The first two authors contributed equally to this paper.
Cancer Cell Biology
Expression of the coxsackie and adenovirus receptor in human astrocytic tumors and xenografts
Article first published online: 6 DEC 2002
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 103, Issue 6, pages 723–729, 1 March 2003
How to Cite
Fuxe, J., Liu, L., Malin, S., Philipson, L., Collins, V. P. and Pettersson, R. F. (2003), Expression of the coxsackie and adenovirus receptor in human astrocytic tumors and xenografts. Int. J. Cancer, 103: 723–729. doi: 10.1002/ijc.10891
- Issue published online: 3 JAN 2003
- Article first published online: 6 DEC 2002
- Manuscript Accepted: 16 OCT 2002
- Manuscript Revised: 15 OCT 2002
- Manuscript Received: 17 JUN 2002
The sensitivity of human tissues and tumors to infection with type C adenoviruses correlates with the expression of the human coxsackie B- and adenovirus receptor, hCAR. HCAR is heterogeneously expressed in various tissues and types of human cancer cells, which has implications for the use of adenoviruses as vectors in cancer gene therapy. Using immunoblotting, real-time PCR, FACS-analysis and sensitivity to infection with adenovirus-lacZ, we analyzed the expression level of hCAR in glioma Grade IV cell lines. With real-time PCR, we also analyzed hCAR expression in primary human astrocytomas of different malignancy grades, as well as in their xenograft derivatives. Analysis of a set of 10 cell lines showed great variation in hCAR expression. Susceptibility to Ad5lacZ correlated well with hCAR expression, whereas no correlation was observed with the expression of αvβ3/αvβ5 integrins, proposed to function as co-receptors for adenoviruses. A great variation of CAR expression was also observed in primary astrocytomas of different malignancy grades. The mean value of CAR expression was significantly lower in 22 Grade IV tumors as compared to the values for 6 Grade II (p = 0.01) and 6 Grade III (p = 0.01) tumors. When the hCAR expression in 11 xenografts derived from Grade IV gliomas were compared to the levels detected in the original parental tumors, a mean 12-fold higher expression was seen in the xenografts (P = 0.01). Two xenografts with low hCAR expression grew considerably faster than the hCAR-expressing cells. Our results have relevance for the use of adenoviruses in gene therapy against astrocytomas. © 2002 Wiley-Liss, Inc.
High-grade malignant astrocytomas (anaplastic astrocytomas and glioblastoma multiforme) belong to the group of human cancers with poor prognosis. Because traditional treatment modalities such as surgery, chemotherapy and radiation therapy only marginally affect survival, there is a great need for new alternative therapeutic strategies.1, 2 Based on the local, non-metastatic growth pattern of gliomas, gene therapy has been considered as a possible new approach for the treatment of these tumors. Vector systems used in gene therapy for gliomas include replication-defective retroviruses, adenoviruses (Ad), herpes simplex viruses (HSV) and hybrid vectors derived from them.3, 4, 5, 6, 7, 8, 9, 10 These vectors have been used to deliver a therapeutic (suicide) gene expressing e.g., the herpes simplex virus thymidine kinase (HSVtk) followed by treatment of the patient with ganciclovir. To be useful as an alternative or a complement to traditional treatments, however, better vector systems are needed to achieve a more efficient delivery of the transgenes into the tumor cells.3, 6, 8, 9, 10 This is also true for suicide gene therapy, where transgenes, such as HSVtk, are used and where a bystander effect is obtained.8, 9, 10, 11, 12
Replication-defective recombinant adenoviruses are the most widely used vectors for experimental in vivo cancer gene therapy applications. This is mainly due to their potential to infect a broad range of dividing and non-dividing cells and because they can be produced in large quantities and to high titers.10, 12, 13 In cancer gene therapy trials, adenovirus vectors have been used to deliver therapeutic genes that can act to directly or indirectly, kill or block the growth of the tumor cells, inhibit angiogenesis, stimulate immune responses to tumor antigens or block tumor cell invasion.12, 13 In addition, replicating, tumor cell-selective, oncolytic adenoviruses have been explored as toxic agents.14 Adenovirus Type 2 and 5, the serotypes so far most commonly used in gene therapy, utilize the knob domain of the fiber to bind to its cellular receptor, CAR.15, 16 Upon virus attachment to CAR, viral internalization by endocytosis appears to be facilitated by αvβ3 and αvβ5 integrins.17 CAR is a 46 kDa glycoprotein composed of an extracellular part containing 2 Ig-like domains, a transmembrane region and a cytoplasmic tail.15, 16 Structural analysis has revealed that the Ad fiber knob interacts with the N-terminal Ig domain of the receptor.18, 19 CAR belongs to the CTX-group within the immunoglobulin superfamily20 and seems to be expressed in many different tissues.16, 21, 22, 23
In human tumors and tumor cell lines, CAR expression at quite variable levels has been detected.24, 25, 26, 27, 28, 29, 30 Tumor cells lacking or expressing low levels of CAR are resistant to adenovirus infection and to efficient oncolysis by replicating Ad.31 This has led to new approaches such as tumor targeting, using modifications of the fiber knob to direct the attachment of Ad to cell surface molecules other than CAR.25, 26, 32, 33, 34, 35, 36, 37
The malignant progression from low-grade astrocytomas to high-grade glioblastomas is accompanied by an increasing number of genetic and epigenetic abnormalities.38 Many of these abnormalities center around genes/proteins playing critical roles in the regulation of the cell cycle. Lowered or lost expression of tumor suppressors or increased activity of oncoproteins, both results in cell cycle disturbances and increased proliferation rates.
We have used a standardized quantitative real-time RT-PCR method complemented with other techniques to analyze CAR expression in glioma cell lines, primary tumors and xenografts grown subcutaneously in nude mice.
MATERIAL AND METHODS
The human glioma cell lines U251MG, U373MG, T98G, A172, U118MG, U87MG, U138MG, TP365MG, TP483MG and H4 were originally established from tumors classified as glioblastoma multiforme (malignancy Grade IV). Their sources have been described previously.39 Normal human foreskin BJ-1 fibroblasts were kindly provided by Dr. Jerry Shay, Department of Cell Biology, University of Texas, Dallas, Texas. HeLa cells and human embryonic kidney (HEK) 293 cells were obtained from American Type Culture Collection (ATCC). All cell lines were grown in DMEM (Life Technologies Inc., Renfrewshire, Scotland) supplemented with 10% FCS (Life Technologies) and 1% each of penicillin, streptomycin and L-glutamine during culturing and experimental procedures.
Tumor material and xenografts
Tumor tissue and non-tumor tissue removed at surgery for brain tumors was collected as described.40 All tumors were classified according to the WHO classification of tumors of the central nervous system.41 The xenografts were generated directly upon surgical removal, by subcutaneous transplantation to the flanks of female nude mice of the Balb/C background (nu/nu, Balb/c, Harlan Sprague-Dawley, Indianapolis, IN).42 The studies were approved by the regional ethical committee for animal experimentation (N229/96) and the human research ethical committees of the Karolinska and Addenbrooke's Hospitals.
Cells to be analyzed by immunofluorescence were plated on coverslips in 6-well plates, grown to 60–70% confluency and fixed for 20 min in 3% paraformaldehyde. After fixation, cells were incubated with the monoclonal antibody, Rmcb 43, washed with PBS and incubated with a secondary TRITC-conjugated anti-mouse IgG (whole molecule (Sigma Immunochemicals, St. Louis, MO). Coverslips were mounted onto slides and photographed with an Axiophot microscope (Zeiss), equipped for immunofluorescence.
Adenovirus infections and counting of β-galactosidase positive cells
The recombinant virus Ad5CMVlacZ was obtained from Quantum Biotechnologies Inc. (USA) and was amplified and purified under standard conditions. The virus titer was determined by plaque assay in 293 cells. To study the susceptibility to adenovirus infection of the different cell lines, 1 × 106 cells of each cell line were seeded onto 6-well plates. The following day, adenovirus was adsorbed onto cells in serum-free medium for 60 min at 37°C, with a multiplicity of infection (MOI) of 10 or 50 pfu/ml. The virus-containing serum-free medium was then replaced by normal growth medium and cells were incubated for another 24 hr at 37°C. Staining with the X-gal substrate was then carried out and positive cells were counted at low magnification in five different optical fields, using a Nikon TMS light microscope.
For immunoblotting analysis, cells were lysed in solubilization buffer (1% Triton Tx-100, 20 mM Tris-HCL pH 8.0, 150 mM NaCl, 5 mM EDTA pH 8.0, 100 kIE/ml aprotinin (Trasylol, Bayer AG, Germany). After the removal of cell nuclei by centrifugation, protein concentrations were measured and equal amounts of protein from each cell lysate were loaded onto a 12.5% SDS-PAGE. Proteins were blotted onto nitrocellulose membranes (Hybond ECL, Amersham Life Science), which were incubated with either the mouse, Rmcb, monoclonal antibody (for detection of CAR in the cell lines, Fig. 3b) or a rabbit antibody to the IG1 domain of the receptor (for detection of CAR in the xenografts, Fig. 6b). A rabbit polyclonal antiserum against the endoplasmic reticulum protein, calnexin,44 was used as an internal standard. After washing steps, the membranes were further incubated with secondary anti-mouse or anti-rabbit IgG conjugated to horseradish peroxidase (Amersham) for detection of CAR and calnexin, respectively. Enhanced chemiluminescence (ECL) was carried out under standard conditions. Chemiluminescent signals were quantified using the LAS1000 system (Fuji Photo Film Co. Ltd., Japan) and the ratio between CAR and calnexin expression levels was calculated for each cell lysate.
The surface expression of CAR and the integrins αvβ3 and αvβ5 on different cell lines was analyzed by FACS. Cells grown as monolayers were detached by treatment with 10 mM EDTA in phosphate buffered saline (PBS), washed and incubated with monoclonal antibodies against CAR (Rmcb), αvβ3 (clone 23C6, Pharmingen, San Diego, CA) or αvβ5 (clone P1F6, Life Technologies) for 60 min at 4°C. After washing and centrifugation steps, cells were incubated for 30 min at 4°C with FITC-conjugated goat anti-mouse immunoglobulins (F(ab′)2 fragment, DAKO A/S, Denmark) and washed before cytofluorometric analysis. Cells were analyzed on a FACScaliber flow cytometer and CELLQUEST software version 3.1f (both from BD Biosciences-Life Science Research, San Jose, CA).
Human glioma samples were obtained directly after surgical removal. Total RNA was isolated as described previously.45 Each tumor piece was histologically examined; the majority of specimens contained around 90% tumor cells. Total RNA was treated with RNase free DNase I (Roche) to remove any contaminating genomic DNA. Three micrograms of genomic DNA-free total RNA from each sample was used for cDNA synthesis (100 μl) using SuperScript II RT (GIBCO ERL).
Real Time quantitative PCR
Real Time quantitative PCR was carried out on the LightCycler (Roche) using DNA master SYBR Green I (Roche Molecular Biochemicals) according to the manufacture's protocol. PCR fragments from each target gene were cloned and used at known concentrations to establish standard curves. In each reaction 1 μl of cDNA was used to determine the copy number of the target gene transcript. Each assay was carried out in duplicate. Additionally, the 18S transcript level was used to normalize the cDNA concentration for each sample. The relative expression level of each gene transcript in the tumor samples was compared to normal brain.
Because of the non-normal distribution of the samples and the small size of some subsets of tumors, the statistical evaluation was carried out using nonparametric tests. Comparison of the levels of hCAR mRNA expression, detected by real-time RT-PCR, in the different groups of tumors and xenografts was carried out using the Mann-Whitney U-test.
Correlation between hCAR expression and susceptibility to adenoviruses in glioma cell lines
To correlate CAR expression with sensitivity to infection with adenovirus Type 5, we determined the level of CAR expression in a panel of 10 glioma cell lines using immunoblotting, immunofluorescence and FACS analysis. The cell lines also allowed us to standardize the real-time RT-PCR method for the subsequent quantification of CAR expression in primary astrocytomas. Because HEK293 cells are known to express high levels of hCAR and are easily infected by adenoviruses, this cell line served as a positive control. Human foreskin BJ-1 fibroblasts, based on PCR analysis, expressed undetectable levels of CAR, were included as a negative control. When stained with the monoclonal antibody Rmcb, non-permeabilized 293 cells were clearly positive, displaying a punctate surface staining (Fig. 1a). In contrast, BJ-1 cells were completely negative (Fig. 1f). Representative CAR-positive and -negative glioma cell lines (see below) are shown in Figure 1b–e. U251MG and U373MG cells clearly stained positive for the receptor (Fig. 1b,c), whereas very low or no staining was observed on T98G and Tp365MG cells (Fig. 1d,e).
To correlate hCAR expression with susceptibility to adenovirus infectability, the cell lines were infected with recombinant adenoviruses infection either β-galactosidase (Ad5lacZ) or green fluorescent protein (Ad5GFP). As shown in Figure 2, the cell lines expressing levels of hCAR detectable by immunofluorescence were easily transduced with Ad5lacZ (Fig. 2a–c), whereas CAR-negative cells could not be infected with the virus (Fig. 2d–f). Similar results were obtained when Ad5GFP were used to infect the cell lines (data not shown).
Because we were not able to reliably quantify hCAR protein expression in primary tumors, we used the glioma cell lines to standardize the quantitative real-time RT-PCR method and to evaluate the quality of the data. CAR expression in the 10 cell lines was first studied by immunoblotting (Fig. 3a,b). Two cell lines (U251MG, U373MG) showed high, 2 (A172, H4) low to intermediate and 6 (TP336MG, TP365MG, T98G, U118MG, U87G, U138MG) no, to barely detectable levels of CAR expression.
The relative protein expression levels were then correlated with those of the mRNA by quantitative real-time RT-PCR. In general, the relative protein and mRNA levels in the different glioma cell lines correlated well with each other (Fig. 3a).
Finally, we analyzed the sensitivity of the 10 cell lines to infection by Ad5lacZ by counting the fraction of β-galactosidase-positive cells, 24 hr after infection. Using a multiplicity of infection (MOI) of 50 pfu/cell, the different glioma cell lines were very heterogeneously infected (Fig. 3a). U251MG and U373MG cells were the most easily infected ones, displaying about 90% positive cells. The six CAR-negative cell lines were resistant to adenovirus infection at this MOI, whereas A172 and H4 were partially sensitive, showing about 30–40% positive cells. Thus, there was a good correlation between CAR expression and sensitivity to adenovirus infection.
Expression of αvβ3/αvβ5 integrins in glioma cell lines does not correlate with susceptibility to adenovirus infection
Because it has been shown that αvβ3/αvβ5 integrins can serve as a co-receptor for adenovirus 5 entry into cells,17, 46 we determined the expression of CAR and the integrins of 5 of the cell lines by FACS analysis using the corresponding antibodies. HeLa and 293 cells served as positive controls. As shown in Figure 3c, the expression of αvβ3 was low and variable in the 5 cell lines, whereas all cell lines were clearly positive for αvβ5. Notably, 3 of the cell lines (TP336MG, TP365MG and T98G) that were resistant to adenovirus infection expressed significant amounts of αvβ5. In regard to CAR expression, the results from the FACS analysis correlated well with those from the immunoblotting and RT-PCR analyses.
Quantitative measurement of hCAR mRNA levels in primary astrocytomas
Having standardized the quantitative real-time RT-PCR method using the cell lines, we next analyzed hCAR expression in primary astrocytic tumors. Six low-grade astrocytomas (AII), 6 anaplastic astrocytomas (AAIII) and 22 glioblastomas (GB) were included in the study (Fig. 4). CAR expression varied substantially within each group and between groups. Among the Grade II and III astrocytomas, expression ranged from almost undetectable (A27, AA51) to high (AA93) or very high (A9). With few exceptions, the Grade IV tumors expressed undetectable to low levels of CAR. Statistical analyses indicated that the relative mean expression level of hCAR was significantly lower in the Grade IV tumors (GB) as compared to Grade II (A; p = 0.01) and to Grade III (AA; p = 0.01) tumors (Fig. 5). For comparison, 7 normal human tissues were also included in the study (Fig. 4). High expression of hCAR was detected in kidney and adrenal gland, intermediate levels in small intestine and low, but detectable levels in testis. No or very low expression was found in cerebellum, liver and pancreas. The relative expression level of each tumor and normal sample was compared to the mean value of 10 normal brain samples.
HCAR mRNA expression in tumor xenografts
Eleven matched pairs of xenografts/glioblastomas were analyzed for hCAR expression by real-time PCR. The results showed a strongly elevated level of hCAR expression in 2 of 11 xenografts as compared to the matching primary tumor tissues (Fig. 6). In these 2 xenografts, A925 and A907, the levels of hCAR were 170 and 230 times higher than in the primary tumors, respectively. In 4 of the 11 xenografts; A894, A1024, A923 and A1174, hCAR expression was 12–42 times higher and in 3 cases, A940, A1222 and A1043, the levels were elevated 2–4 times. Two of the 11 xenografts (A830 and A948) displayed no difference in CAR expression. In total, the calculated mean expression levels were 12-fold higher in the xenografts compared to the tumors (p = 0.01, Fig. 7). The mean induction level for all 11 matched pairs was 46-fold. Immunoblotting analysis, using a rabbit polyclonal antibody against the IG1 domain of CAR, on 5 of the xenografts showed an excellent correlation between the expression of hCAR at the mRNA and protein levels in the different samples (Fig. 6b). Four normal brain samples were also analyzed and were, in agreement with the results obtained from the real-time PCR analysis, found to not express hCAR (Fig. 6b).
When comparing glioblastomas that grew as subcutaneous xenografts in nude mice with those that did not grow, no statistically significant difference in hCAR expression levels were seen (Fig. 4).
We finally analyzed whether upregulation of CAR expression correlated with the rate of growth of the xenografts in nude mice. As shown in Table I, the 2 xenografts (A830 and A948) expressing very low levels of CAR grew much faster than the ones with moderate to high CAR expression. These 2 tumors had to be passaged once a month as compared to once every 2–4 months for the others.
|Xenograft||hCAR expression||Xenograft primary tumor||Passaging time (months)|
Replication-deficient recombinant adenoviruses carrying a therapeutic gene is currently the most frequently employed vector system used in experimental and clinical cancer studies. It is therefore of importance to determine the susceptibility of tumor tissues to adenovirus infection. It has been well documented that the critical determinant for susceptibility to adenovirus infection is the expression of CAR, the primary receptor for adenovirus Type 2 and 5. We have a long standing interest in the characterization of the genetic aberrations and biology of human astrocytic tumors with the aim of improving diagnosis and therapy. In our present study we have analyzed the expression of CAR in a panel of 10 glioblastoma cell lines, a set of 33 primary Grade II–IV astrocytomas and 10 glioblastomas grown as xenografts in nude mice. We found that the level of CAR expression varied markedly between different cell lines as determined by immunoblotting, quantitative PCR and FACS analysis. Sensitivity to adenovirus infection correlated well with the level of CAR expression, whereas no correlation was observed with the expression of the αvβ3/αvβ5 integrins.
Our results are in good agreement with those now obtained for a range of tumor cell lines. The extreme variability of CAR expression levels has been found for e.g., bladder,24 prostate28 and melanoma27 cell lines. In some tumor cell lines, such as rhabdomyosarcoma30 and gastrointestinal cancers,29 CAR expression has been found to be low or undetectable. In regard to glioma cell lines, several groups have documented highly variable levels of CAR expression.25, 47, 48 In our study, we found that 2 of 10 cell lines expressed a high level of CAR, 2 a low to moderate level, whereas 6 were essentially CAR-negative. Some of the cell lines included in our study have also been analyzed previously by other groups. In most cases, the results are in excellent agreement. For the cell line U87MG, however, Mori et al.47 observed a moderate susceptibility to adenovirus infection despite a low level of CAR expression as determined by quantitative PCR. Likewise, Miller et al.,25 observed moderate binding of radiolabeled Ad5 to U87MG and a low level of surface CAR as determined by FACS analysis. In contrast, in the present and a previous49 study, we were unable to detect CAR expression or infect U87MG with a moderate MOI of Ad5. This discrepancy may relate to the passage history of this cell line. Most glioma cell lines have been kept in culture for years or even decades. During such a long history, expression levels of a number of genes may change substantially. In agreement with previous studies25, 29, 47, 48 we found that the susceptibility to adenovirus infection did not correlate with the expression of β3/αvβ5 integrins.
Similarly to the situation in glioma cell lines we found that primary astrocytic tumors also expressed very variable levels of CAR. Glioma Grade II and III showed a statistically significantly higher expression level than glioblastomas (Grade IV). To date there is almost no information in the literature on the expression of CAR protein or mRNA, in primary astrocytic tumors. This is partly due to the inherent problems of reliably quantifying CAR protein expression from tumor tissue. Miller et al.25 demonstrated highly variable susceptibility to adenovirus infection of primary glioma cultures after short-term in vitro culturing. Interestingly, they found that the susceptibility did not correlate with the original histological type of the tumor. Even a few in vitro passages of tumor material may select subsets of cells and result in altered gene expression, however, such as that of CAR. Combined with our results this information should be of importance in regard to clinical trials in which adenovirus vectors are being used for the treatment of glioblastomas.3, 10 From the point of view of practical gene therapy, the expression levels of CAR in the primary tumors is more relevant than that in established cell lines. Future work should thus focus on determining CAR expression in primary tumors and analyzing their susceptibility to adenovirus infection.
The observation that Grade IV tumors expressed lower levels of CAR than the Grade II and III tumors may point to a possible growth advantage for tumors displaying low CAR expression. Alternatively, CAR may be secondarily down-regulated in fast growing cells. The normal cellular function of CAR is not known. Some recent results, however, suggest that CAR may serve as a cell–cell adhesion molecule.50 Other results suggest that CAR expression may inversely correlate with cell growth. Bladder carcinoma cell lines expressing elevated levels of CAR showed increased cell-cell aggregation, displayed a slower growth rate and antibodies to the CAR ectodomain prevented the inhibitory effect of CAR on growth.51 Furthermore, ectopic expression of hCAR in a prostate cancer cell line resulted in a lowered proliferative potential of these cells, both in vitro and in vivo, indicating a tumor suppressive role for CAR.28 Our results also show that absence of CAR expression may contribute to an increased growth rate of glioma cells, because the 2 xenografts displaying no CAR expression grew considerably faster in nude mice than the 9 CAR-positive tumors.
We made the interesting observation that CAR expression in 9 of 11 cases was upregulated in glioblastoma xenografts as compared to their parental, primary tumors. The reason for this is unclear, but may suggest that the gene expression of CAR is induced by factor(s) produced in the mice. So far little is known regarding signaling that leads to regulation of the CAR promoter. Analysis of CAR regulation will have to await a detailed definition of the CAR promoter region and the development of a reporter system for analyzing factors regulating CAR expression.
In conclusion, we have found great variations of CAR expression levels in glioma cell lines and primary astrocytomas of different malignancy grades, as well as xenografts from Grade IV tumors. The implication of these results for adenovirus-mediated gene transfer to patients remains to be investigated. A thorough study correlating CAR levels with susceptibility to adenovirus seems warranted.
We wish to thank A. Bergström for excellent technical assistance.
- 1Cancer of the nervous system. Oxford, UK: Blackwell, 1997. 11., .
- 8A phase III clinical evaluation of herpes simplex virus type 1 thymidine kinase and ganciclovir gene therapy as an adjuvant to surgical resection and radiation in adults with previously untreated glioblastoma multiforme. Hum Gene Ther 2000; 11: 2389–401..
- 41Tumours of the nervous system. Lyon: IARC, 1997. 66., .