Introduction

The liver is a common site for developing metastatic disease. Although any malignancy can spread to the liver, the direct passage of blood from the gastrointestinal tract to the liver via the portal circulation results in a high rate of liver metastasis from gastrointestinal tract tumors. For example, liver metastases occur in 60% of patients with colorectal cancer. Surgical resection is feasible in less than 20% of patients and results in a 5-year survival rate for only about 30% since the majority of patients relapse after hepatic resection. Systemic chemotherapy produces response rates of 15–30% with a median survival of 10–12 months. The inefficiency of classical methods for treatment of liver metastases has promoted the development of gene therapy approaches using modified viruses for oncolysis or stimulation of anti-tumor immune responses (for a review see¹). Phase I and II clinical trials with adenoviruses (Ads) have been conducted in patients with liver metastases². The results have shown that, although adenovirus-mediated gene therapy has been well tolerated and toxicity has been low, the clinical benefit has so far been disappointing³. Potential reasons for this include inaccessibility of tumor nests due to anatomical barriers^4,5 and inefficient infection of tumor cells^6,7,8. To address these problems, improved oncolytic vectors and adequate animal tumor models for testing these vectors must be developed.

Two main types of animal tumor models are currently used: immunocompetent mice carrying endogenous, induced, or transplanted syngeneic mouse tumors and immunodeficient mice with transplanted human tumor cells (xenograft models). The dilemma with mouse models using syngeneic tumors is that human Ad type 5-based vectors do not productively replicate in mouse cells, which diminishes the value of these models for predicting the performance and/or safety of Ad5-based oncolytic vectors in humans. Xenograft models, although they do not take into account the interaction between the oncolytic vector and the host's immune system, are a more adequate tool to study Ad transduction of tumor cells and viral replication in tumors. The improvement of initial tumor cell transduction is a major task in tumor gene therapy and therefore, xenograft models have seen widespread use in testing new oncolytic vectors. Most current xenograft tumor models are based on subcutaneous transplantation of established human tumor cell lines into immunodeficient mice and subsequent intratumoral vector injection. These models do not mimic the clinical situation of liver metastases that require systemic vector application for treatment. Other models that involve direct injection of tumor cells into liver lobes (see⁹ for an example) do not consider that the vast majority of liver metastases are of a hematogenic nature. More adequate mouse models for metastatic cancer involve orthotopic implantation of breast cancer cells into the mammary fat pad, with spontaneous metastasis to the lungs (see¹⁰ for an example). Our group has established mouse models for liver metastasis by injection of human tumor cells into the portal vein through a permanently placed catheter¹¹. We have used this model to test the efficacy of oncolytic adenoviruses^5,12,13,14 (P. Sova et al., manuscript submitted). In this study, we evaluate whether the morphology of liver metastases in our models is clinically relevant and which structural features of the tumor influence transduction with Ad vectors.

The microenvironment or stroma immediately surrounding tumor cells consists of a three-dimensional extracellular matrix (ECM) and stromal cells such as fibroblasts and inflammatory cells. Basement membranes (BM) are thin sheets of specialized ECM consisting mainly of type IV collagen, laminin, heparan-sulfate proteoglycans, and nidogen/entactin (for a review see¹⁵). Benign disorders are characterized by continuous BM, whereas in malignant tumors stromal invasion is associated with loss or fragmentation of BM^16,17,18. The last phenomenon may be due to either increased breakdown [for example by matrix metalloproteinases (MMPs)] or decreased synthesis of BM components. BM components are produced by tumor cells, stromal cells such as fibroblasts, and, in the case of hepatic tumors/metastases, hepatic stellate cells. Tumor cells also secrete enzymes that change the structure of the basement membrane and host connective tissue. As a result of ECM degradation, factors that stimulate angiogenesis are released, such as vascular endothelial growth factor, basic fibroblast growth factor, or platelet-derived growth factor. Vascular endothelial-cell proliferation leads to sprouting of preexisting microvessels. These vessels invade the ECM and form tubes and, ultimately, the tips of these tubes connect to create loops that are capable of conducting blood flow¹⁹. Recently it was also suggested that the bone marrow contributes cells to the tumor stroma²⁰ and that a subfraction of bone-marrow-derived cells can differentiate into vascular endothelial cells²¹. Newly formed tumor vasculature is structurally immature, exhibiting poor association between endothelial cells and supporting cells, and is often leaky and hemorrhagic²².

The ECM is considered to be an anatomical barrier for transduction with viral vectors. For example, the ECM blocks transduction of mature myofibers by viral vectors, including Ad-²³ and HSV-1-based vectors²⁴. Normal urothelium is surrounded by continuous BM and is relatively refractory to Ad transduction; however, gene delivery to malignant human urothelial cells grown in organoid cultures is relatively efficient because the BM is disrupted in these tumors²⁵. We also found in our previous studies that the BM represents a barrier in Ad transduction of hepatic metastases derived from human MDA-MB-435 breast cancer cells⁵.

Here we compare the morphology, vascularization, and BM distribution in hepatic metastases derived from cervical, colon, breast, and liver cancer cell lines with human tumors. Furthermore, we analyze the correlation between tumor cell transduction after systemic Ad vector administration and structural features of the various models for metastasis. Finally, we test a number of methods to trace Ad transduction of metastases and viral replication in tumor cells in vivo.

Results and Discussion

Xenograft Models for Liver Metastasis Are Histologically Similar to Human Tumors

To mimic hematogenic metastasis to the liver, we injected 2 10⁶ human tumor cells into the portal vein of immunodeficient C.B-17/lcrCrl-scid-bg/BR (CB17) mice through a permanently placed catheter¹¹. The catheter technique allows for reliable establishment of hepatic metastases in 100% of animals. Tumor engraftment was less efficient in nude and SCID mice (data not shown). The cell lines used were HeLa (cervical carcinoma), LoVo (colon carcinoma), MDA-MB-435 (breast cancer), and Hep3B (hepatoma). We selected the cell lines from a series of models established in our lab based on differences in tumor morphology. All cell lines are readily transducible with Ad5 and fiber-chimeric Ad5/35-based vectors^{5,13,14,26,27}. In preliminary studies, we sacrificed mice at different time points to assess the kinetics of tumor growth. Metastases became macroscopically visible, with a diameter of 2 to 3 mm, at days 18, 23, 28, and 30 after transplantation of HeLa, MDA-MB-435, LoVo, and Hep3B cells, respectively, which reflects differences in the growth kinetics of tumors.

Liver metastases derived from HeLa cells grow as multiple small spherical nests (Fig. 1A). The tumors start to appear at the edge of the liver lobes. Histologically, the metastases are composed of relatively large cells that differ in size. Large metastases show central necrosis (see arrow in Fig. 1B). A fragmented fibrous BM detected by silver staining was seen only in the peripheral region of the tumor [see arrows, Fig. 1C (black)]. The overall tumor morphology is comparable to that observed in tumor sections from patients with poorly differentiated, invasive adenocarcinoma of the cervix, in which BM are present around the tumor nests (Figs. 1D and 1E).

Figure 1.

Histopathology of metastatic tumors in mouse models with human tumors. Xenograft models for liver metastasis: Tumors are derived from HeLa (first column; A–C), MDA-MB-435 (second column; F–H), LoVo (third column; K–M), and Hep3B cells (fourth column; P–R). (First row) Gross examination of tumor-bearing livers; (second row) H&E staining of paraffin sections of liver metastases; (third row) silver staining of ECM on paraffin sections of liver metastases. Histopathology of human tumors: (D and E) Poorly differentiated, invasive adenocarcinoma of the cervix; (I and J) well-differentiated ductal mammary carcinoma; (N and O) poorly differentiated colon carcinoma; (S and T) well-differentiated hepatocellular carcinoma. (Fourth row) H&E staining of paraffin sections of human tumor sections; (fifth row) silver staining of ECM on paraffin sections of human tumors. Note that both cervical carcinoma sections (D and E) are stained for ECM. Original magnification: 20.

Full figure and legend (804K)

MDA-MB-435 tumors form multiple nodules (Fig. 1F). Morphologically the tumor is composed of small, well-separated tumor nests with loosely connected monomorphic cells (Fig. 1G). A dense extracellular matrix is found surrounding most tumor nests, particularly small tumors (see arrows in Fig. 1H). The histopathology of a well-differentiated ductal mammary carcinoma shows a similar nodular tumor growth (Fig. 1I), with compartmentalized, individual tumor nests separated by BM (see arrows in Fig. 1J). In this context it is notable that early stage breast cancer (intraductal carcinoma) has a continuous BM (and is positive for fibronectin and laminin), whereas only a fraction of invasive tumors had BM and were laminin positive²⁸.

Hepatic metastases derived from LoVo colon cancer cells macroscopically display characteristic villi-like structures (Fig. 1K). Tumor-bearing livers are markedly enlarged, indicating that tumors grew inside the liver lobes. Histologically, the tumor cells are relatively uniform in size and shape (Fig. 1L). BM detected by silver staining is visible throughout the tumor (see arrows in Fig. 1M). LoVo tumors display areas with glandula-like arrangement of tumor cells (see arrows in Fig. 1L), which are very similar to those seen in poorly differentiated human colorectal carcinoma (Fig. 1N). A delicate intervening fibrovascular stroma is seen throughout both the tumor model and the human colon cancer sample (see arrows in Fig. 1O).

Livers with Hep3B cell-derived metastases are enlarged (Fig. 1P). Histologically, tumor cells are arranged as nests and show trabecular structures, resembling liver plates. Some tumor nests have a central lumen filled with blood (Figs. 1Q and 1R). The nests of tumor cells are separated by bundles of extracellular matrix (arrows, Figs. 1Q and 1R). Glycogen that appears as black dots by silver staining is present in the Hep3B tumor cells as well as in hepatocytes (Fig. 1R). A similar ECM and glycogen pattern is visible in the section of a hepatocellular carcinoma (see arrows in Figs. 1S and 1T). The histopathology of the human tumor also shows an abortive attempt at formation of a central vein.

Considering the genetic instability of tumors and the fact that the cell lines used in this study have undergone many passages since their establishment, one would expect that they might significantly differ genetically and phenotypically from the tumor they were derived from. It was therefore interesting to observe that when these tumor cell lines were grown in situ as hepatic metastases they still retained key anatomical features of the corresponding primary tumor, including the specific interaction with the ECM.

Relationship between Vascularization, ECM, and Adenoviral Transduction in Different Models

In an earlier study we obtained preliminary data indicating that the transduction of MDA-MB-435 cell-derived metastases critically depended on direct access of tumor cells to blood supply. Here we used the different tumor models to perform a more systematic analysis of the relationship between vascularization, ECM, and tumor cell transduction. Treatment of metastatic cancer requires systemic intravascular application of anti-tumor vectors. In this study, we infused 5 10⁹ pfu of Ad.RSV-Gal (a first-generation Ad vector expressing -galactosidase) through the tail vein into mice with preestablished metastases. We performed Ad infusion at a time point when metastases became macroscopically visible (see above). Three days after Ad.RSV-Gal infusion, we sacrificed the mice and analyzed sections of tumor-bearing livers by immunohistochemistry. Blood vessels were visualized using an antibody against CD31, a platelet/endothelial cell adhesion molecule 1. We observed different patterns of tumor vasculature in different tumor models (Figs. 2A, 2D, 2G, and 2J). Double labeling with anti-CD31 (green) and antibodies against the BM key component laminin (red) showed the relation between BM and vascularization (Figs. 2B, 2E, 2H, and 2K). Finally, triple-label immunohistochemistry with antibodies against CD31 (green), laminin (red), and -galactosidase (blue) revealed the relationship between vascularization, ECM, and tumor cell transduction (Figs. 2C, 2F, 2I, and 2L).

Figure 2.

Relationship between tumor vascularization, extracellular matrix, and transduced tumor cells in mouse tumor models. Sections are from (first column, A–C) HeLa-derived metastases, (second column, D–F) MDA-MB-435-derived metastases, (third column, G–I) LoVo-derived metastases, and (fourth column, J–L) Hep3B-derived metastases. (First row) Detection of blood vessels by immunohistochemistry using anti-CD31 antibodies (green). T, tumor (original magnification: 10). (Second row) Distribution pattern of blood vessels and extracellular matrix in mouse tumor models shown by double labeling of anti-CD31 (for blood vessels, green) and anti-laminin (for basement membrane, red) (original magnification: 20). (Third row) Relationship of transduced tumor cells, extracellular matrix, and blood vessels in mouse tumor models shown by triple labeling with anti-Gal (for transduced cells, blue), anti-laminin (for basement membrane, red), and anti-CD31 (for blood vessels, green) (original magnification: 40).

Full figure and legend (452K)

In HeLa-derived tumors, a small number of blood vessels that labeled with anti-CD31 (Figs. 2A, 2B, and 2C, green) were present in the tumor periphery and did not reach the center of the tumor, where we noticed necrosis. The ECM labeled with anti-laminin antibodies formed a network around groups of tumor cells (Figs. 2B and 2C, red). Only the tumor cells in the peripheral region that had contact with blood vessels could be transduced by the viral vector (Fig. 2C, blue). In MDA-MB-435 tumors, blood vessels were visible between the tumor nodules; however, they did rarely penetrate into the tumor nests (Figs. 2D, 2E, and 2F, green). Most tumor nests formed by MDA-MB-435 tumors were lined by a continuous BM that blocked contact between tumor cells and blood vessels (Figs. 2E and 2F, red). Only MDA tumor nests that had contact with blood vessels could be transduced (Fig. 2F, blue). In LoVo tumors, blood vessels were found in the stroma throughout the whole tumor (Figs. 2G, 2H, and 2I, green). Strands of ECM appear as a loose network (Figs. 2H and 2I, red). Labeling for CD31 and laminin often overlaps (resulting in yellow signals), because, as expected, tumor vasculature is embedded in the ECM. Ad-transduced tumor cell clusters had close contact with blood vessels and were seen throughout the tumor (Fig. 2I, blue). Hep3B-derived tumors were vascularized and displayed small amounts of fragmented ECM strands in between the tumor nests (Figs. 2J, 2K, and 2L, green). Some blood vessels were embedded in the ECM (Figs. 2K and 2L, red). Again, only the tumor cells that had contact with blood vessels were transduced (Fig. 2L, blue).

Because the liver metastases model is most relevant for studies on colon cancer, we included an additional colon cancer cell line, HT-29, in our studies (Fig. 3). As seen in the study with LoVo cells, the overall histologies of human (Figs. 1N and 1O) and HT-29 model tumors (Figs. 3B and 3C) were similar. We found transduced cells predominantly in the proximity of blood vessels, and transduced tumor nests were not surrounded by extracellular matrix (Figs. 3D, 3E, 3F, and 3G). Costaining with monoclonal antibodies against Gal and rabbit polyclonal antibodies directed against a human-specific mitochondrial protein (Chemicon, Temecula, CA, USA) demonstrated that human tumor cells were transduced (Fig. 3H).

Figure 3.

Liver metastases derived from HT-29 colon cancer cells. Tumor morphology and relationship between tumor vascularization, extracellular matrix, and transduced tumor cells are shown. Overall tumor morphology (A) and sections from HT-29-derived metastases (B through H) are shown. (B) H&E staining. (C) Silver staining for ECM (see arrows). (Note that mucus in the lumen of glandula-like tumor structures also stained positive.) (D) Staining for blood vessels (anti-CD31, green) and (anti-Gal, blue). (E) Staining for blood vessels (anti-CD31, green). (F) Staining for blood vessels (anti-CD31, green) and ECM (anti-laminin, red). (G) Staining for blood vessels (anti-CD31, green), ECM (anti-laminin, red), and (anti-Gal, blue). (H) Staining for a human mitochondrial marker protein (Chemicon, Temecula, CA) (red) and Gal (blue). (Note that "red" and "blue" staining overlaps, resulting in "purple" signals, which indicate Gal expression in human tumor cells.)

Full figure and legend (261K)

The figures are representative of the whole tumor (which was serially sectioned). In the vast majority, transduced tumor cells had direct contact to blood supply. We did not find that tumor morphology changed when analyzed at different time points after transplantation. This is in contrast to a recently reported carcinogen-induced murine HCC model, in which BM staining was found to become more intense with increasing tumor size⁴. Because, in our models, tumors were analyzed when they became macroscopically visible, it cannot be excluded that ECM deposition is less in early (smaller) tumors. In summary, tumor cells were transduced only if there was direct contact between blood vessels and tumor cells. Tumor cells separated from blood vessels by a continuous basement were not transduced.

Different Methods to Detect Transduced Tumor Cells and Ad Replication

We used the different tumor models to test a series of techniques to trace Ad-mediated transgene expression or viral replication on sections of metastases. In mice with HeLa metastases, in addition to staining with anti-Gal antibodies on cryosections, we detected -galactosidase expression after systemic Ad.RSV-Gal transduction by X-Gal staining of cryosections (Fig. 4A, blue). As outlined above, we found transduced tumor cells in HeLa metastases only in the tumor periphery in proximity with blood vessels. We injected mice bearing MDA-MB-435 metastases with Ad5/35.CMV-GFP and analyzed GFP expression 3 days after injection on paraffin sections by immunohistochemistry with anti-GFP antibodies using the ABC method (Fig. 4B, brown). This method delivers more reliable results than analysis of GFP fluorescence, which is often difficult to distinguish from the high background autofluorescence present in liver tissue. Fig. 4B corroborates the finding described above that only some MDA-MB-435 metastases can be transduced by Ad vectors.

Figure 4.

Demonstration of different methods for detection of transgene expression or viral replication. (A) X-Gal staining for -galactosidase-expressing tumor cells in HeLa metastases 3 days after injection of Ad.RSV-Gal. Positive staining appears blue (original magnification: 10). (B) Immunohistochemistry for detection of GFP-expressing tumor cells on a section of MDA-MB-435 metastases 3 days after injection of Ad5/35.CMV-GFP. Positive staining appears brown (original magnification: 40). (C) Immunohistochemistry for detection of tumor cells expressing viral hexon (as an indicator of viral replication) on sections of LoVo metastases 14 days after injection of Ad.IR-codA/upp. Hexon staining appears green. The basement membrane appears red (original magnification: 40). (D) Staining for alkaline phosphatase expression (which is activated only upon viral replication) in sections of Hep3B tumors 10 days after injection of Ad.IR-E1a/AP. AP-positive cells appear purple. Hep3B hepatoma cells are visualized with anti-AFP antibodies and appear brown (original magnification: 40).

Full figure and legend (173K)

We have also tested two indirect methods to detect Ad replication in tumor cells. The first method uses the specific kinetics of viral hexon expression from the Ad major late promoter, which is significantly activated only upon initiation of viral DNA replication²⁹. This method has been used before by our group and others^14,30. Fig. 4C shows hexon immunocytochemistry using FITC-conjugated anti-hexon antibodies on cryosections from livers bearing LoVo metastases 14 days after Ad injection. The basement membrane was labeled with anti-laminin antibodies (Fig. 4C, red). We found tumor cells containing replicating Ad (Fig. 4C, green) throughout the tumor along the stroma.

The second method for detecting viral replication uses a system for replication-activated transgene expression (Ad.IR) that was recently developed and tested in our lab^12,13,14. In this system, a promoter is linked to the 5' end of a transgene only upon homologous recombination within the Ad genomes, which in turn depends on viral DNA replication. In Ad.IR-E1a/AP, expression of the bicistronic cassette containing the adenoviral E1a gene and the reporter gene alkaline phosphatase (AP) correlated with viral DNA replication¹³. E1a expression from this vector further enhances viral DNA replication and allows for progeny virus production in tumor cells. In this study, we injected Ad.IR-E1a/AP into mice bearing Hep3B metastases and 10 days later analyzed viral replication on cryosections or paraffin sections based on AP activity using the substrates 4-nitroblue tetrazolium chloride (NBT) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP). AP-positive tumor cells formed clusters indicating viral spread. Transduced tumor cells were close to blood vessels or along the path of blood vessels (Fig. 4D, purple). Hep3B hepatoma cells stained positive with antibodies against -fetoprotein (AFP; Fig. 4D, brown). Overall, we found that the quality and tissue morphology of stained paraffin sections were better than those of cryosections.

In summary, the tumor models differ in morphological features and in the accessibility to virus transduction. The LoVo and Hep3 models, with extensive tumor vascularization and contact between blood vessels and tumor cells, allow for relatively efficient tumor cell transduction. Particularly, the LoVo model is of importance because metastatic colon/colorectal carcinomas are resistant to classical treatment options, and development of gene therapy might have a major clinical impact. The HeLa model displays early central necrosis and is therefore not optimal for studying gene delivery into large tumor masses. The MDA-MB-435 model, in which viral tumor cell transduction is largely blocked by ECM, can be used to develop new methods that allow anti-tumor drugs, viral vectors, or immune cells to penetrate the ECM and gain access to tumor cells. In this context approaches to genetically modifying (autologous) bone marrow-derived vascular endothelial cells that, after transplantation into cancer patients, home to tumor stroma appear to be very promising. For example bone marrow cells could be stably transduced ex vivo to express MMPs inducibly and be used as a "Trojan horse" to permeabilize the ECM for subsequent chemo- and immunotherapy as well as for treatment with oncolytic viruses. However, this approach as well as strategies that involve vasoactive compounds (such as histamine, angiotensin, and the nitric oxide donor nitroglycerin)⁴ might also favor further metastasis.

It is known that primary and secondary hepatic tumors depend mainly on arterial supply rather than on portal blood. Based on this, we reasoned that virus injection into the hepatic artery might allow for better transduction of liver tumors. A series of reports, however, has disproved the effectiveness of this strategy^4,31. This seems to be plausible in the light of recently published data, and results obtained in this study, showing that anatomical barriers formed by the ECM and contact to blood vessels are the primary limitations for tumor cell transduction. In addition, other problems remain to be resolved; for example, high intratumoral pressure, particularly in tumors surrounded by a fibrous ECM capsule, might affect virus diffusion throughout the tumor. Tumor models like those investigated in this study will help to solve these problems and bring gene therapy for metastatic cancer closer to reality.

Materials and Methods

Cell lines

Tumor cell lines were obtained from the American Type Culture Collection (Manassas, VA, USA). 293 cells were obtained from Microbix (Toronto, Canada). 293 cells, HeLa cells (ATCC CCL-2), and Hep3B cells (ATCC HB-8064) were cultured in DMEM, LoVo cells (ATCC CCL-229) were cultured in F12K medium, and MDA-MB-435 cells (ATCC HTB-129) were cultured in L15 medium. All media were supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 U penicillin/ml, and 100 g streptomycin/ml.

Adenovirus vectors

All Ad vectors used in this study have the E1A and E1B genes (nucleotides 342–3523) and the E3 genes (nucleotides 28,133–30,818) deleted. Ad.RSV-Gal³² contained the Escherichia coli lacZ gene under the control of the Rous sarcoma virus LTR promoter. Ad.IR-codA/upp¹⁴ (used in the LoVo model) is a replication-activated vector that expresses cytosine deaminase/uracil phosphoribosyltransferase. Ad5/35.CMV-GFP⁵ (used in the MDA-MB-435 model) expresses GFP under the control of the CMV promoter and possesses a chimeric Ad5 capsid with Ad35 fibers. Ad.IR-E1a/AP¹³ (used in the Hep3B model) is a replication-activated vector for expression of E1a and alkaline phosphatase.

All vectors were banded in CsCl gradients as described elsewhere³³ and stored at -80°C in 10% glycerol, 10 M Tris–Cl, 1 M MgCl₂. Titers were determined by OD measurement and plaque titering on 293 cells. Real-time PCR was performed to assess contamination of Ad vector preparations with E1⁺ replication-competent adenovirus¹³. Only virus preparations that contained less than one E1⁺ (wild-type) viral genome in 1 10⁷ genomes were used.

Human tumor sections

Paraffin sections of human tumor samples were obtained from the Department of Pathology, University of Washington. The specimens were from primary cervical carcinoma, colon adenocarcinoma, breast carcinoma, and hepatocellular carcinoma. Generally, key histological features are similar in primary tumors and corresponding metastases.

Animal models

All experimental procedures involving animals were conducted in accordance with the institutional guidelines set forth by the University of Washington. Eight to ten-week-old immunodeficient CB17 mice (Charles River, Wilmington, MA, USA) were housed in a specific-pathogen-free facility. To generate the mouse models with liver metastases derived from human tumor cells, the mice were infused through a permanently placed portal vein catheter with 2 10⁶ tumor cells as described elsewhere^11,12,13,14. Systemic adenovirus applications were performed 2 to 3 weeks after tumor cell transplantation. Ad vector (5 10⁹ pfu per mouse) was injected through the tail vein. Liver specimens were embedded in OCT compound (Sakura Finetek, USA, Inc., Torrance, CA, USA) and cryosectioned or fixed in 10% neutral formalin and then embedded in paraffin for histopathology, histochemistry, and immunohistochemistry.

Histopathology

Paraffin sections of mouse livers with metastases and the corresponding human tumors were stained with hematoxylin and eosin (H&E). The basement membrane of the tumors was visualized using Jones' methenamine silver staining method³⁴ or periodic acid–Schiff staining.

Histochemistry and immunohistochemistry

Ad.RSV-Gal transduction was detected by X-Gal staining for -galactosidase expression on 8-m OCT sections as described elsewhere¹². -Galactosidase expression was also detected by immunohistochemistry on cryosections or paraffin sections using a mouse anti-Gal antibody (1:200; Promega Corp., Madison, WI, USA).

GFP expression after infusion of Ad5/35-CMV-GFP virus was visualized on paraffin sections by immunohistochemistry with mouse anti-GFP antibody (1:200; Clontech, BD Biosciences, San Diego, CA, USA) using the ABC method (Vector Laboratories, Burlingame, CA, USA).

Viral hexon protein was detected on cryosections with FITC-conjugated polyclonal rabbit anti-hexon antibodies (1:50) (Chemicon International, Temecula, CA). Viral hexon expression from the major late promoter is activated only upon viral DNA replication and therefore serves as an indicator for viral replication.

For detection of Ad.IR-E1a/AP-mediated AP expression, frozen or paraffin sections (after deparaffinization) were fixed in 0.5% glutaraldehyde/PBS for 30 min at room temperature. After being washed with PBS, slides were incubated at 65°C for 1 h to inactivate endogenous AP. Staining was performed in 0.1 M Tris, pH 9.5, 0.1 M NaCl, 0.5 mg/ml NBT (Roche Diagnostics, Indianapolis, IN, USA), and 0.1875 mg/ml BCIP (Roche Diagnostics). The reaction was stopped with PBS/1 mM EDTA. Hep3B hepatoma cells were visualized with anti-AFP antibodies (1:100) using the ABC method²⁷.

The distribution of tumor-feeding blood vessels and extracellular matrix was analyzed by immunohistochemistry on cryosections fixed in 4% paraformaldehyde. Blood vessels were labeled with rat anti-mouse CD31 (1:50; PharMingen, San Diego, CA, USA) followed by anti-rat IgG conjugated with Alexa Fluor 488 (green) (1:200; Molecular Probes, Inc., Eugene, OR, USA). Basement membrane was labeled with rabbit anti-laminin (1:200; DakoCytomation, Carpinteria, CA, USA) followed by anti-rabbit IgG conjugated with Alexa Fluor 568 (red) (1:200; Molecular Probes, Inc.).

To study the relationship between transduced tumor cells, basement membrane, and blood vessels, triple immunohistochemical labeling was employed. The cryosections were fixed in 4% paraformaldehyde and transduced tumor cells were labeled with anti-Gal antibody (1:200; Promega Corp.) followed by anti-mouse IgG conjugated with AMCA (blue) (1:100; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). Blood vessels and basement membrane were labeled as described above.

All photographs were taken on a Leica DMLB microscope.

References

1. Mayer-Kuckuk, P., Banerjee, D., Kemeny, N., Fong, Y. and Bertino, J. R. (2002). Molecular therapies for colorectal cancer metastatic to the liver. Mol. Ther. 5: 492–500. | Article | PubMed |
2. Havlik, R., Jiao, L. R., Nicholls, J., Jensen, S. L. and Habib, N. A. (2002). Gene therapy for liver metastases. Semin. Oncol. 29: 202–208. | Article | PubMed | ChemPort |
3. Kruyt, F. A. and Curiel, D. T. (2002). Toward a new generation of conditionally replicating adenoviruses: pairing tumor selectivity with maximal oncolysis. Hum. Gene Ther. 13: 485–495. | Article | PubMed | ISI | ChemPort |
4. Bilbao, R., et al. (2000). A blood–tumor barrier limits gene transfer to experimental liver cancer: the effect of vasoactive compounds. Gene Ther. 7: 1824–1832. | Article | PubMed | ISI | ChemPort |
5. Shayakhmetov, D. M., Li, Z. Y., Ni, S. and Lieber, A. (2002). Targeting of adenovirus vectors to tumor cells does not enable efficient transduction of breast cancer metastases. Cancer Res. 62: 1063–1068. | PubMed | ISI | ChemPort |
6. Li, Y., et al. (1999). Loss of adenoviral receptor expression in human bladder cancer cells: a potential impact on the efficacy of gene therapy. Cancer Res. 59: 325–330. | PubMed | ISI | ChemPort |
7. Okegawa, T., et al. (2000). The dual impact of coxsackie and adenovirus receptor expression on human prostate cancer gene therapy. Cancer Res. 60: 5031–5036. | PubMed | ISI | ChemPort |
8. Kanerva, A., et al. (2002). Targeting adenovirus to the serotype 3 receptor increases gene transfer efficiency to ovarian cancer cells. Clin. Cancer Res. 8: 275–280. | PubMed | ISI | ChemPort |
9. Herynk, M. H., et al. (2003). Down-regulation of c-Met inhibits growth in the liver of human colorectal carcinoma cells. Cancer Res. 63: 2990–2996. | PubMed | ChemPort |
10. Fabra, A., et al. (2001). Intravascular injections of a conditional replicative adenovirus (adl118) prevent metastatic disease in human breast carcinoma xenografts. Gene Ther. 8: 1627–1634. | Article | PubMed | ChemPort |
11. Vrancken Peeters, M. J., Perkins, A. L. and Kay, M. A. (1996). Method for multiple portal vein infusions in mice: quantitation of adenovirus-mediated hepatic gene transfer. Biotechniques. 20: 278–285. | PubMed | ChemPort |
12. Steinwaerder, D. S., et al. (2001). Tumor-specific gene expression in hepatic metastases by a replication-activated adenovirus vector. Nat. Med. 7: 240–243. | Article | PubMed | ISI | ChemPort |
13. Bernt, K., et al. (2002). A new type of adenovirus vector that utilizes homologous recombination to achieve tumor-specific replication. J. Virol. 76: 10994–11002. | Article | PubMed | ISI | ChemPort |
14. Bernt, K. M., et al. (2002). Enzyme-activated prodrug therapy enhances tumor-specific replication of adenovirus vectors. Cancer Res. 62: 6089–6098. | PubMed | ChemPort |
15. Kalluri, R. (2003). Basement membranes: structure, assembly and role in tumour angiogenesis. Nat. Rev. Cancer. 3: 422–433. | Article | PubMed | ISI | ChemPort |
16. Birembaut, P., Caron, Y., Adnet, J. J. and Foidart, J. M. (1985). Usefulness of basement membrane markers in tumoural pathology. J. Pathol. 145: 283–296. | Article | PubMed | ChemPort |
17. Richards, C. J. and Furness, P. N. (1990). Basement membrane continuity in benign, premalignant and malignant epithelial conditions of the uterine cervix. Histopathology. 16: 47–52. | PubMed | ChemPort |
18. Willebrand, D., Bosman, F. T. and de Goeij, A. F. (1986). Patterns of basement membrane deposition in benign and malignant breast tumours. Histopathology. 10: 1231–1241. | PubMed | ChemPort |
19. Ingber, D. E. (2002). Cancer as a disease of epithelial–mesenchymal interactions and extracellular matrix regulation. Differentiation. 70: 547–560. | Article | PubMed |
20. Studeny, M., et al. (2002). Bone marrow-derived mesenchymal stem cells as vehicles for interferon-beta delivery into tumors. Cancer Res. 62: 3603–3608. | PubMed | ChemPort |
21. Rafii, S., Lyden, D., Benezra, R., Hattori, K. and Heissig, B. (2002). Vascular and haematopoietic stem cells: novel targets for anti-angiogenesis therapy? Nat. Rev. Cancer. 2: 826–835. | Article | PubMed | ISI | ChemPort |
22. Yancopoulos, G. D., et al. (2000). Vascular-specific growth factors and blood vessel formation. Nature. 407: 242–248. | Article | PubMed | ISI | ChemPort |
23. Cao, B., Mytinger, J. R. and Huard, J. (2002). Adenovirus mediated gene transfer to skeletal muscle. Microsc. Res. Tech. 58: 45–51. | Article | PubMed | ChemPort |
24. Huard, J., et al. (1996). The basal lamina is a physical barrier to herpes simplex virus-mediated gene delivery to mature muscle fibers. J. Virol. 70: 8117–8123. | PubMed | ISI | ChemPort |
25. Chester, J. D., Kennedy, W., Hall, G. D., Selby, P. J. and Knowles, M. A. (2003). Adenovirus-mediated gene therapy for bladder cancer: efficient gene delivery to normal and malignant human urothelial cells in vitro and ex vivo. Gene Ther. 10: 172–179. | Article | PubMed | ChemPort |
26. Steinwaerder, D. S., Carlson, C. A. and Lieber, A. (2000). DNA replication of first-generation adenovirus vectors in tumor cells. Hum. Gene Ther. 11: 1933–1948. | Article | PubMed | ISI | ChemPort |
27. Ye, X., et al. (2003). Insulation from viral transcriptional regulatory elements enables improvement to hepatoma-specific gene expression from adenovirus vectors. Biochem. Biophys. Res. Commun. 307: 759–764. | Article | PubMed | ChemPort |
28. Gorczyca, W., Holm, R. and Nesland, J. M. (1993). Laminin production and fibronectin immunoreactivity in breast carcinomas. Anticancer Res. 13: 851–858. | PubMed | ChemPort |
29. Shenk, T. (1996). Adenoviridae. In (Fields B. N., Knipe D. M., Howley P. M., Eds.), Fields Virology. Vol. 2 , pp. 2111–2148. Lippincott–Raven, Philadelphia.
30. Chillon, M., et al. (1999). Group D adenoviruses infect primary central nervous system cells more efficiently than those from group C. J. Virol. 73: 2537–2540. | PubMed | ISI | ChemPort |
31. Maron, D. J., et al. (2001). Intra-arterial delivery of a recombinant adenovirus does not increase gene transfer to tumor cells in a rat model of metastatic colorectal carcinoma. Mol. Ther. 4: 29–35. | Article | PubMed | ISI | ChemPort |
32. Carlson, C. A., Shayakhmetov, D. M. and Lieber, A. (2002). Restoration of a functional open reading frame by homologous recombination between two adenoviral vectors. Mol. Ther. 6: 99–105. | Article | PubMed | ChemPort |
33. Carlson, C. A., Steinwaerder, D. S., Stecher, H., Shayakhmetov, D. M. and Lieber, A. (2002). Rearrangements in adenoviral genomes mediated by inverted repeats. Methods Enzymol. 346: 277–292. | PubMed | ChemPort |
34. Jones, D. B. (1951). Inflammation and repair of the glomerulus. Am. J. Pathol. 27: 991–1007. | PubMed | ChemPort |

Acknowledgements

We thank Qinghua Feng for providing the tissue sections of human tumors. We are grateful to Daniel Stone for critical discussion of the manuscript. This work was supported by NIH Grant R01 CA80192 and by a grant from the Cystic Fibrosis Foundation.