|Home | About | Journals | Submit | Contact Us | Français|
Cancer cells can develop an attenuated immunogenicity and/or create an immunosuppressive microenvironment to prevent tumor eradication by host immune system, the so-called “cancer immunoediting” hypothesis. The aim of the present study was to find evidence for this hypothesis by using a rat orthotopic bladder cancer model. Fisher rats were inoculated with AY-27 cells (a Fisher rat bladder cancer cell line). Cultured cancer cells, rat and human bladder cancer tissues, and publicly available microarray data from human bladder cancer were analyzed by means of bioinformatics and morphology. Results showed that 12 of 24 differentially expressed pathways were concordant in connection to cell cycle and proliferation between rats and humans (both non-muscle-invasive and muscle-invasive tumors) and that 11 of the 24 pathways, including major histocompatibility complex, were related to host immunosurveillance with activations of T cells and natural killer cells in rats. The altered pathways and morphogenesis of this rat model corresponded more closely with those of human muscle-invasive rather than non-muscle-invasive tumors. A unique ultrastructure displaying microvillus-formed niches was found in small areas within the tumor of both rats and humans. These niches were interconnected with desmosomes between cancer cells and without infiltration of lymphocytes. The expression of E-cadherin, selectins, PGP9.5, vascular endothelial growth factor, caspase-3, CD133, Oct-4, nestin, CD3, and CD45RA was lower in the tumor than in the adjacent normal epithelium. We suggest that the microvillus-formed niche that harbors a few implanted cancer cells might be the compartment that prevents the tumor eradication by the host immune system.
Ideal animal cancer models should, in principle, reflect tumor biology in humans. Like humans, immunocompetent animals are capable of recognizing and eliminating syngeneic cancer cells not only before the development of a tumor but also after tumor formation, the so-called “cancer immunosurveillance” [1,2]. However, tumors in immunocompetent individuals still occur and progress, which is called tumor immunoescape. Conceivably, cancer cells (most likely a subpopulation of the cancer cells) can develop an attenuated immunogenicity and/or create an immunosuppressive tumor microenvironment that effectively prevents tumor eradication by the host immune system, called “cancer immunoediting” [3,4].
Recently, we have shown that the host immune system was fully active in the form of immunosurveillance, and yet, the tumors still progressed in a syngeneic orthotopic rat bladder cancer model . Moreover, this syngeneic orthotopic bladder cancer model showed a similar but more rapid progression compared with chemical carcinogen N-butyl-N-(4-hydroxybutyl)nitrosamine (BBN)-induced bladder cancer model [5,6] (own unpublished data). A recent study also showed that BBN-induced rodent (mouse and rat) models of bladder cancer shared genetic similarities with pathways relevant to the development of muscle-invasive bladder cancer in humans . In the present study, we wanted to test the cancer immunoediting hypothesis by using the syngeneic orthotopic rat bladder cancer model in comparison with human bladder cancer. First, we sought the gene expression profile for immunoediting by gene expression profiling and pathway analysis in comparisons between rat bladder cancer cells in vitro (AY-27 cell line) and in vivo and between the rat and human bladder cancers. Second, we sought morphological evidence for the immunosuppressive microenvironment in both rat and human bladder cancers by transmission electron microscopy and immunohistochemistry.
A syngeneic rat bladder cancer cell line (AY-27) was originally established as a primary bladder cancer developed in Fischer 344 rats exposed to N-[4-(5-nitro-2-furyl)-2-thiazolyl]formamide) and was kindly provided by Prof S. Selman, Department of Urology, Medical College of Ohio. The cells were cultivated on culture dishes with six wells in Dulbecco's modified Eagle medium (Sigma, St Louis, MO), 10% fetal bovine (Sigma), 1% penicillin-streptomycin (5000 U/ml of penicillin-G and 5 mg/ml of streptomycin; Sigma); and 0.5% L-glutamine (Sigma). For trypsinization, trypsin-EDTA solution (Sigma) was used. The cells were maintained in an incubator in a humid environment at 37°C and 5% CO2. Doubling time of the AY-27 cells in the pure culture was 27 hours (E. Larimar and D. Chen, unpublished data).
An orthotopic rat bladder cancer model has been described previously . In the present study, female Fisher inbred rats (2-month-old F344; Möllegaard, Skensved, Denmark) were used. The animals were housed in plastic cages with hardwood chips at room temperature of 22± 1°C, humidity of 45% to 55%, and 12 hours of light/dark cycle. Microbiological status was conventional and pathogen-free. The animals had free access to commercial standard rat food pellets (RM1; Scanbur BK AS, Karlslunde, Denmark) and tap water. For establishing bladder cancer, the animals were anesthetized by subcutaneous injection of 0.02 ml/kg of a solution containing 2.5 mg/ml fluanisone, 0.05 mg/ml fentanyl, and 1.25 mg/ml midazolam (Hypnorm [ Janssen Animal Health, Buckinghamshire, UK] or Dormicum [Alpharma AS, Oslo, Norway]). The amount of 4.0 x 105 AY-27 cells per rat was inoculated with an 18-gauge plastic intravenous cannula. A series of rats were subjected to intravesicle bladder cancer cell inoculation and killed at 1, 2, and 3 weeks (three to nine rats at each time point), respectively. A group of nine age- and sex-matched normal rats was used as controls. Entire bladders from six rats at the time point of 2 weeks after inoculation together with entire bladders from six controls were additionally used for microarray analysis. Approval for the experiments was obtained from the Norwegian Animal Research Authority.
Three patients whose conditions were diagnosed as muscle-invasive urothelial cancer (grade 3 according to the World Health Organization) underwent radical cystectomy at StOlav's University Hospital, Trondheim, Norway. Immediately after the bladder was removed, several small tissue specimens (1 x 1 x 1 m3) were taken from defined tumor areas and immersed in a fixative for electron microscopic analysis (see below). Approval for the analysis was obtained from the local ethics committee and from the patients.
The cultured cells and the bladder tissue samples from both rats and patients were collected for both light and transmission electron microscopic evaluations. In brief, the tissue samples were immersed in 4% formaldehyde for 8 to 12 hours at 4°C, dehydrated and embedded in paraffin blocks, and cut at a 4-µm thickness through the whole bladder. Every fifth section was thawed onto glass slides and subjected to a routine hematoxylin-eosin-safron staining for evaluation under a microscope (Nikon Eclipse 55i; Instrument AB, Oslo, Norway). For immunohistochemistry, the tissue sections (from the rats only) were treated with 3% H2O2 to quench endogenous peroxidase, washed several times, and blocked with 10% normal goat serum. The sections were incubated with a series of primary antisera for 48 hours at 4°C (Table 1). Afterward, the sections were incubated in biotinylated antirabbit or antimouse immunoglobulin G for 1 hour (Envision Kit K4007 or K4011; Dako, Glostrup, Denmark). After several washes in Tris-buffered saline, sections were incubated in 0.05% diaminobenzidine/0.001% H2O2 solution and washed at least two times with Tris-buffered saline. Sections were counterstained, dehydrated with absolute alcohol followed by xylene, coverslipped with Permount (The Science Company, Denver, CO), and examined under a microscope (Nikon).
For electron microscopy, many small tissue specimens were taken from different sites of the rat bladder and defined tumor sites of the patient's bladders and were immediately immersed in a mixture of 2% glutaraldehyde in 0.1MPBS, pH7.2. After 6 hours, the specimens were transferred to 1% OsO4 and postfixed for 1 hour, dehydrated in graded acetone, and embedded in Epoxy (TAAB Laboratories, Berkshire, UK). Ultrathin sections of 60- to 80-nm thickness were cut on an LKB MK III Ultratome (LKB, Bromma, Sweden), contrasted with uranyl acetate and lead citrate, and examined under a transmission electron microscope ( Joel, Tokyo, Japan).
Cultured AY-27 cells and the entire bladder from each rat were taken and immediately frozen in a bath of liquid nitrogen and stored at -80°C until isolation of RNA using the RNeasy Fibrous Midi Kit and RNase-free DNase Set (Qiagen, Valencia, CA). Isolated RNA was analyzed by Affymetrix GeneChip (Affymetrix, Santa Clara, CA) with a RAE 230 A chip that includes more than 15,000 rat genes. The arrays were scanned using Affymetrix GeneChip Scanner 3000 7G controlled by GCOS 1.3. Data were normalized using robust multiarray averaging, and data quality was evaluated by principal component analysis (PCA). Differentially expressed genes were identified, P values were corrected for multiple testing, and genes with P < .05 were taken as significant. Individual genes were functionally annotated using EGON v2.0 program (NTNU, Trondheim, Norway). Pathway analysis with Kyoto Encyclopedia of Genes and Genomes (Kyoto, Japan) was performed. In addition, publicly available microarray data from human bladder tumors  were downloaded from ArrayExpress and normalized using the same procedure as for the rat data.
PCA revealed different gene expression profiles among the normal bladder, bladder cancer in vivo, and cultured cancer cells (Figure 1, A and B). In the tissue microenvironment, 69% of differentially expressed genes were functionally related to cell communication, and 16% of genes were related to cell adhesion. In the cultured cancer cells, 39% of genes were annotated to cellular process, 34% were annotated to cellular physiological process, and 7% to 10% were annotated to cell death, cell cycle, or signal transduction. In the rat bladder cancer tissue, major histocompatibility complex (MHC I and II) pathways involving CD8- and CD4-evoked T cells, and KIR-evoked natural killer cells were upregulated. It should be noted that most genes in MHC I and all genes in MHC II were upregulated, and none of the genes in either MHC I or MHC II was downregulated. T-cell receptor signaling pathways in CD4 and CD8 T cells and natural killer cell-mediated cytotoxicity in natural killer cells were activated (Figure 2).
The comparison between this rat model and humans (both non-muscle-invasive and muscle-invasive tumors) revealed that 12 of 24 differentially expressed pathways were concordant in connection to cell cycle and proliferation. In the rat model, it was noted that 11 of 24 pathways were related to host immunosurveillance. Moreover, significantly differentiated pathways of this syngeneic orthotopic rat bladder cancer model corresponded much more closely with those of muscle-invasive human tumor (eight pathways) rather than with those of non-muscle-invasive tumors (four pathways; Table 2).
Minimal tumor formation was found 1 week after cancer cell inoculation. Inflammation characterized as lymphocyte infiltration was often observed within the mucosa (Figure 3A). At 2 weeks, carcinoma in situ and microinvasion into the lamina propria were observed. The tumors appeared surrounded by lymphocytes. Notably, small blood vessel density increased in the lamina propria adjacent to but not within the tumor (Figure 3B). At 3 weeks, both number and size of tumors increased, and the tumors invaded deeply into the lamina propria. Interestingly, in some areas, large tumors were noted with only moderate inflammation, and concurrently, small tumors were surrounded by intensive inflammatory responses (Figure 3, C and D). These histological features were also found in the human tumors (at stage T2 or 3 with grade 3; Figure 3, E and F ).
The AY-27 cancer cells in culture had a typical malignant ultrastructure characterized by irregular, pleomorphic microvilli (Figure 4A). Three weeks after being inoculated into the bladder, the cancer cells appeared to be crowded together, had smaller cell profile size, and had a more undifferentiated appearance than those in the culture (Figure 4B). Concurrently and very occasionally, a small area of the tumor was found to display a unique ultrastructure characterized by a microvillus-formed mesh (Figure 4C ). At higher magnification, the microvilli were tightly connected to each other by desmosomes (Figure 4D). Lymphocytes were not observed inside the mesh. In the human bladder cancer specimens, a similar ultrastructure displaying the desmosome-connected microvilli was also found (Figure 4, E–H).
Figure 5, A-O, shows a panel of immunostained tissue sections of the bladder cancer versus adjacent normal tissue with the antibodies to identify markers for the desmosome-connected microvilli, immune cells, cancer stem cells (CSCs), and stem cells in general. As expected, the cell proliferation marker proliferation cell nuclear antigen (PCNA) was increased, and the apoptosis marker caspase-3 was reduced in the tumor area. However, E-cadherin, one of cadherins of the desmosomes, and selectins (including E-, L-, and P-selectins) were found to be expressed in the normal epithelia but not in the tumor. Desmoglein1 + 2 was negative in the bladder but positive in other organs (e.g., stomach; data not shown). Markers of CSCs and/or stem cells, namely, CD133, Oct-4, and nestin, were positive in normal epithelial cells but slightly positive or not in the tumors. Vascular endothelial growth factor (VEGF), a tumor angiogenesis-promoting factor, was found to be less expressed in the tumor than in the adjacent normal tissue. Protein gene product 9.5 (PGP9.5), a tumor suppressor gene protein , was highly expressed in the normal epithelia but absent in the tumor area. In addition, markers for lymphocytes, namely, CD3 and CD45RA, were found more in the adjacent areas of the tumor than in the areas within the tumors.
Carcinogenesis involves the interaction between host and tumor cells in three phases: elimination of tumor cells by host (immunosurveillance), equilibrium between host and tumor cells, and escape of tumor cells fromhost immune response (immunoediting) [1–5,11,12]. In the present study, we have found that the microvillus-formed niche that harbors a few implanted cancer cells might be the compartment that prevents the tumor eradication by the host immune system.
In the present study, 4.0 x 105 AY-27 cells (approximately 0.5 mm3 in total cell volume with 27 hours of cell doubling time) were inoculated into each rat. Three weeks later, macrotumors (1–5 mm3) were found. Conceivably, most cancer cells did not survive or were eliminated after implantation by immunosurveillance, a response possibly mediated by the MHC I and II pathways that are associated with activations of CD8- and CD4-mediated T cells and KIR-mediated natural killer cells. Apparently, some of the cancer cells were capable of escaping from the immunosurveillance, possibly because of the loss or down-regulation of surface “foreign” antigens within the MHC I complex, which “veils” the tumor tissue and, in a sense, makes it “invisible” to immune cells . However, we were unable to specifically identify these cancer cells by immunohistochemistry using currently available antibodies.
In the present study, we found the differences in gene expression profiles among the bladder cancer cells in vitro, the in vivo normal bladder, and the bladder with implanted tumor. It is obvious that the cellular composition of each of these samples is markedly different. To take these differences (called different experimental conditions) into account, we have applied PCA, a statistical technique for determining the variables in a multidimensional data set. In the present study, both experimental conditions and genes were considered as variables. A set of principal gene components (Figure 1A, red, green, and black) reveals the features of genes that best explain the experimental conditions (Figure 1, A and B: N, B, and C). Namely, the bladder cancer (B) is located in a completely different spatial component compared with the normal bladder (N) or the cultured cells (C). If this difference had only been due to the cellular composition of the experimental conditions, B would be located between N and C. Theoretically, an artificial mixture of cultured cancer cells and normal bladder tissue would be located in the area between N and C. It will be of interest to use laser capture microdissection technique to obtain pure populations of normal epithelia and tumor cells for the gene expression profiling. Unfortunately, it is still difficult to dissect the pure tumor cells because the tumors are heterogeneous, and there are no available staining markers specifically for the tumor cells. The fact that the gene expression profile of the bladder cancer differs from that of the normal bladder and the cultured cancer cells is most likely due to the interactions between the host and the cancer cells. Indeed, morphology (including ultrastructure) of the cancer cells in tissue was different from the cancer cells in the culture (Figures 1B and and4),4), and the bladder with cancer had a high degree of inflammation (see also Figures 1B and and3).3). The difference between in vivo bladder cancers (B) and in vitro bladder cancer cells (C) is unlikely due to the difference between the three-dimensional and two-dimensional cultures. A previous study comparing two-dimensional and three-dimensional cultures of mouse embryonic stem cells failed to find any differentially altered signaling pathways .
In the present study, the differentially expressed genes found in vitro were functionally related to cellular process, cell cycle, cell death, and differentiation, and those genes found in vivo were related to cell communication and cell adhesion. Moreover, the significantly differentiated pathways of this syngeneic orthotopic rat bladder cancer model, like the BBN-induced rat bladder cancer  (own unpublished data), corresponded closely with those of muscle-invasive human tumor. Interestingly, the host response pathways, that is, MHC pathways and related immune cell response pathways, were found to be actually upregulated in this syngeneic and orthotopic rat bladder cancer model (see also Arum et al. ).
Morphologically, we found that a very small proportion of the tumor seemed to survive despite being surrounded by inflammatory cells. Analysis by electron microscopy revealed that these foci of cancer cells (approximately 20–50 cells) were compact and intimately interconnected to each other by a microvillus-formed mesh and that there was no appearance of angiogenesis within these foci, which was confirmed by immunohistochemistry for VEGF. Thus, we propose that the microvillus-formed mesh may enclose the so-called “CSC niche” that excludes lymphocyte infiltration and vascularization, creating a hypoxic and low-nutrient microenviroment within the tumors, thereby resisting chemotherapy and/or radiotherapy.
The concept of niche biology has been proposed as an attempt to understand how stem cells that reside in a special ultrastructure (i.e., “niche”) reproduce or self-renew [15,16]. It has been suggested that the niches are composed not only of stem cells but also of a diverse gathering of neighboring differentiated cell types . It has also been hypothesized that the stem cell niche may be “hijacked” by the CSCs for tumor initiation, progression, invasion, metastasis, and recurrence [18,19]. The present study is certainly not conclusive, but it does lend evidence in support of our hypothesis that indeed a subpopulation of the cancer cells elude immunosurveillance within the microvillus-formed niches and that the putative CSCs may be located within this niche. This is also the case for CSCs as well as CSC niche in general in that conclusive evidence is still lacking. CSCs have been defined on the basis of their ability to seed tumors in animal hosts, to self-renew, and to spawn differentiated progeny (non-CSCs) [20,21]. For instance, a recent Nature Medicine article by Gupta et al. has postulated a novel hypothesis that a dynamic equilibrium may exist between CSCs and non-CSCs within the tumor microenvironment. Moreover, this equilibrium may be shifted in one direction or another by contextual signals within the tumor microenvironment that influence the probability of interconversion between the CSC and non-CSC compartment . On the basis of our observations and in consideration of the hypothesis of Gupta et al. , we suggest that the microvillus-formed niche that harbors both CSCs and non-CSCs cells might act as this compartment and, in effect, prevents tumor eradication by the host immune system.
The model of the CSC niche proposed in the present study differs from those of the previous studies that have arisen primarily from studies in which human tumor cells are transplanted into immunodeficient mice. It is known that xenotransplantation is problematic because the growth of tumor cells requires an intricate network of interactions with diverse support cells (including fibroblasts, endothelial cells, macrophages, mast cells, and mesenchymal stem cells) and many of the cytokines and receptors required for these two-way interactions are incompatible between mice and humans. Furthermore, whether many human tumor cells can home efficiently to an appropriate niche in the mouse is unclear [16,18,19,22,23]. Figure 6 shows a comparison of these two CSC niche models. According to the current model (Figure 6A), anticancer targets include cancer cells per se (not cancer specific), CSCs (not available because of lacking biomarkers), angiogenesis (for large tumors only), adhesion molecules (difficult to increase adhesion for preventing metastasis), and immune system activation (compromised immune response after chemotherapy). According to our model (Figure 6B), the molecular target should be the microvillus-formed CSC niches. In the present study with immunohistochemistry, we found E-cadherin, one of cadherins of the desmosomes, and selectins (including E-, L-, and P-selectins) to be expressed in the epithelia but not in the tumor and desmoglein1 + 2 to be negative. It will be a great challenge in the future to identify the specific molecular markers for the niches, more precisely, the connections, i.e., desmosomes, between the cancer cells rather than the cancer cells and/or the CSCs per se. It will also be of interest to investigate in the future whether this microvillus-formed niche exists in other types of neoplasia. We anticipate that targeting the tumor microvillus-desmosomes could potentially break the niche, allowing host response cells (e.g., lymphocytes, natural killer cells), angiogenesis, and circulating anticancer drugs access to the CSCs, at all stages of the cancer disease.
1This study was supported by grants from the joint program of the Medical Faculty of Norwegian University of Science and Technology and St Olav's Hospital, the St Olav's Hospital Foundation for Cancer Research, the regional platform (Midt-Norge) of Norwegian Functional Genomics, and the Research Council of Norway.