Tumor lymphangiogenesis in head and neck squamous cell carcinoma

A morphometric study with clinical correlations




Tumor metastasis to regional lymph nodes via the lymphatic system represents the first step of dissemination in head and neck squamous cell carcinoma (HNSCC) and serves as a major prognostic indicator for disease progression and as a guide for therapeutic strategies. In the current study, the authors investigated whether tumor lymphangiogenesis may be related to the risk of lymph node metastasis and to clinical outcome in patients with HNSCC.


Immunostaining for the lymphatic marker D2-40 was used, and lymphangiogenesis was quantified within the tumor and in the peritumoral area in 52 HNSCC specimens using computer-assisted morphometric analysis.


Lymphatic vessels were found to be significantly more numerous and larger in the peritumoral area compared with within the tumor, and the number and relative area of intratumoral and peritumoral lymphatics was significantly higher in HNSCC cases with lymph node metastasis. Multivariate analysis demonstrated that high peritumoral lymphangiogenesis (above the median value) was associated with an increased risk of developing lymph node metastasis. No correlation was found between tumor lymphangiogenesis and the disease-free or overall survival in the current series.


The results indicate that peritumoral lymphangiogenesis may be an indicator of the risk of lymph node metastasis in patients with HNSCC. Cancer 2004. © 2004 American Cancer Society.

Tumor metastasis to regional lymph nodes via the lymphatic system represents the first step of dissemination in head and neck squamous cell carcinoma (HNSCC) and serves as a major prognostic indicator for disease progression and as a guide for therapeutic decision-making. Although the relevance of tumor angiogenesis and hematogenous spread in HNSCC has been documented by several molecular and clinical studies, to our knowledge comparatively little is known regarding the mechanisms through which tumor cells gain entry into the lymphatic system. It has been suggested that the lymphatic vessels play only a passive role during this process and lymphatic invasion occurs when tumor cells infiltrate preexisting peritumoral lymphatics.1 However, recent evidence suggests that the formation of tumor-associated lymphatic vessels plays an active role in the metastatic spread of several human malignancies, including HNSCC.2–4 Although an increasing number of clinicopathologic studies have demonstrated that the expression of lymphangiogenesis factors such as vascular endothelial growth factor (VEGF)-C and VEGF-D takes place through tumor cells and metastatic spread, the availability of specific markers for lymphatic endothelium has provided evidence of the existence of intratumoral lymphatics.

The goal of the current study was to investigate tumor lymphangiogenesis in a series of HNSCC patients and to assess whether the extent of tumor lymphangiogenesis, determined within the tumor and in the peritumoral area, is correlated with the presence of lymph node metastasis and with the patients' clinical outcome.


Patients and Tumor Samples

The study included 52 consecutive patients with HNSCC who were diagnosed and surgically treated in the Division of Otolaryngology and Head and Neck Surgery of the Department of Oto-Neuro-Ophtalmologic Surgery at the University of Florence between April 1995 and April 1998. There were 41 men (78.8%) and 11 women (21.2%) with a mean age of 62.3 years (range, 50–74 years). The primary tumor was located in the oral cavity in 18 patients, in the oropharynx in 12 patients, and in the larynx in 22 patients. Seventeen patients (32.7%) had histologically confirmed lymph node metastasis at the time of diagnosis, whereas the remaining 35 patients (67.3%) were found to have no clinical or histopathologic evidence of cervical lymph node involvement. The median follow-up time for the 52 patients was 41.5 months (range, 3–98 months). Twenty patients (38.5%) had developed disease recurrence by the time of last follow-up, whereas 16 patients (30.8%) had died of disease.

Tumor sampling was performed consistently by the same researcher (O.G.) immediately after removal of the surgical specimen. A central section of the entire tumor was taken in each case and a sample including the tumor core and the invasive front of the lesion was immediately fixed in buffered formalin and embedded in paraffin for the histopathologic and immunohistochemical assessments.


Paraffin sections measuring 5 μm in thickness were dewaxed and hydrated; after inactivation of endogenous peroxidase with 3% hydrogen peroxidase for 20 minutes, the sections underwent antigen retrieval with microwave oven heating for 30 minutes in citrate buffer. Sections then were treated with normal goat serum (UltraVision Detection System; Lab Vision Corporation, Freemont, CA), followed by incubation with the monoclonal antibody D2-40 (prediluted) (Signet Laboratories, Dedham, MA) overnight at 4 °C. The D2-40 antibody has been generated against an oncofetal antigen expressed in fetal testis and in testicular germ cell tumors, and has been successively recognized as a selective marker of lymphatic endothelium that does not react with blood vessel endothelium.5–8 This was followed by incubation with the secondary antibodies (UltraVision Detection System, Lab Vision Corporation). The reaction was visualized using 3,3′-diaminobenzidine, and nuclei were lightly counterstained with hematoxylin.

Computer-Assisted Morphometric Analysis of Tumor Lymphatics

To determine the lymphatic vessel density and size within and surrounding the 52 HNSCC sections, those sections stained with the D2-40 antibody were examined using a Leica DM LB microscope (Leica Ltd., Cambridge, UK) at ×200 magnification. Two distinct sets of measurement were performed in each tumor section; 3 fields with the highest lymphatic vascular density (hot spots) were identified 1) within the tumor mass and 2) within an area of 500 μm from the tumor border. In each selected field, the outline of each individually stained vessel was identified and traced using a computer-aided image analysis system (Quantimet; Leica Ltd.) to determine 1) lymphatic vascular density (LVD), defined as the number of vessels per mm2; 2) the average vessel area; and 3) the relative lymphatic vascular area, defined as the percentage of positively stained vascular area in a field.9 The mean values of the measurements in the three fields were used for statistical analysis.

Statistical Analysis

The Mann–Whitney U test or the Kruskal–Wallis test were employed to examine the association between lymphatic vessel parameters and age, smoking habits, alcohol consumption, histologic degree of differentiation, tumor site, disease stage, and lymph node status. Intratumoral and peritumoral lymphangiogenesis parameters were compared within different tissue samples using a Wilcoxon signed rank test. Multivariate logistic regression analysis was employed to determine which clinicopathologic factors were predictive of lymph node metastasis. Overall survival curves and disease-free survival curves were obtained using the Kaplan–Meier method and compared with the log-rank test. A P value < 0.05 was considered significant. All statistical tests were performed using SPSS software (release 10.0; SPSS Inc., Chicago, IL).


D2-40-positive lymphatic vessels were detected in all cases. Consistent with previous reports,2, 3 lymphatic vessels were observed both within the tumor mass and in the peritumoral area, and they were consistently thin walled (Fig. 1). Peritumoral lymphatics were found to have more dilated, open lumina than intratumoral lymphatics. Lymphatic vessels also were found to be associated with mucosal glands, aggregates of lymphoid tissue, nerves, and skeletal muscle. Few lymphatic vessels contained lymphocytes, whereas red blood cells were never observed within positively stained vessels. No immunostaining was identified in the tumor cells.

Figure 1.

Head and neck squamous cell carcinoma specimens stained with the lymphatic marker D2-40. (A) Collapsed vessels are visualized between nests of tumor cells. (B) In the peritumoral area, lymphatic vessels demonstrate open lumina. Red blood cells are present in unstained capillary structures.

Computer-assisted morphometric analysis showed that peritumoral lymphatic vessels were more numerous, had a larger average size, and occupied a greater relative area compared with intratumoral lymphatics (P = 0.002, P < 0.001, and P < 0.001, respectively; Wilcoxon signed rank test). HNSCCs presenting with lymph node metastasis was found to have a significantly higher LVD, both within the tumor mass and in the peritumoral area (P = 0.003 and P = 0.01, respectively; Mann–Whitney U test) (Figs. 2A and 2B). The same positive correlation was observed when the lymphatic vessel area was considered as a parameter with which to describe lymphangiogenesis (P = 0.009 for intratumoral lymphatics and P = 0.01 for peritumoral lymphatics; Mann–Whitney U test) (Figs. 2C and 2D), although there was no significant difference noted with regard to average vessel size between metastatic and nonmetastatic carcinomas, both within and around the tumor (P = 0.7 and P = 0.5, respectively; Mann–Whitney U test). In addition, tumors with a more advanced TNM stage (Stages III and IV) demonstrated a significantly higher lymphatic vascular area, both at the intratumoral (P = 0.003; Mann–Whitney U test) and peritumoral level (P = 0.01; Mann–Whitney U test). The correlation between intratumoral and peritumoral LVD and TNM stage only approached statistical significance (P = 0.07 and P = 0.06, respectively). No significant correlation was apparent between the lymphangiogenesis parameters determined within and around the carcinomas and age, gender, smoking history, alcohol consumption, tumor site, histologic differentiation, and tumor inflammatory infiltrate (Table 1).

Figure 2.

Computer-assisted morphometric analysis of tumor lymphangiogenesis in 52 head and neck squamous cell carcinoma specimens. Tumors with lymph node metastasis (N+) demonstrated significantly higher (A) intratumoral and (B) peritumoral lymphatic vessel density (LVD) (P = 0.003 and P = 0.01, respectively), as well as significantly higher (C) intratumoral and (D) peritumoral lymphatic vessel relative area (LVA) (P = 0.009 and P = 0.01, respectively) compared with tumors without lymph node metastasis (N–;).

Table 1. Relation between Tumor Lymphangiogenesis and Clinicopathologic Parameters in 52 Head and Neck Squamous Cell Carcinoma Specimens
FactorNo. of casesMean intratumoral LVD (± SD)Mean intratumoral LVA (± SD)Mean peritumoral LVD (± SD)Mean peritumoral LVA (± SD)
  • LVD: lymphatic vessel density; LVA: lymphatic vessel area; SD: standard deviation.

  • a

    P = 0.003.

  • b

    P = 0.009.

  • c

    P = 0.01.

Age (yrs)
 < 632633.5 ± 20.30.27 ± 0.1647.9 ± 11.81.0 ± 0.7
 ≥ 632630.9 ± 17.50.28 ± 0.1638.2 ± 10.30.9 ± 0.4
 Male4133.1 ± 19.00.29 ± 0.1843.9 ± 12.21.1 ± 0.6
 Female1124.9 ± 14.50.15 ± 0.0840.8 ± 11.30.6 ± 0.1
Smoking history
 No1025.5 ± 10.20.29 ± 0.1437.6 ± 3.50.9 ± 0.5
 Mild1527.4 ± 20.90.31 ± 0.2239.5 ± 9.61.5 ± 1
 Heavy2736.0 ± 20.60.29 ± 0.1946.0 ± 13.40.9 ± 0.5
Alcohol consumption
 No1619.4 ± 9.90.15 ± 0.1240.8 ± 5.10.6 ± 0.1
 Mild2032.4 ± 20.10.31 ± 0.1746.5 ± 12.91.3 ± 0.4
 Heavy1632.3 ± 19.60.39 ± 0.1743.8 ± 16.01.1 ± 0.9
 Oral cavity1826.2 ± 16.40.19 ± 0.1032.1 ± 4.90.5 ± 0.1
 Oropharynx1244.9 ± 28.70.33 ± 0.1848.9 ± 10.81.2 ± 0.6
 Larynx2231.5 ± 18.00.30 ± 0.1944.0 ± 12.50.5 ± 0.1
T classification
 1–23232.1 ± 20.00.26 ± 0.143.2 ± 13.00.9 ± 0.6
 3–42032.9 ± 16.00.31 ± 0.1843.9 ± 9.81.2 ± 0.4
TNM stage
 I–II2324.5 ± 16.20.18 ± 0.1437.8 ± 7.00.6 ± 0.2
 III–IV2937.5 ± 19.80.36 ± 0.1747.0 ± 14.41.3 ± 0.6
Lymph node metastasis
 Negative3525.4 ± 16.10.22 ± 0.1639.0 ± 8.90.7 ± 0.3
 Positive1745.7 ± 19.0a0.38 ± 0.17b51.8 ± 12.8c1.4 ± 0.7c
Histologic differentiation
 Well1835.6 ± 18.20.30 ± 0.1845.9 ± 16.20.9 ± 0.5
 Moderate2329.5 ± 19.10.29 ± 0.2044.6 ± 8.91.4 ± 0.7
 Poor1132.4 ± 22.40.25 ± 0.1637.8 ± 12.10.6 ± 0.1
Tumor inflammation
 Absent/low2030.6 ± 16.60.32 ± 0.1838.1 ± 3.90.8 ± 0.7
 Moderate/intense3233.4 ± 20.90.30 ± 0.1945.8 ± 12.71.1 ± 0.4

To determine which of the parameters studied best predicted the presence of lymph node metastasis in the current series of HNSCC cases, we performed logistic regression analysis with lymphangiogenesis parameters, T classification, tumor site, and histologic grade as covariates. The results demonstrated that carcinomas with a high peritumoral LVD and lymphatic area (with the median value as the cutoff value) had a significantly higher risk of lymph node metastasis (P = 0.01, odds ratio [OR] = 11.32; and P = 0.001, OR = 14.05, respectively). The other parameters were not found to be associated with an increased risk of lymph node metastasis in this model.

We then explored the effect of lymphangiogenesis parameters on the disease-free interval and overall survival using the median value and the upper quartile as the cutoff values. Carcinomas with higher lymphangiogenesis parameters tended to have a shorter disease-free interval and a worse overall survival, but no parameter was found to reach statistical significance (data not shown).


In the current study, we report a series of 52 HNSCC that were characterized for the presence of lymphatic vessels using immunohistochemistry coupled with computer-assisted morphometric analysis. The existence and biologic function of lymphatic vessels within human tumors is a controversial issue, mainly because of the lack of specific markers for lymphatic endothelium. Recently, antibodies recognizing lymphatic endothelium have become available for immunohistochemical studies of tumor tissue, providing important new insights into the process of tumor-associated lymphatic formation and its possible clinical relevance. The most commonly used lymphatic marker is the LYVE-1 antibody, which recognizes a hyaluronan receptor that is present on normal and tumor-associated lymphatic endothelial cells. Other markers that have been employed in tumor tissue evaluation are the VEGF-3 receptor, podoplanin, and desmoplakin. Considering the relative specificity of these markers to recognize lymphatic vessels in histologic sections, studies employing different markers are needed to elucidate any correlation between lymphangiogenesis and tumor growth, metastases, and prognosis. In the current study, we used the monoclonal antibody D2-40, which has been developed against a testicular oncofetal antigen, but it also has been demonstrated to recognize tumor-associated lymphatic vessels in breast carcinoma, colonic adenomas, and adenocarcinomas.5–8 We noted that intratumoral and peritumoral lymphatics are present in specimens of HNSCC, and that they are more numerous and occupy a larger area in carcinomas with lymph node metastasis compared with carcinomas without lymphatic dissemination. In addition, a higher number and larger vessel area of peritumoral lymphatics were indicative of a greater risk of developing lymph node metastasis.

Two recent studies have examined tumor lymphatics in HNSCC employing the LYVE-1 antibody on paraffin-embedded tumor tissue sections, followed by vessel counting.2, 3 Both reported the presence of intratumoral and peritumoral lymphatics. In one report, a positive correlation was found between intratumoral LVD and lymph node metastasis in the subgroup of oropharyngeal squamous cell carcinomas only,2 whereas in the other report this correlation was observed in the entire series of 71 carcinomas.3 In addition, double immunostaining with antibodies to LYVE-1 and the proliferation marker Ki-67 showed that a proportion of intratumoral lymphatics present in HNSCC are indeed actively proliferating vessels and therefore it is likely that they are not preexisting vessels entrapped in the growth of the tumor.2 In another study employing the PA2.26 marker, the presence of intratumoral lymphatics was found to be correlated with a high risk of locoregional recurrence in patients with early-stage oral squamous cell carcinoma, without cervical lymph node involvement.10 Although to our knowledge these studies have assessed only the number of tumor lymphatic vessels in correlation with clinicopathologic parameters, in the current study we also have observed that lymphatic vessel area also is increased in HNSCC metastatic to the lymph nodes, indicating that enlargement of the lymphatic vessels also may be a relevant aspect of tumor lymphangiogenesis associated with a higher risk of developing lymph node metastasis. Moreover, our analysis indicates that lymphatics present in the peritumoral area most likely are more relevant to the process of lymphatic dissemination, because peritumoral LVD and relative vascular area were found to be independently related to lymph node metastasis on multivariate analysis. Taken together, these results support the hypothesis that tumor-associated lymphangiogenesis is an active process that plays a clinically relevant role as the route for lymphatic dissemination in HNSCC.

A critical point that to our knowledge remains to be elucidated is whether intratumoral lymphatic vessels are relevant for tumor dissemination, as opposed to peritumoral lymphatic vessels. Functional studies in animal models have shown that lymphatic vessels present in the peripheral portion of tumors are enlarged and perfused, whereas those present in the central portion are compressed and nonfunctional.11 This experimental evidence has led to the conclusion that the functional lymphatics in the tumor margin alone may be sufficient for lymphatic metastasis.11 Although the process of lymphatic invasion by tumor cells is likely to be different from interstitial fluid uptake, involving complex molecular interactions between tumor cells and lymphatic endothelial cells, the majority of studies of human tumors indicate a stronger correlation between peritumoral lymphangiogenesis and tumor aggressiveness.9, 12–14 In agreement with this finding, we observed that in HNSCC specimens in the current study, the peritumoral lymphatics were more numerous and larger than the intratumoral lymphatics, and that the number and relative area occupied by the peritumoral lymphatics were the best predictor of lymph node metastasis.

Previous studies have shown that HNSCC cell lines and tumors produce lymphangiogenic factors such as VEGF-C and VEGF-D, and that VEGF-C appears to be frequently overexpressed in tumor tissue samples compared with normal mucosa.15 In addition, VEGF-C overexpression is reported to be correlated strongly with the presence of lymph node metastasis in HNSCC patients.15 Therefore, lymphangiogenic factors produced by tumor cells may induce proliferation and dilatation of peritumoral lymphatics, as well as proliferation of intratumoral lymphatics, thereby favoring the metastatic spread. However, the biologic relevance of peritumoral and intratumoral lymphangiogenesis may be different, and the presence of proliferating intratumoral lymphatics could merely be a sign of the active formation of tumoral lymphatics, even if these structures may not be involved in tumor spread directly.

Evaluation of the impact of tumor lymphangiogenesis on patient survival has led to controversial results. In HNSCC, the presence of intratumoral LYVE-1-positive lymphatic vessels was found to be associated with a significantly higher risk of local disease recurrence and a poorer prognosis,3 but these correlations were not found on multivariate analysis. In the current study, we did not observe any significant correlation between parameters of lymphangiogenesis and disease-free or overall survival, considering both intratumoral and peritumoral lymphatics. However, the results of the current study support the possibility of using the determination of tumor lymphangiogenesis to identify patients with HNSCC who are at risk of developing cervical lymph node metastasis. If a positive correlation between lymphangiogenesis and cervical lymph node involvement is confirmed in further studies, this parameter could be useful for selecting HNSCC patients who are more susceptible to metastatic spread via lymphatic pathways to undergo elective cervical lymph node dissection.