Bevacizumab terminates homeobox B9-induced tumor proliferation by silencing microenvironmental communication

To investigate HOXB9 expression in colorectal cancer, we examined specimens from patients who had p-stage II or III colorectal cancer and had undergone surgery at our institute. HOXB9 mRNA was expressed in approximately 70% of the patients (Figure 1a), and HOXB9 was found to be upregulated in a fraction of tumors, relative to that in the uninvolved adjacent normal mucosa, in both RNA in situ hybridization (Additional file 1: Figure S1a), and immunohistochemical staining (Figure 1b). HOXB9 mRNA expression was significantly higher in poorly differentiated adenocarcinomas than in highly and moderately differentiated types of adenocarcinomas (Figure 1c) and was highly related to liver metastasis (Additional file 1: Figure S1b). Moreover, although HOXB9 did not affect the disease free survival after surgery, increased HOXB9 expression in colorectal cancer was significantly associated with a poor prognosis in overall survival (Figure 1d), as previously reported for breast cancer [14].

We introduced the HOXB9 construct into the HT29 and HCT116 cancer cell lines (HT29-T and HCT116-T, respectively), both of which showed low HOXB9 mRNA expression relative to that in the other human colorectal cancer cell lines (Additional file 1: Figure S1c), in order to evaluate the functional consequence of HOXB9 overexpression in colorectal cancer. HOXB9 overexpression is regulated by a tetracycline repressor (Figure 1e,f and Additional file 1: Figure S1d,e) [15]. Angiogenic factors, including amphiregulin (AREG), epiregulin (ERG), basic fibroblast growth factor (bFGF), interleukin-8 (IL8), and VEGF were significantly upregulated by HOXB9 (Figure 1g). To determine the effect of the loss of HOXB9, we produced short hairpin RNA (shRNA)-expressing HOXB9 lentiviral constructs capable of reducing endogenous HOXB9 expression by up to 99% (Figure 1e). These constructs were used to reduce the expression of angiogenic factors in WiDr and SW480 cell lines (WiDr-S and SW480-S, respectively) that initially had high expression of HOXB9 (Figure 1g). Moreover, the transforming growth factor (TGF) -beta axis was regulated by HOXB9 in colorectal cancer (Figure 1h), similar to the results previously obtained for breast cancer [5]. Next, we investigated the functional significance of these factors in ELISA assays. Significantly increased tumoral secretion of angiogenic factors were confirmed in the supernatants of HOXB9-overexpressing cell lines (Additional file 2: Figure S2a,b). To confirm the role of this acquired angiogenesis of colorectal cancer in vivo, we established human tumor xenografts in BALB/c nude mice. HOXB9-overexpressing xenografts of the HT29-T cell line demonstrated a dramatic increase in tumor burden (Figure 2a,b) and microvessel density (Figure 2c,d and Additional file 2: Figure S2c). The immunohistochemical staining showed higher production of angiogenic factors including IL8 and VEGF in HOXB9 overexpressed tumor (Additional file 2: Figure S2d).

The cell proliferation test for 24 hours revealed almost the same proliferation in HOXB9 overexpressed cells as control (Additional file 3: Figure S3a), and the cell viability test for 7 days showed rather smaller survival rate in HOXB9 overexpressed cells in vitro (Additional file 3: Figure S3b). Therefore, the role of the microenvironment in tumor proliferation under the regulation of HOXB9 was investigated, given that the primary difference between the in vivo and in vitro conditions was the existence of a microenvironment including vessels and fibroblasts. When HUVECs and human dermal fibroblasts (HDFs) were co-cultured with cancer cell lines to mimic the microenvironment (Additional file 3: Figure S3c,d), HOXB9 significantly promoted cell proliferation (Figure 3a). Then, we hypothesized that anti-angiogenic treatment may cause strong regression in tumors with such angiogenic properties as HOXB9, which depends tumor proliferation on enhanced angiogenesis and stromal reaction in the microenvironment. In order to evaluate the impact of anti-angiogenic agents on tumor microenvironment in vitro, the tube formation of HUVECs co-cultured with HOXB9-regulated cancer cells was estimated under administration of bevacizumab. HOXB9 significantly enhanced angiogenesis, and bevacizumab administration significantly reduced angiogenesis of HUVECs co-cultured with HOXB9-overexpressing cancer cells (Figure 3b and Additional file 3: Figure S3e,f).

Next, BALB/c nude mice transplanted with xenografts that had been previously established using HOXB9-overexpressing colon cancer cells were treated intraperitoneally with bevacizumab (5 mg/kg, weekly). Both of HOXB9-induced cancer cells showed accelerated tumorigenesis. However, the bevacizumab treatment dramatically reduced tumor volume and tumor weight both in HT29-T cells (Figure 4a,b) and in HCT116-T (Additional file 4: Figure S4a,b). Tumor gross vascularity was also reduced by bevecizumab (Figure 4c). The growth inhibition mediated by bevacizumab significantly increased in HOXB9-overexpressing tumors (Figure 4d,e and Additional file 4: Figure S4c,d). We further established an intrasplenic injection (ISP) model of HCT116-T cells in BALB/c nude mice to determine whether anti-angiogenic treatment with bevacizumab prolongs survival in angiogenic tumors. HOXB9 overexpression was associated with longer third-quadrant survival upon bevacizumab treatment (Additional file 5: Figure S5a,b). To vertify our hypothesis that the loss of angiogenic factors plays an important role in tumor shrinkage in angiogenesis-driven tumor, sunitinib malate, which is a multiple tyrosine kinase inhibitor with effects including VEGF inhibition, was also administered in HOXB9 overexpressed tumor. Sunitinib also suppressed cell proliferation in HOXB9-overexpressing breast cancer xenografts (Additional file 5: Figure S5c).

To determine the key factors for enhanced tumor growth in microenvironment, we performed comprehensive multiplex assays involving the co-culture supernatants and observed significant elevation of the levels of several cytokines, including interleukin-6 (IL6) (Figure 5a). We then focused on the differences between the secretion of IL6 in monoculture and co-culture systems to elucidate the impact of microenvironment on tumor proliferation. The concentration of IL6 was dramatically higher in the co-culture system than in the mono-cultures (Figure 5b and Additional file 6: Figure S6a), suggesting enhanced IL6 secretion in microenvironment including tumor and stromal cells. To confirm the importance of microenvironmental IL6 secretion, cell proliferation in the presence of an IL6 neutralizing antibody was evaluated. We also used bevacizumab to inhibit the activity of VEGF released from cancer cells (Figure 5c). Both the IL6 neutralizing antibody and bevacizumab produced significant reduction of cancer cell proliferation (Figure 5d,e and Additional file 6: Figure S6b). Moreover, IL6 secretion in the supernatant dramatically decreased on administration of bevacizumab in co-culture systems (Figure 5f).

To clarify the clinical significance of anti-angiogenic treatment for angiogenic colorectal cancer, we conducted a retrospective study of patients who had undergone surgery between 2001 and 2010 at our institute and were treated by chemotherapy, including bevacizumab (n = 39). Specimens obtained from patients before chemotherapy were analyzed by Quantitative reverse transcription-PCR (qRT-PCR). The demographic and clinical data showed no significant differences with regard to age, sex, pathological stage, or chemotherapy between groups with high (n = 13) or low (n = 26) expression of HOXB9 (Additional file 7: Table S1). High expression of HOXB9 mRNA was significantly associated with longer overall survival and progression-free survival after treatment with bevacizumab combined with chemotherapy (Figure 6a,b), and mRNA expression of HOXB9 was significantly correlated with IL6 mRNA expression in clinical specimen (Figure 6c).

Our analysis included colorectal tumors of patients who presented between January 2004 and January 2008 at Keio University Hospital; these data were obtained from the hospital colorectal cancer database. Among the samples of consecutive sporadic colorectal cancer patients who received a diagnosis of p-stages II or III disease, 69 panels of frozen tissue specimens containing cancerous and matched normal mucosal tissue were available for qRT-PCR assessment. All samples were obtained from surgically resected material, immediately frozen in liquid nitrogen, and stored at −80°C. Adequate lymph node dissection was performed in all these cases, and a pathological diagnosis was made according to the TNM classification. HOXB9 mRNA expression was examined at cancer sites and in normal mucosa in consecutive pStages II and III patients who had undergone surgery at our institute during 2004–2008. The expression ratios (cancer site to normal mucosa) were examined using qRT-PCR. RNA was extracted from the cells by using the RNeasy kit (QIAGEN, Valencia, CA). Conditions for semiquantitative amplification of cDNA were 95°C for 2 min, followed by 25 cycles of 95°C for 30 s, 56°C for 30 s, and 72°C for 60 s, with a final extension cycle of 72°C for 10 min. RT-PCR analysis was run in triplicate for each sample on a Light Cycler 480 Real-Time PCR System using SYBR Green 1 Master Mix (Roche). The following program was run: preincubation for 5 min at 95°C, amplification for 45 cycles (10 s of denaturation at 95°C, 10 s of annealing at 57°C, and a 10-s extension at 72°C), and melt-curve analysis. mRNA levels of the target gene were normalized against mRNA levels of GAPDH, which was used as an internal control. The sequences of all the primers used are listed in Additional file 8: Table S2. Written informed consent was taken and the study was approved by the ethics committee of the School of Medicine of Keio University.

The HOXB9 probe for ISH was generated as described previously [5]. Briefly, the plasmid encoding HOXB9 was cleaved with Hinc II and XmaI (New England Biolabs). The acquired 286-bp fragment was outside the homeodomain and spanned nucleotides 26–324 of the HOXB9 transcript. This region had 100% homology to HOXB9, as expected, and had little homology with other members of the HOX family or other HOX genes. After linearization of the plasmid, in vitro transcription was carried out, using 11-digoxogem-UTP (Roche) and 2 units/μL of either T3 of T7 RNA polymerase (Promega) for sense and antisense probes, respectively. Riboprobes were purified by ethanol precipitation, and their size and integrity checked by gel electrophoresis. The tissues were dehydrated, washed twice with PBS, and treated with 3% hydrogen peroxide for 15 min. Following treatment with proteinase K (10 μg/ml) for 9 min, tissues were washed with PBS and refixed with 4% paraformaldehyde in PBS for 5 min. The tissues were pre-hybridized for 1 hr at 65°C before hybridizing with sense or anti-sense digoxigenin-labeled riboprobes overnight at 70°C. After hybridization, samples were placed in 1% sheep serum for 30 min at room temperature, and then incubated with anti-digoxygenin-AP (Roche) antibody for 2 hours at room temperature. BM-Purple precipitation (Roche) was used to detect the signal.

The human colorectal cancer cell lines SW480, WiDr, HT-29, and HCT116 were obtained from the American Type Culture Collection (Rockville, MD). The cell lines HT29 and HCT116 were seeded at a density of 1 × 10⁶/ml in RPMI 1640 medium (Gibco, MA), and SW480 and WiDr cells were seeded a density of 1 × 10⁶/ml in Dulbecco’s modified Eagle medium. Both media were supplemented with 10% FBS (CSL Ltd., Australia), 100 IU/ml penicillin, and 50 mg/ml streptomycin sulfate. HT29 and HCT116 cells with low HOXB9 expression were infected with HOXB9-expressing lentiviral constructs (Invitrogen). For the Tet-regulated system, the entry vector was transferred to the lentiviral pLenti4/TO/V5-DEST vector and used in combination with the Tet-repressor vector pLenti6/TR (Invitrogen) according to the manufacturer’s instructions. The induction of HOXB9 was regulated by the addition of 1 μg/ml tetracycline (Invitrogen). To knock down HOXB9 expression, viral constructs of shRNA were generated, and SW480 and WiDr cells were infected as described previously [5].

DNA synthesis was measured using 5-ethynyl-2´-deoxyuridine (EdU) Flow Cytometry Assay Kits (Life Technologies, Tokyo, Japan). Experiments were performed according to the manufacturer’s instructions. Briefly, 3 × 10⁶ cells were seeded into 10 cm dish and incubated for 24 hours at 37°C with or without 1 μg/ml tetracycline (Invitrogen). Cells were incubated for 2 hours with 20 μM EdU. Subsequently, cells were processed using the Click-iT EdU Flow Cytometry kit. EdU incorporation was measured by Flow cytometer (Gallios Flow Cytometer, Beckman Coulter).

Cell proliferation was measured using the 2,3-bis (2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide inner salt (MTT) assay. Briefly, 5,000 cells per well in 96-well microtiter plates (Sumilon, Sumitomo Bakelite Co., Tokyo, Japan) were incubated for 24 h, followed by continuous exposure to 1 μg/ml tetracycline (Invitrogen) for 6 days. The absorbance in the wells was measured at every 2 days using a NJ-2300 microplate spectrophotometer at 540 and 630 nm (Immuno Reader, Nalge Nunc International, Rochester, NY).

Expression of proteins were detected in western blotting using antibody against HOXB9 (Santa Cruz Biotechonlogies), pAKT (Cell Signaling Technology), AKT (Cell Signaling Technology), VEGF (ABCAM) and GAPDH (ABCAM). The western blot protocols used have been described previously [29]. For immunocytochemistry, cells were plated in 8-well chamber slides (Lab Tek) and treated as described previously [5]. Briefly, after blocking, the cells were incubated with an anti-human HOXB9-specific antibody (rabbit polyclonal antibody; Santa Cruz Biotechnology, Santa Cruz, CA) at a dilution of 1:250 or an anti-CD31 antibody (ab28364; ABCAM) at 1:200 for 1 h at room temperature in incubation buffer (phosphate-buffered saline (PBS) containing 1% BSA). For immunohistochemistry, HOXB9, CD31, CD34, IL8, and VEGF expression was detected using the anti-human HOXB9-specific antibody at a dilution of 1:250 (Santa Cruz, CA), the anti-CD31 antibody at a dilution of 1:200 (ab28364; ABCAM), the anti-CD34 antibody at a dilution of 1:200 (ab8158; ABCAM), anti-IL8 antibody at a dilution of 1:1000 (ab7747; ABCAM), the anti-VEGF antibody at a dilution of 1:1000 (ab46154; ABCAM).

Angiogenesis was evaluated based on the microvessel density of HUVEC cultures. Culture supernatants without FBS or antibiotic supplementation were assessed using an angiogenesis kit (Kurabo, Okayama, Japan). Experiments were performed according to the manufacturer’s instructions. The angiogenesis kit consisted of a 24-well cluster dish in which HUVECs and fibroblasts were seeded under optimal conditions for capillary tube formation. The microvessel density was calculated as the sum of the total capillary length in 6 areas at 200× magnification by using the ImageJ software (NIH, Bethesda, MD). HUVECs and fibroblasts were co-cultured with HOXB9-overexpressing cancer cells and separated from the cancer cells by using micropore membranes. Each well was cultured with serum-free medium corresponding to each cell line, respectively.

Tetracycline was administered to regulate HOXB9 induction after 24 hours of co-culture. Endogenous IL6 was neutralized by application of an anti-IL6 antibody, and VEGF neutralization was assayed in a co-culture system in which 5 × 10³ HOXB9-overexpressing cancer cells were seeded per well in a medium supplemented with 1 μg/ml anti-human IL6 antibody (R&D Systems) and/or 1 μg/ml bevacizumab. HOXB9 expression was controlled by the addition of tetracycline.

Cytokine levels in cell culture supernatants were quantified using ELISA (BD Biosciences Pharmingen for VEGF, HGF, CXCR8, and IL6). Comprehensive proliferative cytokine levels in co- and monoculture supernatants were measured using a Procarta Cytokine Profiling kit (Affymetrix, Santa Clara, CA, USA), which utilizes xMAP multianalyte profiling Luminex technology for simultaneous detection and quantification of multiple protein targets. Fluorescently encoded antibody beads were detected individually in a flow cytometer, and 9 cytokines (bFGF, G-CSF, GM-CSF, HGF, IL1beta, IL6, MCSF, TGFbeta, and TGFalpha) were evaluated.

All the animals were cared for and experiments were performed according to AAALAS guidelines using protocols approved by the institutional review board and the institutional animal care and use committee of Keio University or Massachusetts General Hospital. Six-week-old Balb/c female nude mice were subcutaneously implanted with 4 × 10⁶ cells in 100 μl of PBS (n = 8). If the tumor size was greater than 75 mm³, oral tetracycline administration was started on day 1. Tetracycline (Sigma-Aldrich) was dissolved in distilled water at a concentration of 1 mg/ml. Bevacizumab (5 mg/kg) or PBS was intraperitoneally administered weekly. The tumor volume was calculated from the pi-based ellipsoid volume formula π/6 × length × width × height. Growth inhibition of the tumor was calculated as (1 – (mean tumor volume of treated group/mean tumor volume of vehicle group)) × 100. An intrasplenic injection (ISP) model was established to create liver metastasis as described previously [30, 31]. Briefly, nude mice were anesthetized by ether inhalation, and open laparotomy was performed. HOXB9-overexpressing cancer cells (5 × 10⁶ cells of HCT116-T in 100 μl of PBS) were injected into the spleen, and splenectomy was then performed. Tumor samples were immunostained using an anti-CD31 antibody (Abcam, Cambridge, UK) and then counterstained with DAPI. Vessels were counted using a microscope in areas of the tumor containing the highest number of capillaries and small venules. Highly vascular areas were first identified by scanning tumor sections, and the vessels were counted in 6 such areas at 200× magnification. Bevacizumab was purchased from Chugai Pharmaceutical Co. Ltd (Tokyo, Japan). MCF10A cells expressing activated G12V H-RasV12-expressing lentivirus and either LacZ or HOXB9 were generated using standard protocols. Xenografts were established by injecting 5 × 10⁶ cells in 100 μl of PBS into the mammary fat pad of 6-week-old female Swiss nu/nu mice. Tumor volumes were measured at regular intervals.

Sunitinib (SU11248; Sutent) was purchased from LC Laboratories. MCF10A cells expressing activated G12V H-Ras-expressing lentivirus and either LacZ or HOXB9 were injected into the mammary fat pads of mice (n = 16). After 2 weeks of tumor growth, 8 animals from each group were treated with 40 mg/kg sunitinib orally every day for up to 14 days. Sunitinib was dissolved in 0.5% carboxymethyl cellulose. The remaining 8 animals from each group were treated with the vehicle.

Our analysis included 456 colorectal tumors of patients who presented between January 2006 and January 2008 at Keio University Hospital; these data were obtained from the hospital colorectal cancer database. Patients who showed recurrence after surgery and were treated by chemotherapy combined with bevacizumab were selected from the database. Among the specimens from 77 consecutive patients, 39 specimens were available and were analyzed by qRT-PCR. The study was approved by the ethics committee of the School of Medicine of Keio University.

Statistical analysis was performed using the SPSS 18.0 statistical software (SPSS, Chicago, IL). Analyses of age, pathologic tumor size and weight, mRNA expressions, and total capillary length were performed using the Mann–Whitney U test. The tumor size, weight, and burden were statistically evaluated at the time of sacrifice in xenograft models. Disease-free survival (DFS) and overall survival (OS) curves were drawn according to Kaplan–Meier estimates and compared by log-rank tests. DFS was defined as the absence of local or regional recurrence, distant metastasis, second primary carcinoma, or death from any cause. The clinicopathologic categorical variables were compared using Fisher’s exact test or logistic regression where appropriate. The cutoff value for HOXB9 mRNA expression was determined according to the receiver operating characteristic curve. P < 0.05 was defined as significant.