Purpose: Lymph node (LN) lymphangiogenesis has recently been shown to be important in the premetastatic niche of sentinel LNs. To study its role in the further metastatic spread of human breast cancer, we investigated the association of angiogenesis and lymphangiogenesis in sentinel LN metastases with the presence of nonsentinel LN metastases in breast cancer patients with a positive sentinel LN.
Experimental Design: Angiogenesis and lymphangiogenesis—quantified as endothelial cell proliferation fraction (ECP%) and lymphatic ECP fraction (LECP%)—were assessed in sentinel LN metastases of 65 T1/T2 patients with breast cancer using CD34/Ki67 and D2-40/Ki67 immunohistochemical double stains. Correlations were analyzed between nonsentinel LN status, LECP%, and other clinicopathologic variables (number of involved sentinel LNs, size of the primary tumor and LN metastasis, presence of lymphovascular invasion in the primary tumor, and of extracapsular growth in the sentinel LN metastasis).
Results: Thirty seven out of 65 patients (56.9%) had at least one involved nonsentinel LN. Size of the sentinel LN metastasis (P = 0.001), lymphovascular invasion (P = 0.02), extracapsular growth (P = 0.02), and LECP% (P = 0.01) were correlated with a positive nonsentinel LN status. The multivariate logistic regression model retained high LECP% (odds ratios = 4.2, P = 0.01) and the presence of extracapsular growth (odds ratios = 3.38, P = 0.04) as independently associated with the presence of nonsentinel LN metastases.
Conclusions: Increased sentinel LN metastasis lymphangiogenesis is associated with metastatic involvement of nonsentinel axillary LNs. These are the first data sustaining the hypothesis that sentinel LN lymphangiogenesis is involved in further metastatic spread of human breast cancer.
- Breast cancer
- Sentinel lymph node
Primary breast tumors can metastasize in two different ways: tumor cells can leave the primary tumor and spread by the lymphatic or by the hematogenous route leading to the formation of locoregional lymph node (LN) and distant metastases, respectively (1). In the lymphatic pathway, the sentinel LN is the first metastatic station and a sentinel LN free of carcinoma has a very high negative predictive value for further lymphatic dissemination in patients with breast cancer. Axillary LN dissection and associated morbidity could therefore be avoided in 65% to 70% of patients with breast cancer (2). Even in cases of metastatic involvement of the sentinel LN, metastases in further axillary LNs will only be present in 30% to 60% of the patients. Several authors studied the features of sentinel LN metastasis and of the primary tumor that can predict axillary nonsentinel LN involvement. The size of the sentinel LN metastasis has emerged as a most powerful independent predictor in several studies (3–15). Furthermore, the number of involved sentinel LNs (3, 4, 7, 10, 11, 16), extracapsular growth of the sentinel LN metastasis (3, 6, 15), the size of the primary tumor (7, 9, 10, 13, 14), and the presence of lymphovascular invasion (4, 7, 9, 14, 15) have also been associated with metastatic involvement of nonsentinel axillary LNs. Based on these factors and on other clinicopathologic variables, Van Zee et al. developed a nomogram to predict the likelihood of additional nodal metastases in breast cancer patients with a positive sentinel node biopsy (17).
The role of LN lymphangiogenesis in the further dissemination of breast cancer remains largely unexplored, especially in human samples. Due to a lack of lymphatic specific markers, the process of tumor-induced LN lymphangiogenesis has only recently been described. We have shown the induction of lymphangiogenesis in and around LN metastases in a population of 110 patients with breast cancer (18). Data in animal models have been focusing on the role of LN lymphangiogenesis in premetastatic vascular LN remodeling. In two comparable transgenic mouse models of chemically induced skin cancer, the induction of lymphangiogenesis in LNs draining the primary tumor has been shown and is suggested to be driven by vascular endothelial growth factor (VEGF)-A (19) and VEGF-C (20). Remarkably, the induction of lymphangiogenesis in the sentinel LN even started before tumor cells arrived. The same phenomenon of premetastatic vascular remodeling in tumor-draining LNs, with both angiogenesis and lymph sinus remodeling/dilation, has recently also been described in spontaneous metastases models of nasopharyngeal carcinoma (21) and of melanoma (22), and in samples of patients with breast cancer (21). Tumor-induced (lymph)angiogenesis in LNs is thus involved in the metastatic pathway of breast cancer by preparing future sites for tumor growth. Whether LN lymphangiogenesis is also involved in further dissemination of the disease with formation of secondary LN metastases, remains to be elucidated. In primary breast tumors, the role of lymphangiogenesis in tumor spread has been shown (23–26). Although Guidi et al. showed that intense angiogenesis in LN metastases was associated with decreased survival (27), to the best of our knowledge, no data in human samples exist on the role of sentinel LN lymphangiogenesis in the further lymphatic dissemination of breast cancer. Therefore, the aim of this study was to quantify angiogenesis and lymphangiogenesis in sentinel LN metastases of patients with breast cancer and to investigate the association of both processes with the involvement of nonsentinel axillary LNs.
Materials and Methods
Patients and samples. The medical records of all patients with T1 or T2 operable breast cancer that underwent a sentinel node biopsy (Tc99 radioisotope or patent blue dye if no or weak radioactive signal could be detected) by the same gynecological oncologist (P. van Dam) were reviewed. Only patients that had at least one metastatically involved sentinel LN on intraoperative frozen section examination were included in this study. All these patients had a complete axillary dissection. At our institution, as stated by international guidelines (28), sentinel LNs that were uninvolved on intraoperative examination were extensively investigated by step-sectioning at 200-μm levels. When a sentinel LN metastasis is then shown in one or more of the extra sections, no paraffin-embedded tissue was left for additional immunohistochemical examination of the archival material. Therefore, patients with a sentinel LN that was uninvolved on intraoperative examination but was involved on extensive examination were not included because no tissue was available for the extra analyses. All protocols were reviewed and approved by the ethical committee of the General Hospital St.-Augustinus. One paraffin block from the involved sentinel LN was selected per patient and consecutive 5-μm sections were cut for immunohistochemical analysis. If more than one sentinel LN was involved, the largest metastasis was selected. If the metastasis was no longer present on the immunohistochemistry-stained slides, those patients were excluded. Table 1 summarizes the clinicopathologic data of the final study population (n = 65). Age, tumor size, histologic type, tumor grade, estrogen and progesterone receptor status, Her2/neu and p53 status, size of the primary tumor, and the presence of lymphovascular invasion in the primary tumor were recorded by review of the pathology reports. The presence of nonsentinel LN metastases was also recorded from the pathology report: at our institution, nonsentinel LNs were grossly cut in 2-mm-thick slices and examined on standard H&E stains.
Immunohistochemistry. For the quantification of proliferating vascular and lymphatic endothelial cells, immunohistochemical double stains for Ki67/CD34 and Ki67/D2-40 were done on serial sections as previously described (18, 29). In brief, sections were first incubated with the Ki67 (dilution 1:150, clone MIB-1; Dako) primary antibody which was visualized with the Envision + dual link system (Dako). A second primary antibody against CD34 (dilution 1:50, clone QBend10; Dako) or podoplanin (dilution 1:20, clone D2-40; Dako) was then applied and visualized with the Envision G/2 system/AP (Dako). Figure 1 gives an overview of these immunohistochemical stains.
Assessment of angiogenesis, lymphangiogenesis, size, and capsular invasion. Endothelial cell proliferation fraction (ECP%) and lymphatic ECP fraction (LECP%) were assessed by counting Ki67-positive and Ki67-negative vascular and lymphatic endothelial cells, respectively. As previously described, lymphatic endothelial cells are D2-40–negative and CD34-negative or -positive, whereas blood vessel endothelial cells are D2-40–negative and CD34-positive (30). ECP% and LECP% assessments were done in the areas of highest vascular density, which were identified at low magnification (×10 ocular and ×10 objective). ECP% and LECP% were then assessed on high magnification (×10 ocular and ×40 objective). For ECP% and LECP%, we aimed at counting 200 and 100 endothelial cells, respectively. All assessments were done by two independent observers (G. Van den Eynden and M. Vandenberghe), who were blinded from other variables. A high correlation was found between the results of both observers (Fig. 2 ). Therefore, the mean of the results from the two observers was used for further statistical analysis. Furthermore, the size of the sentinel LN metastasis and the presence of extracapsular growth of the sentinel LN metastasis were assessed on H&E slides.
Statistical analysis. Statistical analyses were done using SPSS 12.0 software (SPSS). P ≤ 0.05 was considered statistically significant and 0.5 < P < 0.1, a trend towards statistical significance. Normality was tested with a Kolmogorov-Smirnov test assuming normality of data if P > 0.1. Because continuous data (e.g., ECP%, LECP%, size) were not normally distributed, correlations were analyzed with Spearman's correlation statistics. The equality of medians was tested with a Mann-Whitney U test. For some of the analyses, ECP%, LECP%, and size of the sentinel node metastases were dichotomized according to the median (ECP%, 1.5%; LECP%, 2.5%; size, 9 mm). For analyzing associations between categorical variables (e.g., extracapsular growth, lymphovascular invasion, dichotomized ECP%, LECP%, and size), the χ2 test or—when the assumptions of the χ2 test were not met—the Fisher's exact test, were used. To study the independent contribution of different factors, a multivariate linear regression analysis was used.
Characteristics of the sentinel LN metastases. The median number of sentinel LNs resected per patient was 1 (range, 1-7). In 31 of 65 cases (47.7%), more than one sentinel LN was resected. Metastatic involvement of 1, 2, 3, or 4 sentinel LNs was found in 45 (69.2%), 11 (16.9%), 8 (12.3%), and 1 (1.5%) patient, respectively. Extracapsular growth of the sentinel LN metastasis was found in 36 of 65 cases (55.4%). Table 2 shows ECP%, LECP%, and size of the sentinel LN metastases. There was a weak, but significant, correlation between ECP% and LECP% (r = 0.27, P = 0.03) and both ECP% (r = 0.30, P = 0.01) and LECP% (r = 0.34, P = 0.005) correlated with the size of the sentinel LN metastases. Furthermore, sentinel LN metastases with extracapsular growth were significantly larger than metastases without extracapsular growth (P = 0.009). When samples were dichotomized according to median ECP% and LECP%, no association was found between ECP% or LECP% and extracapsular growth.
Metastatic involvement of nonsentinel axillary LNs: univariate analysis. The median number of nonsentinel axillary LNs resected was 13 (3–29). Metastatic involvement of at least one nonsentinel axillary LN was found in 37 of 65 (56.9%) patients, with the median number of involved nonsentinel axillary LNs being 3 (range, 1-18). Neither the size of the primary tumor, nor Nottingham grade, nor estrogen receptor, progesterone receptor, or Her2 status, nor the involvement of more than one sentinel LN, were correlated with the presence of nonsentinel LN metastases. In 25 of 36 (69.4%) patients with extracapsular growth compared with 12 of 29 (41.4%) patients without extracapsular growth of the sentinel LN metastasis, involvement of the nonsentinel axillary LNs was shown (P = 0.02). Furthermore, involvement of nonsentinel axillary LNs was more frequent in primary tumors with (lympho)vascular invasion (P = 0.02). In patients with tumor involvement of the nonsentinel LNs, the median size of the sentinel LN metastases and median LECP% were, respectively, 11.0 mm and 5.0%, compared with 6.5 mm and 1.1% in patients without metastases in the nonsentinel axillary LNs (Psize = 0.001; PLECP% = 0.01). There was no difference in median ECP% of the sentinel LN metastases between patients with and without metastases in nonsentinel LNs. Figure 3 compares ECP%, LECP%, and size of the sentinel LN metastases between patients with and without involvement of additional axillary LNs. Sentinel node metastases with a high LECP% (P = 0.002) or with a size >9 mm (P = 0.02) were more frequently associated with metastatic involvement of the nonsentinel axillary LNs. For ECP%, this was not the case (Table 3 ). When only patients with metastatic involvement of nonsentinel LNs were taken into account, none of the abovementioned factors were correlated with the number of involved nonsentinel LNs.
Metastatic involvement of nonsentinel axillary LNs: multivariate analysis. To study the association of each individual factor with the status of the nonsentinel axillary LNs, a multivariate logistic regression model was built. The model included all variables univariately associated with involvement of the nonsentinel axillary LNs (lymphovascular invasion, extracapsular growth, size of the sentinel LN metastasis, LECP%). In order to be able to calculate odds ratios, sentinel LN metastasis size and LECP% were included as a dichotomized (according to median) variables. Table 4 shows the results. In this model, LECP% showed the strongest association (odds ratios = 4.63, P = 0.01) with the presence of metastases in the nonsentinel axillary LNs. The presence of extracapsular growth (odds ratios = 3.38, P = 0.04) was also independently associated, and for lymphovascular invasion in the primary tumor, a trend to statistical significance was found (odds ratios = 2.88, P = 0.08).
To study a possible role of sentinel LN angiogenesis and lymphangiogenesis in lymphatic dissemination in patients with breast cancer, we have investigated the association of these processes—quantified as ECP% and LECP%, respectively—with the presence of nonsentinel LN metastases in breast cancer patients with a positive sentinel node biopsy. Increased lymphangiogenesis, not angiogenesis, was associated with an increased frequency of involved nonsentinel LNs. In the multivariate model, LECP% was independently associated with the presence of nonsentinel LN metastases.
These findings support the hypothesis that sentinel LN lymphangiogenesis is involved in further lymphatic spread of human breast cancer. As previously mentioned, most animal studies on sentinel LN lymphangiogenesis focused on its role in the premetastatic remodeling of the sentinel LN (premetastatic niche). However, some data from these models indeed suggest that sentinel LN lymphangiogenesis is also involved in further metastatic dissemination once sentinel LN metastases have been formed (19, 20, 22). Hirakawa et al. showed that lymphangiogenesis in sentinel LN metastases of VEGF-A– or VEGF-C–overexpressing mice was higher than in control mice and that the percentage of mice with sentinel LN metastases that also developed nonsentinel LN metastases was also significantly higher in VEGF-A or VEGF-C transgenic mice (19, 20). In these animal models, LN lymphangiogenesis also seemed to be involved in distant blood-borne metastasis (20). To the best of our knowledge, our data is the first to confirm that LN lymphangiogenesis is also involved in further spread of human breast cancer. The fact that lymphangiogenesis is involved in metastasis to and from axillary LNs, makes it a possible therapeutic target in patients with breast cancer. Instead of a mechanistic relationship between sentinel LN lymphangiogenesis and further metastatic spread, another explanation for our findings might be that sentinel LN lymphangiogenesis is just a marker of underlying aggressive tumor biology. The difference is difficult to determine with a translational approach. However, the fact that lymphangiogenesis, and not angiogenesis, predicted nonsentinel LN involvement and the fact that there was no correlation between extracapsular growth and angiogenesis or lymphangiogenesis favors a specific mechanistic relationship. In oral squamous cell carcinoma, gene expression profiles associated with aggressive tumor behavior are also associated with the presence of extracapsular spread in LN metastases (31). Furthermore, the fact that extracapsular growth of the sentinel LN metastases was also related univariately and multivariately to an increased risk of involvement of nonsentinel LNs, suggests that further lymphatic spread from sentinel LN metastases might also use a combination of newly formed and preexisting perinodal lymph vessels.
Although valuable from a tumor biological point of view and although LECP% was retained as the strongest predictor of nonsentinel LN metastases in the logistic regression model, our data do not support the use of sentinel LN LECP% as a predictive factor or as a biomarker for further axillary involvement in patients with breast cancer and a positive sentinel LN. The small, selected, and very specific population of sentinel LN–positive patients with breast cancer and the design as a tumor biology and not a tumor biomarker study do not allow conclusions about the use of LECP% in this context. As stated in the REMARK criteria, tumor marker studies need to be designed and reported specifically for that purpose (32). Because additional sections were needed for the immunohistochemical analysis and assessment of ECP% and LECP%, we had to select sentinel LN metastases from which enough material was available. Therefore, patients with isolated tumor cells, micrometastases, and even very small macrometastases could not be included. This resulted in a population with an overrepresentation of large and extensive sentinel LN metastases, explaining the high frequency of involvement of the nonsentinel LNs in comparison to other studies (4). The overrepresentation of large sentinel LN metastases might also be responsible for the fact that the size of the sentinel LN metastasis and the number of involved LNs were not retained in our multivariate model, although it is one of the most powerful independent predictors of further axillary LN involvement in most other studies (3–15).
In conclusion, we show that high sentinel LN lymphangiogenesis is associated with the presence of nonsentinel LN metastases in breast cancer patients with a positive sentinel LN biopsy, both in univariate and in multivariate analyses. Because our data are in line with findings in animal models and because sentinel LN lymphangiogenesis and not angiogenesis was associated with further axillary LN involvement, this suggests a mechanistic role for LN lymphangiogenesis in the lymphatic spread of breast cancer. To be able to use sentinel LN lymphangiogenesis as a therapeutic target in patients with breast cancer, further studies of the molecular mechanisms involved are needed. Although the molecular determinants of (lymph)angiogenesis in primary breast tumors have extensively been studied, the microenvironmental differences between primary tumors and LN metastases might lead to different interactions between tumor and stroma cells and to a different molecular basis of (lymph)angiogenesis at different tumor growth sites (33).
We thank Liliane Schelfout and the technical staff of the laboratory for pathology of the General Hospital St.-Augustinus for expert technical assistance.
Grant support: Fund for Scientific Research Flanders: G. Van den Eyden is a research assistant and grants L.3.058.06N and 60-10004N.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received May 21, 2007.
- Revision received June 29, 2007.
- Accepted July 2, 2007.