LS174T colorectal cancer (12) and SK-BR-3 breast cancer cells (13) were obtained from American Type Culture Collection (ATCC) (Manassas, VA, USA). MKN45 gastric cancer cells (14) were from RIKEN BioResource Center (Tsukuba, Japan). C57BL/6J and C57BL/6-Tg-CAG-EGFP mice were from SLC (Shizuoka, Japan). CAnN.Cg-Foxn1^nu/CrlCrlj nude mice were from Charles River Laboratories (Yokohama, Japan). ELISA kits and recombinant human VEGFs were from R&D Systems (Minneapolis, MN, USA). Antibody against mouse LYVE-1 was from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-mouse CD34 antibody was from Hycult Biotech (Uden, The Netherlands). Alexa Fluor 594-conjugated anti-rabbit secondary antibody, SuperScript VILO cDNA Synthesis kit and 4′,6-diamidino-2-phenylindole (DAPI) were from Invitrogen (Carlsbad, CA, USA). Histopaque-1083 was from Sigma (St. Louis, MO, USA). PCR primers and SYBR Premix Ex Taq II were from Takara Bio, Inc. (Osaka, Japan). RNeasy Plus Mini kit was from Qiagen (Tokyo, Japan). Type I collagen-coated dishes were from Iwaki (Tokyo, Japan). Tissue-Tek OCT compound was from Sakura Finetek (Tokyo, Japan). EBM-2 medium was from Lonza (Walkersville, MD, USA). RPMI-1640 medium, antibiotic-antimycotic and fetal bovine serum (FBS) were from Life Technologies (Grand Island, NY, USA). Other chemicals were of the highest quality available.

MKN45 gastric cancer, LS174T colorectal cancer and SK-BR-3 breast cancer cells were cultured in RPMI-1640 medium supplemented with 10% FBS and 1% antibiotic-antimycotic.

The culture media of human adenocarcinoma cells were collected after 72 h in culture. Contents of VEGF-A, VEGF-C, VEGF-D, epidermal growth factor (EGF) and hepatocyte growth factor (HGF) were determined by species-specific ELISA kits as previously described (15).

All animal procedures were carried out according to a protocol approved by the Animal Care and Use Committee of Gifu University Graduate School of Medicine. GFP-positive bone marrow cells were obtained from age-matched 6 week-old C57BL/6-Tg-CAG-EGFP mice by flushing the femurs and tibias with Hanks’ balanced salt solution, and mononuclear cells were isolated by density gradient centrifugation using Histopaque-1083 (16). The recipient nude mice, CAnN.Cg-Foxn1^nu/CrlCrlj, were lethally irradiated with 8.0 Gy, and received 6×10⁶ donor GFP-positive bone marrow mononuclear cells intravenously (11). Bone marrow cells from 6 week-old C57BL/6J mice were used as a negative control for GFP-positive donor cells. At 4 weeks after bone marrow transplantation, mice were used for tumor implantation models. To determine the transplantation efficiency of GFP-positive bone marrow, tail peripheral blood (10 μl) or bone marrow from recipient nude mice was collected 3 weeks after tumor implantation, and GFP-positive cells spread on cover slides were analyzed by fluorescent microscopy.

At four weeks after bone marrow transplantation, 2×10⁶ cells were subcutaneously injected into chimeric nude mice. At 3–8 weeks after tumor cell injection, mice were anesthetized, perfused with 0.1 M phosphate-buffered saline (PBS; pH 7.40) followed by 2% paraformaldehyde (PFA) in PBS and sacrificed. The tumor area was removed, fixed for 2 h with 2% PFA in PBS, and stored frozen in OCT compound (11).

The cryostat sections (6 μm) of tumor tissues were immunostained with the primary anti-body, rabbit anti-mouse LYVE-1 (1:100). The sections were next incubated with Alexa Fluor 594-conjugated anti-rabbit secondary antibody (1:200) and DAPI. Images were analyzed using a BIOREVO immunofluorescence microscope (Keyence, Osaka, Japan). To evaluate lymphatic vessel density or the proportion of GFP/LYVE-1 double-positive cells, we used the inverted-gray scale images of all data (nine fields in each sections), and calculated the total area positive for LYVE-1 or the area double-positive for GFP/LYVE-1 using NIH ImageJ software. To elucidate the histological structure positively stained with LYVE-1 or GFP, light hematoxylin staining was performed on the sections near the immunostained sections. Each result was obtained in at least 4 separate experiments.

Bone marrow mononuclear cells were collected from C57BL/6J mice as described above. Cells plated onto type 1 collagen-coated dishes and cultured in EBM-2 medium in the absence or presence of recombinant human VEGF-A (50 ng/ml), human VEGF-C (150 ng/ml) or mouse VEGF-D (150 ng/ml). For immunofluorescent staining, cells were fixed with 2% PFA for 20 min at room temperature and were stained with rabbit anti-mouse LYVE-1 (1:100) primary antibody for 2 h at room temperature. Cells were next incubated with Alexa Fluor 594-conjugated anti-rabbit secondary antibody (1:200) at room temperature for 2 h, and also counterstained with DAPI for 2 min. Images were analyzed using a BIOREVO immunofluorescence microscope.

Total RNA was extracted from bone marrow mononuclear cells using an RNeasy Plus Mini kit and 0.5 μl of RNA was reverse transcribed to cDNA using a SuperScript VILO cDNA Synthesis kit. cDNA was amplified with SYBR Premix EX Taq II and the following primers: murine LYVE-1 (sense, 5′-CAAAGCCTATTGCCACAACTCATC-3′ and antisense, 5′-AGTAGGCGCTGCTGACA-3′) and mouse ACTB/β-actin. ACTB/β-actin transcripts were measured simultaneously in all reactions as internal controls. Up to 40 PCR cycles with an annealing temperature of 60°C were performed for cDNA amplification. Relative expression values were calculated from the cycle threshold (Ct) as previously described (17).

Each result was obtained in at least 5 separate experiments. All values are expressed as means ± SEM. Values were analyzed using one-way analysis of variance, and then the significance of differences in multiple comparisons was determined using Scheffe’s multiple comparison test. P<0.01 was considered statistically significant.

We developed a tumor xenograft experimental model using chimeric nude mice with GFP-positive bone marrow cells in order to identify the contribution of marrow-derived cells in tumor lymphangiogenesis (11). Human colorectal cancer LS174T cells were subcutaneously implanted in this chimeric nude mouse model. LS174T cells formed tumor mass (Fig. 1A) three weeks after inoculation and a tumor lymphatic-rich microenvironment as detected by LYVE-1 (Fig. 1B and C). LYVE-1-positive tumor lymphatics were located not only in peritumoral tissues but also in intratumoral tissue. Consistent with the locations of tumor lymphatics, GFP-positive cells from bone marrow were detected in peritumoral and intratumoral (Fig. 1D) tissues. Nearly 60% of LYVE-1-positive cells co-expressed GFP (Fig. 1E). These results suggest the possible involvement of bone marrow-derived cells in tumor lymphangiogenesis.

In the same chimeric nude mouse model, SK-BR-3 human breast cancer cells were subcutaneously injected. At 8 weeks after inoculation, SK-BR-3 human breast cancer cells formed a tumor mass (Fig. 2A) and a tumor lymphatic-rich microenvironment as detected by LYVE-1 (Fig. 2B and C). GFP-positive cells from bone marrow were also located at peritumoral lymphatic tissues (Fig. 2D). Nearly 50% of LYVE-1-positive cells co-expressed GFP (Fig. 2E). These data also support the hypothesis that bone marrow-derived lymphatic endothelial progenitor cells are involved in the tumor lymphangiogenesis in addition to pre-existing lymphatics.

In order to identify the growth factor(s) involved in the lymphangiogenesis in the peritumoral tissues, human gastric cancer MKN45, colorectal cancer LS174 and breast cancer SK-BR-3 cells were cultured in vitro and growth factors in the culture medium were measured by specific ELISA assay. Among the growth factors tested, these three cell lines predominantly produced VEGF-A at 72 h after incubation (Fig. 3). The well-characterized lymphangiogenic factors (2,3,6–8), VEGF-C and VEGF-D, were below the detectable levels in the present ELISA assay. Furthermore, secretion of other growth factors, which are implicated in lymphangiogenesis and angiogenesis (18,19), EGF and HGF, was not detected in these cells.

The present and previous (11) results suggest that VEGF-A-rich human adenocarcinomas recruit lymphatic epithelial progenitor cells from bone marrow. We further examined whether VEGF-A directly induced the differentiation of bone marrow mononuclear cells into lymphatic epithelial progenitor cells. First, bone marrow mononuclear cells were cultured in the absence or presence of VEGF family proteins and the relative expression of LYVE-1, a specific lymphatic endothelial marker, was analyzed. VEGF-A significantly increased LYVE-1 mRNA in bone marrow cells at 6 h after administration (Fig. 4). At the same time, VEGF-C and VEGF-D tended to increase LYVE-1 mRNA, yet the increase was not statistically significant. To further confirm VEGF-A-dependent differentiation of bone marrow mononuclear cells into lymphatic endothelial lineage cells, bone marrow cells were incubated with VEGF-A and immunostained with antibody against LYVE-1. In the control culture without VEGF family proteins, most of the attached cells were positive for CD34, a marker of hematopoietic and vascular endothelial progenitor cells (data not shown), but negative for LYVE-1 (Fig. 5A). The well-characterized lymphangiogenic factors, VEGF-C (Fig. 5B) and VEGF-D (Fig. 5C), stimulated the expression of LYVE-1 in bone marrow mononuclear cells in this analysis system after incubation for 40 h; 50–60% cells were positive for LYVE-1. VEGF-A also clearly induced the increase of LYVE-1 positive bone marrow mononuclear cells (Fig. 5D). Nearly 60% of cells became positive for LYVE-1 at 40 h after incubation. These results suggest the differentiation of bone marrow mononuclear cells into lymphatic endothelial lineage cells in response to VEGF-A.