Introduction

With the increased understanding of virology and tumor-cell biology, new targeted and armed adenoviruses have been generated with improved safety and tumor-cell specificity profiles. Recent preclinical and clinical studies with replicating viruses have underscored the safety and the potential efficacy of such new viral oncolytic agents, but also revealed several factors that reduce the efficiency of oncolytic viral therapy. These include downregulation of viral receptors at the tumor-cell surface,1 the limited distribution of the virus after intratumoral administration, the physical barriers formed by the tumor stroma2, 3, 4 as well as humoral immunity against viral therapeutics.5 These factors hamper efficient tumor cell infection and frustrate effective therapy.

Tumor-cell selectivity can be achieved by cancer-cell specific entry of the virus particle and by restricting replication or transcription of viral genes to tumor cells with tumor-specific promoters. In addition, several strategies have been proposed to enhance the spread of the virus within tumors, such as expression of transgenes involved in the degradation of the extracellular matrix6, 7 or mutations of viral genes leading to enhanced release of the virus.3, 8 Whereas some information on the performance of such modified viruses can be obtained from in vitro experiments in cultured tumor cells grown in monolayers, the true anticancer efficiency can only be evaluated in genuine tumor models. The standard assays for evaluating the performance of oncolytic viruses involve human tumor xenografts in immunodeficient mice. Preclinical assessment of viral vectors requires a substantial number of tumor-bearing animals, as there are still no other ways of identifying the best candidates and determination of their optimal intra-tumoral spread. However, mouse xenograft models are expensive and time-consuming assays to test anti-tumor potency of introduced modifications.

As a complementary approach, we have explored the possibility of using tumors grown on the chorioallantoic membrane (CAM) of embryonated chicken eggs to test the performance of modified oncolytic viruses in solid tumors. This system would allow a screening of the vectors and reduce substantially the number of animals used for the mandatory in vivo experiments.

The CAM is a highly vascularized membrane located at the periphery of the chicken embryo, easily accessible by opening a hole in the egg shell. The CAM has been intensely used to study angiogenesis and is currently the most widely used model for testing pro- or anti-angiogenic drugs.9 Modified oncolytic viral vectors have also been used to study the transduction of chicken endothelial cells in a CAM model.10, 11 Owing to the immature immune system of the chicken embryo, the CAM has also been successfully used to graft tumor explants or to grow suspensions of cancer cells of different origins. Especially, the model is very attractive to study the steps involved in metastasis formation.12, 13, 14, 15 In the field of cancer therapy, the CAM tumor model has been used to evaluate the efficiency of anticancer drugs,16, 17, 18 but has never been used so far for testing oncolytic vectors.

One of the important factors that hamper effective oncolytic virus therapy is inefficient spread of the viral vectors within tumors. The CAM tumor assay could be of interest as a tool to study the early spread of modified viral vectors in a solid tumor mass.

To improve intra-tumoral spread of adenoviral vectors, we exploited the property of fusogenic membrane glycoproteins encoded by enveloped viruses to induce cell–cell fusion, leading to the formation of syncytia (giant multinucleated cells).19, 20 Ability to express such fusogenic protein could facilitate the propagation of human adenovirus type 5 (HAdV-5) vectors into the solid tumor tissue, as a transduced cell can fuse with non-transduced cells, and further release the viral progeny from the whole syncytium. We have engineered hyperfusogenic proteins derived from the fusion protein of parainfluenza virus type 5 (F-PIV5)21 and inserted the fusogenic cassettes into replication-competent adenoviral vectors. In this study, a replication-competent HAdV-5 fusogenic vector was tested in melanoma tumors grown on CAM for its ability to replicate and to form syncytia in solid human-like tumors.

Materials and methods

Cell lines

Human melanoma cell lines Mel2A, MZ2-MEL43, MelJuSo, TW12 (a kind gift from Dr J Portoukalian, Department of Transplantation and Clinical Immunology, Claude Bernard University, Lyon, France), colorectal adenocarcinoma cell line LoVo and human glioblastoma U87-MG and U118 cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen, Breda, The Netherlands) supplemented with 10% heat-inactivated fetal bovine serum (Invitrogen), 2 mM L-glutamine, 100 IU ml−1 penicillin and 100 μg ml−1 streptomycin.

Chicken chorioallantoic membrane assay and grafting procedures

Fertilized white Leghorn chicken eggs (Gallus domesticus) from a local hatchery were placed in a humidified 37 °C incubator without CO2 to induce embryogenesis (embryo development day; EDD 0). On EDD 4, eggs were transilluminated to check the vitality and the localization of the embryo. After cleaning the egg shell with chlorhexidine gluconate, a small hole was made with a 19 G needle in the air sack and a window was cut using forceps under sterile conditions. The shell membrane was humidified with sterile phosphate-buffered saline (PBS) solution and carefully cut out. This window was then sealed with a sterile 3 cm Petri dish and eggs were placed back to the incubator and inspected daily until the day of experiment. At day 7 (EDD 7), viability of the embryos and the vasculature of the CAM were visually inspected and CAM was gently lacerated with a sterile cotton swab to create a blood spot. Tumor cells were collected by trypsinization, washed with culture medium and pelleted by gentle centrifugation. After removing the medium, 5 × 106 cells were resuspended in 30 μl ice-cold Matrigel (Growth-Factor Reduced Matrigel; Becton-Dickinson, Breda, The Netherlands) and inoculated on the CAM at the site of the blood spot. Eggs were then sealed and placed back into the incubator. Tumor growth was inspected on a daily basis. On day 5 post-grafting (EDD 12), their size ranged from 3 to 5 mm in diameter with visible neoangiogenesis. At EDD 20, eggs were placed on ice for at least 2 h to euthanize the chick embryos by hypothermia and tumors were cut out and stored in 4% paraformaldehyde for further processing.

Adenoviral vectors

Wild-type HAdV-5 was amplified on HeLa cells, purified by double cesium chloride gradient and titrated by plaque assays on A549 cells.22 In a previous study, we engineered hyperfusogenic fusion glycoproteins of the paramyxovirus type 5.21 A fusogenic adenoviral vector was constructed by inserting the gene encoding the hyperfusogenic glycoprotein Fus8, driven by the cytomegalovirus promoter into the E3-deleted region of the HAdV-5. This vector contained the wild-type E1 region and was replication-competent. A control non-fusogenic, replication-competent vector (Ad.ΔE3) was constructed in parallel. Cloning strategies are available on request. Replication-competent fusogenic vectors were amplified on PER.C6 cell line,23 purified by double cesium chloride gradient and titrated by plaque assays on A549 cells.22 The virus stocks obtained had equal virus titer of 1010 PFU ml−1. The fusogenic phenotype of rAd.Fus8 in cells cultured in vitro and the morphology of the syncytia was identical to that observed in transient expression experiments.21

Injection of adenoviral vectors

A measure of 2 μl of purified virus (corresponding to 8 × 108 PFU for wild-type (wt) Ad5 and 2 × 107 PFU for rAd.Fus8 and Ad.ΔE3) were injected at tumor day 7 (EDD 14) in the center of each tumor mass using a 30 G insulin syringe. A set of 10 eggs (the CAM-tumors) at EDD 14 was used for the experiment with wt Ad5. The set was divided into three groups: group 1 (two tumors without treatment), group 2 (2 tumors mock injected with 2 μl of sucrose solution) and group 3 (6 tumors injected with the virus). For the experiment with the recombinant viruses, a set of 20 eggs (the CAM-tumors) at EDD 14 was divided into three groups: group 1 (8 tumors; 4 tumors without treatment and 4 tumors mock injected with 2 μl of sucrose solution), group 2 (6 tumors injected with Ad.ΔE3 virus) and group 3 (6 tumors injected with rAd.Fus8 virus). For every such experiment, tumors were harvested 6 days later (EDD20).

Histology and immunohistochemistry

Samples were dehydrated and embedded in paraffin. Sections of 3 μm were performed for hematoxylin and eosin staining. Serial sections were made from tumors injected with oncolytic vectors, to find the most representative areas of the infected tumors. For immunohistochemistry, paraffin sections (4 μm) were mounted on poly-L-lysine-coated slides. After deparaffinizing and rehydrating, antigen retrieval was carried out by heating sections at 98 °C for 45 min in 10 mM citrate buffer, pH 6.0, for tyrosinase and adenovirus immunostainings, or in 1 mM Tris-EDTA buffer, pH 9.0, for MelanA, matrix-metalloprotease 9 (MMP-9) and F-PIV5 immunostainings. The following antibodies and dilutions were used for immunohistochemistry: anti-PS100 (1/200; Dako, Trappes, France), anti-MelanA (1/10; Dako), anti-tyrosinase (1/50; Neomarkers, Francheville, France), anti-MMP-9 (1/100; Chemicon, Molsheim, France), anti-adenovirus serotype 5 (polyclonal rabbit serum obtained by subcutaneous injection of whole virus particles to the rabbit, 1/100) and anti-F-PIV5 (F1a, 1/40; a kind gift from R Randall, University of St Andrews, Scotland). The iVIEW DAB Detection Kit (Ventana, Illkirch, France) and the NEXES apparatus (Ventana) were used to stain the sections.

Immunofluorescence

MelJuSo cells were cultured in 24-well culture plates and infected by rAd.Fus8 at an MOI of 1, incubated at 37 °C for 24 h. Infected cells were fixed with 1% (v v−1) paraformaldehyde in phosphate-buffered saline (PBS), washed twice and allowed to immobilize on coverslips overnight. They were incubated with monoclonal F1a antibody against F-PIV5 at 1/10 in PBS for 3 h, washed and labeled with anti-mouse immunoglobulin G-Alexa 633 secondary antibody (Invitrogen, Cergy-Pontoise, France) at 1/200 in PBS for 30 min. After washes, cells were incubated for 10 min with 4′,6-diamidino-2-phenylindole (1/1000) and then mounted on microscopic slides with Fluoromount G (Clinisciences, Nanterre, France). Fluorescent optical section images were acquired with a TCS SP2 confocal microscope (Leica, Rijswijk, The Netherlands).

Results

Tumor growth on the CAM

The procedure of the CAM tumor assay is summarized in Figure 1. Briefly, trypsinized tumor cells from melanoma, colorectal carcinoma and glioblastoma cell lines were seeded on the CAM of fertilized chicken eggs (Figures 1a–d). All cancer cell lines evaluated in this study grew successfully on the CAM, and were able to induce rapidly the formation of solid tumors from 5 to 10 mm diameter, depending on the cell line (Figure 1e). The size and morphology of the tumors were reproducible and allowed to obtain a batch of 10–20 similar tumors in a few days for a single experiment (Figure 1f). The reproducibility in growth rate shows that the CAM tumor model is suitable for evaluating anticancer agents.

Figure 1
figure 1

The chorioallantoic membrane (CAM) tumor assay. (a) Overview of the procedure. The development of the chick embryo lasts 21 days after incubation at 37 °C. A hole is made in the shell after 4 days and tumor cells are seeded on the CAM after 7 days. Oncolytic viruses are injected 7 days later, id est at the 14th day of embryo development day (EDD). Embryos are euthanized at EDD 20. (b) At EDD 4, a window is sterilely opened in the shell, revealing the inner shell membrane, which is further carefully removed. (c) At EDD 7, the highly vascularized CAM is visible through the window. (d) Trypsinized tumor cells are seeded onto the CAM after gentle laceration. (e) The development of tumors is daily monitored through the window. (f) Typical batch of melanoma tumors obtained in a single experiment. (g) Low magnification ( × 50) of a tumor section stained by hematoxylin–eosin. The tumor (T) is connected to the CAM (arrow).

Full size image

Histological structure of CAM tumors and comparison with patient samples

All tumors samples cut out from the CAM were used to perform histology. Observation of histological sections showed dense tumors surrounded by the CAM (Figure 1g). For a given cell line, the histological features were similar for all samples (data not shown). To assess the relevance of the model, tumors obtained from CAM assays were compared with human tumor samples from patients using histology and immunohistochemistry. Sections of human metastases of melanoma, colorectal adenocarcinoma and glioblastoma multiforme were compared with sections of tumors grown on CAM using melanoma (Tw12, Mel2A, MZ2-MEL43 and MelJuSo), colorectal cancer (LoVo) and glioblastoma multiforme (U87-MG and U118) cell lines, respectively.

Histological comparison showed a strong similarity between CAM-grown and patient tumors (Figure 2a). Sections of CAM-grown melanomas showed a dense tumor tissue organized in tumor nests. Blood vessels were clearly visible within the tumor tissue, showing that the human tumor cells induced a neovascularization from the chick vasculature, similar to what is observed in patient samples. Of note is the presence of nucleated avian erythrocytes. Colorectal adenocarcinoma tumors grown on CAM from LoVo cells showed a glandular structure, with clearly visible tubular crypts, comparable to those observed in metastases from patients. Between the crypts, stromal cells were visible, demonstrating that the human cancer cells recruited avian stromal cells to reproduce the architecture of the tumor. The connective tissue between the crypts was less dense in the CAM tumors. This may be correlated to the rapid formation of the tumors. Experimental glioblastoma multiforme obtained from U87-MG and U118 cells showed a diffuse pleiomorphic infiltrate of fibrillar and stellate cells, with neoangiogenesis, edema and areas of necrosis, which are the hallmarks of such tumors. However, pseudopalisading necrosis and endothelial cell hyperplasia, which are very suggestive of glioblastoma multiforme, were not observed in CAM-grown tumors.

Figure 2
figure 2

Histological study of tumors obtained from the chorioallantoic membrane (CAM) assay. (a) Comparison of CAM-grown melanoma (MelJuSo), colorectal adenocarcinoma (LoVo) and glioblastoma multiforme (U118) with patient samples from tumors of the same type. Arrows indicate blood vessels and asterisks indicate tumor stroma (hematoxylin–eosin staining, × 100 magnification for melanoma and colorectal adenocarcinoma, × 200 magnification for glioblastoma multiforme). (b) Histological study of the tumor stroma from melanoma (MelJuSo), colorectal carcinoma (LoVo) and glioblastoma multiforme (U118) grown on CAM (hematoxylin–eosin staining, × 400 magnification).

Full size image

Interestingly, the histological characteristics of the tumor stroma were very different according to the tumor type (Figure 2b). In melanoma tumors, stromal cells were rare compared with malignant cells and induced the formation of extracellular matrix between the tumor nests. In colorectal cancer, there were more stromal cells, which were dispersed between the crypts formed by LoVo cells. In the glioblastoma tumors, the tissue harbored many stromal cells, which were organized in a fibrillar pattern, very similar to what is observed in the glial tissue. Altogether, these results show that tumor cells grown on CAM do not multiply anarchically, but behave as in real tumors and are able to induce tissue organization and recruitment of avian endothelial and stromal cells.

Immunohistochemistry of CAM tumors

For further characterization of the tumors, immunohistochemistry was performed on melanoma tumors using antibodies recognizing the differentiation (anti-MelanA, anti-PS100 and anti-tyrosinase) and the invasiveness of the cells (anti-MMP-9). Sections of the tumors obtained from four melanoma cell lines (Tw12, Mel2A, MZ2-MEL43 and MelJuSo) were compared with sections from 10 melanoma metastases randomly issued from the collection of the Department of Pathology of the Centre Hospitalier Lyon-Sud (Pierre-Bénite, France). Results are shown in Table 1. Immunohistochemistry showed that the expression of the melanoma antigens were expressed within the CAM tumor, in a similar manner as observed in metastases from patients (Figure 3). Depending on the differentiation of the cell line, some experimental tumors were not stained by all antibodies, as some patient samples, reflecting the biological heterogeneity of melanoma. For instance, amelanotic melanomas were weakly stained by tyrosinase antibody, either in CAM tumors or in patient samples.

Table 1 Histological and immunohistochemical characteristics of melanoma tumors grown on the CAM from four cell lines, compared with melanoma metastases from 10 patients
Full size table
Figure 3
figure 3

Immunohistochemistry performed on melanoma metastases from patients and melanoma tumors grown on chorioallantoic membrane (CAM). Tumor samples were stained using anti-PS100, MelanA, tyrosinase and matrix-metalloprotease 9 (MMP-9) antibodies, respectively ( × 200 magnification for PS100 and MelanA stainings, × 100 magnification for tyrosinase and MMP-9 stainings).

Full size image

Intra-tumoral injection of wild-type adenovirus 5

To test the feasibility of injection of replication-competent adenoviruses into CAM-established tumors, we first injected intra-tumorally wt HAdV-5 into melanoma tumors obtained from Mel2A cells. At 6 days post-injection, no significant difference in the size of the tumors was observed between treated and control tumors. However, histology revealed a large necrotic cavity in the center of the tumors treated by wt HAdV-5 (Figure 4a). At a higher magnification, necrotic cells were observed all around this cavity (Figure 4b). More distant from that area, most of the tumor cells were alive and showed no cytological sign of infection. A clear demarcation could be observed between necrotic and living cells (Figure 4b). To detect virus-infected cells, immunohistochemistry was performed with a polyclonal antibody targeted against the capsid proteins of wt HAdV-5. Cells in which the viral replication was efficient are stained, as biosynthesis of the capsid proteins occurs at the late stage of the adenoviral cycle. Infected cells also exhibited degenerative changes such as swelling and rounding of the nucleus, typical of the adenovirus-induced cytopathic effect, confirming the specificity of the staining. Only few cells were visibly stained near the site of injection, showing that the efficiency of virus spread to the neighboring cells was low (Figures 4c and d).

Figure 4
figure 4

Injection of HAdV-5 in a melanoma tumor (Mel2A) grown on CAM. A unique injection of 2 μl purified HAdV-5 (8 × 108 plaque forming units) was performed in the center of tumor, which was fixed 6 days later. (a) In the center of the tumor, a necrotic cavity (*) is visible at the extremity of the needle tract (arrows) ( × 20 magnification). (b) Around this cavity, tumor cells are necrotic (nec). The demarcation between dead and living cells (liv) is clearly visible ( × 200 magnification). (c, d) Immunohistochemistry with anti-adenovirus 5 antibody shows that replication of the virus occurs only in few cells at the center of the tumor ( × 40 magnification in e and × 400 in f). Inflammatory cells from the chick embryo are visible among infected cells.

Full size image

Induction of intra-tumoral cell fusion by fusogenic adenoviral vectors

To study whether it was possible to modify the tumor architecture with an armed adenoviral vector, we constructed a replication-competent vector expressing an engineered fusogenic glycoprotein.21 The armed oncolytic vector or the control non-fusogenic replication-competent vector were further injected intra-tumorally in the melanoma MelJuSo tumors grown on the CAM. At tumor day 7, tumor mass reached the size of 4 mm in diameter suitable for manual injection. Two microliters of sucrose/PBS (mock treatment), purified fusogenic adenovirus or purified control virus were injected in the center of the tumor. After 6 days, no macroscopic changes were observed in any of the tumors. However, histological studies revealed large areas of necrosis in tumors treated by the fusogenic vector, compared with tumors treated by the control non-fusogenic virus. No significant necrosis was observed in mock-treated tumors. Moreover, immunohistochemical studies revealed that the number of cells in which viral replication occurred was much higher in tumors treated with the fusogenic virus (Figures 5a and b). Numerous syncytia were observed in many parts of the tumors treated with the fusogenic vector, predominantly along the needle tract (Figures 5c and d) and their morphology was similar to that observed in vitro (Figure 5e). Immunohistochemistry performed with the anti-F-PIV5 antibody showed that the fusogenic protein was strongly expressed within the syncytia (Figure 5f).

Figure 5
figure 5

Transduction of chicken chorioallantoic membrane (CAM)-grown melanoma tumors with a replication-competent fusogenic adenoviral vector. Melanoma MelJuSo tumors were obtained on CAM and treated with a single injection of either the replication-competent fusogenic vector rAd.Fus8 or the control vector rAd.ΔE3. (a, b) Representative anti-adenovirus immunohistochemistry of MelJuSo tumors transduced by non-fusogenic (Ad.ΔE3, a) or fusogenic (rAd.Fus8, b) adenoviral vectors are shown ( × 16 magnification). (c) Syncytia (arrows) visible along the necrotic cavity after treatment with rAd.Fus8 ( × 100 magnification). (d) High magnification of an intra-tumoral syncytium ( × 1000). (e) Immunohistochemistry of the syncytia using anti-F1a antibody targeted against the fusogenic protein of rAd.Fus8 ( × 100). (f) Contrast-phase microscopy (left) and immunofluorescence with F1a antibody (right) of a syncytium induced by rAd.Fus8 in cultured MelJuSo cells.

Full size image

Discussion

The development of new anticancer therapies requires the use of pre-clinical models to assess their effect on tumor cells. Among these, the most simple is the in vitro testing on cells in monolayer cultures. However, this model does not reflect the complexity of the real tumor environment, and probably overestimates the anti-tumor efficiency of oncolytic viruses, whose spread is facilitated by a two-dimensional environment devoid of physical barriers, blood flow and immune cells. Three-dimensional models have been developed, such as monotypic and heterotypic spheroids,24 and mammalian models, which include xenografts in immunocompromised animals or more sophisticated transgenic animal models.25 However, these models have their limitations and are expensive, and there is an active push to reduce animal experimentation (Replacement, Refinement and Reduction of animal experiments, also described as the 3Rs principle). The CAM tumor model could allow a pre-screening of oncolytic vectors and subsequently reduce the number of animals used for in vivo experiments.

We have grown human cancer cell lines from different origins on the CAM. All cell lines tested so far grew successfully and induced the formation of solid tumors within days. This makes this assay suitable to test a batch of similar tumors for comparison of anticancer agents. Such results had already been obtained by other groups for prostate cancer,18 glioblastoma multiforme,16 melanoma13 and leukemia17 cell lines. Compared with the rodent model, the CAM assay is much faster, as tumors grow on the CAM in several days, whereas it may take weeks to observe subcutaneous xenografts on immunodeficient mice.26 Overall, the CAM tumor assay is fast, easy to handle and less expensive than the rodent model, as the only materials needed are a humidified incubator and a cell culture lab. However, the CAM tumor model can only reduce the use of animal experiments, but will not replace it in the full pre-clinical assessment of anticancer agents.

Histological analyses of the CAM tumors revealed a well-organized tumor tissue, which was reproducible for a given cell line. The histology of the CAM tumors strongly resembled clinical specimens of human tumors. These data indicate that tumor cells do not grow anarchically in the CAM, but induce formation of an organotypic structure. Thus, the human cells were able to interact with avian cells to reproduce the architecture of the tumor tissue. For instance, neovessels from the chicken were observed within the tumor tissue in the melanoma and glioblastoma multiforme sections, showing that the malignant cells induced angiogenesis from the chicken vasculature. Histology of CAM tumors revealed as well that the organization of the stromal cells was modulated differently, depending on the tumor type. However, in glioblastoma multiforme tumors grown on CAM, pseudopalisading necrosis and endothelial hyperplasia were not observed. These highly specific histological structures of glioblastoma, which might be induced by hypoxia, are absent in most animal models involving human cell lines.27 However, the expression of genes involved in gliomagenesis has been confirmed in the CAM glioblastoma multiforme model.16 In our study, immunostainings of the melanoma tumors showed that the expression of the tumor antigens was very similar to what is observed in patient samples, reflecting the differentiation and the aggressive phenotype of the tumor cells in the CAM tumor model.

Altogether, these data indicate that the CAM model allows the formation of tumors comparable to patient samples, with a degree of fidelity to human disease, which is impossible to reach with other non-animal models. The spheroid model is a common surrogate to in vivo experiments.24 Spheroids are widely used to test the sensitivity of cancer cell lines to chemotherapy, immunotherapy or radiotherapy.24 They have also been used to test oncolysis and infectivity of adenoviral vectors.28 However, their structure is much simpler than a real tumor environment, as only one type of cell is grown in spheroids. Recently, advanced models of mixed spheroids have been developed, using co-culture of tumor cells with HUVEC and/or fibroblasts.24 However, the relevance of these ex-vivo tumors, whose size barely reaches 0.5 mm, is probably less striking than what is observed in CAM-grown tumors.

Studies concerning the propagation of adenoviral vectors in the tumor tissue showed that viral spread is often limited, due to physical barriers such as blood vessels and extracellular matrix.3, 4 The role of stromal cells appears to be of critical importance in the limited spread of adenoviral vectors.29 Given that all these structures are present in CAM tumors, we hypothesized that this model would be useful to test the propagation of oncolytic adenoviral vectors. A single injection of wt HAdV-5 was able to induce necrosis at the center of the tumors 6 days post-infection. At the site of injection, most of the tumor cells were lysed or showed histological signs of necrosis. However, the number of cells in which the virus was replicating 6 days post-injection was very low, as assessed by immunohistochemistry. These data suggest that most of the viral load was lost despite the ability of the virus to replicate in the tumor, and illustrate one of the main hurdles of cancer virotherapy.3 The reason for the inefficient transduction remains to be elucidated. Heterogeneity of the expression of viral receptors, inefficient replication or spread of the virus in the tumor environment, as well as clearance of the vector by the chicken leukocytes are possible explanations for that phenomenon.

Fusogenic membrane glycoproteins are among the most promising transgenes to arm oncolytic vectors, as they harbor a potent bystander effect and improve the dispersion of viral proteins from cell to cell.20, 30, 31, 32 In a previous study, we engineered hyperfusogenic genes derived from the fusion protein of the parainfluenza virus 5.21 One of these genes was then inserted into a replication-competent HAdV-5 backbone and the purified vector was injected into melanoma tumors grown on CAM. Transduction of tumor cells seemed to be much higher for the fusogenic vector than for the control non-fusogenic vector, although the capsid composition and the input dose were the same. Syncytia and expression of the fusogenic protein were observed for the fusogenic vector, but not for the control vector. These data show proof of principle that the CAM model is suitable to study the expression of transgenes delivered by adenoviral vectors in a three-dimensional tumor model, and to assess the histological changes induced by the expression of such transgenes. Many transgenes used to arm adenoviral vectors are designed to modify the tissue architecture. For instance, vectors that degrade the extracellular matrix,6, 7, 33 inhibit the angiogenesis34, 35, 36, 37 or target the stromal cells29 have been constructed. The CAM model is probably relevant to study such vectors.

Although the CAM tumor model is very attractive, the duration of the follow-up period is limited to 7–10 days after the injection of the vector, due to the hatching of the chick 21 days after incubation. Thus, the CAM model is the best suited for studies addressing the transduction efficiency of the tumor cells and the early replication and spread of the vectors. The use of fertilized eggs with longer embryonic development, such as turkey eggs (28 days), could allow observing a regression of the tumors after treatment. Other applications of the CAM tumor model in the field of oncolytic vectors could be considered, such as the study of the transduction of tumor cells by targeted vectors harboring modified capsids, or the study of conditionally replicative vectors in a three-dimensional environment.

In conclusion, the CAM tumor model is a promising tool in the field of anticancer virotherapy, which combines the advantages of an in vivo environment with the simplicity of an in vitro experiment.