Targeting Slit–Roundabout signaling inhibits tumor angiogenesis in chemical‐induced squamous cell carcinogenesis

Buccal pouch squamous cell carcinoma model.  Sixty outbred male Syrian golden hamsters (Mesocricatus auratus), weighing 60–80 g (each) at the beginning of the experiment, were divided randomly into an experimental group (n = 50) and a control group (n = 10). They were housed under constant conditions (22°C, 12:12 h L : D cycle) and fed with tap water and standard Purina laboratory chow ad libitum. The experimental methods were the same as those described previously.^{( 27} , ^{28 )} Hamsters in the control group were untreated throughout the entire experiment. For experimental animals, the left pouch was painted twice with a 0.5% DMBA–acetone solution (Sigma, St Louis, MO, USA) using a no. 4 sable‐hair brush, and the right pouch was painted twice with acetone alone as the solvent control. Approximately 0.1 mL of the respective solution was twice applied topically to the medial wall of each pouch. Hamsters were treated three times per week (Monday, Wednesday, and Friday) for 10 weeks. Animals in the experimental group were killed simultaneously in groups of 10 by inhalation of a lethal dose of diethyl ether after treatment for 6, 8, 10, 12, or 16 weeks, whereas animals in the control group were killed after treatment for 16 weeks. Appropriate animal care and an approved experimental protocol were ensured through animal handling, and all procedures were conducted in accordance with the regulations of the China National Guide for the Care and Use of Laboratory Animals.

Histology.  Histopathological classification of epithelial lesions in the buccal pouches of hamsters following DMBA treatment was scored as follows: (1) epithelial hyperplasia (n = 8); (2) moderate dysplasia (n = 12); (3) squamous cell CIS (n = 10); and (4) squamous cell carcinoma (n = 24). Normal mucosa (n = 8) from the right buccal pouch treated with acetone only was used as the control. Tissue samples were classified according to the World Health Organization guidelines for histological typing of cancer and precancer of the oral mucosa.^{( 34 )}

Informed consent was obtained from patients with OSCC for the histological examination. The formalin‐fixed, paraffin‐embedded specimens (117 in total) were obtained as surgical or biopsy samples, including normal cheek mucosa (n = 19), epithelial hyperplasia (n = 18), epithelial dysplasia (n = 20), CIS (n = 20), and carcinoma (n = 40). Specimens were collected between September 1999 and June 2003 from the archives of the Oral Pathology Laboratory, Shanghai 9th Hospital, China. Tissues were selected initially on the basis of the original pathological diagnosis. Following hematoxylin–eosin (HE) staining of the tissue sections, they were reevaluated by two experienced pathologists, according to World Health Organization classification guidelines.^{( 34 )} Specimens from patients who had not previously received radiotherapy were included in the current study to avoid potential complications from the pathological changes caused by radiation injury. Five serial sections (4 µm thick) from specimens collected from patients or animals were prepared and used for HE or immunohistochemical staining.

Immunohistochemical staining.  The buccal pouches of hamsters were dissected, excised, and fixed in 10% formalin solution for 24 h followed by paraffin embedding. OSCC specimens were also fixed and paraffin‐embedded as described above. Tissue sections (4 µm thick) were subsequently dewaxed, and endogenous peroxidase was quenched with 3% H₂O₂ in methanol. Prior to staining, non‐specific binding was blocked by incubation with 10% bovine serum albumin in 50 mM Tris‐HCl (pH 7.6), 0.15 M NaCl, and 0.02% NaN₃ (TBS) at 37°C for 30 min. Tissue sections were incubated with 10 µg/mL of a rabbit preimmune IgG, a rabbit polyclonal antibody to von Willebrand factor (vWF; Peninsula Diagnostics, San Carlos, CA, USA), or a rabbit anti‐Slit2 polyclonal antibody,^{( 24 )} and a rabbit anti‐VEGF polyclonal antibody (Peninsula Diagnostics, San Carlos, CA, USA) in TBS containing 1% bovine serum albumin at 4°C overnight. The tissue sections were then treated with a horseradish peroxidase‐conjugated goat antirabbit IgG antibody (Sigma) at 37°C for 1 h. Color development was carried out by incubation with 0.05% 3‐3′diamenobenzidine solution (Sigma) in TBS supplemented with 2 mM H₂O₂ for 3 min. All incubations were followed by three washes in TBS over 15 min. Tissue sections were finally counterstained with hematoxylin. Semiquantitation of vWF, Slit2, and VEGF was determined according to previously published methods.^{( 35} , ^{36 )} Statistical analysis was carried out using the Mann–Whitney U‐test and one‐way analysis of variance (ANOVA). Association among variables was assessed using the Spearman bivariate correlation test. P‐values < 0.05 were considered statistically significant. Results are the mean ± SD of the MVD, or mean ± SE of Slit2 and VEGF expression.

Microvessel density counts.  The MVD counts were determined in terms of the method described previously by Weidner et al.^{( 37 )} After the areas containing dense vascularity, either within the tumor or during different pathological stages, were selected using a microscopic field with ×100 power, all of the brown‐stained vessels were counted under three fields with ×200 power. The mean value of the vessels in the three fields was taken as the MVD count of a tumor. Results are the mean ± SD of the MVD.

R5 treatment protocol.  Eighty male Syrian golden hamsters (6 weeks old) were treated with DMBA as described above. Briefly, the left buccal pouch was painted with a 0.5% solution of DMBA in acetone, and the right buccal pouch was painted with acetone alone for 10 weeks. Animals were weighed, and the size of the tumor was measured twice per week with a microcaliper. The tumor size was calculated by using the formula: volume = 0.52 × length × width².^{( 24 )} In the animal model, the tumor size varied among animals because of individual differences. Of the 80 animals treated with DMBA, 48 hamsters with similar tumor sizes were selected. The rest of the 32 hamsters with too large or too small tumor sizes were eliminated from the experiments. Forty‐eight animals were divided randomly into three groups: R5 treatment group (n = 18), control antibody treatment group (n = 16), and sham group (n = 14).

At the beginning of the fourteenth week (there were 4‐week intervals between the DMBA and R5 treatments), an intraperitoneal injection of R5 or its isotype‐matched control IgG_2b (1.2 mg/twice per week) was given for 4 weeks. For measurement of cell proliferation, intraperitoneal administration of bromodeoxyuridine (BrdU, 50 mg/kg; Sigma) was given 2 h prior to the death of the animals. Blood obtained retrovitreously was used for counting leukocytes. Normal and pathogenic mucosa of the bilateral buccal pouch was dissected and fixed for histological and immunohistologic studies as described above. For measurement of BrdU‐incorporated nuclei, a rat anti‐BrdU antibody (Roche Diagnostics, Basel, Switzerland) was used at a concentration of 5 µg/mL. More than three non‐contiguous, randomly selected fields (at least 2000 cells) in each sample were counted under 400× magnification. The proliferative index (%) was calculated as the number of positively stained cells divided by the total number of tumorous cells counted.^{( 29 )} All data were analyzed with SPSS 10.0 statistical software using Student's t‐test and one‐way ANOVA. P‐values < 0.05 were considered statistically significant. Results are the mean ± SD of the MVD and proliferative index, or the mean ± SE of VEGF expression.

Levels of vWF, Slit2, and VEGF in chemical‐induced carcinogenesis.  We then studied variations in the levels of vWF (an indicator of MVD), Slit2, and VEGF during different pathological stages in the hamster buccal pouch model of squamous cell carcinogenesis induced by DMBA. Tissue samples were stained with HE and classified into different pathological stages, including normal mucosa, hyperplasia, dysplasia, CIS, and carcinoma (Fig. 2a), according to World Health Organization classification guidelines.^{( 34 )} Blood vessels were stained immunohistochemically with an antibody to vWF. The presence of the Slit2 and VEGF proteins was determined immunohistochemically with an anti‐Slit2 antibody and an anti‐VEGF antibody, respectively. Notably, Slit2 was found only in a minority of animals with normal mucosa (1/8 hamsters, 12.5%) and an increasing number of animals with hyperplastic tissue (2/8 hamsters, 25%) and dysplastic tissue (5/12 hamsters, 41.7%). In contrast, Slit2 expression was increased in the tissue of all animals with CIS (10/10 hamsters, 100%), and carcinoma (24/24 hamsters, 100%). Analogously, VEGF expression was increased with varying frequency in tissues from normal mucosa (1/8 hamsters, 12.5%), hyperplasia (1/8 hamster, 12.5% positive), dysplasia (2/12 hamster, 16.7%), CIS (8/10 hamsters, 80%), and carcinoma (22/24 hamsters, 91.7%). Notably, staining for vWF also increased at advanced stages of squamous cell carcinogenesis. Quantitative analysis of our data illustrated that the stage of disease progression correlated positively with the degree of vWF, Slit2, and VEGF levels (Fig. 2b–d; r = 1.0, P < 0.01, for correlations between MVD vs Slit2, MVD vs VEGF, and Slit2 vs VEGF). Our findings indicate that Slit2 is present during multiple stages of DMBA‐induced squamous cell carcinogenesis, and imply that Slit–Robo signaling might play a role throughout disease progression.

Neutralization of the Slit2–Robo1 interaction suppresses tumor angiogenesis and growth.  To explore the pathological significance of Slit2–Robo1 signaling in DMBA‐induced hamster buccal pouch squamous carcinogenesis, we examined whether treatment with R5 inhibited tumor angiogenesis and growth in vivo. Macroscopically, the malignant lesions induced by DMBA were highly dysplastic with the profound architectural and cytological distortion characteristic of human cancer (Fig. 3a). Compared to the isotype‐matched control IgG_2b, intraperitoneal administration of R5 altered the appearance of the papillary protuberance (normally an ulcerative, erosive, reddish, and fragile solid tumor with microhemorrhaging) into a smooth, pale, and encapsulated tumor mass, suggesting a reduced blood supply to the surface of the tumor. Microscopic observation revealed profound cribriform tumor growth with areas of necrosis and prominent desmoplastic stromal reaction suggestive of invasion of the lamina propia (Fig. 3b). Compared with the control IgG group, R5 treatment clearly decreased lesions of blood vessels that were dilated irregularly, swollen, incontiguous, and microhemorrhagic. Immunohistochemical staining with anti‐vWF antibody revealed less neovasculature in cancerous tissue from hamsters treated with R5 compared with hamsters treated with control IgG (Fig. 3c,f). BrdU staining showed reduced proliferation in the R5‐treated tissues compared with control IgG‐treated tissues (Fig. 3d,f). The R5 group also had a significantly smaller tumor volume when compared with the blank and control IgG groups (Fig. 3e). Statistical analysis further demonstrated a reduction of MVD in the R5 group when compared with the MVD in the blank and control IgG groups (Fig. 3f). Notably, leukocytes and VEGF levels were not reduced in the R5‐treated group, thus ruling out the possibility that R5 acts on leukocytes or VEGF. Taken together, our data indicate that Slit–Robo signaling plays a causal role in tumor angiogenesis and growth in chemical‐induced squamous carcinogenesis.

Levels of vWF, Slit2, and VEGF in human OSCC.  To evaluate our in vivo findings in humans, we examined the variation in protein levels of vWF, Slit2, and VEGF in a total of 117 samples of human cheek mucosal squamous cell carcinoma at varying stages of disease progression. The pathological stages were classified into normal mucosa, hyperplasia, dysplasia, CIS, and carcinoma according to World Health Organization classification guidelines (Fig. 4a). Notably, Slit2 was rarely found in the tissue of normal mucosa (1/19 cases, 5% positive). It appeared with increasing frequency in samples of hyperplastic tissue (4/18 cases, 22% positive), dysplastic tissue (7/20 cases, 35% positive), CIS (9/20 cases, 45% positive), and carcinoma (34/40 cases, 85% positive). The basal VEGF level was moderately higher than the Slit2 levels in all disease stages, but a similar VEGF profile was observed (3/19 cases, 16% positive for normal mucosa; 11/18 cases, 61% positive for hyperplasia; 14/20 cases, 70% positive for dysplasia; 17/20 cases, 85% positive for CIS; 37/40 cases, 93% positive for carcinoma). Quantitative analysis of data on the levels of vWF, Slit2, and VEGF revealed a positive correlation between vWF, Slit2, and VEGF levels and the pathological stages of human OSCC (Fig. 4b–d; r = 1.0, P < 0.01, for the correlations between MVD vs Slit2, MVD vs VEGF, and Slit2 vs VEGF). Analogous to VEGF, the importance of Slit2 for tumor angiogenesis and growth is supported by our findings that Slit2 is present with increasing frequency in advanced stages of human OSCC.