Possible contribution of CD44 variant 6 and nuclear β-catenin expression to the formation of budding tumor cells in patients with T1 colorectal carcinoma

Authors


Abstract

BACKGROUND

In an earlier study, the authors demonstrated that tumor budding was useful for predicting lymph node metastasis in patients with early invasive (T1) colorectal carcinoma. This study was undertaken to clarify the associations between tumor budding, E-cadherin-catenin complex, and CD44 variant 6 abnormalities.

METHODS

In 51 T1 colorectal carcinomas, tumor budding (the number of dedifferentiation units at the invasive margin) on hematoxylin and eosin-stained slides was counted under light microscopy. Immunostaining for E-cadherin, α-catenin, β-catenin, and CD44 variant 6 was performed on formalin fixed, paraffin embedded sections. The associations between locoregional failure (lymph node metastasis or local recurrence) and tumor budding and clinicopathologic parameters and immunoreactivity were examined statistically.

RESULTS

In univariate analysis, tumor budding and nuclear β-catenin expression were associated significantly with locoregional failure (P = 0.004, 0.01). Multivariate analysis showed that tumor budding alone was associated significantly with locoregional failure (P = 0.02), and the association between nuclear β-catenin expression and locoregional failure was marginally significant (P = 0.07). Analysis of variance showed that lymphatic invasion alone was associated significantly with tumor budding (P = 0.02), and there was a significant interaction effect for tumor budding between CD44 variant 6 expression and nuclear β-catenin expression (P = 0.01). There was a significant correlation between expression patterns of these two molecules and locoregional failure (P = 0.01).

CONCLUSIONS

The current results suggest that the up-regulation of CD44 variant 6 through nuclear β-catenin activation may contribute to the formation of tumor budding, and immunostaining of these two adhesion molecules may be useful in identifying those at high-risk for locoregional failure among patients with T1 colorectal carcinoma. Cancer 2001;92:2539–46. © 2001 American Cancer Society.

Although local excision or endoscopic resection has been accepted widely as the treatment of choice for most patients with early invasive (T1) colorectal carcinoma,1–6 nearly 10% of these patients have regional lymph node metastases, and radical surgery with lymph node dissection is necessary for these tumors. In the last 20 years, many researchers have advocated histologic criteria that include massive submucosal invasion, vascular invasion, or poorly differentiated histology when considering additional surgery after patients undergo endoscopic resection of T1 colorectal carcinoma.1, 4, 7–9 However, the positive predictive value of these histologic features is rather low, about 20%, and a more useful risk factor is mandatory in the clinical setting.10 Recently, along with several Japanese researchers, we reported that dedifferentiated histology at the invasive margin (tumor budding) is useful in predicting lymph node metastasis or hematogenous metastasis in patients with colorectal carcinoma.10–12

Cell-cell adhesion is regulated mainly by the E-cadherin-catenin complex system,13 and abnormalities in this system have been reported in a number of tumors, including esophageal,14 gastric,15 breast,16 colorectal,17, 18 pancreatic,19 bladder,20 and prostatic21 carcinoma. Because tumor budding is one of the histologic features of loss of cell-cell adhesion and abnormal epithelial differentiation, it is assumed that E-cadherin-catenin complex abnormalities are associated with this histologic feature.

CD44 is a transmembrane glycoprotein, and functions as a major cell-surface receptor for hyaluronate. It also plays an important role in cell-cell and cell-matrix adhesion.22 The earlier studies showed the positive association between CD44 variant 6, one of the alternatively spliced variants, and colorectal tumor progression and poor prognosis.23–26 Therefore, we attempted to clarify the associations between tumor budding and E-cadherin-catenin complex or CD44 variant 6 expression in this immunohistochemical study.

MATERIALS AND METHODS

Patients

Fifty-one early invasive colorectal carcinomas from 50 patients treated by local excision or radical surgery at the First Department of Surgery, Kyorin University, between 1994 and 2000 were examined retrospectively. Thirteen patients underwent local excision initially (endoscopic mucosal resection in 6 patients, transanal resection in 4 patients, transsacral resection in 3 patients), and an additional bowel resection with lymph node dissection was attempted in 7 patients after histologic examination of the resected specimens. The remaining 37 patients underwent initial radical surgery (bowel resection with lymph node dissection). Thirty-six patients were men, and 14 patients were women. The mean age was 69 years, with a range of 40–87 years. The depth of invasion was submucosal in all tumors. The tumor location was the rectum in 18 lesions, the left colon in 17 lesions, and the right colon in 16 lesions. The mean follow-up was 24 months, with a range of 1–79 months. Colonoscopy or barium enema study to check for intraluminal recurrence and chest X-ray, abdominal ultrasonography, and computerized tomography scans to check for regional and distant metastasis were performed annually. Patients' clinical records and pathology reports were reviewed with special attention to the presence or absence of lymph node metastases, local recurrence, and distant metastases.

Tissues

From each resected specimen, one archival formalin fixed, paraffin embedded tissue block containing most deeply invading portion of the tumor was selected for immunohistochemical study. Serial 5-μm-thick paraffin sections were cut from each block and mounted on lysine-coated slides.

Immunohistochemistry

Immunostaining for E-cadherin, the two types of catenin (α-catenin and β-catenin), and CD44 variant 6 was performed using a standard avidin-biotin peroxidase technique. Sections were deparaffinized in xylene and hydrolyzed with 100%, 90%, 80%, and 70% ethanol. Antibodies used were 6F9 (dilution, 1:20) for E-cadherin (mouse monoclonal; PROGEN, Heidelberg, Germany), C-19 (prediluted) for α-catenin, C-18 (prediluted) for β-catenin (goat polyclonal; Santa Cruz Biotechnology, Santa Cruz, CA), and 2F10 (dilution, 1:1000) for CD44 variant 6 (mouse monoclonal; Serotec, United Kingdom). For the antigen-retrieval procedure for E-cadherin, tissues were digested in 0.1% trypsin solution for 60 minutes at 37 °C. Prior to α-catenin and β-catenin staining, the sections were heated in 10 mM citrate buffer solution, pH 6.0, twice for 5 minutes each at 95 °C. Prior to CD44 variant 6 staining, the sections were heated in a microwave oven in citrate buffer, pH 6.0, three times for 7 minutes each. Endogenous peroxidase was blocked with 0.3% H202 in PBS for 30 minutes, and the sections were rinsed in PBS. Nonspecific staining was eliminated by 30 minutes of incubation with normal horse serum. The sections were incubated at 4 °C overnight with primary antibodies, at 37 °C with biotinylated secondary antibodies for 30 minutes, and with avidin-biotin complex for 30 minutes. Peroxidase reaction was visualized with a solution of diaminobenzidine tetrahydrochloride. The sections were counterstained with hematoxylin, dehydrated, and mounted.

Evaluation of Immunostaining

Evaluation of E-cadherin, α-catenin, and β-catenin immunostaining was performed by two independent observers (T.M. and A.G.) as follows: Localization of positive staining was classified into membranous, cytoplasmic, or nuclear staining. Staining patterns were classified into uniformly positive (+), heterogeneous (±), and uniformly negative (−). In case of disagreement, the slides were reviewed, and a consensus view was achieved. Tumors were considered positive when > 10% of tumor cells were stained with CD44 variant 6 antibody according to a previous report.23

Evaluation of Clinicopathologic Parameters

A consecutive section from each specimen was stained with hematoxylin and eosin. Tumors were evaluated for predominant grade, lymphatic vessel invasion, blood vessel invasion, regional lymph node involvement, surgical margin status, the presence or absence of adenomatous component, the presence or absence of dedifferentiated histology at the invasive margin, and the finding of a single tumor cell or a solitary trabecular form along the entire invasive margin, which was identical to budding proposed by Hase et al.11) (Fig. 1). For objective evaluation, the number of dedifferentiation units along the entire invasive margin was counted under light microscopy according to a previous publication.12 The depth of submucosal invasion was graded into three levels, as described previously.10

Figure 1.

Representative photomicrograph of tumor budding at the invasive margin. Small nests of tumor cells with poorly differentiated histology are seen at the invasive margin of the tumor (arrows) (hematoxylin and eosin staining; original magnification, ×50).

Statistical Analysis

The associations between lymph node metastasis or local recurrence and the expression of E-cadherin, α-catenin, β-catenin, or CD44 variant 6 and clinicopathologic parameters were examined using a chi-square test, Fisher exact test, and logistic regression analysis (software: SPSS for Macintosh, version 6.1; SSPS, Inc., Chicago, IL). The associations between budding and clinicopathologic and immunohistologic parameters were examined using a Student t test, the Kruskal–Wallis one-way analysis of variance, and a multiple analysis of variance. A P value < 0.05 was accepted as statistically significant.

RESULTS

Demographic Features of Seven Patients with Adverse Outcomes

Table 1 shows that lymph node metastases were found in radically resected specimens from three patients, and distant metastases were found in three patients. One patient had local recurrence, although he had undergone potentially curative transanal resection for a 20-mm rectal tumor 1 year before. It was assumed that this recurrence was due to residual lymphatic vessel involvement. For the purpose of data analysis, these four tumors with lymph node metastases or with the suspicion of lymphatic vessel involvement were classified as potential risk of locoregional failure (LRF) if treated by endoscopic resection or local excision alone.

Table 1. Demographic Details of Seven Patients with Adverse Outcomes
GenderAge (yrs)LocationSize (mm)ShapeTreatmentHistologySm-InvLyVAdenomaLN metLocal recDistant met
  1. M: male; F: female; P: pedunculated; S: sessile; U: ulcerated; LE: local excision; RS: radical surgery; W: well differentiated; P: poorly differentiated; Sm-Inv: level of submucosal invasion; Ly: lymphatic invasion; V: venous invasion; Adenoma: presence/absence of adenomatous component; LN: lymph node; met: metastasis; rec: recurrence; +: positive; −: negative.

M47Rectum20SLEW1+
M75Right colon8PLE+RSM2++
M71Rectum17URSP3++
F74Left colon40PRSW1++
M73Rectum30PRSW1Liver (8 months)
M61Rectum15SLEW3+Bone, lung (2 yrs)
M73Right colon12SRSW3Lung (4 yrs)

Tumor Budding and Lymph Node Status

Tumor budding in LRF positive tumors was significantly greater than that in LRF negative tumors (mean ± standard error:; 31.5 ± 10.7 vs. 8.9 ± 2.0, respectively; P = 0.004).

Staining of Normal Colonic Mucosa

Internal control normal adjacent colonic mucosa showed exclusively intense membranous staining of E-cadherin, α-catenin, and β-catenin throughout the epithelium (Fig. 2).

Figure 2.

Representative photomicrographs of E-cadherin-catenin immunostaining. (a) Diffuse, membranous expression of β-catenin is seen in normal colonic mucosa (original magnification, ×200). E-cadherin and α-catenin expression in normal colonic mucosa were also membranous without exception. (b) Diffuse, membranous expression of E-cadherin in early invasive colorectal carcinoma (original magnification, ×200). (c) Heterogeneous cytoplasmic expression of α-catenin in early invasive colorectal carcinoma (original magnification, ×200). Membranous expression was completely absent. (d) Heterogeneous nuclear expression of β-catenin in colorectal carcinoma (original magnification, ×200). Membranous expression was completely absent.

Staining of T1 Colorectal Carcinomas

E-cadherin

Table 2 shows that E-cadherin expression was uniformly positive in 21 tumors (42%), heterogeneous in 19 tumors (38%), and uniformly negative in 10 tumors (20%). Localization of E-cadherin was exclusively membranous in both the uniformly positive and heterogeneous groups (Fig. 2). There was no significant correlation between E-cadherin expression pattern and LRF (P = 0.25).

Table 2. E-cadherin, Catenins, CD44 Variant 6, and Locoregional Failure
ImmunostainingaLocoregional failureP value
PresentAbsent
  • a

    In immunostaining for E-cadherin, α-catenin, and β-catenin, +, ±, and − stand for diffusely positive, heterogeneous, and diffusely negative, respectively.

  • In CD44 variant 6 staining, + and − stand for > 10% and < 10% of tumor cells positive, respectively.

E-cadherin
 +120
 ±3160.25
 −010
α-Catenin
 Membranous
  +00
  ±1160.71
  −331
 Cytoplasmic
  +–±429
  −0180.23
 Nuclear
  +–±00
  −447
β-Catenin
 Membranous
  +03
  ±1260.34
  −318
 Cytoplasmic
  +–±436
  −0110.31
 Nuclear
  +–±418
  −0290.01
CD44 variant 6
 +320
 −1270.21

α-Catenin

Table 2 shows that membranous α-catenin expression was lost in 34 of 51 tumors examined (67%) and heterogeneous in 17 tumors (33%). Cytoplasmic α-catenin expression was observed in 33 of 51 tumors (65%; Fig. 2); however, none of 51 tumors examined showed a nuclear pattern. There was no significant correlation between α-catenin expression pattern and LRF.

β-Catenin

Membranous β-catenin expression, as shown in Table 2, was reduced or absent in 48 of 51 tumors (94%). Conversely, cytoplasmic β-catenin expression was up-regulated (heterogeneous or diffusely positive) in 40 of 51 tumors (78%). There was no significant correlation between membranous or cytoplasmic β-catenin expression and LRF. Nuclear β-catenin expression was up-regulated in 22 of 51 tumors (43%) (Fig. 2). There was a significant association between nuclear β-catenin expression and LRF (P = 0.01).

CD44 variant 6

Table 2 shows that CD44 variant 6 expression was positive in 23 of 51 tumors (45%) (Fig. 3). There was no significant correlation between CD44 variant 6 expression and LRF (P = 0.21).

Figure 3.

Representative photomicrographs of CD44 v6 expression in T1 colorectal carcinomas. (a) The intense, membranous CD44 variant 6 (v6) staining pattern (> 90% of tumor cells were positive) (original magnification, ×72.5). (b) The focal and faint, membranous CD44 v6 staining pattern (< 10% of tumor cells were positive) (original magnification, ×72.5).

Clinicopathologic Factors and Lymph Node Status

Univariate analysis showed that tumor budding and nuclear β-catenin expression were associated with LRF with statistical significance (P = 0.004, 0.01) (Table 3). Tumor histology had marginal significance (P = 0.05). Multivariate logistic regression analysis using tumor budding, histology, and nuclear β-catenin expression as independent variables confirmed that tumor budding alone was associated significantly with LRF (P = 0.02), and the association between nuclear β-catenin expression and LRF was marginally significant (P = 0.07) (Table 3).

Table 3. Clinicopathologic Factors and Locoregional Failure
UnivariateaMultivariate variablesP valueOdds ratio95% CI
FactorP value
  • Ly: lymphatic invasion; V: venous invasion; Sm-Inv: level of submucosal invasion; adenoma: presence or absence of adenomatous component; (m): membranous pattern; (c): cytoplasmic pattern; (n): nuclear pattern; 95% CI: 95% confidence interval.

  • a

    Student t test and chi-squared test.

  • b

    Location is categorized into right colon, left colon, and rectum.

  • c

    Shape is categorized into pedunculated, sessile, and ulcerated.

  • d

    Histology is categorized into well, moderately, and poorly differentiated.

  • e

    Level of submucosal invasion is classified into three levels according to Masaki and Muto.10

Age0.71
Gender0.91
Locationb0.81
Tumor size0.95
Tumor shapec0.76
Histologyd0.05Histology
Ly0.39
V0.24
Sm-Inve0.36
Adenoma0.14
Budding0.004Budding0.021.071.01–1.13
Immunostaining
 E-Cadherin0.25
 α-Catenin (m)0.71
 α-Catenin (c)0.23
 β-Catenin (m)0.34
 β-Catenin (c)0.31
 β-Catenin (n)0.01β-Catenin (n)0.0712.060.75–188.8
 CD44 variant 60.21

Tumor Budding and Clinicopathologic and Immunohistologic Parameters

Lymphatic invasion alone, as shown in Table 4, was significantly associated with tumor budding (P = 0.002), and the associations between tumor budding and histologic differentiation, membranous α-catenin expression, and CD44 variant 6 expression were marginally significant (P = 0.05, P = 0.08, and P = 0.07, respectively). Analysis of variance showed that lymphatic invasion alone was associated significantly with tumor budding (P = 0.02), and there was a significant interaction effect for tumor budding between CD44 variant 6 expression and nuclear β-catenin expression (P = 0.01) (Table 5).

Table 4. Clinicopathologic Factors and Tumor Budding: Univariate Analysisa
FactorNo.BuddingP value
  • M: male; F: female; Ly: lymphatic invasion; V: venous invasion; Sm-Inv: level of submucosal invasion; −: negative; +: positive; Adenoma, presence or absence of adenomatous component; (m): membranous pattern; (c): cytoplasmic pattern; (n): nuclear pattern.

  • a

    Student t test or Kruskal–Wallis one-way analysis of variance.

  • b

    Diffusely positive (+), heterogeneous (±), and diffusely negative (−) immunostaining are indicated.

Gender (M/F)37/1411.8 ± 2.8/7.9 ± 2.50.42
Location (right colon/left colon/rectum)16/17/1811.2 ± 4.3/5.8 ± 1.9/14.8 ± 4.30.47
Shape (pedunculated/sessile/ulcerated)18/21/127.6 ± 3.1/13.2 ± 3.8/11.0 ± 4.20.4
Histology (well/moderate/poor)42/7/28.3 ± 1.9/19.7 ± 8.3/30.5 ± 19.50.05
Ly (−/+)45/68.3 ± 1.8/28.6 ± 9.40.002
V (−/+)39/128.9 ± 2.1/16.5 ± 5.70.13
Sm-Inv (level 1/2/3)13/8/2910.6 ± 3.7/10.8 ± 6.8/11.0 ± 2.90.58
Adenoma (−/+)11/4012.7 ± 4.8/10.2 ± 2.40.63
Locoregional failure (−/+)47/49.1 ± 1.9/37.0 ± 13.00.002
Immunostainingb
 E-Cadherin (+/±/−)21/19/1011.0 ± 3.4/10.7 ± 3.5/11.2 ± 5.40.97
 α-Catenin (m) (±/−)17/346.1 ± 3.0/13.0 ± 2.80.08
 α-Catenin (c) (+/±/−)21/12/1810.5 ± 3.2/13.8 ± 4.3/8.9 ± 3.90.29
 β-Catenin (m) (+/±/−)3/27/212.0 ± 2.0/10.0 ± 2.6/12.9 ± 3.90.45
 β-Catenin (c) (+/±/−)19/21/119.5 ± 2.8/11.2 ± 3.3/11.8 ± 6.30.84
 β-Catenin (n) (+–±/−)22/2913.6 ± 3.8/8.5 ± 2.40.24
 CD44 variant 6 (+/−)23/2815.0 ± 3.8/7.2 ± 2.10.07
Table 5. Clinicopathologic Factors and Tumor Budding: Multivariate Analysis of Variance
VariableP value
  1. Ly: lymphatic invasion; V: venous invasion; (m): membranous pattern; (n): nuclear pattern; ±: heterogeneous immunostaining.

Histology (well/moderate/poorly differentiated)0.08
Ly (−/+)0.02
V (−/+)0.09
α-Catenin (m) (±/−)0.74
β-Catenin (n) (+–±/−)0.47
CD44 variant 60.65
CD44 variant 6 by β-catenin (n)0.01

Table 6 shows that CD44 variant 6 positive/nuclear β-catenin positive tumors had a greater amount of tumor budding than other categories. Table 7 shows that three of four LRF positive tumors (75%) were CD44 variant 6 positive/nuclear β-catenin positive, and there was a significant correlation between the expression patterns of these two molecules and LRF.

Table 6. Tumor Budding and CD44 Variant 6/Nuclear β-Catenin
CD44 variant 6Nuclear β-catenin
NegativePositive
  • a

    P = 0.03 (CD44 negative/β-catenin negative and CD44 positive/β-catenin positive).

  • b

    P = 0.05 (CD44 negative/β-catenin positive and CD44 positive/β-catenin positive). All other P values were nonsignificant.

Negative8.5 ± 3.1a4.9 ± 2.2b
Positive8.5 ± 4.121.0 ± 6.0
Table 7. Nuclear β-Catenin/CD44 Variant 6 Expression and Locoregional Failure
CD44 variant 6Nuclear β-catenin (no.)
NegativePositive
  • a

    Patients with locoregional failure/number of patients in each category (chi-square test, P = 0.05; Mantel–Haenszel test, P = 0.01).

Negative0/18a1/10
Positive0/113/12

DISCUSSION

Tumor budding is a histologic feature showing small nests of tumor cells with poor differentiation at the invasive margin of the tumor. This multivariate analysis showed that tumor budding alone was associated significantly with LRF (lymph node metastasis or local recurrence) in patients with T1 colorectal carcinoma, a finding consistent with our previous publication.10 Several researchers reported previously that this histologic finding is useful for predicting adverse outcome and for decision making after patients undergo endoscopic polypectomy of T1 colorectal carcinomas.27 However, the molecular mechanisms in the formation of budding cells have been ill defined. Abnormalities of expression and function of E-cadherin, α-catenin, and β-catenin have been shown in a number of tumors and cell lines, and it is speculated that the loss of normal expression or function of any component of the cadherin-catenin complex plays a crucial role in tumor invasion and metastasis.13–21 Multivariate analysis in the study by Hugh et al.28 revealed that, among immunohistochemical parameters, widespread nuclear β-catenin expression was an independent predictor of short survival. However, the implication of widespread nuclear β-catenin expression in tumor metastases is unclear, because tumor recurrence patterns were not demonstrated in that study. In our study, univariate analysis showed that nuclear β-catenin expression was associated with LRF with statistical significance (P = 0.01), and this association was marginally significant in multivariate analysis (P = 0.07). Analysis of variance showed that there was a significant interaction effect for tumor budding between CD44 variant 6 expression and nuclear β-catenin expression. Tables 6 and 7 show that CD44 variant 6 positive/nuclear β-catenin positive tumors had greater amounts of tumor budding than other categories, and three of four LRF positive tumors (75%) were CD44 variant 6 positive/nuclear β-catenin positive. This implies that the collaboration of CD44 variant 6 and nuclear β-catenin in patients with T1 colorectal carcinoma may contribute to tumor aggressiveness through the formation of budding cells, and immunostaining using these antibodies may facilitate the identification of patients who are at high risk for LRF. Coordinated expression patterns of several adhesion molecules have been studied in patients with prostate carcinoma,29 bladder carcinoma,30 and breast carcinoma,31 and down-regulation of E-cadherin-β-catenin with or without up-regulation of CD44 variant expression was associated with adverse outcome or dedifferentiation. Compared with these observations, our finding was unique, in that not the ordinary cadherin-catenin interactions but the collaboration of nuclear β-catenin and CD44 variant 6 in the Wnt signaling pathway may play a significant role in the formation of budding cells at the invasive margin. A previous study using APC and Tcf mutant mice has shown that CD44 expression (standard form and variant 6) is controlled by the β-catenin/Tcf-4 pathway,32 suggesting that CD44 is one of the target genes of β-catenin. CD44 is a transmembrane glycoprotein and functions as a major cell-surface receptor for hyaluronate. It plays an important role in tumor cell-matrix adhesion.22 The results of this study may suggest that up-regulation of CD44 variant 6 through nuclear β-catenin activation may promote the dissociation of tumor cells from the invasive margin and the invasion of scattered tumor cells into the stroma. Further molecular research focusing on budding tumor cells at the invasive margin will be mandatory to solve this issue.

Acknowledgements

The authors thank Drs. Y. Shimada, K. Nunomura, and S. Uchiyama for providing tissue blocks and Ms. N. Sato and K. Ooshima for expert technical assistance.

Ancillary