Metastatic colorectal cancer cells maintain the TGFβ program and use TGFBI to fuel angiogenesis

Colorectal cancer (CRC) cells are traditionally considered unresponsive to TGFβ due to mutations in the receptors and/or downstream signaling molecules. TGFβ influences CRC cells only indirectly via stromal cells, such as cancer-associated fibroblasts. However, CRC cell ability to directly respond to TGFβ currently remains unexplored. This represents a missed opportunity for diagnostic and therapeutic interventions. Methods: We examined whether cancer cells from primary CRC and liver metastases respond to TGFβ by inducing TGFβ-induced protein ig-h3 (TGFBI) expression, and the contribution of canonical and non-canonical TGFβ signaling pathways to this effect. We then investigated in vitro and in vivo TGFBI impact on metastasis formation and angiogenesis. Using patient serum samples and an orthotopic mouse model of CRC liver metastases we assessed the diagnostic/tumor targeting value of novel antibodies against TGFBI. Results: Metastatic CRC cells, such as circulating tumor cells, directly respond to TGFβ. These cells were characterized by the absence of TGFβ receptor mutations and the frequent presence of p53 mutations. The pro-tumorigenic program orchestrated by TGFβ in CRC cells was mediated through TGFBI, the expression of which was positively regulated by non-canonical TGFβ signaling cascades. TGFBI inhibition was sufficient to significantly reduce liver metastasis formation in vivo. Moreover, TGFBI pro-tumorigenic function was linked to its ability to stimulate angiogenesis. TGFBI levels were higher in serum samples from untreated patients with CRC than in patients who were receiving chemotherapy. A radiolabeled anti-TGFBI antibody selectively targeted metastatic lesions in vivo, underscoring its diagnostic and therapeutic potential. Conclusions: TGFβ signaling in CRC cells directly contributes to their metastatic potential and stromal cell-independence. Proteins downstream of activated TGFβ, such as TGFBI, represent novel diagnostic and therapeutic targets for more specific anti-metastatic therapies.

Representative images for each insert were taken at a 5X magnification and migrating cells were quantified by densitometry using the ImageJ software (National Institute of Health, USA, public access). Three wells (technical replicates) per condition were counted.

Colony formation assay
For colony formation, HT29 cells were seeded in 6-well plates (500 cells/well) in the presence of absence of SB202190 (10 µM) or BAY11-7082 (5 µM). shNT or shTGFBI SW1222 cells were seeded at a density of 1000 cells/well. After 9 days of culture, cells were washed twice in 1X PBS, fixed, and stained with 0.5% crystal violet in methanol for 30 min.
Colonies were counted using the ImageJ software. Experiment were done in triplicate and repeated three times.

Proteomic analyses
Protein extracts from cells were prepared as described for western blot analysis (see Materials and Methods). Conditioned media were collected from cancer cells, centrifuged to remove debris, and concentrated using Amicon ultra-filtration devices with 3 kDa cutoff (Merck-Millipore; cat. no. UFC900308). Then, 50 mg of protein extracts or concentrated conditioned media were reduced by adding 20 mM DTT at 60 °C for 30 min, followed by alkylation using 50 mM 2-chloroacetamide at RT for 30 min. Proteins were precipitated using the 2D Clean-Up Kit (GE Healthcare,Chicago,IL,USA;cat. no. 80648451 The peptide samples were analyzed using the 1D-nano-HPLC-Q-TOF 6600 system (Sciex, Framingham, MA, USA). One microgram of sample was injected in the C18 column (Acclaim® 75 µm x 150 mm, p/n: 162224; Dionex, California, USA). Peptides were resolved with a gradient of 0-40% phase B (90% acetonitrile, 9.9% water and 0.1% formic acid) for 100 min at the flow rate of 0.3 µl/min. Two acquisition modes were used: data-dependent (DDA) for the library, and SWATH for the individual samples. In the DDA mode the setting was as follows: one full scan in the mass range 400 to 1600 m/z, followed by up to 30 MS/MS scans of the most intensive peptides bearing +2 or +3 charges. The acquired data for each fraction of the library sample were merged and used for MS/MS database search with the Protein Pilot software (Sciex). For the SWATH acquisition, the DDA method was adapted using the automated method generator embedded in the Analyst software (Sciex). Proteins were quantified using the SWATH algorithm in the Peak View software and the previously generated protein library. Further data analysis was conducted using R. SWATH data were normalized based on the total protein load, estimated by the sum of all MS intensities reported for all proteins found in a given sample. The mean values of 3 replicates per condition were calculated, and ratios of treated versus control samples were calculated. Proteins that showed changes higher than 2-fold were retained for further analysis.

Network analysis using the STRING software
Protein-protein interaction analysis was performed using the online STRING tools, version 10 (www.string-db.org).

Production and validation of anti-TGFBI antibodies (10G9A10 and 4G6B10)
Custom-made murine monoclonal anti-TGFBI antibodies were produced (Diaclone, France) by immunizing 5 mice by injection in the footpads of 1 µg/footpad of recombinant TGFBI (Targetome SA, Belgium). Cells were collected from lymph nodes, pooled and fused to generate the myeloma X63/AG.8653 (Diaclone), and distributed in 96-well plates. TGFBI reactivity was screened by ELISA. Briefly, 96-well plates were first coated with goat antimouse IgG (Diaclone), saturated in PBS/5% BSA. and then incubated with 10 µl/well of hybridoma supernatants. Following extensive washings, 2 ng of biotinylated TGFBI or biotinylated POSTN was added into each well, followed by incubation at RT for 1h. Finally, plates were washed and streptavidin-HRP (Europa Bioproducts; cat. no. PZCJ30H) was added to each well. The signal was revealed using TMB according to the manufacturer's recommendations (1-Step Ultra TMB-ELISA, Thermo Scientific; cat. no. 34028).
Purified anti-TGFBI antibodies (clones 10G9A10 and 4G6B10) were also evaluated using HEK-293 cells transfected with TGFBI or POSTN (Targetome) and cultured in the presence of brefeldin A (Sigma-B7651) for 5h. Unlabeled isotype controls were from Diaclone. Cells were incubated first with Cytofix/Cytoperm buffer (BD Biosciences; cat. no. 554722) at 4 °C for 20 min, and then with the antibodies under study at different dilutions (1 µg to 0.01 µg/well) in BD Perm/Wash buffer (BD Biosciences; cat. no. 554723) at 4 °C for 30 min.
Binding of anti-TGFBI antibodies was evaluated by surface plasmon resonance (SPR) analysis and immunofluorescence. Binding of the antibodies 4G6B10 and 10G9A10 to immobilized TGFBI was assessed using a BIAcore X-100 apparatus (Cytiva, Marlborough, MA, USA). TGFBI (20 µg/ml in 10 mM sodium acetate, pH 4.0) was allowed to react with a flow cell of a CM5 sensor chip previously activated with a mixture of 0.2 M N-ethyl-N'-(3dimethylaminopropyl)-carbodiimide hydrochloride and 0.05 M N-hydroxysuccinimide (35 µl, flow rate 10 µl/min). After ligand immobilization, matrix neutralization was performed with 1.0 M ethanolamine (pH 8.5) (35 µl, flow rate 10 µl/min) and activated/deactivated dextran was used as reference (control) system. Increasing concentrations of 4G6B10 and 10G9A10 (from 18.75 to 600 nM) were injected over the TGFBI-coated sensor chip and the response was recorded by tracking the SPR intensity change upon binding progression. Injection lasted for 2 min (flow rate 10 µl/min) to allow the association with immobilized TGFBI and was followed by 10 min of dissociation; each run was performed in HBS-EP buffer (Cytiva) and the sensor chip was regenerated with glycine, pH2. The equilibrium (plateau) values of the SPR sensorgrams were used to build the binding isotherms, after normalization. Binding isotherm points were fitted with the Langmuir equation for monovalent binding to evaluate the mass surface dissociation constant, Kd. The best-fitting procedure was performed with the SigmaPlot 11.0 software package (Systat Software Inc.).
For immunofluorescence analysis, 5 mm sections from fresh frozen human CRC-LM samples were cut using a cryostat (CM3050, Leica) and fixed in cold methanol. Sections were then incubated with 0.5% Triton X-100 in PBS at RT for five minutes, followed by blocking in 5% BSA at RT for 1h and incubation with the anti-TGFBI antibodies (clones 10G9A10 and 4G6B10) (1:500 dilution in 1% BSA) at 4 °C overnight. Following extensive washing, sections were incubated with the secondary goat anti-mouse antibody coupled to Alexa  Table S1: Clinical data of the patients whose serum samples were used for TGFBI measurement by ELISA. Table S2-S3: Clinical data of the patients whose primary CRC (S2) and CRC-LM (S3) samples were used for immunohistochemistry and immunofluorescence analyses.