Adenoviral-mediated gene transfer is suboptimal in human glioma and limits in vivo gene therapy approaches. There is a need for targeted vectors able to enhance gene transfer into the tumor as well as to lower the viral load in the surrounding normal tissues. We evaluated primary human tumor samples by immunohistochemistry and fluorescence-activated cell sorter for expression of the Coxsackie-adenovirus receptor and other antigens with potential utility to redirect adenoviruses (Ads) to gliomas. In the majority of the samples, Coxsackie-adenovirus receptor expression was low. This correlated with inefficient gene transfer in vitro. Epidermal growth factor receptor (EGFR) and αvβ5 integrins were often highly, but heterogeneously, expressed. We hypothesized that these receptors, overexpressed in tumor but not in normal brain, could serve as independent binding sites for alternative pathways of infection with targeted Ads. We examined this, using Ads that expressed the luciferase reporter gene under the cytomegalovirus promoter. Targeting to the EGFR was performed with a single-chain bispecific antibody directed against the human EGFR and against the fiber knob of the Ad. Targeting to the αv integrins was performed by insertion of an integrin-binding sequence, RGD-4C, in the HI-loop of the Ad. Increased luciferase gene transfer in primary glioma cells was observed in 8 of 13 samples with EGFR-targeting (2–11 times enhancement; median, 6) and in all of the samples with RGD-targeting (2–42 times enhancement; median, 12). Combining the two targeting motifs further enhanced the gene transfer in primary glioma cells in an additive manner (3–56 times; median, 20). The double-targeted Ads also strongly augmented gene transfer into organotypic glioma spheroids. Conversely, gene transfer into normal brain explants was reduced dramatically using Ads targeted to the tumor. Our findings demonstrate the feasibility and benefit of binding multiple ligands to the adenoviral fiber knob. These vectors have a great potential for clinical use in the context of tumors that are usually heterogeneous for target antigen expression at the single-cell level.
Nervous system tumors are one of the leading causes of cancer-related death, especially in children. For patients with malignant glioma, present treatment options mostly consist of surgery and irradiation. Chemotherapy and immunotherapy have not brought significant improvement of survival in these patients. Recent advances have made it possible to consider using gene transfer techniques for the treatment of cancer. Because gliomas are relatively refractory to current treatments, they are good candidates for gene therapy. A variety of different vectors and delivery techniques are being studied to transfer genes in the central nervous system (1) . Although several studies have shown successful treatment of brain tumor in animal models (2) , these strategies have not been translated successfully to the clinic thus far (3 , 4) . The main problem encountered in most of the trials was related to the inefficient delivery of the vector to the tumor cells. Moreover, analysis of the long-term survivors in the animal studies pointed out that chronic toxicity attributable to transduction of normal cells with suicide genes may be substantial (5) . Taken together, these data call for better vectors, more specific for tumor cells with enhanced gene transfer efficiency.
Adenoviral vectors have been widely used to transfer genes because of their unique ability to accomplish efficient gene delivery in various tissues. The initial and limiting step of Ad3 infection depends on the binding of the virus to the CAR (6) . Several tumor cell lines have shown relatively low CAR expression, e.g., head and neck carcinomas (7) , melanoma (8) urological tumors (9) , and gliomas (10) , thus limiting the efficiency of Ad-mediated gene transfer.
Ads are able to infect glioma cells but also normal brain cells, including ependymal cells, astrocytes, and neurons in animal models and in humans (2 , 11) . Expression of CAR is high in normal human brain tissue (12) . Thus, in vivo, the recombinant virus may be sequestered by high CAR-expressing normal cells, whereas the true target cells, if low in CAR, are poorly transduced. Therefore, it is important to provide adenoviral vectors with an alternative, and more effective, entry pathway into the tumor cells. In recent years, adenoviral genetic targeting toward specific receptors became possible through ligand insertion in the adenoviral fiber [Refs. 13 , 14 ; reviewed by Krasnykh et al. (15)] . Alternatively, indirect targeting can be performed through a bispecific molecular bridge between the Ad and the target cell. In the latter approach, a neutralizing antibody against the fiber knob blocks the interaction with CAR, and the other binding moiety redirects the virus to a novel receptor (16, 17, 18) . We have shown previously that redirecting Ads to the EGFR via a bispecific antibody can dramatically enhance gene transfer to various tumor cell lines, including glioma cell lines (10 , 18) . Because the expression of EGFR is restricted to glioma cells and virtually absent in the normal brain, this target is relevant for glioma-specific gene therapy. In addition, insertion of an integrin-binding RGD-4C peptide in the fiber of the vector was previously shown to extend its tropism for target integrins, especially those containing the αv chain (14) . Because these integrins are expressed on glioma cell lines (10) and primary glioma (19) but not in the normal brain (20) , these genetically modified vectors may also have interesting targeting potency.
In glioma patients, however, tumor architecture and tumor-specific antigen expression are heterogeneous (21, 22, 23) notably for the EGFR (24 , 25) . Thus, it can be postulated that combined targeting approaches toward more than one antigen may provide better gene transfer based on the provision of a compensation mechanism for such heterogeneity.
On this basis, the aim of this study was to evaluate: (a) the availability of adenoviral receptors and alternate targets on primary glioma cells and tumors; (b) the gene transfer efficacy of untargeted and targeted Ads into primary glioma cells in vitro; and (c) the feasibility and usefulness of combined targeting strategies.
MATERIALS AND METHODS
Human cell lines A431 (vaginal carcinoma) and U373 MG (anaplastic astrocytoma) were obtained from the American Type Culture Collection (Manassas, VA). Human glioma-derived cell line U118 MG (glioblastoma multiforme) was obtained from Dr. J. T. Douglas (Gene Therapy Center, University of Alabama at Birmingham, Birmingham, AL). Human glioma cell line Gli-6 was obtained from Dr. S. Leenstra (Neurosurgery, AMC, Amsterdam, the Netherlands). All of the cell lines were grown in DMEM supplemented with 10% FCS (Life Technologies, Grand Island, NY) and antibiotics.
Primary Glioma Cell Cultures and Spheroids.
We used fresh material collected during brain tumor surgery from adult patients. Primary glioma cell cultures were obtained after mechanical dissociation, according to the technique described by Darling (26) . The cells were cultured in DMEM or HAM-F12 supplemented with 10% FCS and antibiotics. Glial origin was confirmed by morphology and by staining with anti-GFAP MoAb clone 6F2 (Dako, Glostrup, Denmark). All of the experiments on these cells were done before passage 5, usually between passage 1 and 2. From tumor sample VU-1, two morphologically and phenotypically different subsets could be derived after two passages, VU-1.1 growing in HAM-F12 and VU-1.2 growing in DMEM. They were used as distinct samples in the experiments.
Organotypic multicellular glioma spheroids were grown from small explants of fresh human tumor in 48-well plates coated with agarose, one spheroid per well, according to the technique originally described by Bjerkvig et al. (27) . After checking for viability by morphology and trypan blue exclusion, spheroids of similar diameter (400–500 μm) were used for gene transfer experiments. Spheroids made from the Gli-6 cell line served as control. In addition, explants of normal brain of patient VU-15 could be made and cultured for 1 week.
Mouse anti-CAR monoclonal antibody, RmcB (28) , prepared as ascites fluid, was obtained from Dr. R. L. Crowell (Hahnemann University, Philadelphia, PA). Anti-αvβ5 integrin MoAb clone P1F6 was purchased from Life Technologies (Breda, the Netherlands). Anti-αvβ3 integrin MoAb clone LM609 was purchased from Chemicon (Temecula, CA). Supernatant of the 425-hybridoma (29) culture, purchased from the American Type Culture Collection, was used as a source for anti-EGFR MoAb.
Cultured cells were trypsinized, washed with PBS, and centrifuged. Cells (500,000) were incubated with first antibody diluted in PBS containing 0.1% BSA for 1 h on ice. Concentrations of the first antibody were 10 μg/ml, 2.5 μg/ml and 10 μg/ml for RmcB, P1F6, and LM609, respectively. For the 425 MoAb, undiluted supernatant of the hybridoma was used at a concentration of 1 μg/ml of IgG1 as measured by ELISA. The irrelevant first antibody used was a mouse monoclonal IgG1 antibody 323A3 directed against an epithelial marker not expressed on glioma cells at a concentration of 10 μg/ml Subsequently, the cells were washed and incubated with FITC-conjugated rabbit antimouse antibody (Dako) for 30 min on ice and in the dark. After washing in PBS, the cells were resuspended in 500 μl of PBS containing 1% formaldehyde. Analysis was performed on a FACScan (Becton Dickinson, Erembodegem-Aalst, Belgium). Human glioma-derived cell line U373 MG served as positive control for CAR, EGFR, αvβ3 and αvβ5 integrins because high positivity for these four antigens has been shown previously, and U118 MG served as negative control for CAR (10) .
The fluorescence intensity of each sample was quantified as the ratio between median fluorescence intensity of the stained sample and median fluorescence intensity of the irrelevant antibody control. These ratios were compared with those of the positive control cell line, U373, and expressed as a percentage of the latter value. Levels of the target antigens were considered as high if they were at least 50% of that of the positive control cell line.
IHC was done on frozen sections of 4 μm fixed with acetone. We used a frozen pellet of 293 cells and normal human liver as positive controls for CAR. Normal human skin was used as positive control for EGFR and normal human tonsil as positive control for αvβ3 and αvβ5 integrins. Nonspecific antibody binding was blocked by incubation with 2% normal rabbit serum. Tumor slides and positive controls were incubated for 1 h at room temperature in blocking buffer (1% BSA in PBS) with the same mouse monoclonal antibodies that were used for cytometry. Concentrations were 1 μg/ml for 425 and 10 μg/ml and for RmcB and for LM609 and P1F6, respectively. Two negative controls were used for each staining. These negative controls were treated either with blocking buffer only or with normal mouse IgG1κ. Specimens were then washed in PBS and incubated with biotinylated rabbit antimouse IgG antibody (Dako A/S Glosdrup, Denmark) diluted 1/500 in PBS for 30 min. Second antibody binding was visualized with a peroxidase conjugated streptavidin ABC kit (Dako A/S) diluted 1:200 in PBS for one h. The slides were counterstained with hematoxylin. Immunoreactivity was graded (−) if no tumor cells were positive, (+) if less than 5% of tumor cells were positive, (++) if 5–50% of tumor cells were positive, (+++) if 50–80% of tumor cells were positive, and (++++) if >80% of cells were positive.
A recombinant E1-deleted Ad expressing the luciferase reporter gene under the CMV promoter, AdCMVLuc, was provided by Dr. R. D. Gerard (University of Texas Southwestern Medical Center, Dallas, TX). A similar Ad that also expresses the luciferase from the CMV promoter but contains an integrin-targeting peptide within the HI-loop, Ad5lucRGD, was generated as described previously (14) . The presence of the RGD-binding motif in the HI-loop was confirmed by PCR and restriction enzyme analysis. The functionality of the RGD-binding motif was checked by the ability of the vector to enhance gene transfer in CAR-negative cell lines (CHO and U118 MG). Recombinant Ads were propagated on the permissive 293 cell line and were purified using cesium chloride gradient banding. Subsequently, they were titered in parallel in particles by the A260 method and in plaque-forming unit by end point titration on 293 cells. The titers were 3.3 × 10+12 particles/ml and 4.1 × 10+12 particles/ml for AdCMVLuc and Ad5lucRGD, respectively. The ratio infectious particles:particles was 1 in 4 and 1 in 10 for AdCMVLuc and Ad5lucRGD, respectively. When the two viruses were compared, the same number of particles was used.
Bispecific scFv antibodies were produced in mammalian cells as described previously (18) . The 425-S11 fusion protein recognizes on one side the EGFR (30) , and on the other side the fiber knob (31) .
Supernatant of transiently transfected COS-7 cells was used for the targeting experiments. The optimal ratio between COS-7 cell supernatant and viral particles was determined by targeting experiments on A431 and U118 MG (18) . One single batch of supernatant of transfected COS-7 cells was used for all of the experiments.
Gene Transfer Assays.
To assess adenoviral infection on cell lines, 10+5 cells/well were plated in a 24-well plate in triplicate and were incubated overnight in 1 ml of culture medium to allow adherence. Before infection, 10+8 vp of Ad were incubated with 25 μl of the 425-S11 supernatant for 30 min at room temperature. Next, the mixture was diluted in DMEM containing 2.5% FCS to a concentration of 5× 10+7 vp/ml, and 200 μl were added to each well, i.e., 100 vp/cell. The cells were incubated at 37°C in 5% CO2 for 1 h, washed with PBS, and then supplemented with 1 ml of DMEM containing 10% FCS. Twenty-four h after infection, the cells were assayed for luciferase expression.
To assess adenoviral infection on primary glioma cells, a similar procedure was used but with 10+4 cells per well in a 96-well plate in quadruplicate. The viruses were five times more diluted than for the cell lines and only 100 μl of the solution were added to each well, i.e., 100 vp/cell.
To assess adenoviral infection on spheroids, each spheroid was plated in a separate well of a 96-well plate coated with agarose, in 150 μl of DMEM with 10% FCS. Ten+7 vp of the conjugated viruses or of the control viruses, diluted in 50 μl of DMEM containing 2.5% FCS, were added to the wells.
Luciferase assay was performed after 24-h incubation at 37°C in 5% CO2.
To measure luciferase activity in the cells, we used the luciferase assay system (Promega, Madison, WI). After removing the culture medium, lysis buffer was added to the wells and the whole plate was snap-frozen on dry ice. After thawing, the luciferase activity was measured during 10 s immediately after initiation of the light reaction in a Lumat LB 9507 luminometer (EG&G Berthold, Bad Wildbad, Germany). Values were normalized per number of viable cells (trypan blue exclusion).
To measure luciferase activity in the spheroids, they were removed from the wells, rinsed with PBS and resuspended individually in 100 μl of lysis buffer. After two cycles of freeze-thawing and vortexing to insure complete lysis of the spheroids, the luciferase measurement was performed using 25 μl of the sample. Values were expressed per spheroid.
IHC on Primary Human Gliomas.
Fresh tumor samples of 13 consecutive patients with glioma, operated on between November 1999 and April 2000, were obtained for analysis.
Table 1⇓ summarizes the results of IHC analyses for expression of CAR, EGFR, and αvβ3 and αvβ5 integrins, performed on 12 primary tumors. Fig. 1⇓ shows, as an example, the primary tumor of patient VU-10, which was positive for all of the four antigens tested. In one case (VU-9), the frozen sample contained only a few tumor cells infiltrating the normal brain, and no conclusion could be made from the IHC study.
In only two cases, VU-7 and VU-10, more than 50% of the tumor cells were positive for CAR. The positivity for CAR was membranous and cytoplasmic in these cases (see Fig. 2⇓ for VU-10). The staining was weak and accentuated around the nucleus in the six other positive tumors, in which less tumor cells were also positive (Table 1)⇓ .
Nine tumors were positive for EGFR. In five of the positive cases, more than 80% of tumor cells were positive (see Fig. 1⇓ for VU-10). The staining was mostly cytoplasmic but membranous staining was observed as well in five cases.
In all of the cases but one (VU-1), there were isolated tumor cells positive for the αv integrins spread throughout the tumor. However, more than 10% of positive tumor cells were observed in only three cases for αvβ3 and in only four cases for αvβ5. As expected, the tumor vessels were largely positive for αv integrins (Fig. 1)⇓ .
The reactive astrocytes in the normal brain invaded by the tumor were strongly positive for CAR in a typical stellate pattern but not for the other three antigens tested.
In conclusion, the staining pattern of this unselected sample of gliomas in adults was heterogeneous both within one given tumor and among the whole group of tumors. The adenoviral receptor CAR was absent in a significant number of primary tumors and could thus lead to inefficient Ad-mediated gene transfer. On the basis of these findings, we chose to retarget the Ads toward both EGFR and the integrins to be able to infect as many tumor cells as possible.
Flow Cytometry on Primary Human Glioma Cells.
Flow cytometry analysis was done on early passages (before passage 5 but usually passage 1 or 2) of the short-term cultures from patient’s gliomas to quantify the expression of CAR, EGFR, and αv integrins.
Some of the samples exhibited intrasample heterogeneity in the expression of these target antigens. Subsets of cells expressing different levels of target antigens than the mainstream population were observed in some cases. For example, only a few VU-5 cells expressed CAR and few VU-10 cells expressed αvβ5 integrins. The heterogeneity of these cell populations was also visible on light microscopy (data not shown). In addition, the two subpopulations of primary cells grown from patient VU-1, VU-1.1 and VU-1.2, showed a markedly different pattern of EGFR and integrin αvβ5 expression (Fig. 2)⇓ .
Fig. 2⇓ shows the level of expression of the four antigens tested on 12 different short-term cultures. For VU-2 and VU-12, we could not grow enough cells to perform FACS analyses.
CAR expression was scored high in only 3 of the 12 samples but was always lower than in the positive control human glioma cell line U373 MG except for one primary cell culture, VU-6. αvβ3 integrin expression was found high in only 3 of 12 samples and was lower than in U373 MG except for VU-7. Conversely, αvβ5 integrin was highly expressed more often, i.e., in 8 of 12 cases, and the level of expression of αvβ5 integrin was similar or higher for most of the positive samples than in the positive control U373 MG. EGFR levels varied widely, ranging from negative to higher than the positive control.
These observations confirmed that the latter two candidate receptors on the primary human glioma cells might serve as useful targets for gene transfer.
Gene Transfer with Untargeted Ads.
Human primary glioma cells were infected with untargeted adenoviral vector, AdCMVLuc, (100 vp/cell). Twenty-four h later, mean luciferase activities measured with a chemiluminescent assay differed almost two log between different samples (Fig. 3)⇓ . In about one-half of the samples, gene transfer was poor. For 11 samples, both FACS data and gene transfer data were available. When median gene transfer level and median CAR-expression were used as cutoff, two distinct populations of samples could be identified: on one hand, those with low CAR-expression and low gene transfer and, on the other hand, those with higher CAR-expression and better gene transfer (Fig. 3)⇓ .
Thus, in this unselected sample of primary glioma cells, resistance to adenoviral-mediated gene transfer could be largely explained by CAR deficiency. This important observation calls for improved gene transfer via CAR-independent pathways.
Feasability of a Combination of Immunological and Genetic Targeting.
We have shown previously that targeting of Ads to the EGFR with 425-S11 bispecific scFv markedly enhances gene transfer efficiency into various human cell lines via the EGFR pathway (18) . These experiments were done using adenoviral vectors without capsid modifications. Here, we explore the possibility to similarly target vectors with a modified fiber knob carrying a RGD motif in the HI loop that mediates CAR-independent gene transfer through binding to integrins (14) . Fig. 4⇓ shows how the different adenoviral vectors used in this study were derived using both immunological and genetic targeting strategies.
Fig. 5⇓ shows the additive effect of the two targeting strategies on U118 MG, a human glioma cell line lacking CAR. Enhancement of gene transfer with targeted vectors was 51-fold and 53-fold for AdCMVLuc with 425-S11 and Ad5lucRGD alone, respectively. The combination resulted in an additive 84-fold augmentation of gene transfer compared with untargeted Ads. This additive effect of double-targeting was also observed on two human carcinoma cell lines, A431 and OVCAR-3 (data not shown). We conclude that by providing the fiber knob with different ligands targeted to distinct receptors, the chance for successful binding of the Ads to cells carrying these receptors is increased.
Gene Transfer with Targeted Ads into Primary Glioma Cells.
Targeting experiments were done on early passages of 13 short-term human glioma cultures (patients VU-1 to 12 and patient VU-15) with 425-S11 bispecific scFv immunological targeting, with RGD-mediated genetic targeting, or with the combination of both targeting approaches (Fig. 6A)⇓ . For the sample VU-9, we could not grow enough cells to perform the targeting experiments.
The ratio of gene transfer with EGFR-targeted AdCMVLuc compared with gene transfer with untargeted AdCMVLuc ranged from 0.9 (i.e., decreased gene transfer) to 11, with a median of 2.7-fold increase. Eight (62%) of 13 samples showed improved gene transfer when the vector was redirected to the EGFR. We could evaluate the targeted gene transfer relative to the expression of the target in 11 samples in which both FACS data and gene transfer data were available. When the samples were divided into two groups; one with high levels of EGFR receptor (n = 6) and one with low or negative levels (n = 5), we could observe that EGFR-targeted gene transfer was correlated with the level of expression of the EGFR (Fig. 6B)⇓ . Only one sample with low levels of EGFR but also low levels of CAR exhibited increased gene transfer with EGFR-targeted Ads. The gene transfer with integrin-targeted Ad5lucRGD was augmented 2- to 42-fold, with a median of 11.6-fold increase. There was no correlation between the levels of αv integrin expression and the efficiency of gene transfer with RGD-targeted Ads, probably because most of the samples exhibited high amounts of αvβ5 integrin (Fig. 2)⇓ . An alternative explanation could be that other integrins on the surface of the glioma cells that we did not test for, could also bind the RGD-motif (14 , 32) .
The combination of Ad5lucRGD with 425-S11 bispecific scFv was usually better than each of the two targeting approaches alone. The effect of the combination was, in general, additive, but in five cases, the gene transfer with double targeting was better than the addition of each targeting. Gene transfer enhancement, compared with the untargeted Ads, ranged from 3- to 56-fold (median, 21.1-fold increase). The differences between the four groups were highly significant (P = 0.002). The largest improvement of gene transfer was observed for the cells that were difficult to transduce with untargeted Ads. Whereas the absolute gene transfer values with untargeted Ads varied widely over almost a two-log range, the distribution of double-targeted gene transfer values was confined to only a single log range. Hence, double-targeting improves also the reproducibility of gene transfer to human gliomas.
Gene Transfer with Targeted Ads into Glioma Spheroids.
Short-term cultures of glioma cells, albeit closer to a patient’s tumor than cell lines, still represent a selection of cells able to attach and grow in the tissue culture plates. Therefore, we wanted to confirm the potential of the double-targeted vectors on organotypic spheroids. Spheroids derived from patient’s material can be cultured on agarose for several weeks and retain most of the characteristics of the primary tumor (tumor cell heterogeneity, cellular composition, histology, GFAP immunoreactivity; 33 ). IHC on established organotypic spheroids derived from patient material showed the maintained expression of CAR, αv integrins, and EGFR for at least 1 month after start of culturing (data not shown).
Organotypic spheroids from the tumor of eight patients and from the adjacent normal brain of one patient were infected either with untargeted Ads (AdCMVLuc) or with double-targeted Ads (Ad5lucRGD + 425-S11 conjugate). As shown in Fig. 7⇓ , double-targeting enhanced the gene transfer on all of these primary gliomas cultured as spheroids (median, 5.9-fold; range, 3.3- to 7.6- fold). Conversely, in the explants from the normal brain of patient VU-15, gene transfer with the double-targeted vector was markedly decreased compared with the untargeted vector. For this particular patient, the double-targeted Ads increased gene transfer 6.6-fold into the tumor and decreased gene transfer in the normal brain 18.3-fold. The ratio of gene transfer in tumor:normal brain was 0.7 with untargeted Ads and 90.6 with double-targeted Ads. Hence, double-targeting of the Ad modified the therapeutic index from an unfavorable better transduction of the brain tissue with untargeted viruses into a high tumor transduction together with a low normal brain infection.
Ad targeting is a main issue for brain tumor gene therapy, both for safety and efficacy purposes. Targeted vectors are designed to selectively localize viral transduction and gene expression to the tissue of interest (34) . Gene therapy in this context is likely to be safer. Normal parenchyma will be spared from the direct toxicity of the vector and its transgene and from the immune/inflammatory responses essentially directed to the infected cells (5) . The therapeutic ratio can be further increased if the infection via the alternate pathway is enhanced compared with the one via the common CAR/integrin-mediated pathway, thus allowing viral dose-reduction to limit toxicity. We and others have shown enhanced gene transfer on tumor cell lines with Ads targeted to the folate-receptor (16) , to the fibroblast growth factor receptor (35) , to the αv integrins (14) , to the EGFR (10 , 18) , and to the pan-carcinoma antigen, EpCAM (17) . In this study, we have explored the basis of glioma refractoriness to adenoviral vector infection, as well as two strategies to overcome the limitation of these vectors by targeting to tumor-specific antigens, and the usefulness of their combination.
We studied an unselected sample of patient materials from which we derived short-term cultures and organotypic spheroids. The presence of CAR, αvβ3/5 integrins and EGFR were assessed by IHC in the tumor and spheroids or by flow cytometry on the isolated cultured cells. Both in primary tumors and in tumor-derived cells, CAR deficiency was frequently found. In addition, CAR staining on tumor tissue was often restricted to the reactive astrocytes. The gene transfer efficiency of unmodified adenoviral vectors correlated with the presence of CAR on these glioma cells in short-term cultures, as has been previously shown on several tumor cell lines (7, 8, 9, 10 ,, 32 , 36 , 37) . Thus, one can assume that unmodified adenoviral vectors injected in the tumor may infect preferentially those CAR-positive astrocytes and infect less the CAR-negative tumor cells. In this situation, the therapeutic ratio may be unfavorable with normal tissue being at higher-risk for the toxic effect of the Ad and its transgene. It is thus understood that any modification of the adenoviral vector to increase gene delivery to the tumor cells, without enhancement of the gene transfer into the normal surrounding tissue, will have important impact on gene therapy of glioblastoma. EGFR, and αv integrins were more widely expressed on tumor cells than CAR. However, the percentage of positive cells varied largely for each of the antigens tested. This intratumoral heterogeneity may predict that a vector directed to only one of these alternate receptors may not be able to infect all of the tumor cells. We thus postulated that adenoviral vectors, with redirected tropism toward both EGFR and αv integrins, could enhance gene transfer into glioma cells more than either targeting approach alone could achieve.
As a first component to derive such a double-targeted vector, we used an adenoviral vector capable of achieving CAR-independent gene transfer via integrins (14) . In glioma cell lines and in cultured primary glioma samples, we observed a markedly enhanced gene delivery compared with unmodified adenoviral vectors. In general, more than one log of enhancement was observed.
As a second component to derive the double-targeted vector, we used a bispecific scFv mini-antibody to bridge the Ad to the EGFR (18) . Using this immunological approach, gene delivery was enhanced in 8 of 13 samples of short-term glioma cell cultures studied, and this improvement could reach one log differential in some cases.
The combined targeting approach was first tested on the human glioma cell line U118 MG. This cell line was used as a surrogate for primary glioma cells because its pattern of expression of the target receptors is similar to that of most primary human glioma cells, i.e., negative for CAR, but heterogeneous for integrins and EGFR. As a whole, gene transfer could be enhanced by almost 100-fold with combined targeting compared with untargeted vectors. This effect led us to evaluate this double-targeting approach on primary human glioma cells.
The combination of immunological targeting of EGFR with genetic targeting of αv integrins improved gene transfer into primary glioma cells in short-term cultures by almost two orders of magnitude. The effect of the combination was usually better than the effect of each of the targeting approaches alone. As expected, only in cases in which EGFR targeting did not enhance gene transfer, its combination with RGD-targeting did not enhance the effect of RGD-targeting alone. In general, the effect of the combination was additive, but in some cases, we observed a supra-additive effect. This suggests that there may exist a cooperation between the two different entry pathways. Clustering of the integrins together with the binding to the internalizing receptor EGFR could thus enhance adenoviral entry into the target cells. Furthermore, the double-targeted vectors also enhanced gene transfer into primary human glioma spheroids. In contrast, gene transfer into the normal brain explants was markedly reduced. The combination, thus, represents an improvement over available targeting approaches.
In conclusion, in this study, analyzing unselected patient material, we confirm the potential impact of targeted adenoviral vectors for gene therapy of glioma. Moreover, we show that the therapeutic index can be improved by providing the vector with multiple targeting ligands. The achieved enhancement of gene transfer into the tumor cells while sparing the normal brain may permit the lowering of the viral dose required to achieve adequate tumor transduction. This may have a major impact on the dose-related in vivo toxicity of these vectors. In the context of brain tumors, in which one of the goals of any therapy is to spare the normal tissue more than in any other organ, targeted vectors such as those described here are likely to have a high level of significance for realizing the full therapeutic potential of gene therapy.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 Supported by a grant from the Association pour la Recherche sur le Cancer (to J. G.), by the Spinoza Award (to H. M. P.), and by National Cancer Institute Grants RO1 HLDK-50277, RO1 CA-74242, and RO1 CA-68245 (to D. T. C.). This work was presented in part at the 91st Annual Meeting of AACR in San Francisco, CA (April 1–5, 2000).
↵2 To whom requests for reprints should be addressed, at Division of Gene Therapy, Department of Medical Oncology, University Hospital-Vrije Universiteit, Postbus 7057, 1007 MB Amsterdam, The Netherlands. Phone: 31-20-444-8423; Fax: 31-20-444-8168; E-mail:
↵3 The abbreviations used are: Ad, adenovirus; FACS, fluorescence-activated cell sorter; CAR, Coxsackie-Ad receptor; EGFR, epidermal growth factor receptor; MoAb, monoclonal antibody; CMV, cytomegalovirus; vp, viral particle(s); scFv, single-chain Fv; IHC, immunohistochemistry.
- Received August 15, 2000.
- Revision received December 4, 2000.
- Accepted December 14, 2000.