Timely detection of colorectal cancer metastases may permit improvements in their clinical management. Here, we investigated a putative role for bone marrow–derived cells in the induction of epithelial-to-mesenchymal transition (EMT) as a marker for onset of metastasis. In ectopic and orthotopic mouse models of colorectal cancer, bone marrow–derived CD11b(Itgam)+Jagged2 (Jag2)+ cells infiltrated primary tumors and surrounded tumor cells that exhibited diminished expression of E-cadherin and increased expression of vimentin, 2 hallmarks of EMT. In vitro coculture experiments showed that the bone marrow–derived CD11b+Jag2+ cells induced EMT through a Notch-dependent pathway. Using neutralizing antibodies, we imposed a blockade on CD11b+ cells' recruitment to tumors, which decreased the tumor-infiltrating CD11b+Jag2+ cell population of interest, decreasing tumor growth, restoring E-cadherin expression, and delaying EMT. In support of these results, we found that peripheral blood levels of CD11b+Jag2+ cells in mouse models of colorectal cancer and in a cohort of untreated patients with colorectal cancer were indicative of metastatic disease. In patients with colorectal cancer, the presence of circulating CD11b+Jag2+ cells was accompanied by loss of E-cadherin in the corresponding patient tumors. Taken together, our results show that bone marrow–derived CD11b+Jag2+ cells, which infiltrate primary colorectal tumors, are sufficient to induce EMT in tumor cells, thereby triggering onset of metastasis. Furthermore, they argue that quantifying circulating CD11b+Jag2+ cells in patients may offer an indicator of colorectal cancer progression to metastatic levels of the disease. Cancer Res; 73(14); 4233–46. ©2013 AACR.
Metastatic disease is a major cause of cancer-associated mortality. Despite significant advances in the treatment of primary tumors, metastases remain a significant clinical problem, likely reflecting our limited knowledge of the mechanisms governing this complex process (1). It is accepted that metastasization follows a series of interrelated steps, each of which can be rate-limiting. These steps include: local invasion by tumor cells, entry into systemic circulation (intravasation), invasion of the target organ (extravasation), and finally proliferation and growth of the secondary tumor (2). One of the major processes regulating local invasion in epithelial tumors is termed epithelial-to-mesenchymal transition (EMT; refs. 3, 4). EMT is a transcriptional regulated transdifferentiation process characterized at the tumor cell level, by a decrease in epithelial markers such as E-cadherin, loss of cell–cell adhesion, apical–basal polarity, and acquisition of mesenchymal markers such as vimentin associated with an increase in cell motility and invasion capacity (5–8). EMT has been positively correlated with increase breast and colon cancer metastasis and decreased patient survival (3, 9, 10).
In the last decade there has been increasing evidence suggesting that tumor metastasis is also regulated by nonmalignant cells of the tumor microenvironment, namely by bone marrow–derived cell populations (11). In fact distinct bone marrow–derived populations such as tumor-associated macrophages (12, 13), premetastatic niche cells (14), and endothelial progenitor cells (15) have been shown to enhance metastasization via multiple processes. Nevertheless, a direct role of bone marrow–derived cells in promoting EMT at the primary tumor has not been described, and was the focus of the present study.
Using ectopic and orthotopic colorectal cancer models in mice, we show that a population of bone marrow–derived myeloid (CD11b+F4/80+) expressing Jagged2 (Jag2) is actively recruited into colon tumors and accumulates in tumor areas undergoing EMT. Detailed analysis of this tumor: bone marrow–derived cell interaction shows the latter induce EMT via Notch activation on the tumor cells. Importantly, in vivo depletion of CD11b+ cells in ectopic colorectal cancer models reduced the recruitment of CD11b+Jag2+ cells into the tumors and significantly decreased EMT. Quantification of circulating (peripheral blood) and tumor-derived CD11b+Jag2+ cells in patients with colorectal cancer was significantly correlated with the presence of metastatic disease. Together, the data presented here reveal a novel undisclosed role for bone marrow–derived cells in inducing EMT in primary colorectal cancer and identifies a bone marrow–derived cell population that may be targeted and studied as a biomarker for colorectal cancer metastases formation.
Materials and Methods
Human peripheral blood samples collection and processing
Peripheral blood samples of patients with sequential colorectal cancer evaluated at diagnosis by the Multidisciplinary Colorectal Cancer Team were collected at the Gastroenterology Department at Instituto Português de Oncologia (IPO, Lisbon, Portugal) after informed consent and Institutional Review Board approval (IPO), in accordance with the Declaration of Helsinki. Patients were included if they had a pathology exam showing colorectal adenocarcinoma. All patients previously submitted to endoscopic, surgical, or medical treatment for colorectal cancer were excluded. Peripheral blood samples were collected in 4 EDTA-coated tubes to a total volume of 12 mL. Samples were centrifuged at 4°C for 8 minutes at 1,500 rpm. Plasma was collected and stored at −80°C. The remaining fraction was lysed using 50 mL of red cell lysis buffer for 20 minutes at room temperature. The resulting mononuclear cell fraction was washed in PBS EDTA 2 mmol/L + 0.5% bovine serum albumin (BSA) and used for further analysis. Staging of patients with colorectal cancer was done according to the American Joint Committee on Cancer (AJCC) Staging System.
Mouse strains, bone marrow transplants, ectopic, and orthotopic colon carcinoma model
Animal experiments were carried out with the approval of the Animal Care Committee and Review Board at the Instituto Gulbenkian de Ciência (Oeiras, Portugal). In vivo experiments were carried out on 4- to 8-week-old female nude mice (C57/BL6 background). For bone marrow transplants, nude mice received a whole body lethal irradiation (800–950 rads) and 24 hours later received an intravenous injection of 2–3 × 106 bone marrow mononuclear cells collected from Actin-GFP male mice (C57/BL6 background). Mice were allowed to recover for 2 to 4 weeks. After this period peripheral blood samples were collected from the facial vein in EDTA-coated tubes (Multivette 600, Sarstedt) and analyzed by flow cytometry for GFP+ cells. Mice were considered suitable for further experiments when the percentage of GFP+ cells in the peripheral blood was more than 80% of total cells. Xenografted ectopic colon carcinoma tumors were induced by inoculation of 5 × 106 HCT15, HCT116, DLD-1, or HT-29 cells (human colorectal carcinoma cell lines; these were obtained from American Type Culture Collection, in 2012, and were not passaged for more than 6 months in our Laboratory) subcutaneously in nude mice. Tumors were allowed to grow and at specific time points (1–3 weeks) mice were sacrificed and tumors were collected. Tumors were fixed (10% formalin or paraformaldehyde) and included (paraffin or gelatin, respectively), frozen at −80°C for further RNA isolation, or digested for fluorescence-activated cell sorting (FACS) analysis. Xenografted orthotopic colorectal carcinoma tumors were induced by inoculation of 1 × 106 HCT15 or HCT116 cells, into the visceral cecal wall of nude mice. Peripheral blood samples were collected from the facial vein at different time points and further processed for flow cytometric analysis. CD11b-neutralizing antibody in vivo administration was conducted as follows: briefly, 500 μg anti-CD11b (clone 5C6) neutralizing monoclonal antibodies against CD11b were administered intraperitoneally every 3 days into tumor-bearing mice, starting on day 5 postinoculation.
Isolation of bone marrow–derived cells from the tumors
Tumor samples were mechanically fragmented into 2 × 2 mm2 pieces and then digested with collagenase (Sigma-Aldrich, 2 mg/mL in serum-free Dulbecco's Modified Eagle Medium (DMEM) for 1 hour at 37°C and 5% CO2. After digestion, tumor cell suspensions were passed through a mesh and washed in sterile PBS. Further isolation of tumor cell population was conducted by cell sorting using FACSaria (BD Biosciences). Isolation of tumor-derived GFP+ population was done without the use of any antibody staining, whereas isolation of Jag2+ and CD11b+ population required previous staining with fluorochrome-conjugated antibodies (PE anti-mouse Jag2, 131007, BioLegend and FITC anti-mouse CD11b, 101205 BioLegend). Antibodies used were diluted 1:100 in PBS + BSA 0.5% and incubated in the dark with rotation at 4°C for 45 minutes.
In vitro co-culture assays
HCT15 cells were cultured at 1 × 104/cm2 cell density in DMEM-supplemented medium (GIBCO) with 10% FBS (Sigma-Aldrich). After 24 hours, medium was changed to DMEM-supplemented medium with 2% FBS and 1 × 105 tumor-associated bone marrow–derived cells were added (GFP+, GFP+Jag2+, GFP+Jag2−, or Jag2+ CD11b+ cells). Coculture was maintained for 48 hours. After this period, tumor-associated bone marrow–derived cells were gently washed from the culture and HCT15 cells collected for mRNA extraction or fixed with 2% paraformaldehyde for 15 minutes for further immunocytochemistry staining. γ-Secretase inhibitor (GSI; DAPT, Sigma-Aldrich) was added to cocultures at a final concentration of 10 μmol/L and respective controls received DMSO (Sigma-Aldrich).
Peripheral blood mononuclear cell (PBMC) fractions derived from patients with colorectal cancer were stained for CD11b and Jag2. Briefly, 5 × 105 cells were blocked for 10 minutes at 4°C with FcR fragment in a 1:100 dilution and then incubated with anti-CD11b (APC anti-human CD11b, 301409 BioLegend) and Jag2 (PE anti-human Jag2, 346903 BioLegend) antibodies in a 2.5:100 dilution in PBS + BSA 0.5% for 45 minutes in the dark with rotation at 4°C. Characterization of tumor-associated bone marrow–derived cells was conducted by staining digested tumor samples for Jag2 (PE anti-mouse Jag2, 131007, BioLegend), CD11b (FITC anti-mouse CD11b, 101205 BioLegend), Gr-1 [APC/Cy7 anti-mouse Ly-6G/Ly-6C (Gr-1), 108423 BioLegend], F4/80 (FITC anti-mouse F4/80, 123107 BioLegend), Sca-1 (FITC anti-Mouse Ly-6A/E, 557405 BD Pharmingen), c-Kit [APC/Cy7 anti-mouse CD117 (c-kit), 105825, BioLegend], CD3 (FITC anti-mouse CD3, 100203 BioLegend), and CD19 (Alexa Fluor 488 anti-mouse CD19, 115524 BioLegend). Antibodies used were diluted 1:100 in PBS + BSA 0.5% and incubated in the dark with rotation at 4°C for 45 minutes. Flow cytometry was conducted on FACSCalibur and analyzed with FlowJo 8.7 Software (Tree Star, Inc. 1997–2012).
Quantitative reverse-transcription PCR
RNA extraction (TRIzol, Invitrogen) and cDNA synthesis [Reverse-transcription with Superscript II reverse transcriptase (Invitrogen)] were conducted following standard protocols. Reverse-transcription PCR (RT-PCR) was conducted with Power SYBR Green PCR Master Mix in 7900HT Fast Real-Time PCR System (both from Applied Biosystems). Primer sequences for hE-Cadherin, hVimentin, hHey1, hHey2, mDll1, mDll4, mJag1, and mJag2 are shown in Table 2. The housekeeper gene used to normalize human samples was h18s and to normalize mouse samples was m18s. RT-PCR data were analyzed by DataAssist software (Applied Biosystems).
Tumor histocytochemistry procedure and analysis
Human colorectal cancer tumor paraffin-included samples were provided by the Pathology department at IPO-Lisboa. Samples were serially sectioned (3 μm), adsorbed into slides and then subjected to antigen retrieval (PT Link, Dako). After being deparaffinized, sections were blocked in PBS + 10% goat serum for 30 minutes at room temperature and then incubated with primary antibody at room temperature for 1 hour. Counterstaining was conducted using Mayer hematoxylin. Serial sections were stained for Jag2 (sc-56041 Santa Cruz, 1:15), CD11b (HPA002274 Sigma, 1:300), E-cadherin (18-0223 Invitrogen, 1:50), and cytokeratin-19 (CK19; M0888 Dako, 1:50). Slides were analyzed and photographed in a Leica DMD108 microscope. Concerning CD11b+ Jag2+ cell quantification, tissue sections were screened at low power field (×100 magnification), and the 10 areas with the most intense staining for CD11b (hot spots) were selected. Jag2 staining in these areas was confirmed in the serial slide. Counts of the hot spots were conducted at high power field (HPF; ×400). The mean number of positive cells in the 10 hot spot areas was expressed for each condition. E-cadherin quantification on each slide was conducted using the ImmunoMembrane software version 1.0i (16). Results were further validated by a pathologist at IPO-Lisboa. Mouse tumors samples were included in either gelatin or paraffin. Paraffin tumor sections were further subjected to antigen retrieval protocols (PT high). Tumor cryosections were blocked with a 5% FBS/0.1% BSA solution in PBS for 30 minutes. Slides were then covered with primary antibodies: anti-mouse PECAM (553369, Pharmingen), anti-mouse CD11b (550282, Pharmingen), anti-human E-cadherin (M3612, Dako); anti-human CK19 (M0888, Dako), anti-human (M0851, Dako), anti-human C-Kit (A4502, Dako), anti-mouse B220 (553085, Pharmingen), and anti-human Jag2 (AbCam). After overnight incubation at 4°C, tumor sections were washed in PBS and incubated with secondary antibodies from Invitrogen (anti-rat-FITC, anti-mouse-Alexa568, anti-rabbit-Alexa488, respectively) for 2 hours at room temperature. For the quantification of GFP+ lineage+ cells in tumor samples, the total number of GFP+ lineage+ cells was quantified in 5 HPFs (×400) and then divided by total GFP+ cells. Quantification of vimentin+ cells was preformed by direct count of vimentin-expressing cells in 5 HPFs (×400). Quantification of E-cadherin expression on tumor samples was performed by direct quantification of staining intensity and area using ImageJ software.
Results are expressed as mean ± SEM. Data were analyzed with GraphPad software (GraphPad Software, Inc; v4.0b) using unpaired two-tailed student t test or Mann–Whitney test when indicated. P values of less than 0.05 were considered statistically significant.
Bone marrow–derived cells infiltrate ectopic colorectal tumors and localize in regions undergoing EMT
To characterize the contribution of bone marrow–derived cells during colorectal cancer progression, we started by inoculating HCT15 (human colon carcinoma cell line) subcutaneously into nude mice that had been transplanted with actin-GFP+ BM (Supplementary Fig. S1). Tumors were allowed to grow for 1 to 3 weeks (termed early and late tumors, respectively), after which the mice were sacrificed and tumors collected. As shown in Fig. 1, bone marrow–derived cells (GFP+ cells) actively infiltrate early and late tumors, and are found at higher frequency in late tumors (Fig. 1A). We also observed that as tumors grew, there was a significant decrease in E-cadherin expression and a significant increase in the expression of the mesenchymal marker vimentin (Fig. 1B and C), suggestive of EMT. Considering that bone marrow–derived cells' infiltration into tumors and EMT were associated with tumor growth, we analyzed possible correlations between GFP+ cell infiltration and E-cadherin and vimentin expression in tumor tissues. There was an inverse correlation (P = 0.0097; R2 = 0.7678, 7 samples) between GFP+ and E-cadherin+ areas, but a direct correlation (P = 0.0297; R2 = 0.6448, 7 samples) between GFP+ areas and the number of vimentin+ cells in tumor tissues (Fig. 1D). Detailed observation of tumor sections showed that CK19-positive tumor cells in close proximity with GFP+ cells acquire the expression of vimentin, while tumor cells further away do not (Fig. 1E, bottom, white arrows). Consistent with the EMT phenotype, tumor cells in proximity of GFP+ cells also show decreased E-cadherin expression (Fig. 1E bottom right, white arrow). These data suggest that colorectal cancer growth is associated with an increased frequency of bone marrow–derived cells infiltration, which in turn correlate with the onset of EMT. Considering this, we hypothesized bone marrow–derived cells could induce EMT on epithelial tumor cells.
Bone marrow–derived cells induce EMT on colorectal carcinoma cells in a Notch pathway-dependent manner
To test the ability of bone marrow–derived cells to induce EMT on tumor cells, we conducted in vitro coculture experiments with HCT15 cells and bone marrow–derived (GFP+) cells isolated (sorted out) from late subcutaneous tumors. As depicted in Fig. 2A in the presence of bone marrow–derived (GFP+) cells isolated from tumors (TGFP+ cells), HCT15 lose E-cadherin expression and gain vimentin expression, as measured by quantitative RT-PCR (qRT-PCR). This can also be observed in the quantification of E-cadherin–expressing HCT15 cells by immunostaining after coculture (Fig. 2B), suggesting that TGFP+ cells induce vimentin expression and decrease E-cadherin expression at both transcriptional and translational level. Next, we looked into the molecular pathways regulating EMT induced by TGFP+ cells. The TGF-β and the Notch signaling pathways (19) have been implicated in EMT, thus we tested whether these pathways could be involved in bone marrow–derived cell-induced EMT on colon carcinoma cells. In agreement, we conducted coculture assays by adding a TGF-β blocking antibody and a Notch pathway inhibitor (GSI) and determined its effect on HCT15 cells E-cadherin and vimentin expression. TGF-β inhibition had no effect (data not shown), whereas addition of GSI inhibited the TGFP+ cell-induced decrease in E-cadherin and increase in vimentin expression (Fig. 2A). Furthermore, confirming the involvement of the Notch pathway is the observation that HCT15 cells express Notch receptors 1 and 4 (data not shown) and more importantly the Notch pathway downstream targets Hey 1 and 2 are upregulated in HCT15 cells undergoing EMT, in the presence of TGFP+ cells (Fig. 2A). These data suggest that bone marrow–derived cells induce EMT via Notch pathway activation on tumor cells.
Bone marrow–derived cells expressing Jag2 induce EMT on colorectal cancer cells
Having shown TGFP+ cells induce EMT on HCT15 cells and that this process involved Notch pathway activation, next we determined which ligands of the Delta:Notch family were expressed by the tumor-infiltrating bone marrow–derived cells. We determined and compared the expression of the ligands Delta-like ligand 1 (Dll1), 4 (Dll4), and Jag1 and 2 (Jag1, Jag2) in bone marrow–derived cells (GFP+) collected from late tumors and from the corresponding bone marrow samples. Jag2 was the most abundantly expressed ligand on TGFP+ cells, its expression being significantly higher on TGFP+ cells than on those isolated from bone marrow (Fig. 3A). Detailed flow cytometry analysis of tumor and bone marrow samples of tumor-bearing mice revealed a population of GFP+ Jag2+ cells in both tissues, although at a higher frequency in the tumors, representing an average of 5.5 ± 1.12% cells (Fig. 3B). Taking into account the high frequency of GFP+ Jag+ cells in tumors, next we investigated whether this population was able to induce EMT on HCT15 colon carcinoma cells. GFP+ Jag2+, and also GFP+ Jag2− cells were sorted from late tumors (Fig. 3B) and cocultured with HCT15 in the presence or absence of GSI. As depicted in Fig. 3C and D, GFP+ Jag2+ but not GFP+ Jag2− cells were able to reduce E-cadherin transcription in a Notch pathway-dependent manner. In accordance, the number of HCT15 E-cadherin+ cells is significantly reduced when these are cocultured in vitro with tumor GFP+Jag2+ (Fig. 3C). This phenotype is reversed upon the addition of GSI or when HCT15 cells are cocultured with GFP+ Jag2 negative cells (Fig. 3E). Taking together these results suggest that bone marrow–derived cells modulate EMT via Jag2 mediated Notch-pathway activation on tumor cells.
Bone marrow–derived CD11b+Jag2+ cells induce EMT via Jag2-mediated Notch pathway activation
Having shown that bone marrow–derived Jag2+ cells are responsible for tumor cell EMT via Notch pathway activation, next we sought to determine the identity of these cells more precisely. The majority of TGFP+ Jag2+ cells were CD11b+ (85 ± 1.0%), F4/80+ (64.1 ± 0,55%), and Sca-1+ (52.5 ± 1.55%) (Fig. 4A, B, and C). Considering the high frequency of CD11b+ cells within the TGFP+ Jag2+ population, we tested whether tumor-derived CD11b+ Jag2+ cells (CD11b+Jag2+) could induce HCT15 EMT in vitro. Accordingly, coculture of HCT15 tumor cells with tumor-derived CD11b+Jag2+ cells led to a significant decrease in E-cadherin expression at both the mRNA and protein level (Fig. 4D and E), in a Notch pathway-dependent manner, as GSI treatment reverted this effect. Altogether these data suggest that CD11b+Jag2+ bone marrow–derived cells actively infiltrate colorectal cancer tumors and induce EMT on tumor cells in a Notch pathway-dependent manner.
CD11b+Jag2+ cells are mobilized to PB and home to tumor tissues in different colorectal cancer models
Considering the observed effect of bone marrow–derived CD11b+Jag2+ cell population in inducing EMT in HCT15 cells in vitro and in vivo, next we quantified the recruitment of this population in other colon cancer models. For this purpose, we developed subcutaneous colon carcinoma xenotransplants using 3 human colorectal cancer cell lines, HCT15, DLD-1, and HT-29, and evaluated the presence of CD11b+Jag2+ cells in the peripheral blood and in the tumor tissues. There was a significant increase in the frequency of CD11b+Jag2+ cells in the peripheral blood compared with control mice (no tumor) in all models (Fig. 5A), this being particularly evident on HCT15-xenotransplanted mice, which showed a significant increase on days 7 and 14 after tumor inoculation. Moreover, FACS analysis of tumor samples showed that CD11b+Jag2+ cells are present in tumor xenotransplants derived from all cell lines at a frequency ranging from 4.5%–8.5% of total cells (Fig. 5B and C). To further validate the biologic significance of CD11b+Jag2+ cells in colorectal cancer growth, we developed orthotopic models of colorectal cancer using HCT15 and HCT116 cell lines and evaluated their presence in the peripheral blood and in tumor tissues, as above. Mice bearing orthotopic HCT15 and HCT116 tumors showed increased frequency of peripheral blood CD11b+Jag2+ cells compared with control mice on weeks 4 and 6, after tumor inoculation (Fig. 5D). Moreover, FACS analysis of tumor samples shows that CD11b+Jag2+ cells are present in orthotopic tumors at a frequency ranging from 0.5% to 6% of total tumor cells (Fig. 5E and F). Together, these data show that in different murine models of colorectal cancer, CD11b+Jag2+ cells are mobilized to the peripheral blood and are actively recruited into tumors, which is suggestive of the biologic significance of this population in colorectal cancer growth and EMT onset.
CD11b-neutralizing antibodies reduce CD11b+Jag2+ cell recruitment, resulting in decreased tumor growth and EMT in vivo
To further test the role of CD11b+Jag2+ in colorectal cancer progression and EMT, we treated mice bearing ectopic HCT15 tumors with neutralizing monoclonal antibodies against CD11b. CD11b neutralization caused a significant reduction in tumor growth (Fig. 6A) and also on the peripheral blood levels of CD11b+ and CD11b+Jag2+ cells compared with mice treated with vehicle alone (Fig. 6B). Moreover, flow cytometry and immunostaining-based quantification of CD11b+ and CD11b+Jag2+ cells in tumor tissues showed that CD11b neutralization led to a significant reduction in both cell populations (Fig. 6C–E). Having shown that CD11b neutralization leads to a significant reduction in the number of peripheral blood and tumor-infiltrating CD11b+Jag2+ cells, we determined its effect in the onset of EMT. As shown in Fig 6, while control (untreated) mice show the expected decrease in E-cadherin expression (Fig. 6F and G), this is sustained and even slightly increases in tumors of mice treated with CD11b-neutralizing antibody. Analysis of Notch pathway downstream targets Hey 1 and Hes 1 expression showed the reduced EMT onset in tumors treated with anti-CD11b–neutralizing antibody is accompanied by decreased activation of the Notch pathway (Fig. 6F), which is in accordance with our in vitro data. Taken together, these data suggest that treatment of ectopic colon cancer-bearing mice with neutralizing monoclonal antibodies against CD11b reduces infiltration of CD11b+Jag2+ cells into the tumors, reducing EMT and reducing Notch pathway activation.
CD11b+Jag2+ PB and tumor levels correlate with lower E-cadherin expression and metastatic disease in patients with colorectal cancer
Having shown the importance of bone marrow–derived CD11b+Jag2+ cell population in inducing EMT in murine colorectal cancer models, and the easy quantification of these cells in peripheral blood of colon cancer-bearing mice, next we tested the feasibility and usefulness of quantifying these cells in the samples of patient with colorectal cancer. We quantified the levels of CD11b+Jag2+ cells in peripheral blood samples of 40 patients with colorectal cancer diagnosed at different stages (according to the AJCC Staging System). For the analysis, patients were classified as: TxN0M0 (patients with colorectal cancer without lymph node or distant metastasis, stages I–II of the AJCC), TxNxM0 (patients with colorectal cancer with lymph node metastasis and without distant metastasis, stage III of the AJCC), and TxNxMx (patients with colorectal cancer with distant metastasis, stage IV of the AJCC). We observed a significant correlation between PB CD11b+Jag2+ cell levels and colorectal cancer stages, with higher stage colorectal cancer patients showing higher peripheral blood CD11b+Jag2+ levels (Fig. 7A and B). Moreover, there was an inverse correlation (P = 0.0197; R2 = 0.4708, 11 patients) between peripheral blood CD11b+Jag2+ cell levels and tumor E-cadherin expression (Fig. 7C). We were also able to investigate the relation between CD11b+Jag2+ tumor levels and colorectal cancer staging in 12 patients. We observed that TxNxM0 and TxNxMx stage patients had significantly higher numbers of CD11b+Jag2+ cells in tumor samples relatively to TxN0M0 patients (Fig. 7D and E). We also observed a strong inverse correlation (P = 0.0007; R2 = 0.7011, 12 patients) between CD11b+Jag2+ numbers in tumors and tumor E-cadherin expression levels (Fig. 7F). Importantly, in our patient cohort, there was no correlation between tumor size and the presence of peripheral blood or tumor CD11b+Jag2+ cells (Supplementary Fig. S2). Together, these data strongly suggest that quantification of both CD11b+Jag2+ cells in the PB and in the tumor samples of patients with colorectal cancer correlates with colorectal cancer staging, and thus may be used as a prognostic or diagnostic marker. These data also validate our data obtained from colorectal cancer mouse models, where we showed CD11b+Jag2+ cells infiltrate colorectal cancer tumors and induce EMT.
Over the last decade, there has been an increase in our understanding of the contribution of tumor stromal components in the regulation of tumor progression and metastasis formation (17). EMT has been shown to be essential for carcinoma progression, namely in breast, colon, and prostate carcinoma, preceding invasion and metastasis formation (3, 18). One of the major molecular regulators of EMT is E-cadherin. E-cadherin is a member of the cadherin family of homophilic cell adhesion molecules and is essential for the maintenance of adherens junctions that confer physical integrity and polarization to epithelial cells. Targeted disruption of E-cadherin during tumor progression resulting in decreased intercellular adhesiveness is one of the most common alterations in human cancers (19, 20). In fact, E-cadherin functional inactivation represents a critical step in the acquisition of invasive capacity by epithelial tumor cells. Accordingly, abolishing E-cadherin function in vitro confers invasive properties to noninvasive cells and conversely, introduction of E-cadherin into invasive epithelial cell lines abrogates their invasive potential. Not surprisingly, then, loss of E-cadherin expression is a defining feature of EMT (9). Although many molecular regulators of EMT have been identified, the cellular interactions between tumor cells and tumor stromal cells responsible for tumor EMT are still unclear.
Bone marrow–derived cells, have been shown to directly impact tumor pathophysiology regulating multiple aspects of the metastasization process by modulating tumor angiogenesis, promoting tumor cell invasion, extravasation, intravasation, and micrometastasis establishment, and growth and vascularization (21). However, a role for bone marrow–derived cells in the first steps of metastases formation, namely in EMT, had not been addressed and was the subject of this study. In detail, we investigated a putative role for bone marrow–derived cells in regulating EMT in metastatic colorectal cancer. We show in several ectopic and orthotopic colorectal cancer models that tumor progression is associated with an increased accumulation of bone marrow–derived cells, mainly of myeloid origin. Furthermore, we observe that in late tumors there are regions highly infiltrated by bone marrow–derived cells that show evidence for EMT, namely a robust decrease in E-cadherin expression and a significant increase in vimentin expression in CK19+ tumor cells. It was highly suggestive from our analysis of serial tumor slides that tumor cells (CK19+) in close proximity with bone marrow–derived cells (GFP+) showed the described hallmarks of EMT. This observation led us to speculate that a direct interaction between bone marrow–derived and tumor cells could modulate EMT in the latter. To test this hypothesis, we used in vitro coculture systems, where colorectal cancer cell lines are cultured with bone marrow–derived tumor stromal cells (GFP+). We used this simple system to investigate the molecular regulation of EMT induction by bone marrow–derived tumor-associated cells. Although TGF-β pathway activation has been extensively described as an inducer of EMT (22), inhibition of TGF-β with neutralizing antibodies in our coculture experiments failed to inhibit the EMT-inducing capacity of bone marrow–derived cells. On the other hand, the Notch pathway has also been associated with EMT and metastasis formation in pancreatic (23) and colorectal cancer (24). Moreover, Notch pathway activation was shown in breast cancer brain metastasis (25). Another study shows that in patients with breast cancer, increased expression of either Jag1 or Notch1 is predictive of poor overall survival (26). Moreover, in colon cancer, Notch 1 overexpression has been shown to correlate with pathologic grade, progression, and metastasis (27). Considering this, we tested the effect of GSI (Notch pathway inhibitor) in reversing the EMT induction by bone marrow–derived cells on colorectal cancer cell lines. GSI addition inhibited EMT induction by bone marrow–derived cells, accompanied by a significant reduction in the expression of Notch pathway downstream targets Hey 1 and 2. Although we did not address Notch downstream effectors regulating EMT, recent studies suggest the mechanisms by which Notch may induce EMT might involve activation of E-cadherin transcriptional repressor Snail2 (28, 29).
Importantly, we show that the bone marrow–derived tumor-infiltrating cells, which are mainly of myeloid origin (CD11b+), increased expression of Jag2, but not of other Notch ligands. Interestingly, Jag2 expression in tumor cells has been recently described as a major regulator of EMT and metastases in lung adenocarcinoma via a miR-200–dependent downstream mechanism (30). Furthermore, quantification of the GFP+ Jag2+ populations present in the bone marrow and in the tumors revealed a 10-fold increase in the percentage of this population in tumors, suggesting that bone marrow–derived Jag2+ cells are actively recruited into tumors. The signal(s) responsible for the selective recruitment of this specific bone marrow–derived cell population into colon tumors remain undisclosed; we measured the most commonly studied chemokines including SDF1α, MCP1, but these were not actively secreted in our colon cancer models (data not shown). Moreover, recent studies also showed that microparticles/microvesicles/exosomes released by tumors may communicate and “educate” the bone marrow microenvironment, resulting in selective recruitment of bone marrow–derived cells to sites of metastases to form the premetastatic niche (31). Whether specific colon cancer-derived exosomes selectively induce the mobilization of CD11b+Jag2+ cells is undisclosed. Concerning the molecular mechanisms regulating Jag2 expression, recent data indicate its expression on breast cancer cells is regulated by hypoxia and is directly involved in breast cancer metastases (32). It is therefore legitimate to speculate that during colorectal cancer growth, recruited bone marrow–derived CD11b+ cells enter the tumor site and are exposed to signals from the tumor microenvironment including hypoxia that regulate the expression of Jag2. As shown here, once the infiltrating cells express Jag2, these are capable of inducing EMT on subsets of colon cancer cells. The relative contribution of Jag2 expressed by infiltrating bone marrow–derived cells and by tumor cells proper for the onset of EMT is still undisclosed and will be the subject of future studies in our laboratory.
These findings led us to hypothesize that selective targeting of the CD11b+Jag2+ bone marrow–derived population might delay the onset of EMT, a first step in the metastasis cascade. As we do not presently have access to specific monoclonal-neutralizing antibodies to Jag2, we sought to impede the recruitment of CD11b+ cells (which include the Jag2+ population) into the tumors. Several studies followed a similar approach and concluded that targeting the CD11b+ population of infiltrating cells may be of therapeutic benefit (33, 34). Moreover, Toh and colleagues (35) recently showed myeloid-derived suppressor cells (also CD11b+) promote EMT, although they did not identify the mechanism involved. In our colon cancer models, CD11b neutralization led to a significant decrease in CD11b+Jag2+ cell levels in peripheral blood and tumor, delayed EMT, and reduced Notch pathway activation in the tumors. Together, these data strongly suggest that bone marrow–derived CD11b+Jag2+ cells represent a targetable cell population, which is actively recruited into tumors and induces EMT in a Notch pathway-dependent manner.
Having shown the biologic significance of CD11b+Jag2+ cells in mouse models of colorectal cancer, we investigated the involvement of this cell population in human colorectal cancer progression. We observed that peripheral blood levels of CD11b+Jag2+ cells correlate with colorectal cancer stages, with higher levels of CD11b+Jag2+ cells being observed in colorectal cancer patients with metastatic disease (mostly consisting of liver metastases), whereas lower CD11b+Jag2+ levels are detected in patients with colorectal cancer without metastatic disease. Furthermore, we found a significant correlation between CD11b+Jag2+ cell peripheral blood levels and the levels of E-cadherin expression in the corresponding (i.e., patient-specific) tumor tissues. Taking this into account, we suggest the CD11b+Jag2+ cells are biologically relevant in human colorectal cancer and that measuring the peripheral blood levels of CD11b+Jag2+ cells in patients with colorectal cancer may be used as a biomarker to support colorectal cancer staging.
Our data highlight the importance of looking at the initial steps of the metastasis cascade also in a systemic manner. In detail, we show a bone marrow–derived CD11b+Jag2+ population can activate Notch signaling on colon cancer cells, resulting in EMT induction in vitro and in vivo. Given the recent excitement from clinical trials using Notch pathway inhibitors for the treatment of tumors, and particularly aiming at blocking tumor angiogenesis (36, 37), the data presented here paves the way for the use of Notch pathway inhibitors to impede metastases onset, by impeding EMT induction at a primary tumor site.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were declared.
Conception and design: F. Caiado, I. Rosa, H. Yagita, P. Fidalgo, A.D. Pereira, S. Dias
Development of methodology: F. Caiado, T. Carvalho, H. Yagita
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Caiado, I. Rosa, A. Costa, B. Heissig, K. Hattori, J.P. da Silva, P. Fidalgo
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Caiado, A. Costa, P. Fidalgo, A.D. Pereira, Sérgio Dias
Writing, review, and/or revision of the manuscript: F. Caiado, I. Rosa, H. Yagita, P. Fidalgo, A.D. Pereira, S. Dias
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Caiado, L. Remedio, J. Matos
Study supervision: H. Yagita, A.D. Pereira, S. Dias
The authors thank the members of the Angiogenesis Lab for their input and suggestions.
This study was supported by Fundaçao para a Ciencia e Tecnologia (FCT, Portuguese Government) grants (S. Dias) and fellowships.
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.