marked by asterisks (*). AAV, adeno-associated virus; EYS, Eyes Shut Homolog; IRES, internal ribosomal entry site; pA, polyadenylation signal; T2A, T2A “self-cleaving” peptide.

The 5′ vector contains a photoreceptor-specific GRK1 promoter (or a ubiquitously active CMV promoter for in vitro validation), followed by the first EYS segment (1,644 bp), a splice donor (SD) site, a modified lox site (loxJT15), a bicistronic Cre recombinase expression cassette (either T2A “self-cleaving” peptide- or IRES-based) with an embedded intron to prevent prokaryotic expression, and a bovine growth hormone (BGH) polyadenylation (polyA) signal. The middle vector carries a loxJTZ17 site (compatible with loxJT15 in the 5′ vector), a splice acceptor (SA), the second EYS segment (3,903 bp), a second SD, and a lox15:2272 hybrid lox site (compatible with the lox17:2272 site in the 3′ vector). The 3′ vector contains a lox17:2272 hybrid site, an SA, the third EYS segment (3,888 bp), and a BGH polyA signal.

The loxJT15 and loxJTZ17 pair contains the canonical loxP spacer (GCATACAT) (Fig. 1B), whereas the lox15:2272 (hybrid of loxJT15 and lox2272) and lox17:2272 (hybrid of loxJTZ17 and lox2272) pair has the lox2272 spacer (GGATACTT) (Fig. 1C).^17,19,20 Because these two spacers are noncompatible, cross-recombination between the two pairs does not occur. This incompatibility prevents excision of the floxed EYS second segment from the middle vector as well as from the reconstituted expression cassette in the presence of Cre. In addition, loxJT15 and loxJTZ17 are reaction equilibrium-modifying variants of loxP. ¹⁹ These lox sites have mutations in one of the two 13-bp inverted repeats (left element [LE] and right element [RE]), which serve as Cre binding sites. The use of these modified lox sites suppresses reverse reactions, rendering the recombination events essentially irreversible.

Because lox sites are directional (i.e., asymmetrical) and Cre-mediated recombination is sequence-specific, this design enforces ordered and orientation-correct assembly of the three EYS segments. This remains the case even when AAV vector genomes are initially assembled incorrectly. For instance, nonhomologous end joining (NHEJ) and spontaneous recombination between AAV genomes can result in concatenation of vector genomes in random orders and orientations,^21–23 but subsequent Cre-mediated recombination can correct improperly assembled DNAs (Supplementary Fig. S1). The SD and SA sites in the reconstituted expression cassette facilitate RNA splicing of the transcribed pre-mRNA. This splicing process removes any intervening non-coding sequences (including the recombined lox sites), thereby generating a single, continuous mRNA molecule that encodes full-length EYS protein.

The presence of a transcriptional termination signal at the 3′ end of the 5′ vector, together with the absence of promoters in the middle and 3′ vectors, renders the EYS segments in the latter two vectors transcriptionally inactive (regardless of their relative order and orientation within concatenated DNAs) until they are correctly recombined with the 5′ vector via Cre-mediated recombination. This design prevents production of partial EYS proteins, with the exception of transient expression of the EYS N-terminal fragment from the 5′ vector during the prereconstitution phase. Furthermore, the use of a bicistronic element obviates the need for a separate promoter for Cre expression, and the placement of a lox site before the bicistronic element allows self-inactivation of Cre after the reconstitution of therapeutic cassettes by separating the Cre CDS from its promoter.

Reconstitution and expression of full-length EYS in 293T cells

To validate the feasibility of the tripartite vector system in vitro, HEK293T cells were transfected with the three AAV constructs (Fig. 2). The 5′ vector employed the CMV promoter and a T2A peptide for bicistronic Cre expression. Control conditions included individual plasmids (lanes 1–3), omission of one component of the tripartite system (lanes 4–6), and a full-length EYS expression construct (lane 8). Western blot analysis using an anti-EYS antibody raised against residues C795-R861 revealed a ∼350 kDa band corresponding to full-length EYS only in cells transfected with all three vectors simultaneously (lane 7). The reconstituted EYS band comigrated with the band observed in cells transfected with the full-length EYS expression plasmid, confirming accurate assembly and translation. Notably, omission of the 3′ vector (lane 6) did not yield detectable partial proteins corresponding to recombination between the 5′ and middle vectors (∼206 kDa), likely due to the absence of a polyA signal and consequent instability of the transcribed RNAs. Cre protein was detected in lysates from cells receiving the 5′ vector. Expression of β-actin was consistent across all lanes, serving as a loading control. These results provide strong proof-of-principle that the tripartite AAV system can reconstitute the full-length EYS CDS and produce correctly sized protein in mammalian cells.

OPEN IN VIEWER

Figure 2.

In vitro reconstitution and expression of full-length EYS protein in HEK293T cells. HEK293T/17 cells were transfected with individual plasmids of the tripartite AAV-EYS system (lanes 1–3), combinations of two plasmids with one component omitted (lanes 4–6), or all three plasmids (lane 7). A full-length EYS expression plasmid was included as a positive control (lane 8). Cell lysates were collected 3 days posttransfection and subjected to SDS-PAGE followed by Western blotting using an anti-EYS antibody. β-Actin served as a loading control. Molecular weight markers are indicated on the right.

Reconstitution of the EYS expression cassette in mouse retinas

To evaluate the feasibility of the tripartite vector system in vivo, we used mice as a cost-effective surrogate model despite the absence of an EYS ortholog in the murine genome. Tripartite AAV-EYS vectors were packaged into AAV8 capsids and delivered into the subretinal space of 1-month-old wild-type C57BL/6J mice. Two versions of the 5′ vector—one containing a T2A-based Cre cassette and the other containing an IRES-based cassette—were tested, both driven by the GRK1 promoter for photoreceptor-specific expression. Each vector was injected at 3 × 10⁹ GC per eye, for a total dose of 9 × 10⁹ GC.

Three weeks postinjection, eyes were collected for episomal DNA extraction and PCR analysis to assess reconstitution of the EYS expression cassette. Junction-specific primers were designed to amplify across the predicted synthetic introns formed by recombination between the 5′ and middle vectors (F1 and R1 primers) and between the middle and 3′ vectors (F2 and R2 primers) (black arrowheads in Fig. 1A). PCR products with the expected sizes (624 bp for the 5′-middle junction and 510 bp for the middle-3′ junction) were detected exclusively in eyes that received the triple AAV-EYS vectors, regardless of whether the T2A-based (lanes 2–3) or IRES-based (lanes 4–5) 5′ vector was used (Fig. 3A,B). No products were detected in uninjected control eyes (lane 1). A full-length EYS expression plasmid (pcDNA3.1-hEYS) was used as a positive control (lane 6). The PCR products amplified from the control plasmid were slightly smaller (171 bp for the 5′-mid junction product and 166 bp for the mid-3′ junction product) than the corresponding amplicons from the AAV-injected eyes due to the absence of synthetic introns in the EYS expression plasmid. In contrast, PCR performed with primers F1 and R2 yielded no detectable amplicons under the same conditions (Fig. 3C), confirming that no unintended recombination occurred directly between the 5′ and 3′ vectors. In addition, PCR assays designed to detect randomly assembled AAV genomes failed to produce discernible amplicons (Supplementary Fig. S2).

OPEN IN VIEWER

Figure 3.

In vivo validation of EYS gene reconstitution in mouse retinas. (A–C) PCR analysis of junctional regions formed after vector recombination. Wild-type mice were subretinally injected with the complete AAV-EYS vector set at 1 month of age. Episomal DNAs were extracted 3 weeks postinjection from injected and uninjected eyes and used as templates for PCR amplification of the (A) 5′-middle junction (primers F1 and R1), (B) middle-3′ junction (primers F2 and R2), and (C) 5′-3′ junction (primers F1 and R2). The expected sizes of the 5′-middle and middle-3′ amplicons are 624 and 510 bp, respectively. No amplicons were detected for the 5′-3′ junction, confirming the absence of unintended recombination. Lane 1: uninjected eye; lanes 2–3: mice injected with the triple AAV-EYS set (with the T2A-based 5′ vector); lanes 4–5: mice injected with the triple AAV-EYS set (with the IRES-based 5′ vector); lane 6: positive control (full-length EYS plasmid). PCR products from the control plasmid are slightly smaller due to the absence of synthetic introns. (D, E) Sanger sequencing of PCR amplicons confirms accurate junction formation. Sequences contain the expected loxP (D) and lox2272 (E) spacers flanked by loxJT15 LE and loxJTZ17 RE, indicating precise Cre-mediated recombination at both junctions. Upper strands were sequenced using primers F1 (D) and F2 (E), and reverse complementary (rev. comp.) strands were sequenced using primers R1 (D) and R2 (E).

To further validate recombination accuracy and amplicon identity, PCR products were subjected to Sanger sequencing using the same primers employed for amplification (Fig. 3D,E). Sequence analysis revealed that the 5′-middle junction contained the canonical loxP spacer flanked by the loxJT15 LE and loxJTZ17 RE (i.e., the lox15/17 double mutant; Fig. 1B). Similarly, the middle-3′ junction contained the lox2272 spacer flanked by the same loxJT15 LE and loxJTZ17 RE (i.e., the lox15/17:2272 double mutant; Fig. 1C). Sequencing chromatograms exhibited no ambiguous or noisy peaks downstream of the spacer regions. As randomly concatenated and unrecombined AAV genomes would retain the original lox sites present in the individual vectors (i.e., loxJT15 in the 5′ vector, loxJTZ17 and lox15:2272 in the middle vector, and lox17:2272 in the 3′ vector), the lack of mixed or noisy signals after the spacer regions suggests that such unrecombined intermediates were absent or rare. Collectively, these results demonstrate the fidelity and efficiency of Cre-mediated recombination and verify that the orthogonal lox site pairs functioned as designed in vivo.

Expression of full-length EYS in mouse retinas

We next evaluated the expression of full-length EYS in mouse retinas by RT-PCR (Fig. 4). Total RNA was isolated from uninjected control eyes and eyes injected with the tripartite AAV-EYS system at 3 weeks postinjection and subjected to RT-PCR analysis. The full-length EYS expression plasmid (pcDNA3.1-hEYS) served as a positive control. To allow amplification of large cDNA fragments, an extended PCR extension time (2 min) was used.

OPEN IN VIEWER

Figure 4.

RT-PCR analysis demonstrates expression of full-length EYS mRNA in mouse retinas. Total RNA was extracted from uninjected control eyes and eyes injected with the tripartite AAV-EYS vector set (containing either T2A- or IRES-based 5′ vectors) at 3 weeks postinjection. A full-length EYS plasmid was used as a positive control. (A–C) PCR amplification using junction-specific primers confirms correct assembly and splicing of the reconstituted EYS transcript. Primer pair F1/R1 (spanning the 5′-middle junction) yielded the expected 453-bp amplicon (A), and primer pair F2/R2 (spanning the middle-3′ junction) yielded the expected 344-bp amplicon (B). (C) PCR using primer pair F1/R2 produced the expected 4.35-kb product containing the EYS middle segment (CDS2). (D,E) PCR reactions performed to detect potential mRNAs derived from randomly assembled concatemers. Primer pairs F1/F2 (detecting a 5′-reversed middle vector junction) (D) and F1/F3 (detecting a 5′-reversed 3′ vector junction) (E) yielded no amplification products, suggesting that aberrant transcripts from random concatemers are rare or absent. (F) Gapdh RT-PCR served as a loading control, demonstrating comparable RNA input across all samples (except for the EYS expression plasmid control).

As expected, PCR performed on cDNAs from triple AAV-injected eyes using primer pairs spanning the 5′-middle and middle-3′ vector junctions generated amplicons of the predicted sizes (453 and 344 bp, respectively) (Fig. 4A and B). PCR using primers F1 and R2 yielded the expected 4.35-kb amplicon encompassing the EYS middle segment (CDS2) (Fig. 4C). No additional products were detected. Because the synthetic introns present at the vector junctions are removed during RNA splicing, amplicons derived from triple AAV-injected eyes were identical in size to those amplified from the EYS expression plasmid.

To assess potential mRNA production from randomly assembled concatemers, additional PCR reactions were performed with primer pairs F1/F2 (detecting a 5′-reversed middle vector junction) and F1/F3 (detecting a 5′-reversed 3′ vector junction). No amplification products were observed in these reactions (Fig. 4D and E). Although the presence of randomly assembled concatemers cannot be ruled out, these results suggest that transcripts derived from such concatemers are likely rare.

Protein expression was subsequently evaluated by Western blot analysis of retinal lysates harvested at 3 and 6 weeks postinjection. Consistent with the in vitro results obtained in 293T cells, a ∼350 kDa EYS band was observed in all mouse eyes receiving the complete tripartite AAV-EYS vector set (Fig. 5, black arrowheads). Both T2A- and IRES-based 5′ vectors supported EYS expression, indicating that either bicistronic configuration is sufficient to drive Cre expression and enable cassette reconstitution. In contrast, EYS protein was undetectable in uninjected control eyes (Fig. 5A, lanes 1–2) or in eyes injected with only the 5′ vectors (Fig. 5B, lanes 1–4). Notably, partial EYS proteins such as ∼144 kDa (derived from the middle vector), ∼206 kDa (from the 5′ + middle vector fusion), and ∼287 kDa (from the middle + 3′ vector fusion) species were not detected, indicating that partial protein production from unintended random assemblies is absent. In addition, Cre recombinase was readily detectable at 3 weeks postinjection but was absent or barely detectable at 6 weeks in eyes receiving the complete tripartite vector set, suggesting depletion of unrecombined 5′ vector genomes and self-inactivation of Cre prior to the 6-week postinjection time point. A consistent nonspecific band (open arrowhead) was observed across samples, and β-actin served as a loading control.

OPEN IN VIEWER

Figure 5.

In vivo expression of full-length EYS protein in mouse retinas. Western blot analysis of retinal lysates collected at 3 weeks (A) and 6 weeks (B) postinjection. (A) Lanes 1–2: uninjected controls; lanes 3–5: eyes injected with the complete tripartite AAV-EYS vector set containing the T2A-based 5′ vector; lanes 6–8: eyes injected with the complete tripartite AAV-EYS vector set containing the IRES-based 5′ vector. (B) Lanes 1–2: eyes injected with the T2A-based 5′ vector alone; lanes 3–4: eyes injected with the IRES-based 5′ vector alone. Full-length EYS protein (∼350 kDa) is indicated by a black arrowhead, and a consistent nonspecific band is marked with an open arrowhead. Each lane represents an independent biological replicate.

Lastly, we performed immunostaining on retinal sections from mice injected with tripartite AAV-EYS vectors (IRES set) to assess the expression and subcellular localization of AAV-delivered EYS in the mouse retina (Supplementary Fig. S3). Consistent with the use of the photoreceptor-specific GRK1 promoter, EYS immunoreactivity was detected in photoreceptors. However, contrary to the reported subcellular localization of EYS near the connecting cilium,^13,14 AAV-delivered EYS appeared as punctate structures within the ellipsoid region of the inner segment. The precise identity of these structures remains to be determined (see the Discussion section).

Taken together, these results demonstrate that the tripartite vector system enables successful delivery, recombination, and expression of the full-length EYS gene in mouse photoreceptors, validating the feasibility of this approach in vivo.