Tumor budding correlates with poor prognosis and epithelial-mesenchymal transition in tongue squamous cell carcinoma

Abstract

BACKGROUND

Tumor budding is a readily detectable histopathological feature and has been recognized as an adverse prognostic factor in several human cancers. However, the prognostic value of tumor budding in tongue squamous cell carcinoma (TSCC) has not been reported. The purpose of this study is to assess the correlation of tumor budding with the clinicopathologic features, and the known molecular biomarkers (E-cadherin and Vimentin), as well as to evaluate its prognostic significance for TSCC.

METHODS

Archival clinical samples of 230 patients with TSCC were examined for tumor budding. Immunohistochemistry analyses were performed to examine the expression of E-cadherin and Vimentin. Statistical analyses were carried out to assess the correlation of tumor budding with clinicopathologic parameters and patient survival. The potential association between tumor budding and alterations of E-cadherin and Vimentin expression was also assessed.

RESULTS

Of the 230 TSCC cases examined, tumor budding was observed in 165 cases (71.7%), with a mean tumor bud count of 7.5 (range from 1 to 48 buds). High-intensity budding (≥ 5 tumor buds) was observed in 111 cases (48.3%). Statistical analysis revealed that tumor budding was associated with tumor size (P<0.05), differentiation (P<0.05), clinical stage (P<0.05), lymph node metastasis (P<0.01), and correlated with reduced overall survival. In addition, significant associations were observed among tumor budding and the deregulation of E-cadherin (P<0.001) and Vimentin (P<0.001).

CONCLUSIONS

Tumor budding, which associates with epithelial-mesenchymal transition, is a frequent event and appears to be an independent prognostic factor in TSCC.

Introduction

Tongue squamous cell carcinoma (TSCC) is one of the most common cancers within the oral cavity. According to the American Cancer Society, an estimated 10,990 new cases of tongue cancer are expected each year, accounting for approximately 30% of all oral cavity and pharynx cancers [1]. TSCC is significantly more aggressive than other forms of oral cancer, with a propensity for rapid local invasion and spread [2], and a high recurrence rate [3]. The major causes of TSCC-related deaths are local/regional relapse and metastasis. It has been reported that 40% of all TSCC patients have neck metastasis at the time of diagnosis and 20–40% of patients with early-stage TSCC (T1/T2N0) showed occult nodal metastasis [4–7]. These facts point to the immediate need for new diagnostic/prognostic strategies and additional biomarkers will improve the clinical decision-making and the management of patients with TSCC.

It has been suggested that cancer cells located in the invasive tumor front (ITF) are more aggressive in terms of metastatic potential [8]. For oral cancer, a histopathological grading system of the ITF was firstly introduced by Bryne et al in 1992 [9, 10]. This approach have been utilized in a number of studies for evaluating invasive tumor margins, and its values as a prognostic marker for oral cancer patients were well recognized [11–15]. Based on this approach, Brandwein-Gensler et al [16, 17] developed a novel histologic risk assessment system to evaluate the aggressiveness of head and neck squamous cell carcinoma. Recently, a number of unique features of cancer cells located in ITF have been defined, including a cell morphological switch characterized by the epithelial-mesenchymal transition (EMT), which is associated with increased cell motility and invasiveness. Differential expression of several crucial EMT-related genes, including E-cadherin/β-catenin [18], Vimentin [19], claudins [20], laminin 5 and γ2 chains [21–22] in oral cancer cells located in the ITF have further confirmed the existence of EMT cells in ITF.

Tumor budding is defined as the presence of either isolated single cells or small cell clusters (up to four) scattered in the stroma ahead of the ITF [23]. Budding represents two malignant features: cellular discohesion and active invasion. The presence of tumor buds has been considered to be characteristic of aggressive cancer. Tumor budding has previously been demonstrated as a valuable prognostic marker for colorectal cancer patients [24–28]. More recently, it has also been reported as a significant prognostic marker for patients with esophageal cancers [29–31], lung and ampullary adenocarcinoma [32, 33] and laryngeal cancers [34]. An attractive feature of a tumor budding-based assessment is that, as it is based on the examination of the H&E-stained specimens, it can be performed easily as part of routine histopathological examination and the results are fairly reproducible [24]. However, to our knowledge, the effectiveness of this relatively straight-forward histopathological assessment and its prognostic value for TSCC have not been investigated so far. The aims of this study are to investigate the possible association of tumor budding and clinicopathologic features in TSCC, as well as to determine its prognostic significance. In addition, the EMT status of the cancer cells in the tumor buds will also be investigated.

Material and methods

Patients and tissue specimens

Archived tissue samples from 230 cases of TSCC from the Department of Oral Pathology, Hospital of Stomatology and the Second Affiliated Hospital, Sun Yat-sen University were utilized in this study. Clinical characterization of the TSSC patients is summarized in supplemental Table 1. All patients received curative surgery (resection of the primary tumor and radical neck dissection) between May 1996 and June 2005. None of the patients received any form of adjuvant therapy prior to their surgery. The tumor extent was classified according to the TNM system by UICC, and the tumor grade was classified according to the WHO classification of histological differentiation. Survival was calculated based on the date of surgery and the date of the last follow-up (or death). Among 230 cases of TSCC that we examined, follow-up results were available on 133 cases (57.8%). Median duration of follow-up was 65 months (range 3–120). This study was approved by the ethical committee of Sun Yat-Sen University.

Histo- and Immunohistopathological analyses

Paraffin sections (4µm) were routinely prepared and stained with H&E. The immunohistochemical assay was performed as previously described [18, 19]. The slides were immuno-stained with anti-E-cadherin antibody (1:100, BD, USA), anti-Vimentin antibody (1:100, Cell signaling, USA), and pan-cytokeratin antibody (1:100, Cell signaling, USA), respectively. 3,3’-diaminobenzidine (DAB, Sigma-Aldrich) was used for visualization of E-cadherin and Vimentin staining. For pan-cytokeratin staining, AEC chromogen was used for visualization. Negative controls were included by substituting non-immune serum for primary antibodies. All sections were assessed by 2 independent observers who were unaware of the clinical data. Tumor budding was defined as the presence of isolated single tumor cells or small clusters (< 5 cancer cells) ahead of the invasive front as previously described (Figure 1A and B) [23]. Tumor specimens were initially scanned at the ×4 objective lens (and ×10 ocular) to select the areas with the highest density of budding. Tumor budding in the selected areas was then counted using the ×20 objective lens, and the highest count per slide was used as the number of budding. The intensity of tumor budding (budding index) was arbitrary classified as low (< 5 buds/field) or high (≥ 5 buds/field). To test the reproducibility of the tumor budding assessment (for both intra- and inter-observer), 50 randomly selected cases were subjected for a second review by the same investigators one month after the initial review. Excellent agreement was obtained between the initial and the second reviews by both observers (κ=0.880 and 0.838, respectively, based on a hierarchical kappa test [35]). Similarly, agreement between two observers was also achieved (κ=0.717), demonstrating good inter-observer reproducibility.

Figure 1. Histopathological analyses of tumor budding in tongue squmous cell carcinoma.

Histopathological analyses were performed as described in Material and Methods to identify tumor buds in the invasion front of the TSCC. (A) Tumor budding at the invasive front in tongue squmous cell carcinoma (H&E, 20×). (B) Tumor budding with hyperchromatic nuclei in an H&E section (40×). Tumor buds were identified with arrowheads.

The expression of E-cadherin in ITF, budding and central/superficial tumor parts was compared with that of adjacent non-cancerous epithelium, which served as an internal positive control in the same cases. When the positivity of E-cadherin was more than 90%, the case was designated as having preserved expression; otherwise the case was designated as having reduced expression, as described previously [36]. For the Vimentin expression, no detectable or < 10% positive staining of tumor cells was served as negative, while ≥ 10% positive staining of tumor cells was defined as positive [37].

Statistical analysis

Spearman Correlation Coefficient was used to assess correlations among the gene expression and clinical and histopathological parameters. Kaplan-Meier plots were constructed to present the survival outcomes. Cox regression was used for both univariate and multivariate analysis. For multivariate analysis, tumor size, pathological T-stage (pT), pathological N-stage (pN), clinical stages and budding index were considered as co-variates. For all statistical analyses, P< 0.05 was considered statistically significant.

Results

Intensity of tumor budding and its correlation with clinicopathological parameters

H&E staining was performed on 230 TSCC cases. As illustrated in Figure 1, tumor buds can be readily identified based on standard H&E staining. A second staining for pan-cytokeratin was also performed on 30 of the TSCC cases (13%) to assist the identification of tumor buds (Supplementary Figure 1). The tumor budding counts ranged from 0 to 48 buds (mean 7, median 4). Among 230 TSCC cases that we examined, 111 cases (48.36%) revealed high-intensity tumor budding (≥ 5 tumor buds) ahead of the invasive front; 119 cases (51.74%) revealed low-intensity budding (<5 tumor buds). Of the 119 low-intensity budding cases, no tumor bud was observed in 65 cases.

Correlations were tested among tumor budding and clinical parameters of the TSCC cases (Table 1). As expected, correlations were observed among tumor size, pT, pN, and Clinical stage. Strong correlations were observed between the intensity of tumor budding and tumor size (P<0.05), cell differentiation (P<0.05), clinic stage (P<0.05), and lymph node metastasis (P<0.01). Interestingly, correlation between sex and tumor size (and pT) was also observed (p<0.05).

Table 1.

Correlations among clinical and histopathological features of primary TSCC ^#

	Age	Sex	Tumor size	Grade	pT stage	pN stage	Clinical stage	Tumor budding
Age		0.08456	−0.038	0.06086	−0.0728	−0.0016	−0.0294	−0.0019
Gender			0.15257*	0.11917	0.14313*	0.02571	0.1179	−0.0539
Tumor size				0.10703	0.89649**	0.41054**	0.80261**	0.22149*
Grade					0.03091	0.07592	0.07389	0.1543*
pT stage						0.36165**	0.83668**	0.11509
pN stage							0.74313**	0.3178**
Clinical stage								0.22814*
Tumor budding

^#Spearman correlation coefficients were presented.

pT: pathological T-stage; pN: pathological N-stage

^*P<0.05;

^**P<0.01

The prognostic value of tumor budding for TSCC patients

Among 230 cases of TSCC that we examined, follow-up results were available for 133 cases (57.83%). Median duration of follow-up was 65 months (ranges from 3 to 120 months). Among these 133 cases, 74 cases (55.6%) were low-intensity or had no detectible tumor bud, and 59 cases (44.4%) had high-intensity tumor budding. As illustrated in Figure 2, a striking difference in 5-year survival rate was observed between the high-intensity budding group (79.7%) and the low-intensity or no budding group (49.2%). Both univariate and multivariate analysis demonstrated the adverse effect of high-density budding for prognosis (Table 2). Based on univariate analysis, the effects of tumor size, pT, pN and clinical stages on prognosis were also observed. Based on multivariate analysis pT was also identified as an independent prognostic factor.

Figure 2. The effects of tumor budding on prognosis.

Kaplan-Meier plots of overall survival in patient groups defined by intensity of tumor budding (High-intensity tumor budding: ≥5 tumor buds, Low-intensity or No tumor budding: < 5 tumor buds or no tumor bud detectable). The difference in survival rates is statistically significant (p<0.001).

Table 2.

The effects of clinical and pathohistological parameters on prognosis^*

		Univariate analysis			Multivariate analysis
		HR	95%CI	P value	HR	95%CI	P value
Gender	Female	1
	Male	1.470	0.740 – 2.921	0.2711
Age	≤ 55	1
	> 55	1.386	0.751 – 2.555	0.2963
Grade	Well	1
	Mediate	1.935	0.959 – 3.904	0.0652
	Poor	2.359	0.854 – 6.516	0.0979
Tumor Size	≤ 4 cm	1			1
	> 4cm	2.681	1.488 – 4.830	0.0010	0.806	0.251 – 2.588	0.7177
pT	pT1–2	1			1
	pT3–4	3.395	1.872 – 6.158	<.0001	3.990	1.283 – 12.404	0.0168
pN	Negative	1			1
	Positive	2.546	1.410 – 4.597	0.0019	1.209	0.457 – 3.200	0.7022
Clinical stage	I–II	1			1
	III–IV	3.427	1.766, 6.651	0.0003	1.510	0.401 – 5.683	0.5426
Tumor budding	Low	1			1
	high	3.350	1.774, 6.323	0.0002	3.029	1.535 – 5.977	0.0014

^*Analysis was done with Cox proportional hazard regression. HR: hazard ratio; 95% CI: 95% confidence interval.

High-intensity tumor budding is associated with reduced E-cadherin expression and enhanced Vimentin expression in TSCC

Distinct membrane staining of E-cadherin protein was detected in the adjacent non-cancerous epithelium. The expression pattern of E-cadherin in the center/superficial tumor parts was similar to that in the adjacent non-cancerous epithelium (Figure 3A, B). In contrast, an intensive reduction in membranous E-cadherin expression was observed in the ITF and tumor budding, when compared to that in the center/superficial tumor parts (64% vs 90%, P<0.05). In particular, at the tumor budding site ahead of ITF, a loss of E-cadherin expression was frequently observed (Figure 3C). Cytoplasmic Vimentin expression was observed in the stromal cells of the adjacent non-cancerous tongue tissues, but not in the epithelium (Figure 3D). This is in agreement with the previous observation made in OSCC [19]. No staining or weak staining (<10% of tumor cells) was found in the center/superficial tumor parts (Figure 3E). Increased expression of Vimentin was detected in the ITF and tumor budding (40% vs 76%, P<0.05) (Figure 3F). Statistical analysis revealed that the high-intensity tumor budding is associated with reduced E-cadherin expression (P<0.001) and enhanced Vimentin expression (P <0.001) (Table 3).

Figure 3. Immunohistochemistry analyses of E-cadherin and Vimentin expression in the tumor budding cells.

Immunohistopathological analyses were performed as described in Material and Methods to evaluate the expression of E-cadherin and Vimentin expression in tumor budding cells. Strong staining of E-cadherin protein was detected at the cytoplasmic membrane and the intercellular borders in the adjacent non-cancerous epithelium and in the center/superficial tumor parts (A and B, 40×). In contrast, cancer cells at the tumor budding site showed negative or weak (dot-like) membranous immunostaining for E-cadherin (C, 40×, arrow). As shown in D and E, no Vimentin staining was observed in adjacent non-cancerous epithelium or in the center/superficial tumor parts (20×). In contrast, strong Vimentin staining was detected in the ITF and tumor budding cells (F). A subset of cancer cells at budding site also exhibits a spindle-like morphology (40×, arrow).

Table 3.

Association of the tumor budding intensity and the expression of E-cadherin and Vimentin

Genes		Number	Intensity of tumor budding
			Low	High	P value
E-cadherin	Preserved	60	48	12	<0.001
	Reduced	73	26	47
Vimentin	Negative	70	59	11	<0.001
	Positive	63	15	48

Discussion

The molecular heterogeneities of tumor cells reside in different locations of solid tumors (e.g., center/superficial portions, ITF) are well recognized. Accumulating evidence now supports the idea that characteristics of the ITF provide the most useful prognostic information [8]. Accordingly, a malignancy grading of ITF (IFG), or the deep invasive margins of oral cancer, was firstly developed in 1992 [9, 10, 38]. This grading system is based on semiquantification of the following histological parameters of the ITF: 1) degree of keratinization; 2) nuclear polymorphism; 3) pattern of invasion; and 4) infiltration of lymphocytes. Based on this grading system, the prognostic value of ITF has been confirmed by multiple studies [12, 13, 18, 19]. Moreover, Brandwein-Gensler et al [16, 17] combined the pattern of invasion and lymphocytic host response with perineural invasion and developed a novel risk model to assess the aggressiveness of head and neck cancer. This model demonstrated a significant predictive performance. Intriguingly, tumor budding, which usually resides ahead the invasive front, has recently been suggested as a potential index of aggressiveness and poor prognosis for a number of cancer types [23–34]. However, little is known about the prognostic value of tumor budding in patients with TSCC. In this study, we found that the 5-year survival was significantly reduced in patients exhibiting high-intensity tumor budding compared to patients with low-intensity budding. Our observations suggested that high-intensity budding is an adverse prognostic factor in patients with TSCC. This is in agreement with studies on other solid tumors (e.g., larynx and esophageal cancer, colorectal cancer) showing strong associations of tumor budding with a poor prognosis [27, 28, 33,34].

An important advantage of tumor budding-based index as prognostic indicator is the simplicity and reproducible measurement of the budding. It is readily adaptable to routine H&E staining based histopathological examination without the need for additional cost-demanding techniques. This feature is clinically important and may have therapeutic benefits for the patients with TSCC. In this study, in accordance with previous reports in other cancers [23, 24, 39], a good reproducibility for tumor budding evaluation was achieved based on the intra- and inter-observer agreement studies (κ=0.880 or 0.818 and 0.717, respectively). In contrast, the IFG system is more susceptible to intra- and inter-observer disagreement. Several independent groups reported that the inter-observer reproducibility of the IFG system is clinically unacceptable, with a value of κ ranging from 0.193–0.57 [40–42]. In Brandwein-Gensler’s system, the reproducibility seems much better than the original IFG system (κ=0.67). However, it should be noted that a direct comparison between the budding index and the IFG-based grading system or Brandwein-Gensler’s system may not be justifiable. Based on the IFG criteria proposed by Bryne et al [9,10], the tumor budding probably been classified into grade 4 (the pattern of invasion is defined as marked and wide-spread cellular dissociation in small groups and / or in single cells [n<15]). The tumor cells in these small groups showed a high tendency to metastasize to regional lymph nodes compared with those that invade in pushing fronts (IFG, score 1) or in solid cords, bands and / or strands (IFG, score 2 and 3). Same observations were also reported by Brandwein-Gensler’s et al [16, 17] in two consecutive studies in which they found that the worst pattern of invasion 4 (tumor budding can be classified into this group) and 5 significantly associated with lymph node metastasis and overall survival. In agreement with these studies, positive associations of high intensity tumor budding with lymph node metastasis and poor prognosis were reported in these studies. In addition, we observed that it is difficult to identify the budding tumor cells in a subset of poorly differentiated TSCC cases because the tumor budding at the ITF are surrounded by many lymphocytes, cancer associated fibroblasts and other stromal cells. To clearly define the budding status in such cases, we performed additional staining for pan-cytokeratin. This, together with the E-cadherin staining improved the identification of budding tumor cells (as shown in Figure 3C and Supplementary Figure 1 and Supplementary Figure 2). However, it is possible that there are still some unidentified tumor budding cells at the invasive front. For example, some isolated individual cells with multiple nuclei and spindle cells that can’t be distinguished from fibroblasts and/or endothelial cells were not counted. Nevertheless, in agreement with previous reports on other cancer types [23, 24, 39], a good reproducibility for tumor budding evaluation was achieved based on the intra- and inter-observer agreement studies (κ=0.880 or 0.818 and 0.717, respectively).

Microscopically, tumor cells in ITF and tumor buds exhibit distinct morphological features, including de-differentiation and loss of cell-cell adhesion. This fibroblast-like morphological appearance is characteristic of cells undergoing epithelial to mesenchymal transition (EMT), characterized at the molecular level by loss of E-cadherin (a key component of the adherens junctions) and the increase in expression of Vimentin (a ubiquitous mesenchymal intermediate filament). Our results confirmed that the expression of E-cadherin is significantly reduced in cells located in ITF (64%) and cells located in tumor buds (90%), when compared with those located in the central/superficial portions of the tumor samples. The observed reduction in E-cadherin in ITF and budding cells is accompanied by an increase in Vimentin expression. Our observation is consistent with previous observations of reduced E-cadherin expression in ITF of OSCC [18, 19]. Taken together, these findings demonstrate that cancer cells located in the tumor buds underwent EMT, which is associated with enhanced metastatic potential. Additional studies may be warranted to further investigate the molecular events associated with tumor cells that reside in the ITF/budding areas, which will lead to a better understanding of TSCC invasion and metastasis, as well as potential targeted therapeutic strategies.