Coxsackievirus Adenovirus Receptor Expression Predicts the Efficiency of Adenoviral Gene Transfer into Non-Small Cell Lung Cancer Xenografts

Gene delivery to target cells is required for effective gene therapy. Investigations over the last several years have elucidated mechanisms by which Ad² vectors mediate gene transfer into cells in vitro. In cell culture systems, studies indicate that adenovirus initially attaches to a specific high-affinity cell surface receptor [CAR (1, 2)]. This attachment apparently triggers a clustering of cell surface integrins, viral endocytosis using coated pit mechanisms, and cellular actin stabilization, (3, 4, 5). The virus begins to disassemble at the cell surface and, after internalization, disrupts endosomal vesicles to escape into the cytoplasm (6, 7, 8). Endosomal escape, which also uses integrin-mediated processes (9, 10), is followed by microtubule-mediated transport of the residual Ad nucleocapsid to the nucleus (11, 12, 13) and episomal expression of the vector DNA. Despite the elucidation of cellular mechanisms underlying Ad gene transfer, however, successful translation in the clinic is yet to be realized, in part because of inefficient gene delivery. For example, in the experimental treatment of advanced human lung cancer with the use of adenovirus-mediated gene delivery, investigators instilled high doses of adenovirus (up to 7.5 × 10¹² particles) directly into lung tumors by bronchoscopy or transthoracic computed tomography-guided injection (14, 15, 16). Using intratumoral transgene mRNA (by RT-PCR of vector-specific sequences) as a marker of transduction efficiency, Swisher et al. (14) and Nemunaitis et al. (15) reported less than 40% to 50% expression, and Schuler et al. (16) reported transgene expression in 68% of subjects with repeated testing. Unfortunately, it is currently not possible to extrapolate clinical benefits derived from molecular evidence of in situ transgene expression. Collectively, these results suggest that efficient in vivo gene delivery to target cells remains a major limitation for the successful gene therapy of NSCLC (17, 18, 19).

The studies detailed here have focused on the expression and importance of CAR in mediating gene transfer into NSCLC cells in vivo. CAR is a M_r 46,000 transmembrane protein that interacts with the Group C Ad fiber protein with high affinity [with an estimated dissociation rate constant (K_d) of ∼0.5 to 1 × 10⁻⁸ m (20, 21)]. The cellular function of CAR, aside from serving as the attachment receptor for viruses, is largely unknown. Recent reports suggest that the protein serves to mediate homotypic cell-cell adhesion as a component of the tight junction (22). In vitro analyses suggest an absence or paucity of CAR expression in tumor cells (23, 24, 25, 26, 27), leading to the supposition that the usefulness of Ad vectors for cancer gene therapy is likely to be limited in vivo (28). It is unknown, however, whether CAR expression in vitro is a reliable predictor of expression in vivo, and the data are conflicting with respect to whether CAR expression in vivo is sufficient to mediate efficient gene transfer into target cells (29, 30). Moreover, even if CAR expression is a correlate of adenovirus-mediated gene transfer efficiency in vivo, it is not clear whether specific and/or physical barriers in vivo negate the efficiency of Ad transduction that is conferred by CAR expression (30, 31, 32, 33, 34, 35, 36).

Investigating gene transfer to lung cancer, our previous studies provided conclusive evidence that gene transfer into NSCLC cells is heterogeneous (18). Whereas transduction-sensitive cells bind Ad vector with markedly greater efficiency and specificity for the Ad fiber knob, transduction-refractory cells bind and internalize vector by less efficient pathways (18). Thus, these pilot studies suggested a primary difference in CAR expression between transduction-sensitive and transduction-refractory NSCLC cells. Using these data as a platform for further investigation, we report (a) that differences in CAR expression between transduction-sensitive and transduction-refractory NSCLC cells are present in vitro, (b) that these differences in CAR expression persist in xenograft models, and (c) that controlling for adequate vector internalization mechanisms (3), CAR expression is predictive for more efficient gene transfer but insufficient to achieve uniform transduction by Group C Ad vectors in vivo.

NCI-H226, NCI-H1703, NCI-H1437, and NCI-H2122 cells (a gift of Dr. Herbert Oie, National Cancer Institute) were maintained in RPMI 1640 (Irvine Scientific, Santa Ana, CA) with 10% fetal bovine serum (Gemini, Woodland, CA) and penicillin (100 units/ml)/streptomycin (100 μg/ml) [complete growth medium (18)]. Ad vectors were constructed in the Vector Core at the Gene Therapy Center of the University of North Carolina School of Medicine and amplified in the University of California at Los Angeles-Jonsson Comprehensive Cancer Center. Ad5LacZ is E1a/E1b and partially E3-deleted and expresses the reporter LacZ gene under the control of the cytomegalovirus promoter region. Ad vectors were typically purified and concentrated with double CsCl ultracentrifugation and stored at −20°C in a nonfreezing solution containing 25% glycerol, 0.05% BSA, 4 m CsCl, 50 mm NaCl, 0.5 mm MgCl₂, and 5 mm Tris buffer. Immediately before use, vectors were gel filtered (G-50 Sephadex; Boehringer Mannheim, Indianapolis, IN) and eluted into growth medium for transduction studies as described previously (18).

Total RNA was extracted from transduction-sensitive and -resistant cells using TRIzol reagent (Life Technologies, Inc., Grand Island, NY). RT-PCR was performed using the SuperScript One-Step RT-PCR system (Life Technologies, Inc.). Primers were as follows: for CAR expression, 5′-AAACCGCCTACCTGCAGCCG-3′ (sense) and 5′-GAGCTTTATTTGAAGGAGGGACAACG-3′ (antisense) (product size, 747 bp); and for β-actin (control) expression, 5′-CTCGTCGTCGACAACGGCTC-3′ (sense) and 5′-AAACATGATCTGGGTCATCTTCTC-3′ (antisense) (product size, 353 bp). The RT-PCR protocol was as follows: cDNA synthesis (1 cycle: 54°C for 30 min) followed by 94°C for 2 min. DNA amplification followed (38 cycles: 94°C for 15 s; annealing at 55°C for 30 s; and extension at 72°C for 1 min) with a final extension at 72°C for 10 min. A plasmid [pcDNA3 containing the CAR cDNA insert; a kind gift from Dr. David Curiel (University of Alabama at Birmingham Gene Therapy Center)] encoding for CAR was used as positive control for cDNA amplification and for confirmation of the amplicon as CAR by Southern hybridization.

For quantitative real-time RT-PCR, total RNA was extracted from cells in log-phase growth according to the manufacturer’s protocol using the TRI Reagent (Molecular Research Center, Cincinnati, OH). The extracted total RNA was solubilized (in diethyl pyrocarbonate-H₂O) at a concentration of 100 ng/μl, and 500 ng were used for amplification in an iCycler (Bio-Rad, Hercules, CA). Reaction components and cycling conditions were modified from manufacturer’s protocol in the QuantiTech SYBR Green RT-PCR Kit (Qiagen, Valencia, CA). Briefly, a 124-bp CAR amplicon was reverse transcribed and amplified using the following primers: sense, 5′-AGGGACCGCTGGACATCGAG-3′; and antisense, 5′-ACTCGGCCTTTCAGATCTGG-3′. Similarly, a 115-bp amplicon of the control β-actin gene was reverse transcribed and amplified using the following primers: sense, 5′-TGTGGGATCAGCAGAGGAG-3′; and antisense, 5′-CTCGTTTGGGAGCTGTTCA-3′. After reverse transcription (50°C, 20 min), Taq DNA polymerase was activated (95°C, 15 min, 1 cycle), and DNA PCR amplification was performed (45 cycles: denaturation at 94°C for 15 s; annealing at 55°C for 30 s; extension for 15 s). The threshold cycle for CAR expression in each cell line was normalized to β-actin gene expression and standardized to positive control amplification of the CAR cDNA insert in pcDNA3.

Four-week-old female SCID/SCID received s.c. injection in the flanks with 8 × 10⁶ cells/mouse. When tumors reached a cross-sectional area between 0.5–1 × 0.5–1 cm² (2–4 weeks, depending on the cell line), they were directly injected with Ad5LacZ at a MOI of 100 in a volume of 50 μl of PBS. The number of cells in the tumor (for determination of MOI) were estimated based on the following measurements and assumptions: (a) bisecting diameters of the tumors were measured, and the tumor volume was approximated using the formula 0.4 (ab²), where a is the long measured axis of the tumor, and b is the short measured axis (37); (b) the cells within the tumors were assumed to be spherical (volume = 4/3 πr³) with a radius of 10 μm; and (c) the entirety of the tumor mass was assumed to be comprised of tumor cells. Thirty h after vector injections, the tumors were extirpated, minced, and suspended in 4 IU/ml collagenase (type II collagenase; Sigma, St. Louis, MO) in RPMI 1640 at 4°C overnight. The following morning, cells were washed twice in PBS (Irvine Scientific), fixed (0.5% gluteraldehyde, 10 min, 4°C), washed twice in PBS containing 100 mm MgCl₂, and evaluated for LacZ expression by histochemistry (for intracellular 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside hydrolysis; Life Technologies, Inc., Rockville, MD). Tumor cells were identified morphologically, and the percentage of positive (blue) cells was determined by counting a total of at least 400 cells from each group under a hemacytometer as described previously (18).

Soluble recombinant CAR and rabbit antiserum to this protein (anti-recombinant CAR antiserum; Cocalico Biologicals, Reamstown, PA) were gifts from Dr. J. Bergelson [Children’s Hospital of Pennsylvania (35)]. For Western immunoblots, total protein was extracted from cells during log-phase growth using Mammalian Cell-PE LB buffer (Genotechnology, Inc., St. Louis, MO) at 4°C and quantified using the Bio-Rad Assay Kit (Bio-Rad). Fifty μg of protein were suspended in 1× volume of Laemmli sample buffer (Bio-Rad), and the volume of the samples was brought to 50 μl with PE LB buffer for loading. The samples were heated at 95°C for 5 min, resolved by 10% SDS-PAGE, and transferred to immunoblot polyvinylidene difluoride membrane (Bio-Rad) on ice. After blocking the membrane with NAP-Sure Blocker (Genotechnology, Inc.), the membrane was washed three times in PBST over 30 min and probed with rabbit anti-CAR antiserum (1:2500 dilution in PBS, 1 h, room temperature). After three washes in PBST, the primary was labeled with antirabbit-alkaline phosphatase conjugate (1:3000 dilution, 2 h, room temperature; Promega, Madison, WI), and after repeat washes in PBST, specific bands were detected by chemiluminescence (ProtoBlot II AP System; Promega).

RmcB, a murine monoclonal antibody that recognizes the extracellular NH₂ terminus region of CAR (38), was purified from hybridoma (CRL-2379; American Type Culture Collection, Manassas, VA) SCID mouse ascites using protein G affinity chromatography (Amersham Pharmacia Biotech) and used to specifically detect cellular CAR expression by flow cytometry (fluorescence-activated cell sorting), immunocytochemistry, and IHC. To detect CAR using fluorescence-activated cell sorting, 1 × 10⁶ cells were preincubated in 100 μl of 0.1% BSA (in PBS, 20 min, room temperature) before primary antibody (RmcB; mouse IgG1, 1:100 dilution in PBS/0.1% BSA, 90 min, room temperature) was admixed on an orbital shaker. Surface expression for the α_v integrin was accomplished in a similar manner, except the mouse monoclonal MAB1960 (Chemicon, Temecula, CA) was used for primary labeling. Cells were then sedimented and washed three times with PBS/0.1% BSA, incubated with secondary antibody [FITC-conjugated sheep antimouse (Fab′)₂ (Sigma); 1:200 dilution in PBS/0.1% BSA, 30 min, room temperature in the dark], washed three times, and resuspended in 500 μl of PBS for flow cytometry by FACScan using CellQuest software (Becton Dickinson, Mountain View, CA). For all data acquisitions and analyses, gates were based on the forward and side scatter profiles of unstained cells, and surface expression of target proteins was normalized to that of cells that had been incubated with secondary antibody alone. For IHC, xenografted tumors were dissected, fixed in Bouin’s liquid fixative (Fisher Scientific), and embedded in paraffin, and 5-μm sections were sliced. After deparrafinization and serial rehydration, the sections were rinsed for 5 min in tap water. Endogenous peroxidase activity was quenched by incubating the tissue sections in 0.3% H₂O₂ for 30 min. After a PBS rinse, the sections were blocked with 2% goat serum in PBS (30 min, room temperature) and washed, and the primary antibody (RmcB; mouse IgG1) was applied at a 1:100 dilution at 4°C overnight. After three washes in PBST followed by PBS, RmcB was labeled with biotinylated goat antimouse antibody [1:200, 30 min at room temperature (Vector-ABC method for IHC staining; Vector Laboratories Inc., Burlingame, CA)]. After washing with PBS, the tissue was incubated for an hour in horseradish peroxidase-conjugated streptavidin, and the CAR-reactive products were visualized using 0.5% diaminobenzidine and H₂O_2. The tissue was then counterstained with hematoxylin, and the histology was photomicrographed by light microscopy. CAR labeling imparted a brown color.

All transduction and surface labeling experiments were performed using at least three specimens a minimum of two separate times. To determine the statistical significance for differences in transduction efficiency, results were indexed to Ad MOI, and comparisons were made using Kruskal-Wallis ANOVA on ranks, followed by Bonferroni group comparisons. A statistically significant difference was defined as P < 0.05 between the groups compared.

Ad transduction of NSCLC cells is heterogeneous, and these results prompted us to begin evaluating the structural basis for the observed differences in transduction efficiencies. The heterogeneity of adenovirus-mediated gene transfer into NSCLC cells can be exemplified by the disparity in LacZ expression between adenovirus-susceptible [NCI-H226 (□) and NCI-H1703 (○)] and adenovirus-resistant [NCI-H1437 (▪) and NCI-H2122 (•)] cell lines after infection at four different Ad5LacZ-MOIs (Fig. 1). To discern the cause of this heterogeneity and its applicability to gene transfer in vivo, we chose to study these particular cells as models because (a) they required identical conditions for growth in vitro, (b) we had empirically confirmed their capacity to grow as xenografts in immunocompromised mice, and (c) they exhibited similar amounts of α_v integrin expression as measured by flow cytometry (Table 1). The latter observation suggested a functional equivalence between these cells in the capacity to internalize Ad vector after surface binding. Accordingly, these cells enabled us to more precisely define the role of CAR expression in determining Ad transduction efficiency in vitro and in vivo.

To begin with, RT-PCR and quantitative real-time RT-PCR were used to examine differences in CAR expression between the adenovirus-sensitive and adenovirus-resistant NSCLC cells. Semiquantitative RT-PCR (Fig. 2,A) suggested that compared with the constitutive expression of the β-actin gene, the steady-state expression of CAR was reduced in cells that exhibited relatively poor Ad transduction efficiency and was increased in adenovirus-susceptible cells. To confirm this observation, real-time RT-PCR was performed on these cells. The results were congruent (Fig. 2 B), and the steady-state CAR expression in adenovirus-sensitive (NCI-H226 and NCI-H1703) cells was distinctly increased when compared with the transduction-resistant cells (NCI-H1437 and NCI-H2122). The CAR mRNA concentration in extracts of NCI-H226 cells was estimated to be 10 pg/μl, based on a standard curve generated using control amplification of the CAR cDNA insert in pcDNA3, although similar extrapolations could not be made from threshold values of the remaining cells because they fell outside the linear range of the standard curve.

To evaluate the surface expression of CAR in these cells, we performed a series of immunolocalization studies on cytological preparations and protein extracts. CAR-specific immunofluorescent tagging of cytological preparations demonstrated increased membrane labeling in adenovirus-susceptible cells (NCI-H226 and NCI-H1703) versus adenovirus-resistant cells [NCI-H1437 and NCI-H2122 (data not shown)]. Similarly, after surface labeling and evaluation by flow cytometry, CAR expression on NCI-H226 and NCI-H1703 cells was reliably 4-fold and 10-fold higher, respectively, when compared with NCI-H1437 and NCI-H2122 cells (Fig. 3,A). Likewise, Western blots indicated increased CAR expression in NCI-H226 and NCI-H1703 cells as compared with the adenovirus-resistant cells (Fig. 3,B). Thus, by all indications, the general patterns of CAR expression were consistent with the PCR data, and CAR expression was increased in adenovirus-susceptible cells (NCI-H226 and NCI-H1703) as compared with the adenovirus-resistant cells (NCI-H1437 and NCI-H2122). Collectively (Figs. 1,2,3), these data indicate that Ad transduction in vitro positively correlates with CAR expression, provided expression of integrins is a not limiting factor.

The cellular function of CAR is unknown, although extracellular, transmembrane, and cytoplasmic domains have been identified (39, 40), and its putative role as an adhesion receptor that mediates homotypic intercellular contacts has been hypothesized (22). To date, however, it is not known whether CAR expression is affected by culture conditions, or how its expression as a component of tight junctions and/or adherens junctions is regulated. Accordingly, we believed that it was important to ascertain whether the observed differences in CAR expression in vitro between adenovirus-susceptible and adenovirus-resistant NSCLC cells persisted in vivo. To characterize this expression, we established s.c. xenografts and examined CAR expression within these tumors by IHC (Fig. 4). These studies evidenced an expression pattern that was predicted by the in vitro analyses, with adenovirus-susceptible cells (NCI-H226 and NCI-H1703) exhibiting a demonstrably stronger staining pattern for CAR as compared with the adenovirus-resistant cells (NCI-H1437 and NCI-H2122; Fig. 4). Thus, these studies indicated that a pattern of CAR expression exhibited by NSCLC cells in vitro was qualitatively recapitulated in s.c. xenografts implanted in immunodeficient mice in vivo.

It is unknown whether CAR expression is mandatory and/or sufficient to mediate efficient Ad gene transfer in vivo. Indeed, when gene transfer efficiency has been directly examined after the controlled expression of CAR in polarized lung epithelial cells in vivo, the results have been conflicting (29, 30). Similarly, it remains to be determined whether exuberant CAR expression in tumors is a surrogate measure for efficient Ad transduction, and whether the extent of transduction efficiencies achievable in vivo are comparable with those exhibited by the same cells in vitro. To begin to answer this question, we injected AdLacZ, at a MOI of 100, into xenografts established using the adenovirus-susceptible (NCI-H226 and NCI-H1703) and adenovirus-resistant (NCI-H1437 and NCI-H2122) NSCLC cells. A MOI of 100 was chosen based on the observation that the adenovirus-susceptible (NCI-H226 and NCI-H1703) NSCLC cells exhibited uniform transduction at that dose in vitro (Fig. 1). Thirty h after single intratumoral vector injections, the tumors were extirpated, minced, and digested to yield tumor cell suspensions that were evaluated for LacZ expression by histochemistry. Our results, as shown in Fig. 5, indicate that although Ad gene transfer positively correlates with in vivo CAR expression, the efficiency and reliability of transduction of adenovirus-susceptible NSCLC cells may be variable. Thus, whereas NCI-H226 and NCI-H1703 cells exhibit uniform (100%) transduction in vitro at an Ad MOI of 100, the tumors generated using these cells displayed transduction efficiencies of 45.3 ± 26% and 24.1 ± 8.4%, respectively, after a single intratumoral injection in vivo. In contrast, gene transfer efficiency into CAR-deficient cells in vivo remains similar to that observed in vitro and peaks at approximately 5% of cells within the tumor mass. These results indicate that CAR expression is predictive for more efficient gene transfer but is insufficient to achieve uniform transduction by Group C Ad vectors in vivo.

A number of investigators have postulated that poor Ad transduction is caused by poor CAR expression in tumors in vivo (24, 25, 41). That postulate tacitly suggests that robust CAR expression will enable efficient Ad gene transfer and that enhancing (pharmacologically or otherwise) CAR expression will augment Ad transduction efficiency in vivo. To begin to understand the relevance and importance of CAR for mediating intratumoral gene transfer in vivo, we chose to directly investigate the relationship between CAR expression and Ad transduction in defined NSCLC cell model systems. First, we confirmed that the Ad transduction efficiency of NSCLC cells in vitro can be directly related to CAR expression (Figs. 1,2,3). Next, we demonstrated that CAR expression in vitro predicted a comparable pattern of CAR expression in transplanted NSCLC xenografts in vivo (Fig. 4). Finally, whereas increased CAR expression in the NSCLC xenografts may be a correlate, it is not a direct surrogate for predicting Ad transduction efficiency in vivo (Fig. 5). Accordingly, CAR expression alone may not be a guarantee for highly efficient or “therapeutic” gene delivery in vivo, and the efficiency and reliability of in vivo Ad transduction are likely also controlled by parameters other than the expression of attachment receptors for gene transfer vectors.

We have undertaken these studies while taking into account some of the variables that potentially limit an assessment of the relationship between CAR expression and Ad gene transfer efficiency. For example, we demonstrated differential CAR expression both in vitro and in vivo between the various NSCLC cells tested, and we made an effort to use cells that were controlled for the comparable expression of α_v integrin such that postadherence Ad internalization would not be limiting for transduction. For controlling study variables, we chose a methodology that called for a single viral dose administered with a single intratumoral injection. For analysis, we also deliberately chose a technically difficult but methodical approach to quantify transduction in vivo. In this respect, because we wanted to get a global assessment of intratumoral gene transfer after a single controlled vector injection, we systematically counted a sample of transduced cells digested out of extirpated primary tumors. In our view, this method allowed us to better control for the bias introduced by the heterogeneous distribution of transduced cells (e.g., greatest density of transduced cells proximal to the injection site, and a lower proportion of transduced cells distant to the site of injection) within the tumor in situ. Despite these controls, we admit that the system for comparing the relative in vitro versus in vivo transduction efficiency of individual cell types may be somewhat artificial, with many residual confounding variables. For instance, our techniques did not take into account variations in the the intratumoral dispersion of vector, dwell time of vector within tumor bed, and/or clearance of vector by host circulation or reticuloendothelial cells in vivo. Similarly, the comparative doses of virus in vitro versus in vivo were estimates at best, using assumptions (see “Materials and Methods”) that may be disputed. Nevertheless, these measurements do provide an awareness that the vector dose at which cells exhibit uniform transduction in vitro can be estimated to yield approximately 10% to 70% transduction in vivo, when using this intratumoral mode of vector delivery. Indeed, for many gene therapy paradigms with significant bystander effects and/or antitumor immune effects, this degree of gene transfer may be expected to yield clinically therapeutic effects. Thus, pharmacological manipulations designed to up-regulate CAR expression (42, 43) may some have efficacy in terms of improving Ad gene transfer in vivo. In conclusion, this study indicates that prediction of intratumoral Ad gene transfer efficiency needs to account for more than just an examination of Ad entry receptors and that alternative tumor- or host-specific variables may need to be defined for certain gene therapy strategies.

We thank Brian Kuo and Kimberly Atianzar for technical assistance.