Dynamic molecular changes associated with epithelial–mesenchymal transition and subsequent mesenchymal–epithelial transition in the early phase of metastatic tumor formation

Authors

  • Keiju Aokage,

    1. Pathology Division, Research Center for Innovative Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    2. Division of Thoracic Surgery, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    Search for more papers by this author
  • Genichiro Ishii,

    Corresponding author
    1. Pathology Division, Research Center for Innovative Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    • Pathology Division, Research Center for Innovative Oncology, 6-5-1 Kashiwanoha, Kashiwa, Chiba 277–8577, Japan
    Search for more papers by this author
    • Fax: +81-4-7134-6865,

  • Yoichi Ohtaki,

    1. Pathology Division, Research Center for Innovative Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    Search for more papers by this author
  • Yoko Yamaguchi,

    1. Pathology Division, Research Center for Innovative Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    Search for more papers by this author
  • Tomoyuki Hishida,

    1. Division of Thoracic Surgery, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    Search for more papers by this author
  • Junji Yoshida,

    1. Division of Thoracic Surgery, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    Search for more papers by this author
  • Mitsuyo Nishimura,

    1. Division of Thoracic Surgery, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    Search for more papers by this author
  • Kanji Nagai,

    1. Division of Thoracic Surgery, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    Search for more papers by this author
  • Atsushi Ochiai

    Corresponding author
    1. Pathology Division, Research Center for Innovative Oncology, National Cancer Center Hospital East, Kashiwa, Chiba, Japan
    • Pathology Division, Research Center for Innovative Oncology, 6-5-1 Kashiwanoha, Kashiwa, Chiba 277–8577, Japan
    Search for more papers by this author
    • Fax: +81-4-7134-6865,


Abstract

Metastatic tumor formation via vessel route begins with cancer cell extravasation from vessel lumen, migration into the connective tissue surrounding vessels, and invasion into target organ parenchyma. Epithelial–mesenchymal transition (EMT) and mesenchymal–epithelial transition (MET) have been recognized to play an important role in metastatic process, however, how and where these biological changes take place in the early phase of metastatic tumor development has never been clarified. We morphologically evaluated 34 small intrapulmonary metastases formed after cancer cell extravasation from lymphatics (lymphogenic metastasis) and 40 formed in the absence of extravasation (aerogenous metastasis) in human specimens and found that isolated or small clusters of invasive cancer cells (tumor budding) were frequently observed in lymphogenic metastasis (24/34; 71%), but were never observed within aerogenous metastasis. We immunostained 34 lymphogenic metastases for 13 molecular markers of EMT and MET and scored the immunostaining intensity of cancer cells floating in lymphatic vessels (LVs), migrating into the connective tissue surrounding vessels [bronchovascular bundle (BVB)], and growing in lung parenchyma (LP). Cancer cells within BVBs stained more weakly for E-cadherin (p < 0.001), β-catenin (p < 0.001), and Geminin (p < 0.001) and more strongly for MMP-7 (p = 0.046) and Laminin-5 γ2 (p = 0.037) than tumor cells in LVs. However, cancer cells in LP exhibited resurgent E-cadherin (p = 0.011), β-catenin (p < 0.001), and Geminin (p = 0.037) expression and reduced MMP-7 (p = 0.038) and Laminin-5 γ2 (p = 0.001) expression in comparison with cancer cells in BVBs. Our results suggested that in the early phase of metastatic tumor formation cancer cells undergo dynamic phenotypic change associated with EMT and subsequent MET.

The cause of death of most cancer patients is the development of metastases separated from primary tumor. A better understanding of the metastatic process will provide clues that will lead to the development of new therapeutic strategies to control metastasis. However, the metastatic process includes various complex steps; the cellular and molecular mechanisms involved in the process are the topic of constant debate and inspire enthusiasm in many researchers.

Folkman reported that neoplastic tumors first undergo an avascular growth phase to a size not much more than a few millimeters in diameter during tumorigenesis, and the same has been reported to be true of metastatic tumors.1–3 The avascular growth phase includes a process in which tumor cells in vessels migrate into the surrounding tissue and colonize without angiogenesis at distant site. During this phase, a small number of tumor cells including cancer initiating cells, which survive, extravasate and colonize interacting with the stromal cells in the connective tissue surrounding vessels. The tumor cells then invade the parenchyma of the target organ parenchyma followed by the development of metastatic lesions. When considering the molecular mechanisms involved in the metastatic process, it is important to examine the dynamic molecular changes in cancer cells during this process. However, most of the studies of metastasis have targeted cancer cells that form mature metastatic tumors in the vascular phase.

So far, disseminated tumor cells in bone marrow as hematogenous micrometastasis and micrometastasis to sentinel lymph node have been discussed about its identification and prognostic significance.4–6 However, none of the studies have elucidated the precise molecular mechanisms of tumor cell implantation into the target organ. Al-Mehdi et al. examined the steps in early hematogenous metastasis by epifluorescence microscopy in mouse and rat lung7 and reported findings that metastasis was initiated by attachment of tumor cells to the vascular endothelium and that hematogenous metastasis developed from the proliferation of attached intravascular tumor cells. Although it is very important to observe metastatic tumor development in the early phase to identify factors involved in metastatic tumor development, none of the studies have elucidated the dynamic molecular changes in early metastatic tumor and the precise process remains unclear.

The lung is the most common site of metastasis by primary non-small cell lung cancer (NSCLC)8 and three possible routes of metastasis by NSCLC have been postulated: (i) a lymphatic vessel (LV) route, (ii) a blood vessel route and (iii) an airway route.9 Tiny occult intrapulmonary metastatic tumors (micrometastases) are often detected when surgical specimens obtained from patients with primary lung cancer are examined under a light microscope. Most of them undetectable by preoperative chest computed tomography. These tiny metastatic tumors provide an optimal model for investigation of early tumor development.

The discovery of the epithelial–mesenchymal transition (EMT) in tumor metastasis is a relatively recent event in oncology. An EMT is a culmination of transcriptional events and following protein modifications in response to an extracellular stimulus and allows cells to separate, lose the apico-basal polarity and gain motility. Tumor progression involves the occurrence of EMTs in which tumor cells acquire a more invasive and metastatic phenotype.10 Loss of epithelial phenotype seems to be heavily involved in EMT. E-cadherin and β-catenin that E-cadherin mediate hemophilic interactions by connecting to actin microfilaments indirectomry via are emerging as one of the caretakers of the epithelial phenotype. Mesenchymal–epithelial transitions (METs), the reverse phenomenon of EMTs, have also been recognized during tumor progression.11, 12 The disseminated mesenchymal tumor cells must undergo the reverse transition, MET, at the site of metastases, as metastatic tumors recapitulate the pathology of their corresponding primary tumors. There have been only a few reports that have investigated how and where these biological changes take place in the early process of metastatic tumor formation.11, 13 The aim of this study was to determine how the biological features of tumor cells change dynamically during the early metastatic tumor development.

Material and Methods

Patient selection

Examination of the surgical specimens of the 3,161 consecutive patients who underwent surgical resection of a primary lung cancer at the National Cancer Center Hospital East, Chiba, Japan between July 1992 and October 2008 revealed intrapulmonary metastasis by the NSCLC in 222 of them. We extracted the cases of primary adenocarcinoma of the lung alone and evaluated the 233 metastatic tumors in the 129 patients with a pulmonary metastasis (PM) less than 10 mm in diameter and for whom sufficient data were available to analyze in this study. All specimens were collected and analyzed after the subjects gave their written informed consent.

Histological studies

The surgical specimens examined had been fixed in 10% formalin or 100% methyl alcohol. These specimens were sliced in the maximum of primary tumor and all subdivided pieces were embedded in paraffin. We also identified all peripheral subsegmental bronchus and made the divided pieces involving the subsegmental bronchus even in the section without tumor lesions. These were also embedded in paraffin. Median pieces of tissue block were 23 pieces in each case. All serial 4-μm sections were stained with hematoxylin and eosin (HE) method by the Alcian blue-periodic acid-Schiff method for cytoplasmic mucin production and by the Elastica van Gieson (EVG) or Victoria-blue van Gieson (VVG) method for elastic fibers. All histological materials included in the series were initially assessed by pathologists. The materials were subsequently reviewed by two pathologists (K.A. and G.I.) to ascertain the presence of PM and assess the histopathological features of both the primary and metastatic tumors. An intrapulmonary metastasis was defined as an independent tumor having the same histopathological features including growth pattern, cell size and nuclear atypia compared with the primary tumor and was differentiated from synchronous multiple primary lung cancer referring to the criteria established by Martini and Melamed.14 Pathological stage was determined based on the TNM classification of the International Union Against Cancer (UICC).15 Histological typing of the primary tumors was performed based on the World Health Organization classification of cell types, and the types divided into five subtypes: BAC (bronchioalveolar carcinoma) (non-mucinous BAC or mucinous BAC), acinar, papillary, solid adenocarcinoma with mucin production and mixed subtype.16 Lymphatic permeation was evaluated on HE-stained sections fundamentally (34 metastatic tumors were stained with D2-40 to identify lymphatic endothelial cells) and was concluded to be present when tumor cells floating in LV with no supporting smooth muscle or elastic fiber were identified. Vascular invasion was considered to be present when tumor cells in blood vessels were identified on EVG or VVG-stained sections. An intrapulmonary metastasis was divided into: PM1, defined as the presence of a metastatic lesion in the same lobe as the primary lesion, and PM2, defined as the presence of a metastatic lesion in the different lobe as the primary lesion.

Morphological characteristics of metastatic tumor via the lymphatic vessel route

We focused on two metastatic mechanisms; metastasis formed with cancer cell extravasation from lymphatics (lymphogenic) and metastasis formed without cancer cell extravasation (aerogenous). An intrapulmonary metastatic tumor was judged to have formed by the aerogenous mechanism if all of the following criteria were met: absence of lymph node metastasis and no lymphatic/vascular invasion in the primary tumor, the metastatic tumor revealing predominant alveolar replacement growth and no lymphatic permeation within the metastatic tumor.17 An intrapulmonary metastatic tumor was judged to have formed by the lymphogenic mechanism if all of the following criteria were met: both the primary and the metastatic tumor positive for lymphatic permeation and the presence of a metastatic focus in the lung parenchyma (LP) adjacent to a bronchovascular bundle (BVB) involved by lymphatic permeation. As a result, 40 metastatic tumors were diagnosed as aerogenous, and the 34 metastatic tumors were diagnosed as lymphogenic. The other 159 metastatic tumors failed to fulfill the above criteria and were considered to have formed by metastasis via an unknown route.

The extent of stromal fibrosis in the metastatic tumor was evaluated as: none/mild or severe. We also examined for the state of lymphocyte infiltration, the presence of tumor budding, defined as a single dissociated cancer cell or cluster of up to five cancer cells, the presence of a lepidic growth pattern, and front formation, i.e., pathological contact between cancer cells and normal bronchial epithelium.18, 19 The tumor formed by two metastatic route, i.e.,via the LV route and the airway route, were compared in regard with these morphological findings.

Comparison between tumor cell immunophenotypes in lymphatic vessels, connective tissue surrounding vessels (bronchovascular bundles) and the lung parenchyma

The invasion-metastasis cascade via the LV route generally exhibits a complex mechanism and includes a multistep process. In the first step, tumor cells that have separated from the primary tumor intravasate into a LV. The transmigrated cells extravasate from the LV and invade the connective tissue surrounding vessels, which was generally described as BVB. Intrapulmonary metastatic tumors are generated by these cells invaded the LP. Figure 1a shows a schema of the process of development of lymphogenic metastases. We analyzed and compared cancer cell phenotypes in LV, in BVB and in LP.

Figure 1.

(a) Schema of the early phase of metastatic tumor formation. Tumor cells in a lymphatic vessel (LV) extravasate and migrate to the connective tissue surrounding vessels [bronchovascular bundle (BVB)], and then invade the lung parenchyma (LP), where they form a metastatic tumor. (b) Histological appearance of tumor cells in LV, BVB and LP (HE staining). Inset c shows tumor cells in LV. Inset e shows tumor cells in BVB. Inset f shows tumor cells forming a pulmonary metastasis in LP. (c) High-power view of inset c in (b). Many tumor cell clusters are floating in LV. (d) D2-40 immunostaining of lymphatic endothelial cells. (e) High-power view of inset e in (b). Some dissociated tumor cells have invaded the BVB. Black arrowheads point to budding tumor cells. (f) High-power view of inset f in (b). Tumor cells have formed a metastatic tumor in LP.

Antibodies and immunohistochemical staining

Fifteen molecular markers were selected for investigation in this study. The staining procedures were performed according to the manufacturer's protocols. Podoplanin (clone D2-40, Signet, Princeton, NJ) was used to detect LV. CD 31 (clone JC/70A, Dako cytomation, Carpinteria, CA) was used to assess the extent of lymphangiogenesis in the metastatic tumors. The other antibodies consisted of markers for cellular adhesion molecules (E-cadherin; clone 36, BD Biosciences, San Jose, CA, β-catenin; clone 14/β-catenin, BD Biosciences, Laminin-5 γ2; clone D4B5, Chemicon, Temecula, CA, and CD44; clone DF1485, Novocastra, Newcasyle upon Tyne, UK), cell proliferation-related protein (Geminin; clone EM6, Novocastra, Eg5; clone 20, BD Biosciences), a growth factor receptor (EGFR; clone H11, Dako Cytomation), a hypoxia induced protein (CA IX; polyclonal, Novus Biologicals, Littleton, CO), as EMT marker (ZEB-1; polyclonal, Sigma-Aldrich, St. Louis), a tumor-associated macrophage marker (CD204; clone SRA-E5, Trans Genic, Hyogo, Japan), an apoptosis-associated marker (cleaved-caspase 3; polyclonal, Cell Signaling Technology, MA), and others (MMP7; clone 141-7B2, Daiichi Fine Chemical, Toyama, Japan and SPA; clone PE-10, Dako Cytomation). Immunostaining was performed on 4-μm paraffin-embedded tissue sections. The slides were deparaffinized in xylene and dehydrated in a graded ethanol series, and endogenous peroxidase was blocked with 3% hydrogen peroxide in absolute methyl alcohol. After epitope retrieval, they were washed with phosphate-buffered saline and incubated overnight with primary antibodies. The reaction products were stained with diaminobenzidine and counter stained with hematoxylin.

Immunohistochemical scoring

All tissue sections stained were semiquantitatively scored and evaluated independently under a light microscope by two pathologists (K.A. and G.I.), who had no knowledge of the patients' clinicopathological data. Labeling scores were calculated by multiplying the percentage of positive tumor cells per each lesion (0–100%); tumor cells in (i) LV, (ii) BVBs and (iii) LPs forming metastatic lesion, by the staining intensity level (0 = negative, 1 = weak and 2 = strong) except for Geminin, Cleaved caspase-3, and CD 204, which positive tumor cells of the two former antibodies were counted per 100 tumor cells and CD 204 positive cells were counted in high-power micro field (400×; 0.0625 mm2), respectively. We confirmed that positive control tissues were stained by each antibody, and we also performed negative control studies that were made without primary antigen in all antibodies. When their antibody evaluation differed, the observers discussed the results, re-examining the slides if necessary, until agreement was reached.

Statistical analysis

The Mann–Whitney U test or Fisher's exact test was used to test the differences between two groups for statistical significance. All p values reported are two-sided, and the significance level was set at less than 0.05. The analyses were performed with the SPSS 11.0 statistical software program (Dr. SPSS II for Windows, standard version 11.0, SPSS, Chicago, IL) and Graph Pad Prism statistical software program (Prism for Windows, Version 5.02, Graph Pad Software, La Jolla, CA)

Results

Clinicopathologic findings

The clinicopathologic data of the 129 patients are summarized in Table 1. Their median age was 67 years old, and their age range was 30–84 years old. The primary lung tumor of 125 of the patients (96.9%) was pathologically diagnosed as mixed subtype adenocarcinoma. All patients had intrapulmonary metastasis, and in 119 of the patients (90.7%) it was located in the same lobe as the primary lesion (PM1) and the other ten patients (7.8%) it located in a different lobe from the primary lesion (PM2). Both the lymphatic permeation and vascular invasion were observed in 87 patients (67.4%).

Table 1. Patients characteristics
inline image

Morphological features of small lymphogenic metastases

The mean size of the lymphogenic metastases and aerogenous metastases was 2.4 mm and 2.5 mm, respectively, and the difference was not statistically significant (Table 2). The small lymphogenic metastases exhibited significantly less lepidic growth pattern than the aerogenous metastases (lymphogenic: 11/34; aerogenous: 40/40). Significantly less front formation between cancer cells and normal bronchial epithelium was observed in the small lymphogenic metastases than in the aerogenous metastases (lymphogenic: 1/34; aerogenous: 17/40). Tumor budding in the connective tissue was observed only in the small lymphogenic metastases (lymphogenic: 24/34; aerogenous: 0/40). The results of examination for a tumor stromal reaction showed that the small lymphogenic metastases induced lymphocyte infiltration (lymphogenic: 29/34; aerogenous: 0/40) and severe stromal fibrosis (lymphogenic: 11/34; aerogenous: 1/40) more than aerogenous metastases in LP.

Table 2. Morphological comparison between the metastases via lymphatic vessel and airway route (≤10 mm)
inline image

Phenotypic changes of tumor cells in lymphatic vessels, bronchovascular bundles and lung parenchyma

A plausible schema of the process of lymphogenic metastasis is shown in Figure 1a. Tumor cells in a LV extravasate and migrate to a BVB. These tumor cells invade the LP and generate intrapulmonary metastatic lesions. Figure 1b shows a representative small lymphogenic metastasis in a surgical specimen stained with HE method. Figures 1c, 1e and 1f are higher power views of tumor cells in LV (inset c in Fig. 1b), BVB (inset e in Fig. 1b) and LP (inset f in Fig. 1b), respectively. Lymphatic endothelial cells were identified by staining with D2-40, and we identified tumor cells in LV (Fig. 1d). Some single dissociated cells or clusters of up to five cancer cells (tumor budding) were observed in BVB (black arrowhead in Fig. 1e), and, as given in Table 2, these findings were detected in 71% of the lymphogenic metastases.

We evaluated immunohistochemical profiles by staining tumor cells at three sites: LV, BVB and LP. Comparisons of the staining score at the three sites are summarized in Figure 4 according to antibody.

Cellular adhesion molecules

The mean staining score ± standard deviation (S.D.) for E-cadherin in LV, BVB and LP was 1.03 ± 0.45, 0.6 ± 0.32 and 0.82 ± 0.36, respectively. The mean staining score ± S.D. for β-catenin in LV, BVB and LP was 1.09 ± 0.67, 0.44 ± 0.33 and 1.07 ± 0.58, respectively. The E-cadherin and β-catenin expression levels in BVB were significantly lower than in LV (E-cadherin: p < 0.001; β-catenin: p < 0.001) and LP (E-cadherin: p = 0.011; β-catenin: p < 0.001) (Figs. 2a, 2b, 3a, 3b, 4a and 4b). The mean staining score ± S.D. for Laminin-5 γ2 in LV, BVB and LP was 0.07 ± 0.11, 0.3 ± 0.4 and 0.05 ± 0.16, respectively, and the Laminin-5 γ2 expression levels in BVB were significantly higher than in LV (p = 0.037) and LP (p = 0.001) (Figs. 2c, 3c and 4c). The mean staining score ± S.D. for CD44 in LV, BVB and LP was 0.46 ± 0.51, 0.36 ± 0.46 and 0.38 ± 0.55, respectively. There were no significant differences between the levels of CD 44 expression by the cancer cells according to site (Fig. 4d).

Figure 2.

Immunohistochemical staining of tumor cells in lymphatic vessel (LV) and in the connective tissue surrounding vessels [bronchovascular bundle (BVB)]. The dotted line indicates the lymphatic endothelial wall. Black arrowheads point to tumor cells migrating into BVB. Insets are high-magnification images of tumor cells in BVB. (a) Immunostaining for E-cadherin. (b) Immunostaining for β-catenin. (c) Immunostaining for Laminin-5 γ2. (d) Immunostaining for Geminin. (e) Immunostaining for MMP7.

Figure 3.

Immunohistochemical staining of tumor cells in the connective tissue surrounding vessels [bronchovascular bundle (BVB)] and in the lung parenchyma (LP). The dotted line indicates the border between LP and BVB. Black arrowheads point to tumor cells migrating into BVB. Insets are high-magnification images of tumor cells in BVB. (a) Immunostaining for E-cadherin. (b) Immunostaining for β-catenin. (c) Immunostaining for Laminin-5 γ2. (d) Immunostaining for Geminin. (e) Immunostaining for MMP7.

Cell proliferation-related protein

Significantly more Geminin-positive cells per 100 tumor cells were detected in LV (mean ± S.D.; 15.9 ± 11.8) and LP (10.4 ± 8.0) than in BVB (5.9 ± 4.6) (LV vs. BVB: p < 0.001; BVB vs. LP: p = 0.037) (Figs. 2d, 3d and 4e). The mean staining score ± S.D. for Eg5 in LV, BVB and LP was 0.27 ± 0.25, 0.19 ± 0.18 and 0.22 ± 0.21, respectively. There were no significant differences between the levels of Eg5 expression by the tumor cells according to site (Fig. 4f).

Figure 4.

Comparison between the immunohistochemical scores of tumor cells in lymphatic vessel (LV), connective tissue surrounding vessels [bronchovascular bundle (BVB)], and lung parenchyma (LP). (a) E-cadherin, (b) β-catenin, (c) Laminin-5 γ2, (d) CD44, (e) Geminin, (f) Eg5, (g) Cleaved caspase-3, (h) ZEB-1, (i) SPA, (j) MMP7, (k) EGFR, (l) CA-IX and (m) CD204.

Apoptosis-associated protein

The mean number of cleaved caspase-3 positive cells per 100 tumor cells in LV, BVB and LP was 1.28 ± 1.93, 0.78 ± 2.34 and 0.96 ± 1.37, respectively. There were no significant differences between the numbers of cleaved caspase-3 positive cells according to site (Fig. 4g).

Hypoxia-induced protein

The mean staining score ± S.D. for CA-IX in LV, BVB and LP was 0.32 ± 0.41, 0.18 ± 0.29 and 0.22 ± 0.33, respectively. There were no significant differences between the levels of CA-IX expression according to site (Fig. 4l).

Other molecules

The mean staining score ± S.D. of MMP7 in LV, BVB and LP was 0.05 ± 0.12, 0.15 ± 0.26 and 0.05 ± 0.1, respectively, and the MMP7 expression level in BVB was significantly higher than in LV (p = 0.046) and LP (p = 0.038) (Figs. 2e, 3e and 4j). There were significantly fewer CD 204-positive macrophages in LV (mean ± S.D.; 4.9 ± 4.4) than in LP (7.8 ± 4.1; p = 0.014) and BVB (10.2 ± 5.2; p < 0.001) (Fig. 4m). The mean staining score ± S.D. for SPA in LV, BVB and LP was 0.25 ± 0.43, 0.27 ± 0.46 and 0.34 ± 0.55, respectively. The mean staining score ± S.D. of EGFR in LV, BVB and LP was 0.59 ± 0.5, 0.38 ± 0.43 and 0.53 ± 0.53, respectively. There were no significant differences between SPA and EGFR staining according to site (Figs. 4i and 4k). ZEB-1 was expressed in the nucleus of all stromal fibroblasts (data not shown), but it was not stained in tumor cells at all (Fig. 4h).

Discussion

In this study, we focused on lymphogenic intrapulmonary metastasis as a representative mechanism of metastasis through a vessel system for the following reasons: (i) the 34 metastatic tumors were concluded to have formed via the LV route based on the morphological findings, (ii) there were few tumor cells in the blood vessels (as confirmed by staining with anti-CD31 antibody) in or around the metastatic tumors in the surgical specimens that would suggest hematogenous metastasis and (iii) aerogenous metastasis is a lung-specific metastatic mechanism and seems not to be the mechanism that was generally perceived like a hematogenous or lymphogenic metastasis. We considered lymphogenic metastases to be optimal models for analyzing the morphological and dynamic phenotypic changes in early metastasis and that results would be applicable to hematogenous metastasis.

The mean size of the metastatic lesions we examined in this study was 2.4 mm, a size that corresponds to the “avascular phase” described by Folkman. The metastatic tumors in this study displayed less evidence of angiogenesis or lymphangiogenesis, as confirmed by staining with anti-CD31 antibody (data not shown).1–3 However, the morphological findings, including the presence of tumor budding cells, less front formation and high degree of stromal reaction, in the lymphogenic metastases were different from those in the aerogenous metastases, even though all of the metastatic tumors were equally small. This finding suggests that the tumor environment may vary considerably depending on the metastatic route.

When process of generating a small metastatic tumor via the LV route was divided into three stages: first stage, tumor cells floating in a LV, second stage, tumor cells extravasated into the connective tissue surrounding vessels (BVB) and the third stage, tumor cells invaded and grown in LP (Fig. 1a), we detected a single dissociated cancer cells or clusters of up to five cancer cells in BVB in 24 of 34 tumors (70.6%). This phenomenon is generally described as “tumor budding” especially in the field of colorectal cancer.18, 19 Moreover, the level of expression of E-cadherin and β-catenin in the tumor cells in BVB was lower than in the tumor cells in LV and LP, and the lower levels were compatible with the budding phenotype that has been reported previously. By contrast, Laminin-5 γ2 chain and MMP-7 expressions by the tumor cells in BVB was upregulated. Laminins constitute one of the major and ubiquitous families of basement membrane components. The Laminin-5 γ2 chain is one of the subunits composed of Laminin 5. Fragment of the Laminin-5 γ2 chain cleaved by matrix metalloproteinase works as the ligand of EGFR and promotes cell survive or motility.20, 21 It is preferentially expressed in the cytoplasm of cancer cells along the advancing edge of tumors and has been reported to be associated with tumor growth and invasiveness.22, 23 Increased expression in budding tumor cells has also been reported in the field of colorectal cancer.24–26 MMP-7 has been reported to be closely related to the invasiveness of tumor cells and to be strongly expressed in budding cells.27, 28 MMP production by tumor cells themselves is often associated with the expression of other mesenchymal markers and the loss of many epithelial phenotype.29 These morphological and phenotypic changes in tumor cells in BVB correspond to lesions undergoing an EMT, a phenomenon in which cells dissociate from an epithelium and migrate freely, and these contribute to the invasive and metastatic processes.10, 30, 31 EMT has been recognized as the phenomenon by which tumor cells in primary lesions invade the surrounding stroma. In EMT, growth factors including epidermal growth factor (EGF), hepatocytes growth factor (HGF), insulin-like growth factors, fibroblast growth factor and tumor necrosis factor-α upregulate Snail1 and/or Snail2 via signaling through their corresponding receptor tyrosine kinases: phosphinositide-3-kinase, Ras and mitogen-activated protein kinase. These Snail1/2 induce EMT by repressing the transcription of the key adherens junction protein E-cadherin.32–34 The results of this study provided the new insight that intravasated tumor cells use this EMT mechanism to extravasate and migrate into connective tissue.

Tumor cells that had invaded and grown within LP expressed higher levels of E-cadherin and β-catenin and lower levels of MMP-7 and laminin-5 γ2 than tumor cells in BVB and these findings are consistent with the concept of MET reported previously.11, 12 Tumor progression is generally considered to involve spatial and temporal occurrences of EMT, whereby tumor cells acquire a more invasive and metastatic phenotype. Moreover, several recent studies have shown that these cells must undergo the reverse transition, MET, at the site of metastases, because metastases recapitulate the pathologic features of their corresponding primary tumors.11, 12, 35 Yates et al. found that prostate cancer cells that underwent EMT with a gain of autocrine signaling and loss of E-cadherin expression expressed E-cadherin again when co-cultured with hepatocytes in vitro.36 Wells et al. also focused on the role of E-cadherin in metastasis-associated EMT and following mesenchymal–epithelial reverting transitions.13 This concept endorsed that tumor cells that had lost their epithelial characteristics and acquired a mesenchymal-like migratory phenotype in a primary tumor reactivated certain epithelial properties through a MET at the metastatic site reflecting the process in early embryonic morphogenesis that were important in tissue construction in normal development. The results of our study showed that tumor cells that had extravasated and invaded the connective tissue surrounding vessels from within LV underwent an EMT and later underwent a reverse transition, MET, in the early phase of metastatic tumor formation (Fig. 5). This dynamic phenotypic change during the early metastatic process presumably depends on the individual tissue microenvironment.

Figure 5.

Dynamic phenotypic change by tumor cells in lymphatic vessel (LV), connecting tissue surrounding vessels [bronchovascular bundle (BVB)], and lung parenchyma (LP) in regard to three tumor cell properties: cellular adhesion, proliferation and invasiveness.

Geminin is a regular of the process inhibiting DNA replication by interacting with CDt1p and prevents recruitment of the Mcm2-7p complex.37, 38 Its expression is accepted as cell proliferation marker in normal tissues and malignancies, especially during S, G2 and early M phase.39 Higher Geminin expression was seen in tumor cells in LV. Anchorage-independent growth has been used as an indicator of oncogenic transformation. The high-proliferation capability of tumor cells floating in LV suggests that they may acquire the ability to overcome cell cycle inhibitory signaling or apoptosis.40, 41 The proliferative activity of intravascular cancer cells was reported by Zhang et al.42 In contrast, the tumor cells in BVB expressed a low level of Geminin, a finding that is consistent with reports that altered expression of laminin-5 and downregulation of the E-cadherin-β-catenin complex, which is EMT phenotype, is linked focally to a nonproliferating status of budding tumor cells.26

In conclusion, as the schema in Figure 5 demonstrates, a dynamic phenotypic change that includes both EMT and MET occurs in the early phase of metastatic tumor formation. Malignant progression is based on dynamic processes that cannot be solely explained by irreversible genetic alterations and may also be regulated by the tissue microenvironment. The results of this study will raise awareness of the need to analyze not only the biological behavior of tumor cells but also the intravessel and surrounding stromal microenvironment within small metastatic foci. Further understanding of tumor cell behavior during the early metastatic process will provide insight into therapeutic strategies to overcome metastasis.

Ancillary