The Toho Study Group for Cancer Biological Behavior is structured by Kazuto Yamazaki, MD (Department of Pathology, Saiseikai Central Hospital), Reiko Shimokawa, MD (Department of Pathology, Omori Red Cross Hospital), Takahiro Suzuki, MD and Mieko Fujiwara, MD (Department of Pathology, St. Luke's International Hospital), Yoshikazu Kuroda, MD (Department of Gastroenterological Surgery, Kobe University Graduate School of Medicine), Akihiro Ichinose, MD (Department of Plastic Surgery, Kobe University Graduate School of Medicine), Hideko Kiguchi, MD (Department of Pathology, Saiseikai Kanagawa-ken Hospital), Kentaro Ogata, MD (Department of Pathology, Kawasaki Municipal Ida Hospital), Hitoshi Niino, MD (Pathology Division, National Hospital Organization Yokohama Medical Center), and Yoshiro Ebihara, MD (Pathology Division, Toda Chuo General Hospital) in addition to the Department of Pathology, Toho University School of Medicine.
Early colorectal cancer (ECC) is curable by endoscopic local resection; however, 10% of patients with ECC exhibit lymph node (LN) metastasis. In the current study, accurate predictors for LN metastasis in patients with ECC were examined by using immunohistochemistry with the lymphatic endothelial hyaluronan receptor 1 (LYVE-1) antibody to discriminate between lymphatics and blood vessels.
Colorectal tissue specimens obtained from 71 patients with ECC, including 28 patients with regional LN metastasis, were immunostained with antibodies against LYVE-1, β-catenin, claudin-3, claudin-4, and cytokeratin. The significance of the histopathologic variables for LN metastasis in ECC was investigated on the basis of specific histopathologic parameters.
Lymphatic invasion confirmed by LYVE-1 immunohistochemistry was observed mainly in the submucosal area around the primary tumor and rarely was observed in the tumor. Expression patterns of β-catenin, claudin-3, and claudin-4 in cancer cells at the invasive front were irrelevant to LN status. Tumor size, depth of invasion, histologic tumor type, budding formation, and lymphatic invasion were statistically significant to LN status in univariate analysis; however, only 2 factors—lymphatic invasion and budding formation at the invasive front—were independent predictors of LN metastasis in ECC.
Early colorectal cancer (ECC) is defined as adenocarcinoma that is restricted to the mucosa or that invades the submucosa, which includes ECC classified as tumor in situ (Tis) and T1 according to the Tumor, Lymph Node, Metastasis (TNM) classification.1 Patients who have ECC confined to the mucosa do not usually show either lymph node (LN) metastasis or distant metastasis2, 3; however, 10% of patients who have ECC with submucosal invasion (T1) exhibit LN metastasis.4, 5 This suggests that the submucosal invasion of cancer may be the first step in metastasis to the LN. Therefore, currently, endoscopic mucosal resection is considered the first choice as curative treatment for ECC. However, with this method, an additional radical resection of the affected colon or rectum is necessary for patients who have specific risk factors for LN metastasis after histopathologic examination. Therefore, accurate identification of the risk factors for LN metastasis and careful examination of resected colorectal tissues can play crucial roles in reducing cancer recurrence.
Metastasis of colorectal cancer to the regional LN results from cancer cell invasion by the lymphatic route; however, sometimes, it is difficult to discriminate accurately between lymphatics and blood capillaries by using conventional staining methods. However, antibodies against lymphatic endothelial-specific proteins, such as lymphatic endothelial hyaluronan receptor-1 (LYVE-1), podoplanin, prox-1, and D2-40, have been developed.6, 7 Recently, a group of investigators also reported the development of a new type of polyclonal antibody against LYVE-1 that can distinguish lymphatics from blood capillaries in paraffin-embedded tissue sections by using immunohistochemistry.8 In fact, LYVE-1 immunohistochemistry was useful for the detection of lymphatic invasion in formalin-fixed, paraffin-embedded tissue sections of gastric cancer.9 In ECC, to our knowledge, there have been no reports describing the significance of lymphatic invasion in LN metastasis using lymphatic-specific markers to discriminate accurately between lymphatics and blood capillaries.
Conversely, there are various risk factors for LN metastasis in patients with cancers of the bowel, including the role of adhesion molecules expressed by cancer cells, such as β-catenin and the claudins.10–13 It has been postulated that, in addition to metastasis, the loss of β-catenin expression results in the loss of cell-cell adhesion and contributes to tumor cell invasion in the stoma and vessels.10 The relation between the expression of claudins in cancer cells and metastatic behavior has been controversial; however, it has been determined that several members of the claudin family, which includes 24 different types of claudins, are predictors of cancer recurrence and metastatic potential.11–13 However, in ECC, to our knowledge, no reports have demonstrated accurate predictors of LN metastasis in a comprehensive analysis using immunohistochemistry for both lymphatic markers and cell adhesion molecules. In the current study, we investigated predictors of LN metastasis in ECC through a comprehensive analysis of 71 patients with ECC, including 28 patients who had LN metastasis, by using immunohistochemistry against LYVE-1, β-catenin, claudin-3, claudin-4, and cytokeratin in addition to the conventional examinations.
MATERIALS AND METHODS
Colorectal tissue specimens were obtained from 28 patients who had undergone curative resection for colorectal cancer at Toho University Medical Center, Saiseikai Central Hospital, Saiseikai Kanagawa-ken Hospital, Omori Red Cross Hospital, Kobe University School of Medicine, St. Luke's International Hospital, Kawasaki Municipal Ida Hospital, Toda Chuo General Hospital, National Hospital Organization Yohohama Medical Center, and Hiratsuka Municipal Hospital from 1990 to 2006. The 28 histologic specimens exhibited submucosal invasion and had regional LN metastasis, and this group was defined as the LN-positive group. Another 43 patients with colorectal cancer who had submucosal invasion but no metastasis to regional LN were used as the LN-negative group. They had also undergone curative resection at the same hospitals from 1990 to 2006 and were selected from 187 patients with ECC as an age- and sex-matched control group. We defined curative resection as the removal of gross cancer and the demonstration of tumor-negative surgical margins by histopathologic examination of the total circumference. No patients received preoperative chemotherapy, and all patients were free of distant visceral metastases. Written informed consent to use colorectal tissue specimens for this study was provided by each patient at each medical institute. The current study was approved by the Committee on Ethical Standards of Toho University School of Medicine. A summary of all patients who participated in this study is provided in Table 1.
Table 1. Summary of Patients Observed in the Current Study
The surgically resected colorectal tissues were opened longitudinally and pinned onto a cork board. After fixation in 10% formalin, each tumor was cut into 4-mm slices parallel to the major axis of the tumor. The greatest dimension of this major tumor axis was defined as the tumor size for each patient. The number of tissue slices was 1 or 2 for each tumor, because tumor sizes were relatively small. All sliced tissues were embedded in paraffin and cut into 3-μm-thick sections. These thin sections were treated by double staining with Victoria blue (VB) and hematoxylin and eosin (H&E) dyes. VB staining was used to aid in the identification of blood vessel structures, especially with regard to venous invasion of cancer, because elastic fibers, including elastic lamina of blood vessel walls, were stained blue.
The sections obtained from the paraffin-embedded tissues were immunostained with rabbit polyclonal LYVE-1 antibody (1:100 dilution),8 mouse monoclonal von Willebrand factor (vWF) antibody (1:50 dilution; DakoCytomation, Carpinteria, Calif), mouse monoclonal β-catenin antibody (1:50 dilution; clone CAT-5H10; Zymed Laboratories, South San Francisco, Calif), rabbit polyclonal claudin-3 antibody (1:100 dilution; Zymed Laboratories), mouse monoclonal claudin-4 antibody (1:200 dilution; clone 3E2C1; Zymed Laboratories), and mouse monoclonal cytokeratin antibody (1:100 dilution; clone MNFIIb; DakoCytomation). Immunostaining with the LYVE-1 antibody was described in a previous report.9 In summary, the sections were pretreated with citrate buffer solution (10 mmol/L) for 15 minutes at 95°C and then with Proteinase K (DakoCytomation) for 3 minutes at room temperature. After washing, the sections were incubated with LYVE-1 antibody for 1.5 hours at room temperature and treated with the CSA II kit (DakoCytomation) according to the manufacturer's instructions. The immunoproducts were examined by treating slides with diaminobenzidine tetrahydrochloride followed by counterstaining with hematoxylin.
For immunohistochemistry with vWF, β-catenin, the claudins, and cytokeratin, the sections were pretreated with citrate-buffer solution for 15 minutes at 95°C. After washing, they were incubated with the primary antibodies for 2 hours at room temperature; then, they were treated by using the same methods described above for LYVE-1.
Tumor type was assessed by examining sections that were stained with VB and H&E. The tumors were divided into 2 groups—polypoid and nonpolypoid—according to the classification proposed by Shimoda et al.14 Polypoid tumors exhibit intramucosal, polypoid proliferation of adenocarcinoma; whereas nonpolypoid tumors have no intramucosal protuberant growth and exhibit sessile, flat, or ulcerated growth. All specimens that were used in the current study were identified histopathologically as adenocarcinoma. The grade of differentiation that occupied the largest area in the mucosa or superficial region (histologic type of mucosa) was assessed according to the World Health Organization (WHO) classification for tumors of the colon and rectum15 as either low grade or high grade. The same method also was used to assess the grade of differentiation that occupied the largest area at the invasive front of the submucosa (histologic type of invasive front). The “invasive front” in the current study was defined as all regions of the border area between the primary tumor and interstitium in the submucosa. According to this classification, well-differentiated or moderately differentiated adenocarcinoma was defined as low-grade carcinoma, and poorly differentiated or undifferentiated carcinoma was defined as high-grade carcinoma. Mucinous carcinoma was classified as high-grade carcinoma according to the WHO classification.
The existence of budding at the invasive front was determined according to the definition published by Morodomi et al.16 by examining sections with double staining for VB and H&E dyes and with cytokeratin immunostaining. “Budding” is a cancer cell cluster that is immunopositive for cytokeratin and is composed of ≤5 cancer cells or undifferentiated cells, which are isolated without a distinct structure and appear as a bud from a large cancer gland. The sections that had >4 budding formations at the invasive front under light microscopy at ×400 magnification were defined as budding-positive sections. The depth of cancer invasion was measured by using an optical micrometer for light microscopy and was calculated as the distance from the muscularis mucosa to the deepest site of the tumor in the submucosa. When the muscularis mucosa had vanished completely because of ulceration, the depth of tumor invasion was defined as the distance from the ulcer surface to the deepest site of the tumor. Venous invasion was determined by examining sections with double staining for VB and H&E dyes and with vWF immunostaining.
The presence of lymphatic invasion was confirmed by examining sections with LYVE-1 immunostaining. Lymphatic invasion was considered to be present when at least 1 LYVE-1-positive vessel invaded by cancer cells was identified. Immunopositivity for β-catenin, claudin-3, and claudin-4 in the tumor was determined by detailed examination of all areas of the invasive front. The specimens that had nuclear expression of β-catenin in >10% of tumor cells at the invasive front were defined as β-catenin-positive, and the specimens that had cell-membrane expression of claudins in >30% of cancer cells at the invasive front were defined as immunopositive for both claudins.
Tumor size and histopathologic variables were compared statistically between the LN-positive and the LN-negative groups. The Student t test was used for statistical comparisons of the variables tumor size and depth of invasion; the Mann-Whitney U test was used to compare the variables primary tumor location and histologic grade; and the chi-square test was used to analyze the variables budding, lymphatic invasion, venous invasion, and immunoexpression of 3 proteins.
Findings of Lymphatic Invasion and Venous Invasion
LYVE-1 immunohistochemistry clearly demonstrated lymphatic distribution in the normal colonic regions, and lymphatics were abundant in the upper submucosal layer just under the muscularis mucosa but were rare in the mucosa (Fig. 1). Lymphatic invasion by cancer cells was observed in 22 tumors (31%), including 18 LN-positive tumors and in 4 LN-negative tumors after LYVE-1 immunostaining, whereas lymphatic invasion was observed in 19 tumors that were double stained for VB and H&E dyes, including 14 LN-positive tumors and 5 LN-negative tumors. We believe that this discrepancy among the results for lymphatic invasion by different staining methods was caused by an erroneous estimate from ambiguous findings of vessel-like structures because of artifacts or by misjudging true lymphatic invasion as interstitial invasion. Lymphatic invasion confirmed by LYVE-1 immunohistochemistry was distributed in the submucosa in all lymphatic invasion-positive tumors, and only 1 tumor demonstrated lymphatic invasion in the musclaris propria and subserosa in addition to the submucosa. Intratumoral lymphatic invasion was observed in only 4 tumors, whereas extratumoral lymphatic invasion around the tumor was observed in all 22 tumors. In 3 tumors, lymphatic invasion was located in the submucosa just under the muscularis mucosa, distant from the primary tumor (Fig. 2).
Venous invasion was observed in 14 tumors (19.7%), including 8 LN-positive tumors and 6 LN-negative tumors. Veins invaded by cancer cells were located in the tumor in 6 specimens and in the region around the tumor in 8 specimens, and all of them were located in the submucosa (Fig. 2).
Expression of β-Catenin, Claudin-3, and Claudin-4
In the areas of normal colorectal tissue, β-catenin was expressed in the cell membrane of the surface epithelium and the basal crypt epithelium. In tumor tissue, β-catenin was expressed in the cytoplasm or cell membrane of tumor cells at various levels and also was recognized in the nuclei of tumor cells (Fig. 3) located at the invasive front of the submucosa in 29 samples (40.8%).
Claudin-3 and claudin-4 were expressed uniformly in the cell membrane from normal colorectal epithelium. In the tumor, both proteins were immunopositive in the cell membrane in some tumor cells at various levels (Fig. 3A,B). The tumor cells that exhibited poorly differentiated adenocarcinoma usually were negative for both claudin-3 and claudin-4. In the entire group, 49 tumors (69%) were immunopositive for claudin-3 at the invasive front, and 42 tumors (59.2%) were immunopositive for claudin-4.
Comparison of Histologic Variables by LN Status
Differences in histopathologic variables between the LN-positive group and the LN-negative group are shown in Table 2. For tumor type, venous invasion, and expression of the 3 adhesion molecules, statistical analyses revealed no significant difference between the 2 groups. The histologic type of tumor in the mucosa fell into only 2 categories, well or moderately differentiated; however, moderately differentiated tumors were more frequent in the LN-positive group. For histologic type at the invasive front, malignancies in LN-positive group were classified more frequently as high-grade (eg poorly differentiated or mucinous types) compared with malignancies in the LN-negative group. The depth of invasion was significantly greater in the LN-positive group than that in the LN-negative group. The frequency of budding (Fig. 3C,D) and lymphatic invasion was significantly greater in the LN-positive group than in the LN-negative group. However, P values for lymphatic invasion were more significant according to the results obtained by LYVE-1 immunohistochemistry than that by double staining with VB and H&E dyes.
Table 2. Comparison of the Histopathologic Variables With Lymph Node Status in Early Colorectal Cancer
Lymph node status
Positive (n = 28)
Negative (n = 43)
SD indicates standard deviation; LYVE-1, lymphatic endothelial hyaluronan receptor 1; H&E, hematoxylin and eosin.
The results of the multivariate logistic regression analysis of the variables as predictors for LN metastasis are shown in Table 3. Among the 6 parameters, budding and lymphatic invasion by LYVE-1 immunohistochemistry were significantly independent predictors of LN metastasis.
Histologic type (low grade vs. high grade malignancy)
Lymphatic invasion (LYVE-1 immunostaining)
The tumors that had 3 histologic factors confirmed by immunohistochemistry, such as lymphatic invasion, venous invasion, and budding, in all specimens and in the group of LN-positive specimens, are shown in Table 4. Only 6 specimens had all 3 factors, and all of those specimens exhibited LN metastasis. In addition, 23 specimens had 2 of 3 factors, including 17 specimens (73.9%) with LN metastasis. The coexistence of lymphatic invasion and budding was recognized in 19 specimens, including 15 specimens (78.9%) with LN metastasis.
Table 4. The Number of Patients With Lymphatic Invasion, Venous Invasion, and/or Budding
In the current study, 6 variables (tumor size, depth of invasion, histologic type in the mucosa, histologic type at the invasive front, budding, and lymphatic invasion) were examined as predictive factors of LN metastasis in ECC. However, multivariate analysis demonstrated that only budding and lymphatic invasion were independent predictors of LN metastasis. The expression status of β-catenin, claudin-3, and claudin-4 in tumor cells at the invasive front was not related directly to LN status. These results demonstrate that the existence of lymphatic invasion and budding at the invasive front in the submucosa are the most important variables for predicting LN metastasis in ECC.
Various predictive factors for LN metastasis and outcome in patients with ECC have been reported previously, including tumor size,17 type of tumor growth,18 tumor differentiation,19 depth of invasion,4, 19 lymphatic invasion,4, 20 venous invasion,21 and budding formation.19 In the current study, the location of the primary tumor was not a significant predictor of LN status; however, tumor size was significantly greater in the LN-positive group than in the LN-negative group. A previous investigation also demonstrated that tumor size was an independent predictive factor of LN metastasis in early gastric cancer.9 In the stomach and large intestine, the submucosa, just under the muscularis mucosa has a relatively abundant, valved plexus of lymphatics.22 Considering this lymphatic distribution, tumor cells that invade the submucosa from a mucosal origin in the stomach and large intestine may migrate readily through the lymphatic route in this region, and larger mucosal tumors tend to have more frequent contact with the submucosal lymphatics. In fact, the most lymphatic invasion was observed in the submucosa in this series; in contrast, the type of tumor growth was not a predictive factor for LN metastasis. Previous investigators have reported that nonpolypoid-growth ECC is not associated with an adenoma-carcinoma sequence and exhibits rapid growth, aggressive behavior, and a high frequency of LN metastasis compared with polypoid-growth carcinoma.14, 18 This discrepancy between the current results and previous reports may be because of a difference in the tumors examined or because of a recent increase in the detection rate of ECC at an earlier stage, before metastasis.
In the current study, invasion into the submucosa was significantly deeper in the LN-positive group than in the LN-negative group. Because lymphatics under the muscularis mucosa run along with blood vessels toward the muscularis propria,22 it is believed that the deeper the cancer invades into the submucosa, the more frequently it will contact lymphatics. However, the invasiveness of cancer generally is dependent on its grade of differentiation. The current results demonstrated that the histologic type in the mucosa was related significantly to LN status. In addition, the histologic type at the invasive front in the LN-positive group differentiated more frequently toward high-grade malignancy than that in the LN-negative group. Previous studies also have reported that a feature of poor differentiation of the primary tumor is an important indicator of LN metastasis in ECC,19, 23 and focal dedifferentiation at the invasive front is a marker of liver metastasis in colorectal cancer.24 In addition, the existence of budding at the invasive front is a risk factor for LN metastasis in colorectal cancer, including ECC.16, 19, 25, 26 In the current study, “budding” at the invasive front was a vigorous indicator for LN metastasis in ECC. The finding of “budding” at the invasive front in colorectal cancer corresponds with “sprouting” of tumor cells invading into the surrounding stroma25, 26 and frequently has been observed in patients with well-differentiated adenocarcinoma.26 Those reports and our own results suggest that the histopathologic characteristics at the invasive front, rather than the predominant histology of the tumor, are extremely important for determining LN status.
In the this study, lymphatic invasion corroborated by LYVE-1 immunohistochemistry was an independent predictor of LN metastasis in ECC, but venous invasion confirmed by vWF immunohistochemistry in addition to double staining with VB and H&E dyes was not related to LN status. Both factors have been reported as risk factors for LN metastasis in ECC; however, those results were obtained from conventional staining methods without the use of immunohistochemistry.19, 21, 27, 28 To our knowledge, only 2 reports have described lymphatic distribution or lymphatic invasion in colorectal cancer observed with immunohistochemistry by using lymphatic markers, such as podoplanin and D2-40 antibodies.29, 30 D2-40 is an antibody to sialoglycoprotein, which has been used as a selective marker of lymphatic endothelium7 and also reacts with normal mesothelial cells,31 testicular germ cell tumors,32 and some angiosarcomas.7 Conversely, LYVE-1 is an integral membrane glycoprotein that has 41% similarity to CD44 hyaluronan receptor,6 and the LYVE-1 antibody used in the current study is an antisera against amino acids 297 through 308 of the LYVE-1 sequence.8 This antibody is a specific marker for lymphatic endothelial cells and is useful for the detection of small lymphatics distributed in various organs,8, 9, 33 including organs in the gastrointestinal tract. By using the LYVE-1 antibody, 4 specimens were identified as positive for lymphatic invasion in addition to the specimens that had invasion determined by double staining with VB and H&E in the LN-positive group; conversely, 1 specimen was excepted from the invasion-positive tumors in the LN-negative group. In addition, the significant difference (P value) in the frequency of lymphatic invasion-positive specimens between the LN-positive group and the LN-negative group increased from P = .0004 to P<.0001 through LYVE-1 immunohistochemistry. Therefore, it is necessary to detect lymphatic invasion more accurately by using specific markers against lymphatics for the investigation of lymphatic invasion in ECC.
In the current study, the expression of β-catenin and claudins in cancer cells at the invasive front were irrelevant to the LN status. The loss or reduction of β-catenin expression in cancer cells has been associated with the loss of E-cadherin-mediated cell adhesion, which some believe is an early event in the process of tumor metastasis34, 35 and, clinically, is a factor in metastatic behavior or a poor prognosis in oral squamous cell carcinoma, lung cancer, and colorectal cancer.10, 36, 37 In addition, nuclear immunoreactivity for β-catenin in cancer cells was increased in colorectal adenocarcinomas of various stages, including advanced-stage disease with liver metastasis.10 Because of the relation between nuclear immunoreactivity for β-catenin and the metastatic behavior of colorectal cancer, we assumed that the frequency of tumors with nuclear expression of β-catenin at the invasive front in ECC would be greater in the LN-positive group than that in the LN-negative group; however, the results did not indicate a significant correlation. This discrepancy may be the result of differences in tumor stage of colorectal cancer, because the data were obtained only from patients with ECC.
Conversely, the significance of claudin-3 and claudin-4 expression in human cancers has been controversial. Down-regulation of claudin-3 or claudin-4 expression has been related significantly to cancer invasiveness and a poor prognosis in patients with pancreatic and gastric cancers,38, 39 and the up-regulation of claudin-3 or claudin-4 expression has been related inversely to invasiveness, poor prognosis, and carcinogenesis in ovarian, colorectal, gastric, and endometrial cancers.11, 13, 40, 41 However, the current results demonstrated no significant relation between expression patterns of claudin-3 and claudin-4 in cancer cells at the invasive front and LN status in ECC. This discrepancy between previous reports and our results may be because of differences in methodology or tumor stage.
According to our observations of lymphatic invasion in ECC, lymphatics invaded by cancer cells were distributed in the submucosal layer around the primary tumor in all specimens that had lymphatic invasion. This finding suggests that a detailed observation of the submucosa around the primary tumor is important for the detection of lymphatic invasion in patients with ECC. Currently, most patients with ECC may achieve a cure by undergoing endoscopic resection; however, this procedure is insufficient for patients with ECC who have histopathologic risk factors for regional LN metastasis. In the current series, among the patients who had both lymphatic invasion and budding at the invasive front, 78.9% had directly related LN metastasis. Considering these results, the current findings may be applicable to patients who initially undergo endoscopic polypectomy. When such findings are detected in ECC tissues obtained by endoscopic resection, additional therapy, such as adjuvant chemotherapy or a curative LN resection, may be required.