Successful expansion but not complete restriction of tropism of adeno-associated virus by in vivo biopanning of random virus display peptide libraries

Abstract

Targeting viral vectors to certain tissues in vivo has been a major challenge in gene therapy. Cell type-directed vector capsids can be selected from random peptide libraries displayed on viral capsids in vitro but so far this system could not easily be translated to in vivo applications. Using a novel, PCR-based amplification protocol for peptide libraries displayed on adeno-associated virus (AAV), we selected vectors for optimized transduction of primary tumor cells in vitro. However, these vectors were not suitable for transduction of the same target cells under in vivo conditions. We therefore performed selections of AAV peptide libraries in vivo in living animals after intravenous administration using tumor and lung tissue as prototype targets. Analysis of peptide sequences of AAV clones after several rounds of selection yielded distinct sequence motifs for both tissues. The selected clones indeed conferred gene expression in the target tissue while gene expression was undetectable in animals injected with control vectors. However, all of the vectors selected for tumor transduction also transduced heart tissue and the vectors selected for lung transduction also transduced a number of other tissues, particularly and invariably the heart. This suggests that modification of the heparin binding motif by target-binding peptide insertion is necessary but not sufficient to achieve tissue-specific transgene expression. While the approach presented here does not yield vectors whose expression is confined to one target tissue, it is a useful tool for in vivo tissue transduction when expression in tissues other than the primary target is uncritical.

Figure 1. Pathways used for selection of targeted viral capsids by screening random AAV display peptide libraries.

For all selection pathways, genomic DNA containing cap gene fragments from internalized library viruses was extracted from the target cells or tissue. Library inserts were amplified by nested PCR and cloned back into the AAV library backbone plasmid pMT-202-6. The resulting pre-selected plasmid library was used to produce a secondary AAV library by transfection into 293T cells and subsequent superinfection with Ad5. Pre-selected AAV libraries were re-subjected to selection on the target cells in vitro or the target tissue in vivo. Preceding the amplification step, the library selection was done according to one of the following three pathways: Pathway A, in vitro selection: A random AAV display peptide library was incubated on primary breast cancer dissociation cultures derived from female tumor-bearing PymT mice. Non-internalized AAV library particles were removed by extensive washing followed by trypsin digestion prior to DNA extraction and AAV insert amplification. Pathway B, in vivo/ex vivo selection: A random AAV display peptide library was injected intravenously into female tumor-bearing PymT mice. After 24 hours, primary tumor cells of the injected mouse were prepared as in pathway A and grown ex vivo for 96 hours prior to DNA extraction and AAV insert amplification. Pathway C, in vivo selection: A random AAV display peptide library was injected as in pathway B in tumor-bearing mice (for selection of tumor-homing AAV) or wild-type mice (for selection of lung homing AAV), respectively. After 48 hours, the target tissue (tumor or lung, respectively) was removed and lysed, and DNA was extracted for AAV insert amplification.

Figure 2. Vector capsids selected from random AAV display peptide libraries for targeted gene transfer in primary breast cancer cells.

A: Primary PymT breast cancer cells were transduced by recombinant AAV-2 luciferase reporter gene vectors displaying the selected capsid peptide inserts RGDLGLS, RGDMSRE, or DGLGRLV, respectively. Capsids with no (wild-type) or random peptide insert (VRRPRFW) were used as controls. Transduction efficiency was determined after 72 hours by luciferase assay. Luciferase activities are shown in relative light units (RLU) per well. Data represent mean values ± standard deviation (SD) from one representative experiment (out of three) in triplicates (*** = p<0.001 compared to wild-type and random insert controls). B: Breast cancer cell-targeted therapeutic suicide gene transfer using selected capsid mutants. Primary PymT cells were transduced using rAAV-SR39 vectors displaying RGDLGLS or a randomly selected control peptide (VRRPRFW). Four days after initiation of ganciclovir (GCV) treatment, cytotoxic effects were evaluated by MTT assay. Values are shown in % cytotoxicity (i.e., % killed cells). Untreated and untransduced cells served as controls. Data represent mean values ± standard error of the mean (SEM) from nine wells in three independent experiments (*** = p<0.001 selected clone and treated cells vs. all controls).

Figure 3. Kinetics of circulating AAV peptide library particles is similar to wild-type AAV.

A random X₇ peptide library or wild-type AAV-2 viruses were injected intravenously at 1×10¹⁰ vg per mouse. Blood samples were collected after indicated time points and the amount of circulating viral particles in the serum was determined by real-time PCR. Data represent mean values from n = 3 mice per group, analyzed in triplicates ± SD.

Figure 4. Gene delivery by AAV capsid mutants selected for breast cancer transduction in vivo.

AAV luciferase vectors displaying selected peptides or controls (wild-type or VRRPRFW) were injected intravenously into female PymT tumor-bearing mice. After 8 d, representative tissues were harvested and luciferase activities were determined in individual tissues as relative light units (RLU) per mg protein. A: In vivo transduction of tumor tissue in PymT transgenic FVB mice by selected AAV mutants. Bars indicate the median, n = 5 mice per group. * p<0.05 targeted vectors vs. wild-type. # p<0.05 targeted vectors vs. random insert control. B: In vivo transduction of various non-cancerous tissues in PymT transgenic FVB mice by tumor-selected AAV mutants. The dotted line indicates the threshold beyond which luciferase expression data could be reliably delineated from background signal. Data represent mean values ± SEM, n = 5 mice per group. * p<0.05; ** p<0.01 targeted vectors vs. wild-type AAV-2. # p<0.05; ## p<0.01 targeted vectors vs. random insert control.

Figure 5. Targeting of AAV capsid mutants selected on murine lung tissue in vivo.

AAV luciferase vectors displaying selected or control capsids (wild-type or random insert VRRPRFW) were injected intravenously into female FVB mice. Tissue was harvested after 8 or 28 d, respectively, and processed as indicated. A: Evaluation of lung homing. Lung tissue was harvested 8 days after vector injection and the amount of AAV genomes was determined by quantitative PCR. Data represent mean values from n = 3 mice per group, analyzed in triplicates ± SD. B: In vivo lung gene transfer by selected AAV after intravenous injection. Lung tissue was harvested 28 days after vector injection, and luciferase activity was determined as relative light units (RLU) per mg protein. Bars indicate the median value, n = 5 mice per group (** = p<0.001 targeted vectors vs. wild-type and random insert control). C: In vivo transduction of various tissues in mice by AAV library mutants selected for lung transduction. Tissues were harvested and luciferase activity was determined as in 5B. The dotted line indicates the threshold beyond which luciferase expression data could be reliably delineated from background signal. Data represent mean values ± SEM, n = 5 mice per group. * p<0.05; ** p<0.01 targeted vectors vs. wild-type AAV-2. # p<0.05; ## p<0.01 targeted vectors vs. random insert control.