Targeted Expression of bcl-2 to Murine Basal Epidermal Keratinocytes Results in Paradoxical Retardation of Ultraviolet- and Chemical-induced Tumorigenesis¹

The first genes to be associated with tumorigenesis, such as ras, c-myc, or p53 (reviewed in Ref. 1) were regulators of cell proliferation. However, it is now recognized that proteins that prevent apoptosis, a physiological form of cell death, are also highly expressed in many tumors (reviewed in Ref. 2, 3, 4), suggesting that they arise as a result of imbalanced rates of cell proliferation and cell death. Bcl-2, a M_r 27,000 intracellular membrane-bound protein (5), was the first proto-oncogene to be described, the oncogenic activity of which seemed to derive from its ability to inhibit cell death rather than stimulate proliferation (6). Initially found to be constitutively activated in the majority of non-Hodgkin’s lymphomas (7), this protein has since been detected in a wide variety of malignancies (8, 9, 10, 11, 12, 13), including those of the skin. It is highly expressed in melanoma (14, 15) and Merkel cell carcinoma (16), as well as in both malignant and premalignant diseases of keratinocyte origin, such as basal cell carcinoma, basaloid SCC³, and Bowen’s Disease (17, 18). Its oncogenic potential in the context of lymphoma is supported by a number of transgenic mouse models (19, 20, 21), and in the skin, overexpression of bcl-2 in murine keratinocytes of both basal and suprabasal layers led to an increased susceptibility to chemically induced tumors (22). In keratinocytes of normal human skin, however, bcl-2 expression is low and restricted to follicular and interfollicular basal cells (15, 18, 23, 24, 25, 26, 27, 28). Thus, we have developed transgenic mice in which bcl-2 is highly expressed in basal epidermal keratinocytes only (29), under the control of the human keratin 14 promoter (30), and have studied the contribution of this protein to the formation of epidermal tumors induced by either chemical carcinogens or UVB. Our results indicate that constitutive expression of bcl-2 in the epidermal basal layer does not predispose to an increased susceptibility to experimentally induced tumors; rather, it delays the rate of appearance (promotion) or increases the latency period (initiation) of chemically or UVB-induced tumors, respectively.

FVB/N mice transgenic for human bcl-2 under the control of the human keratin 14 promoter (K14/bcl-2 mice) have been described previously (29). Mice transgenic for the v-Ha-ras oncogene under the control of the fetal mouse z-globin promoter (31) were purchased from Charles River Laboratories (Wilmington, MA) and crossed to the K14/bcl-2 transgenic mice at the Brigham and Women’s Hospital animal facility, yielding mice hemizygous for both transgenes. In preliminary experiments, because both founder lines 1 and 7 (29) were indistinguishable with respect to transgene copy number and antiapoptotic properties, only line 1 was used for the carcinogenesis studies.

Hamster antihuman or antimurine bcl-2, clones 6C8 and 3F11, respectively, were purchased from PharMingen (San Diego, CA). Polyclonal rabbit anti-K14 was a kind gift from Dennis Roop (Baylor College of Medicine, Houston, TX), and a mouse monoclonal antihuman CK14, which cross-reacts with mouse cytokeratin 14, was purchased from Neomarker (Fremont, CA). Finally, anti-vimentin antibody, clone VIM 3B4, was obtained from Boehringer-Mannheim (Philadelphia, PA). T lymphocytes were detected with the anti-CD3 monoclonal antibody MCA1477 (Serotec, Oxford, United Kingdom). The appropriate secondary biotinylated reagents for mouse and hamster primary antibodies were from Amersham (Little Chalfont, United Kingdom).

Sections (5 mm) frozen in OCT (Miles, Elkhart, IN) and acetone-fixed, or formalin-fixed, paraffin-embedded sections were used for all of the stainings. For tumor grading, 5-mm thick paraffin sections were stained with H&E according to standard procedures. Staining for bcl-2 was performed on either cryo- or paraffin-embedded sections and, for K14 and vimentin, on paraffin-embedded sections. For K14 and bcl-2, antigen retrieval was achieved by microwaving for 20 min in 0.01 m citrate buffer (pH 6) and, for vimentin, by protease digestion at 37°C for 30 min. All of the sections were neutralized for endogenous peroxidases and blocked with 10% goat serum for 30 min. Subsequently, the sections were covered with primary antibody for 2 h at room temperature and washed with Tris-buffered saline (0.05 m Tris, pH 7.5, in a 0.9% sodium chloride solution), and antibody binding was detected by incubation with the appropriate biotinylated goat secondary reagents, peroxidase ABC kits (Vector Laboratories, Burlingame, CA or DAKO, Carpinteria, CA), and 3,3′-diaminobenzidine substrate (DAKO) for visualization. Counterstaining was with hematoxylin. Antibody dilutions were as follows: antihuman bcl-2, 1:50; antimurine bcl-2, 1:25; anti-K14, 1:100; and anti-vimentin, 1:50.

Tumors were graded on H&E-stained sections based on the system developed by Ruggeri et al. (32) but grouped as highly (grade I) or moderately (grade II to III) differentiated, undifferentiated (spindle cells; grade IV), or a mixture of completely undifferentiated areas together with areas with some degree of differentiation in the same tumor (mix).

Chloroacetate esterase staining to visualize mast cells and neutrophilic granulocytes and immunohistochemistry to identify T-cell infiltrates of the UVB-induced tumors and peritumoral areas have been described in detail elsewhere (33).

Female K14/bcl-2 and FVB/N wild-type mice (n = 20 of each) with an average age of 12 weeks (range 10–13 weeks) were irradiated on their shaved backs with a bank of Philips Ultraviolet B TL40W/12 sunlamps (Philips, Hamburg, Germany) with doses of UVB increasing from 2.5 to 10 kJ/m² for a total of 6 months (34). Tumor development was recorded as before for an additional 4 months. Mice were sacrificed once tumors reached 7 × 7 mm in size, and tumor samples were fixed in paraformaldehyde, embedded in parablast, and analyzed for morphology, K14, bcl-2, and vimentin expression. The degree of inflammation surrounding each tumor was assessed by an inflammation score from one to five, and cellular infiltrates were enumerated by chloroacetate esterase staining to identify mast cells and neutrophils and anti-CD3 immunohistochemistry for T cells as described (33). Using a 40× lens, five 0.2-mm² fields/tumor were counted “blind” and averaged.

Female mice, 20 FVB/N and 25 K14/bcl-2 transgenics, with an average age of 15 weeks (range, 10–20 weeks) were induced for carcinomas as described previously (35). Briefly, mice were initiated on their shaved backs by application of a single dose of 25 μg of DMBA (Sigma Chemical Co., St. Louis, MO) dissolved in acetone and, after 1 week, promoted at weekly intervals with 5 μg of PMA (Sigma) also in acetone for 20 weeks. Papilloma development was monitored from week 10 until week 51, as described (35). When the carcinomas reached a size of about 10 mm in diameter, the mice were sacrificed. All of the tumors and carcinomas, both primary and lymph node metastases as well as some papillomas from each mouse, were excised and analyzed histologically. In a second experiment, hemizygous bcl-2 transgenic mice were bred to homozygous v-Ha-ras mice, yielding populations of v-Ha-ras mice and v-Ha-ras/bcl-2 crosses, hemizygous for both genes in the first generation. PMA (5 μg) diluted in ethanol was applied to the shaved dorsal surface of male and female v-Ha-ras/+, bcl-2/+, and v-Ha-ras/bcl-2 crosses (four to nine mice each) 9 to 16 weeks of age (average, 12 weeks) 3 times/week for 8 weeks, and papilloma development was monitored as above from weeks 8 to 38.

Statistical analyses were performed as before (34). The Kaplan-Meier method was used to calculate the probability of tumor development. Statistical differences for the development of tumors between the two strains of mice in each experiment were determined by using a log-rank test (36) or Wilcoxon two-sample test, and the differences in tumor latent periods were analyzed by a Mann-Whitney U test. Finally, the differences in numbers of cells in the tumor infiltrates and surrounding tissues were analyzed by Student’s t test.

We have shown previously that apoptosis after UVB irradiation was impaired in K14/bcl-2 mice (29). Therefore, we predicted that, as a result of increased survival of mutated epidermal cells, K14/bcl-2 mice would demonstrate an increased susceptibility to UVB-induced carcinogenesis. Surprisingly, however, UVB-irradiated transgenic mice were significantly retarded in tumor formation (Fig. 1; tumor latency was about 175 days for the transgenics, compared with about 100 days for controls). The rate of tumor formation in the wild-type mice was biphasic, being initially fairly low, but with a sudden increase in tumor bearers at about 160 days of irradiation. In contrast, K14/bcl-2 transgenics retained a constant and lower overall rate of tumor development, and 3 of 20 mice remained tumor-free throughout the whole observation period (P < 0.0001).

The numbers and location of tumors was similar in both mouse lines (Table 1). Although more ear than back tumors were observed, they appeared at roughly the same time as back tumors in both lines. Bcl-2 expression was restricted to the basal layers, whereas K14 was expressed throughout the differentiated carcinomas (data not shown). None of the spindle cell tumors induced by UVB exhibited K14 or bcl-2 immunoreactivity but stained uniformly with vimentin (data not shown).

Because in the skin the development of tumors seems to be strongly influenced by the immune system, we analyzed all of the UVB-induced tumors and peritumoral areas qualitatively for degree of inflammation and quantitatively for macrophage, neutrophilic granulocyte, and CD3+ T-cell infiltrates. Although there seemed to be more Mast cells (60/mm² compared with 45/mm²) and neutrophils (75/mm² compared with 56/mm²) infiltrating the tumors of K14/bcl-2 mice compared with wild type, the differences were not statistically significant. Furthermore, there was no significant difference in the inflammation score or in the numbers of T cells located within the tumors. Finally, the three cell types investigated were equally represented in the peritumoral areas in both transgenic and wild-type mice.

To investigate whether the paradoxical retardation of tumor formation associated with overexpression of bcl-2 would be reflected in other carcinogenesis protocols, we subjected K14/bcl-2 and wild-type mice to classical two-stage chemical carcinogenesis. The numbers of papillomas/mouse were counted from weeks 10 through 25 after the beginning of promotion. Lesions were recorded when they were 1 mm or more in diameter and were present for 2 or more consecutive weeks. Papillomas began to appear at week 10 after initiation in both mouse lines (Fig. 2 A), but K14/bcl-2 mice were retarded in further papilloma development, with 50% of the animals presenting with lesions at week 15, compared with week 12 for nontransgenic animals. At week 25, all of the controls had developed papillomas, whereas 2 of the 25 transgenic animals (8%) were still free of lesions. However, the probability of developing papillomas over the whole 25-week observation period was not significantly different between the two strains.

The range of papilloma number (1 to 28 for the transgenic strain and 1 to 29 for the controls) and average number of papillomas/mouse were identical for both strains while PMA was being applied (Fig. 2 B). After promotion was terminated, the mean numbers of lesions/mouse in nontransgenic animals declined, whereas in the transgenic line, the mean papilloma number stayed constant. Because of the large variation of papilloma number/mouse, however, these differences do not achieve statistical significance (Wilcoxon two-sample test).

Analysis of papilloma size (Fig. 2,C) shows that, although the rate of increase in mean numbers of large papillomas (2 to >4 mm) was the same in both strains, beginning at week 18, the transgenic strain had a tendency to more small papillomas (<1–2 mm). These continued to multiply for several weeks after the termination of PMA application, whereas they stayed constant in the controls. Although the difference in numbers of small papillomas between the strains did not achieve statistical significance, Ps declined from 0.5 to 0.19 for <1- to 2-mm papillomas (Wilcoxon two-sample test) from week 20 to week 25. Because the numbers of large papillomas continued to increase, the increase in the numbers of small papillomas must have been attributable to the generation of new lesions and not to size regression of existing lesions. Conversion to carcinomas began at week 25; however, high-level expression of the transgene did not affect the conversion rate (Fig. 2 D).

Carcinomas and lymph node infiltrates were examined histologically and graded according to type and degree of differentiation (Table 2). The percentage conversion rate of papillomas to carcinomas was somewhat greater in the nontransgenic strain, but the difference was not statistically significant. The degree of differentiation of primary tumors was similar in both strains (6 of 11 were highly differentiated in the transgenic strain, compared with 4 of 10 in the nontransgenic strain). However, of the carcinoma-grade lymph node infiltrates, 4 of 6 were highly differentiated SCCs in the transgenic strain, compared with 0 of 6 in the control mice. Thus, all of the lymph node metastases in wild-type mice were of the undifferentiated, spindle cell type.

Fig. 3 shows the expression of the bcl-2 transgene (Fig. 3, A, C, E, and G) and K14 (Fig. 3, B, D, F, and H) in a benign papilloma (Fig. 3, A and B) and in a selection of malignant tumors (Fig. 3, C–H) from transgenic mice. High-level expression of transgenic bcl-2, confined largely to the basal layers, was seen in the papilloma (Fig. 3,A), in all of the squamous cell tumors analyzed (7 of 11; Fig. 3,C), and in some areas of 5 of 11 spindle cell tumors (Fig. 3,E). No staining could be detected in these tissues with the control hamster immunoglobulin (data not shown). All of the areas of tumor that stained with anti-bcl-2 antibody also stained with K14 antibody (compare Fig. 3, C and D, and Fig. 3, E and F, the latter being part of the same tumor as depicted in Fig. 3, C and D), demonstrating that transgenic bcl-2 protein expression is tightly coupled to K14 promoter activity. The K14 reactivity of these spindle cell tumors further indicates an epithelial origin. The remaining 6 of 11 spindle cell tumors were nonreactive with bcl-2 (Fig. 3,G) and cytokeratin 14 antibodies (Fig. 3,H). Despite significant hyperplasia in the epidermis above the spindle cell tumor, with the exception of isolated cells, bcl-2 expression remains essentially confined to the basal layer (Fig. 3,G), whereas CK14 (Fig. 3, B and H) is found throughout the whole epidermis, probably because of the inherent stability of the K14 protein (37). Only 1 of 10 spindle cell carcinoma samples from nontransgenic tumors showed evidence of cytokeratin 14 staining. All of the papillomas and squamous cell tumors examined stained positive with cytokeratin 14 antibodies in both transgenic (Fig. 3, B and D) and nontransgenic (data not shown) mice. Selected tumors representing the various grades of differentiation were stained with antimurine bcl-2 antibody. High-level staining was observed in cells infiltrating the undifferentiated tumors, in the dermal papillae of all of the sections examined, and in isolated cells of the epidermal basal layer. Weaker staining was found in the hair follicle outer root sheath, bulge, and some sections of interfollicular epidermis (data not shown). Tumor tissue of all of the grades evidenced pale, diffuse staining, but this was probably nonspecific, because it was also present in the immunoglobulin controls. Thus, as is the case with human SCC, these murine SCC do not express bcl-2.

Autopsies of some mice from both strains revealed that the kidney, lung, liver, brain, thymus, and ovary were normal. Mice bearing large carcinomas suffered from splenomegaly, but carcinoma-grade infiltrates were only found in lymph nodes adjacent to the carcinomas (usually the inguinal lymph nodes). These occasionally reached about 1 cm in diameter.

Topical application of DMBA to mouse skin induces mutations in codon 61 of the ras oncogene (38) in a certain percentage of cells. To investigate the effect of overexpression of bcl-2 in a setting in which ALL cells carry a ras mutation, we crossed K14/bcl-2 mice with mice that express the v-Ha-ras oncogene under the control of the z-globin promoter (31). The v-Ha-ras mice develop epidermal papillomas in response to stimuli such as abrasion or chemical tumor promotion and, thus, exhibit features of an initiated phenotype. First generation offspring of hemizygous bcl-2 and homozygous v-Ha-ras mice (ras/bcl-2), their littermates hemizygous for v-Ha-ras (ras/+), and K14/bcl-2 mice were subjected to twice weekly PMA promotion, as described in “Materials and Methods.” The numbers of papillomas and conversion frequency to carcinomas were recorded as before.

Papillomas first appeared between 3 and 4 weeks after the start of PMA application in all of the experimental groups except female ras/bcl-2 mice (Fig. 4,A). The tumors became so large in all of the males (data not shown) that they had to be sacrificed at week 13 for humane reasons. In accordance with the previous two experiments, both latency and rate of papilloma development were retarded in ras/bcl-2 females compared with ras/+ female controls (Fig. 4 A). K14/Bcl-2 mice did not develop papillomas under these conditions and were not analyzed further.

PMA application was discontinued at week 8, and the female mice were monitored for a total of 38 weeks. The early sacrifice of the males in this experiment resulted in only four female mice remaining in the ras/+ control group, and three of these each had 20–40 papillomas/mouse. In comparison, six of nine female ras/bcl-2 mice developed seven or fewer papillomas, whereas the remainder presented with 23–38 lesions at the time of maximal papilloma development (weeks 13–15; Fig. 4 B). In contrast to the previous experiment in which papillomas on nontransgenic mice began to regress immediately, both ras/bcl-2 and ras/+ mice continued to develop new papillomas for an additional 6–8 weeks, and the ultimate rate of papilloma decline was indistinguishable in these two strains (data not shown).

There was a considerable range in papilloma size, especially in the ras/bcl-2 group (3–20 mm, largest diameter; compared with 2–7 mm for the ras/+ group at week 27), and the mean papilloma size was consistently larger in the ras/bcl-2 mice (Fig. 4 C). Papillomas were altogether much larger in all of the mice expressing v-Ha-ras compared with the mice from the previous experiment (3–5 mm in the former compared with 2 mm in the latter at week 25 after start of promotion). The time of conversion to carcinomas (week 32 compared with week 20) was much later and the conversion rate much lower in this experiment compared with the previous one. All of the tumors that developed, including the one lymph node infiltrate, were of the spindle cell type, and none stained with anti-keratin 14 antibody (data not shown).

The tumor induction protocols used in this study use carcinogenic agents that, in addition, are potent inducers of apoptosis. The ability of antiapoptotic bcl-2 family members to protect keratinocytes from potentially carcinogenic apoptotic stimuli is well documented. In vivo, overexpression of either bcl-2 or bcl-x_L in murine epidermis dramatically reduced the number of UVB-induced sunburn cells (22, 29, 39), whereas in vitro, bcl-2-overexpressing keratinocytes were more resistant to DMBA- and PMA-induced cell death (22).

We had hypothesized that keratinocytes that were prevented from entering the apoptotic pathway in response to a DNA-damaging signal would remain viable as initiated cells, thereby precipitating tumor formation. Surprisingly, however, overexpression of bcl-2 in the epidermal basal layer resulted in a retardation of tumor formation, an effect that was most pronounced in the UVB induction protocol (Fig. 1, Fig. 2,A, and Fig. 4 A). A similar antitumor activity of bcl-2 has recently been reported for DMBA-treated mice that overexpress this oncogene in the epithelium of mammary glands (40).

We have demonstrated previously (29) that the bcl-2 transgene in these mice displayed the expected antiapoptotic activity after UVB irradiation. Antiapoptotic activity was also evident in the first chemical carcinogenesis experiment, where papillomas continued to appear in transgenic but not control mice after the termination of PMA application (Fig. 2 C, numbers of both large and small papillomas increase, suggesting that the small papillomas are new ones, rather than regressing large ones).

To investigate the mechanism underlying the delay in tumorigenesis in the transgenic mice, we explored the possibility that bcl-2 might indirectly influence the cytokine microenvironment in the peritumoral area, leading to enhanced infiltrates and decreased tumor growth. However, we found no significant differences in the degree of inflammation in the numbers of infiltrating mast cells, neutrophilic granulocytes, or CD3+ T cells in the UVB-induced tumors. Thus, we judged this putative mechanism as unlikely to retard tumor formation in the K14/bcl-2 mice.

In addition to its antiapoptotic properties, bcl-2 is able to regulate the cell cycle in a number of systems (41, 42). In particular, HL60 cells transfected with bcl-2 withdraw into G₀ in response to treatment with TNF-α (43). Because TNF-α has been shown to be secreted by keratinocytes in response to both UVB irradiation (44) and PMA application (45), it is conceivable that after such stimulation bcl-2-overexpressing keratinocytes are delayed in their entry into the cell cycle by a TNF-α-dependent mechanism, thereby retarding the inheritance of mutations and subsequent tumor appearance. Secondly, bcl-2 has been reported to exert antioxidant activity (46, 47, 48). ROS have been implicated in the cellular action of both TNF-α (induced by UVB) and PMA (49). High-level expression of bcl-2 may reduce the concentration of ROS, thereby reducing ROS-mediated DNA damage, resulting in retarded tumor formation. Additional investigations will be necessary to clarify these issues.

Oncogenesis appears to necessitate multiple genetic defects, often in growth regulatory genes (e.g., reviewed in Ref. 50). Although not so well documented, bcl-2 has also been shown to enhance the tumorigenic potential of the oncogenes c-myc (51) and c-Ha-ras (52), possibly by counteracting their apoptotic activity while leaving the mitogenic properties intact (53, 54, 55, 56). Furthermore, mutated ras can specifically up-regulate bcl-2 expression in human melanoma cell lines thereby potentially protecting them from apoptosis (57). Finally, while this manuscript was in preparation, a correlation was found between bcl-2 overexpression and H-ras mutation in naturally occurring murine hepatocellular proliferative lesions (58). However, because we found no accelerated tumor development in double transgenic K14/bcl2/ras mice compared with ras/+ controls, there must still be other factor(s) contributing to chemical-induced tumorigenesis in epidermal keratinocytes.

The majority of UVB-induced tumors (70–75%; Table 1) were composed solely of highly undifferentiated cells that were completely bcl-2 and/or K14 negative and vimentin positive (data not shown). This staining is suggestive of dermally derived fibrosarcomas, which are difficult, however, to distinguish microscopically from highly undifferentiated epidermal tumors that have lost keratin and gained vimentin expression (Ref. 59 and references therein). However, all of the tumors in the transgenic mice arose much later than in the wild type. If the spindle cell tumors did originate from dermal cells, one would have to speculate that bcl-2, overexpressed in the basal layer, somehow also retards dermal tumor formation. This would have to involve a secondary mechanism, because bcl-2 is a nonsecreted, membrane-bound protein (5). Thus, we feel that it is more plausible that the spindle cell tumors are highly dedifferentiated cells of epidermal origin that have undergone an epithelial-mesenchyme transition, with the concomitant change in gene expression. Indeed, about half of the spindle cell tumors arising in DMBA/PMA-treated K14/bcl-2 mice and 1 in 10 in wild-type mice exhibited areas of bcl-2 and/or K14 immunoreactivity, indicating that at least some of these were epidermally derived tumors. Whether the increased numbers of K14/bcl-2-staining spindle cells in the tumors from transgenic mice indicate an effect of the transgene on the degree of differentiation of the tumors or on gene regulation during epithelial-mesenchyme transition cannot be ascertained from the present data.

Two transgenic models similar to ours, in which the antiapoptotic bcl-2 family members bcl-2 or bcl-x_L were overexpressed in murine epidermis (22, 60), displayed, in contrast to our findings, enhanced formation of chemically induced tumors. In both of these models, the antiapoptotic transgenes were expressed throughout the epidermis, whereas our mice expressed bcl-2 in the basal layer only. The notion that deregulated activity of bcl-2 family members enhances tumor formation when overexpressed in the upper layers of epidermis but retards tumor formation when overexpression is restricted to the basal layer is intriguing. The implication may be that the relative importance of cell cycle regulation/DNA damage protection and antiapoptosis activity varies, depending on the level of differentiation within the tissue. In the basal layer, where the initiating events are thought to occur (61), overexpression of bcl-2 seems to delay tumor formation, possibly because of its cell cycle regulatory function and ability to reduce DNA damage by virtue of its antioxidant properties. If cells do become initiated and move into the suprabasal layers in the course of normal skin cell turnover, in our model they will no longer be protected from apoptosis because bcl-2 is restricted to the basal layer. Thus, they would be eliminated at the normal rate, and no tumor enhancement is observed. However, if the initiated cells continue to be protected from PMA or UVB-induced apoptosis because of deregulated bcl-2/bcl-x_L expression in the suprabasal compartment as well, they will survive and continue to accumulate additional mutations. If these cells have also acquired mutations in their growth regulatory genes, then the ultimate outcome may be enhanced tumor formation. Indeed, that cells that have left the basal layer are able to be induced to form tumors has been elegantly shown by the work of Bailleul and et al. (62), who found that mice expressing a ras oncogene under the control of a suprabasal promoter developed papillomas in response to mechanical irritation.

In summary, overexpression of bcl-2 in the basal layer of epidermis, either alone or together with the v-Ha-ras oncogene, paradoxically delays rather than accelerates tumor formation, particularly those induced by UVB. The mechanism(s) underlying this paradox deserve(s) further investigation.

We thank Dr. Ifor Williams for invaluable help with the chemical carcinogenesis experiments, Dr. Thomas Lang (Institut für Medizinische Statistik der Universität Wien, Vienna, Austria) for the statistical analyses, Dr. Angelika Url (Veterinärmedizinische Universität, Vienna, Austria) and Dr. Johannes Pammer (Department of Clinical Pathology, AKH, Vienna, Austria) for confirming the morphological analyses of the tumors, and Barbara Lengauer and Robert C. Kubitza for excellent technical assistance.