METHOD
doi:10.1016/j.ymthe.2003.11.016
Model of Unidirectional Transluminal Gene Transfer Marco A. Arap,1 Johanna Lahdenranta,1 Amin Hajitou,1 Frank C. Marini III,2 Christopher G. Wood,3 Kenneth C. Wright,4 Juan Fueyo,5 Wadih Arap,1,6,* and Renata Pasqualini1,6,* 1
Department of Genitourinary Medical Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas, 77030, USA Department of Bone Marrow Transplantation, The University of Texas M. D. Anderson Cancer Center, Houston, Texas, 77030, USA 3 Department of Urology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas, 77030, USA 4 Department of Veterinary Medicine, The University of Texas M. D. Anderson Cancer Center, Houston, Texas, 77030, USA 5 Department of Neuro-Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas, 77030, USA 6 Department of Cancer Biology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas, 77030, USA
2
1 Abbreviations used: AAV, adeno-associated virus; DMEM, Dulbecco’s modified Eagle’s medium; FBS, fetal bovine serum; GAG, mucopolysaccharide/glycosaminoglycan; GFP, green fluorescent protein; m.o.i., multiplicity of infection; PBS, phosphate-buffered saline; PBS-MK, phosphate-buffered saline containing 1 mM MgCl2 and 2.5 mM KCl; RCA, replication-competent adenovirus; SEM, standard error of the mean.
*To whom correspondence and reprint requests should be addressed. E-mail:
[email protected] (W. Arap) or
[email protected] (R. Pasqualini).
Gene transfer assays in vitro are poor indicators of transduction efficacy observed in vivo. We designed and optimized an intermediate model for assessing and quantifying unidirectional transduction ex vivo. The model enables simultaneous transmucosal evaluation of up to 96 different variables under the same tissue conditions. We show that the model is versatile and suitable for use with different vectors (adenovirus and AAV), different reporter genes (h hgalactosidase and green fluorescent protein), and viscera with various tissue features such as peritoneum and urothelium. Ex vivo transduction assays may correlate better with in vivo gene transfer results. Because the experimental model described here can be performed in small samples, it may enable translational applications in tissues of human origin. Key Words: adenovirus, AAV, h-galactosidase, ex vivo model, gene transfer, green fluorescent protein, transduction
INTRODUCTION Gene transfer has shown promise in the treatment of several diseases [1 – 7]. Potential advantages of gene transfer techniques include the high efficiency of gene transfer to certain cell types and the ability to transduce cells with a low mitotic index [8,9]. However, the efficient transduction often observed in cell culture is difficult to recapitulate in vivo largely because of contextual differences among cell lines used in vitro, animal models, and patient studies. Other reasons invoked to explain these data include species-specific differences and the variability between normal and diseased tissues [10]. Although in vitro studies can address mechanistic questions, there is a relative paucity of less reductionist models that facilitate the assessment of transduction efficiency and reproduce the disease-specific microenvironment found in vivo. Assessment of gene expression in tissue explants immersed into vector suspensions [10] is one of the methods that attempts to simulate in vivo
MOLECULAR THERAPY Vol. 9, No. 2, February 2004 Copyright D The American Society of Gene Therapy 1525-0016/$30.00
conditions. However, it is difficult to determine the depth of penetration of the vectors into stratified epithelium. Furthermore, immersion-type assays do not allow simultaneous evaluation of multiple conditions, procedures, gene therapy vectors, and reporters. Here we adapted a phage display targeting assay [11] to evaluate gene transfer ex vivo. We show that the model allows serial comparison of transluminal and unidirectional transduction in up to 96 equal-size tissue fields. Because the assay can be performed in small samples, it may enable translational experimentation in tissues of human origin.
RESULTS Standardization of ex Vivo Gene Transfer Assays We used a dot-blot apparatus (Fig. 1A) to standardize gene transfer assays ex vivo (Fig. 1B). To optimize and standardize gene transfer and transduction conditions, we used
305
METHOD
doi:10.1016/j.ymthe.2003.11.016
statistically significant correlation (Student’s t test, P < 0.001) of the number of transduced cells per high-power magnification field with gene vector inputs (Fig. 2). Homogeneous transduction was observed in the exposed peritoneum that received the vector but not in the negative control wells (Figs. 3A and 3B). After tissue sectioning and staining for hematoxylin/eosin, we observed intracellular blue color derived from h-galactosidase expression in unidirectional layers up to 50 Am deep into the peritoneal tissue, while the negative control wells did not express the reporter protein (Figs. 3C and 3D).
FIG. 1. Dot-blot chamber and schematic use for gene transfer assays. (A) The dot-blot chamber consists of two main devices (upper and lower) that can be tightened to create up to 96 uniform tissue fields. (B) Schematic use of the dot-blot chamber for ex vivo gene therapy assays. The chamber is represented with a bladder in position. A single well with a temporal sequence of tissue infection and gene transfer is represented in the middle. The bottom represents theoretical serial events culminating with nuclear incorporation of the reporter gene into the DNA of a single cell.
fresh peritoneum as the primary source of tissue and an adenovirus expressing Escherichia coli h-galactosidase (AdLacZ-F/S35) as the reporter gene. Empty microwells were used only to rule out trans-well contamination. One week after transduction, we stained the tissue for h-galactosidase expression. Protein expression was first observable macroscopically at 6 h and was maximal at 24 h after staining, when the peritoneum was removed from the chamber for pathological analysis. We also evaluated the peritoneal epithelium under microscopic view. We observed reproducible transduction rates with positive cells homogeneously distributed throughout the tissue in the wells to which the Ad-LacZ-F/S35 was added. In contrast, no transduced cells were observed in the control wells. We have also quantified the protein expression of hgalactosidase according to the vector input per well. This analysis showed a direct correlation between multiplicity of infection (m.o.i.)1 and h-galactosidase-positive cells within the tissue studied. We observed a direct and
306
Transduction of Tissues with a Mucopolysaccharide/ Glycosaminoglycan Layer Having shown that the model can be used to visualize and quantify transduction of tissues, we set out to determine whether the assay could be used in organs covered by a mucopolysaccharide/glycosaminoglycan (GAG) layer, a barrier to gene transfer in tissues such as the urothelium [11,12]. To establish the versatility of the system, we utilized a different vector (adeno-associated virus; AAV) carrying the cDNA corresponding to a different reporter gene (green fluorescent protein; GFP). We used AAV-GFP for transduction assays in porcine urothelium from the urinary bladder. After surgically removing the excessive connective tissue around the bladder dome, opening the organ, and exposing its mucosa, we removed the GAG layer by the instillation of an acid solution into the wells [13]. We achieved infection by adding 2 109 particles/well of the AAV-GFP vector in a 50-Al DMEM solution. We maintained the chamber at 37jC for 6 h, removed the vector suspension by gentle washing, and added medium supplemented with serum and antibiotics
FIG. 2. Semiquantitative analysis of transduction. Porcine peritoneum was transduced with an Ad-LacZ-F/S35 vector. The mean number of positive cells per high-power magnification field (HPMF) was assessed after h-galactosidase staining. Results are expressed as mean values F SEM of three different experiments. In each experiment, three wells were infected by the addition of 109, 108, or 107 PFU of Ad-LacZ-F/S35 vector in 50 Al of DMEM in each well. Reporter gene expression was progressively higher according to the vector input per well. Data were analyzed using the Student t test, P < 0.001.
MOLECULAR THERAPY Vol. 9, No. 2, February 2004 Copyright D The American Society of Gene Therapy
doi:10.1016/j.ymthe.2003.11.016
METHOD
walled, GAG-layer-covered tissue (urinary bladder); under these conditions, we observed several layers of transduced cells up to 105 Am deep into the tissue. Next, we evaluated whether we would obtain comparable results with a conventional optical microscope. By changing the micrometric focus of the microscope we were able to identify the different layers of transduced cells previously
FIG. 3. Transduction of peritoneum and urothelium. Fresh, sterile, porcine and human specimens were obtained and microdissected and the tissues were positioned in the chamber with their visceral side (porcine peritoneum) or luminal side (urothelium from porcine bladders or human ureters) facing upward. (A) Negative control peritoneum. (B) Peritoneum transduced with Ad-LacZ vector. (C) Hematoxylin and eosin staining of untreated peritoneum. (D) Hematoxylin and eosin staining of peritoneum transduced with the AdLacZ vector. (E) Negative control bladder. (F) AAV-GFP-transduced bladder. (G) Negative control ureter. (H) AAV-GFP-transduced ureter. Arrowheads point to hematoxylin-stained nuclei and arrows point to h-galactosidasepositive cells. Scale bar, 50 Am.
to the wells. We observed a clear and reproducible GFP expression in each well into which AAV-GFP was added and detected no fluorescent signal in the control wells (Figs. 3E and 3F). We were also able to identify unidirectional (i.e., from the luminal to the retroperitoneal side of the organ) transduction of deep layers into the organs, without tissue sectioning. To evaluate multilayer tissue transduction, we used a computer-assisted confocal microscope to identify transduced cells systematically up to 50 Am deep into peritoneum-derived tissue. Next, by using the same approach, we evaluated deep transduction in a thick-
MOLECULAR THERAPY Vol. 9, No. 2, February 2004 Copyright D The American Society of Gene Therapy
FIG. 4. Untargeted and targeted transduction of urothelium compared by computer-assisted confocal microscopy. Porcine bladders were obtained fresh and sterile, microdissected, and positioned in the chamber with the luminal side facing upward. (A) Adenovirus with no reporter gene. (B) Nontargeted adenovirus containing the GFP gene. (C) RGD-targeted adenovirus vector containing the GFP gene. Scale bar, 50 Am.
307
METHOD
seen with the confocal microscope. Moreover, when we measured the deepness of transduction (again by changing the micrometric focus of the microscope), the results for both confocal and regular microscopy were similar (positive cells found up to 105 Am deep). Thus, it is possible to evaluate tissue-deep transduction in laboratories in which a confocal microscope is not available by establishing micrometric focus variation in conventional optical microscopes. We then decided to test if the assay would be suitable for use with limited amounts of incidental human tissue obtained from surgical procedures. Typically, urothelium tissue samples from normal ureter are obtained during the resection of renal cell carcinomas. After obtaining sterile human samples, we microdissected the ureters to remove the external connective tissue around the specimens. We surgically opened the ureter longitudinally, exposed its urothelial mucosa, and removed the GAG layer [13]. Next, we used 2 109 viral particles of AAVGFP for infection and added medium supplemented with serum and antibiotics to the wells after 6 h. We observed homogeneous transduction in the urothelium and detected it in all of the wells into which the vector was added, while no signal was detected in the negative control wells (Figs. 3G and 3H). Finally, we tested whether the model could be used for the evaluation of targeted gene transfer. We compared cell transduction of an adenovirus carrying the GFP reporter gene (Ad-GFP) to an RGD-targeted adenovirus (Ad-F/RGD-GFP) vector. We used an adenovirus with no reporter gene as a negative control to rule out trans-well contamination and evaluate background autofluorescence. We measured transduction by computer-assisted confocal microscopy. We also determined the pixel density of the pictures, to compare objectively the efficacy of the vectors. GFP expression transduced by an RGD-targeted adenoviral vector was significantly higher (f2.5fold) than GFP expression transduced by a nontargeted adenoviral vector (Student’s t test, P < 0.001). Moreover, we did not observe background autofluorescence in neighboring microwells that contained tissues receiving empty adenovirus vector (Fig. 4). Deep-layer transduction was similar to that observed in other experiments, with the deepest positive cells found up to 105 Am for both targeted and nontargeted vectors. Taken together, these data indicate that the ex vivo unidirectional assay described here can be used to compare transduction from specific and nonspecific vectors by using different ligand – receptor systems within the same experiment.
DISCUSSION Assays to assess gene transfer efficiency in vitro often do not translate well under in vivo conditions. To provide a practical solution for this paradox, we sought to develop a simple model to standardize and quantify local gene
308
doi:10.1016/j.ymthe.2003.11.016
transfer in tissues. The model proposed here allows ex vivo assay of up to 96 different conditions in a single tissue sample. We show the feasibility of obtaining reproducible and homogeneous qualitative and quantitative reporter gene transfer data. We used peritoneum for the optimization of the assay because it is a relatively abundant and thin-walled tissue. We evaluated the transduction of fresh porcine peritoneum by a Lac-Z-F/S35 adenoviral vector. Dose-dependent expression of h-galactosidase was observed in all tissue wells, indicating a reproducible semiquantitative reporter protein expression in ex vivo tissue samples. In contrast, the expression of h-galactosidase was not detectable in the peritoneum from the negative control wells. Despite the very close proximity of neighboring wells within the chamber, no trans-well contamination was observed. The model allows a unidirectional (i.e., from the luminal epithelium to the deeper tissue layers) evaluation of transduction. Certain specialized tissues such as the urothelium lining the urinary tract are accessible to local gene therapy trials [14 – 16]. However, the urothelium is a multilayered specialized epithelium with tight cell junctions and a covering GAG layer [4]. Because such physical barriers may prevent transduction to deeper tissue layers and reduce efficacy of gene therapies against transitional cell carcinoma [17], chemical removal of the GAG layer of the bladder has been proposed as a method to increase intravesical gene expression [12,13,18]. Thus, we attempted to simulate these harsh conditions in our ex vivo model. After removal of the GAG layer, we evaluated the transduction of porcine urothelium with an AAV carrying GFP as the reporter. Again, we observed homogeneous reproducible GFP expression, with no trans-well contamination and minimal autofluorescence signal detected in the control wells. The capacity to allow detection of transduced cells and to measure the length of transduction into internal layers, even after tissue manipulation, shows that the model is suitable for studies of thick-walled viscera (such as the bladder). Similar results were obtained with normal human ureters obtained from surgical procedures (incidental tissue in radical nephrectomies from patients with renal cell carcinomas). Thus, the model may have applications in preclinical ex vivo studies in which only minute amounts of tissue are available. Gene therapy remains a promising therapeutic modality for other tumors that spread to the peritoneal cavity (such as ovarian [19] or gastrointestinal cancers [20]) or to meningeal membranes (such as malignant gliomas [21]). Evaluation of effectiveness depends on reliable assays. Thus, ex vivo gene transfer may offer an intermediate step between in vitro and in vivo studies. Ex vivo strategies are based on harvesting cells for gene transfer and subsequent reinjection into the primary organ [22] or immersion of tissue samples into vector suspensions [10]. These techniques are cumbersome, require large samples, and
MOLECULAR THERAPY Vol. 9, No. 2, February 2004 Copyright D The American Society of Gene Therapy
doi:10.1016/j.ymthe.2003.11.016
do not account for unidirectional transduction since both luminal and interstitial sides are in contact with the vector suspension. In contrast, the model described here enables the evaluation of different vectors at varying m.o.i. under the same conditions at the luminal surface of the tissue of interest. In summary, our results establish an ex vivo model for standardization of gene transfer assays as an alternative to current in vitro protocols for serially evaluating transduction in hollow organs. Ultimately, this simple assay may correlate better with transduction observed in vivo and facilitate translation of gene therapy into clinical applications.
MATERIAL AND METHODS Surgical Samples Porcine peritoneum tissue specimens were obtained sterile from fresh surgical samples, washed with sterile DMEM, and placed in the dot-blot chamber (Bio-Rad, Hercules, CA, USA) with the luminal side facing upward. Human ureter samples were obtained from nephrectomies for renal cell carcinomas (as incidental normal tissue) and porcine bladders were obtained sterile from fresh surgical samples. Organs and tissues were maintained sterile in DMEM. All surgical manipulation was performed under sterile conditions. Recombinant Viral Vectors To evaluate transduction, we used four vectors containing different reporter genes (Ad-LacZ-F/S35, Ad-GFP, Ad-F/RGD-GFP, and AAV-GFP) and a control adenoviral vector with no reporter gene (Ad-dL3-12), as follows: Ad-LacZ-F/S35 vector. We used a fiber-modified adenoviral vector, using the fiber portion of the adenovirus serotype 35 (S35). Briefly, a recombinant LacZ adenoviral vector expressing h-galactosidase (Ad-LacZ-F/S35) was created by recombination of the S35 fiber expression cassette [23] and an E1/E3-deleted adenovirus (type 5) backbone expressing LacZ [24]. The vector backbone was transfected into HEK293 cells using Fugene6 (Boehringer Mannheim, San Diego, CA, USA), followed by expansion of a single plaque. At the maximal cytopathic effect, the cells were harvested and pelleted by gentle centrifugation. Ad-LacZ-F/S35 was extracted from the HEK293 cells by three consecutive freeze-and-thaw cycles and amplified by infection of a larger culture of HEK293 cells. Ad-LacZF/S35 vectors were purified by two consecutive cesium chloride gradient ultracentrifugation steps and desalted on exclusion columns (Bio-Rad Laboratories). Adenoviral vectors were stored at 70jC in 10% glycerol, 10 mM Tris – Cl (pH 7.4), and 1 mM MgCl2. The titer of the large-scale purified adenovirus preparation was determined by spectrophotometry and plaque assays on HEK293 cells. Preparations were routinely tested for the presence of replication-competent adenovirus (RCA) by plaque assays on A549 cells. The titer was always less than 1 RCA/1010 viral particles. Ad-GFP and Ad-F/RGD-GFP. Unmodified Ad5-CMV-GFP (an E1/E3-deleted adenovirus expressing GFP, serotype 5) was used as the backbone for these experiments. For creation of RGD fibers, the peptide RGD-4C (ACDCRGDCFC) was cloned into pSK II+[X-K]nb (the adenovirus serotype 5 fiber cDNA) (provided by the Riken Gene Bank; Riken Tsukuba Institute Tsukuba, Ibaraki, Japan) and recombinant fibers were subjected to sequencing to ensure the correct DNA sequences. After PCR amplification, modified fibers were subcloned into the XbaI (nt 30470) to KpnI (nt 33594) site in psK[EcT22-H3]nb, a fiber recombination plasmid. Briefly, modified fiber recombination plasmids and Ad-CMV-GFP DNA were cotransfected into 293 cells as described [23] and recombinant adenovirus rescued after plaque formation. Adenovirus stocks were prepared, and
MOLECULAR THERAPY Vol. 9, No. 2, February 2004 Copyright D The American Society of Gene Therapy
METHOD
viral DNA was isolated and subjected to sequencing with appropriate fiber primer sets to determine the correct fibers and ensure fibers were mutation free. As a negative control, we utilized an empty adenoviral vector lacking the reporter gene sequences (Ad-dL3-12) [23]. AAV-GFP vector. Initially, the recombinant AAV vector was constructed by cloning a 2043-bp EcoRI – HindIII fragment, derived from the pCMVeGFP plasmid (Clontech, Palo Alto, CA, USA), containing the CMV promoter and the enhanced GFP gene cassette into the EcoRI – HindIII site of pAAV-ITR (Stratagene, San Diego, CA, USA). Semiconfluent 293T cells on 15-cm plates were cotransfected with the packaging plasmid pRep-CAP (2.5 Ag) (Stratagene) and vector plasmid pAAV CMV-GFP, by using Fugene6 (Boehringer Mannheim). On the following day, cells were transfected with pHelper (a plasmid expressing adenovirus E4orf6 protein) and incubated for an additional 72 h until cytopathic effect became visible. A clarified cell lysate was prepared by five cycles of freeze and thaw, followed by centrifugation at 6000g for 10 min, to remove cell debris. Viral stocks were incubated at 56jC for 30 min to inactivate the helper virus. The rAAV preparation was further purified by heparin column chromatography [25]. Briefly, we initially removed cellular debris by discontinuous step gradient centrifugation by using Iodixanol (Nycomed, Princeton, NJ, USA). The viral fraction was then applied to a 2.5-ml heparin agarose type I column (Sigma, St. Louis, MO, USA) equilibrated with PBS containing 1 mM MgCl2, 2.5 mM KCl (PBS-MK). The column was washed with 10 ml of PBS-MK and the virus eluted with 3.5 ml of PBS containing 1 M NaCl. To determine functional rAAV titer, 105 293T cells were transduced overnight in serum-free medium with serial dilutions of the rAAV CMV-GFP. Next day, the cells were washed and then cultured for a further 48 h before assessment of transduction efficiency. Flow cytometric determination of GFP expression by 293T cells indicated rAAV CMV-GFP vector transduction efficiency. Recombinant AAV particle numbers were determined by slot-blot analysis, as previously described [26]. Aliquots of rAAV were periodically subjected to polyacrylamide gel electrophoresis and electron microscopy to verify that preparations were free of adenovirus and/or cellular contamination. Transduction of Peritoneum with an Ad-LacZ Vector After the porcine peritoneum was positioned with the visceral side facing upward, the chamber was tightly closed so that 96 fields of mucosa were exposed. Generally, prevention of loss of medium was achieved, trans-well contamination avoided, and transduction efficiency uniformly evaluated ex vivo. Infection was performed by the addition of 107, 108, and 109 PFU (at least three wells per dose) of a LacZ adenoviral vector expressing E. coli h-galactosidase (Ad-LacZ) in 50 Al DMEM solution. DMEM with no vector was added to the same number of control wells. The chamber was maintained at 37jC in a humidified atmosphere with 5% CO2 for 6 h. After the vector-containing solution was removed by gentle washing, DMEM (400 Al) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin G, 100 U/ml streptomycin, and glutamine was then added to all the wells, and the mounted chamber was maintained at 37jC in the humidified CO2 incubator for 7 days. Daily evaluation of the chamber revealed no loss of the medium. After 7 days, wells were washed three times with PBS. Peritoneum was fixed with freshly prepared 0.5% glutaraldehyde in PBS (500 Al in each well) for 5 min at room temperature and the wells were washed once more with PBS. h-Galactosidase staining was then performed by the addition of a prestain solution supplemented with 3 mM potassium ferricyanide, 0.05% h-galactosidase, and 3 mM potassium ferrocyanide. After staining, the chamber was maintained at 37jC in a 5% CO2 incubator, and transduction efficiency was evaluated macroscopically every other hour from 1 to 24 h. Peritoneum was removed from the chamber at 24 h after h-galactosidase development for microscopic analysis. Transduction of Urothelium with an AAV-GFP Vector Excessive external connective tissue from porcine bladder or human ureter specimens was microdissected and removed. Next, the specimens were opened, their mucosa was washed with sterile DMEM, and they were placed in the dot-blot chamber with the urothelium facing upward. The
309
METHOD
chamber was tightly closed to prevent loss of medium and vector suspension, and transduction could be evaluated ex vivo in uniform fields. The GAG layer was removed from ureter and bladder specimens by incubating the mucosa at room temperature with an acid solution [13]. Wells were then washed three times with DMEM and infection was obtained by the addition of 2 109 particles/well of an adeno-associated viral vector carrying the cDNA for green fluorescent protein (AAV-GFP) in 50 Al DMEM solution. The chamber was maintained at 37jC in a humidified atmosphere with 5% CO2 for 6 h. After removal of the vector-containing solution by gentle washing, DMEM (400 Al) supplemented with 10% FBS, 100 U/ml penicillin G, 100 U/ml streptomycin, and glutamine was then added to the wells, and the mounted chamber was maintained at 37jC in the humidified CO2 incubator for 7 days. Daily evaluation of the chamber revealed no loss or significant evaporation of the medium. Fluorescence was evaluated with an Olympus IX70-S8F2 fluorescence microscope (Olympus, Japan) and was initially detected 4 days after infection. The signal progressively enhanced until the 7th day, when it was maximal. Targeted Transduction of Urothelium with an Ad-RGD-GFP Vector External connective tissue from porcine bladder specimens was microdissected and removed. Positioning of the tissue in the chamber and removal of the GAG layer were done as described above. Infection was obtained by the addition of 2 109 particles/well of the adenoviral vectors in 50 Al DMEM solution. We used an adenovirus carrying the cDNA for GFP (Ad-GFP) and we compared transduction to a targeted adenovirus expressing GFP (Ad-F/RGD-GFP). A control vector with no reporter gene was used as a negative control. Vector solutions were maintained for 6 h and then removed by gentle washing and supplemented DMEM was added to the wells. The mounted chamber was maintained at 37jC in the humidified CO2 incubator for 2 days and fluorescence was evaluated with a Zeiss LSM 510 META laser-scanning microscope (Jena, Germany). Pixel density was evaluated in the pictures with the ImageJ software (National Cancer Institute, Bethesda, MD, USA). Statistical Analysis The statistical significance of differential findings between experimental wells and controls was determined by Student’s t test and considered significant if two-tailed P values were
