|Home | About | Journals | Submit | Contact Us | Français|
Cancer associated fibroblasts (CAFs) constitute an abundant stromal component of most solid tumors. Fibroblast activation protein (FAP) α is a cell surface protease that is expressed by CAFs. We corroborate this expression profile by immunohistochemical analysis of colorectal cancer specimens. To better understand the tumor‐contextual role of FAPα, we investigate how FAPα shapes functional and proteomic features of CAFs using loss‐ and gain‐of function cellular model systems. FAPα activity has a strong impact on the secreted CAF proteome (“secretome”), including reduced levels of anti‐angiogenic factors, elevated levels of transforming growth factor (TGF) β, and an impact on matrix processing enzymes. Functionally, FAPα mildly induces sprout formation by human umbilical vein endothelial cells. Moreover, loss of FAPα leads to a more epithelial cellular phenotype and this effect was rescued by exogenous application of TGFβ. In collagen contraction assays, FAPα induced a more contractile cellular phenotype. To characterize the proteolytic profile of FAPα, we investigated its specificity with proteome‐derived peptide libraries and corroborated its preference for cleavage carboxy‐terminal to proline residues. By “terminal amine labeling of substrates” (TAILS) we explored FAPα‐dependent cleavage events. Although FAPα acts predominantly as an amino‐dipeptidase, putative FAPα cleavage sites in collagens are present throughout the entire protein length. In contrast, putative FAPα cleavage sites in non‐collagenous proteins cluster at the amino‐terminus. The degradomic study highlights cell‐contextual proteolysis by FAPα with distinct positional profiles. Generally, our findings link FAPα to key aspects of CAF biology and attribute an important role in tumor–stroma interaction to FAPα.
Solid cancers comprise a microenvironment, which includes different stromal cell types together with secreted proteins and extracellular matrix components. Cancer associated fibroblasts (CAFs) constitute a highly abundant stromal cell type (Kalluri and Zeisberg, 2006), together with further stromal components such as endothelial cells or infiltrating immune cells. They are characterized by their spindle‐shape morphology and expression of specific markers, such as alpha smooth muscle actin (αSMA), vimentin or fibroblast activation protein alpha (FAPα) (Augsten, 2014).
CAFs play a pivotal role in the tumor microenvironment, in particular due to their high capacity of releasing proteins. Thus, CAFs govern key features of tumor: growth, invasion, and metastasis (Augsten, 2014; Han et al., 2015). By secreting versatile signaling molecules such as chemokines, cytokines and growth factors, CAFs activate multiple signaling pathways in tumor cells as well as stimulate the inflammatory response by attracting immune cells (Bremnes et al., 2011; Zhang and Liu, 2013). Given their distinctive features together with high abundance in the tumor mass, CAFs become a prominent target in development of novel therapeutic strategies (Gonda et al., 2010; Kakarla et al., 2012; Ozdemir et al., 2014; Duluc et al., 2015).
CAFs are major sources of proteases and protease inhibitors, which are present at the cell surface and secreted into the tumor microenvironment (Augoff et al., 2014). Proteolysis is an irreversible post‐translational processing that affects every protein (Shahinian et al., 2013; Schilling and Findeisen, 2014). Limited proteolysis leads to stable cleavage products with new or altered biological activities (Rogers and Overall, 2013). Examples include activation of zymogens, precursor processing of structural proteins, and inactivation of chemokines (Lai et al., 2015). Therefore, proteolysis and the interplay of proteases and their inhibitors are of high importance in tumor biology. In this context, the secreted by stromal cells matrix metalloproteases (MMPs) constitute a prominent example (Lunardi et al., 2014). MMP activity has been linked to many types and stages of cancers (Nagase and Woessner, 1999; Overall and Kleifeld, 2006; Radisky and Radisky, 2007) and reported to facilitate tumor progression (Kessenbrock et al., 2010; Sato and Takino, 2010; Taddei et al., 2013; De Bock et al., 2014). However, MMP inhibition has failed as an anti‐tumor therapy; largely due to an incomplete understanding of MMP biology and substrates (Coussens et al., 2002; Overall and Lopez‐Otin, 2002).
FAPα is a cell surface serine protease, expressed by CAFs but not by the actual tumor cells in most solid tumors (Garin‐Chesa et al., 1990; Zi et al., 2015), leading to FAPα being a marker for CAFs (Augsten, 2014). FAPα expression is tightly regulated and in adult tissues is maintained on a very low level (Rettig et al., 1988; Garin‐Chesa et al., 1990; Rettig et al., 1993; Scanlan et al., 1994; Martin, 1997). FAPα is detected during early development in mesenchymal tissues (Niedermeyer et al., 2001), in disorders associated with activated stroma, including healing wounds, cirrhosis and pulmonary fibrosis (Mathew et al., 1995; Acharya et al., 2006; Wang et al., 2008) as well as tumor stroma of epithelial cancers (Garin‐Chesa et al., 1990; Rettig et al., 1993; Kelly, 2005) with particularly elevated expression at the invasive front (Cohen et al., 2008). FAPα expressing CAFs are thought to suppress antitumor immunity (Kraman et al., 2010). FAPα is further expressed by mesenchymal stem cells (MSCs) (Bae et al., 2008) and depletion of FAPα‐positive pluripotent mesenchymal cells results in cachexia (Roberts et al., 2013; Tran et al., 2013). The growing body of literature indicates that CAFs can be derived from FAPα‐expressing MSCs (Mishra et al., 2008; Paunescu et al., 2011; Kidd et al., 2012; Shangguan et al., 2012; Brennen et al., 2013; Jung et al., 2013; Kim et al., 2014b).
Tumor‐promoting or ‐suppressing effects of FAPα expression are conversely discussed (Ariga et al., 2001; Cheng et al., 2002; Huang et al., 2004; Ramirez‐Montagut et al., 2004; Cheng et al., 2005). Several studies have been based on altered FAPα expression in tumor cells, including ovarian cancer cells (Yang et al., 2013b), breast cancer cells (Huang et al., 2011b) and human embryonic kidney cells (Cheng et al., 2005). These settings do however not address the rather stromal expression of FAPα. FAPα deletion in a murine lung cancer model reflects the stromal expression in a more suitable manner. This study supports a pro‐tumorigenic and pro‐angiogenic function of FAPα (Santos et al., 2009).
FAPα predominantly functions as an amino‐dipeptidase, which cleaves specifically after proline in the P1 position (P1Pro). It also exhibits endopeptidase activity, preferentially cleaving after Gly‐Pro sequence motif (P2GlyP1Pro) (Aertgeerts et al., 2005, 2006, 2006; Hamson et al., 2014). FAPα is involved in remodeling of the extracellular matrix (ECM) with native and denatured collagens (gelatin), being reported as FAPα substrates (Aoyama and Chen, 1990; Pineiro‐Sanchez et al., 1997; Park et al., 1999; Ghersi et al., 2002; Christiansen et al., 2007). FAPα is expressed as an active protease without the requirement for zymogen activation. FAP‐α −/− mice lack an overt phenotype (Niedermeyer et al., 2000).
Due to its near‐exclusive expression in tumor stroma, FAPα has become a widely investigated target for antitumor therapy, including vaccination strategies (Loeffler et al., 2006; Gottschalk et al., 2013), pro‐drug conversion (Brennen et al., 2012), and specific delivery of cytotoxic drugs (Ostermann et al., 2008). Several attempts to develop FAPα inhibitors have been reported (Edosada et al., 2006, 2006, 2013), including recently published selective small molecule FAPα inhibitors (Jansen et al., 2014). Earlier, inhibition of FAPα enzymatic activity with the small molecule Talabostat in patients with metastatic, non‐resectable colorectal cancer yielded only minimal clinical benefit (Narra et al., 2007). Application of a humanized antibody against FAPα (sibrotuzumab) in advanced colorectal cancer has also yielded little clinical benefit (Scott et al., 2003). Both clinical studies did however underline clinical safety of FAPα targeting and did not report adverse side effects. FAPα inhibition in less advanced disease settings has not yet been investigated.
In the present study, we aim to investigate how FAPα determines the function as well as the secreted proteome and degradome of CAFs in both FAPα loss‐ and gain‐of function systems. Our findings show that FAPα influences key aspects of the tumor microenvironment, including vessel sprouting and matrix stiffness. Of particular note is a pronounced link between FAPα and transforming growth factor β (TGFβ) signaling.
FFPE tissue specimens from previously well characterized (Lassmann et al., 2009; Herz et al., 2012; Sijare et al., 2015) primary colorectal carcinomas (n = 19) were re‐classified according to the actual WHO Classification of Tumours of the Digestive System as follows: adenocarcinoma NOS (n = 16), mucinous adenocarcinoma (n = 3) and tubular adenoma with high grade intraepithelial dysplasia (neoplasia) (n = 1). Ethical approval was obtained from the local institutional ethics committee (University of Freiburg, Ethik‐Kommission, ID #234/13).
Tissue slices of 2 μm thickness were prepared and stained for FAPα (R&D Systems, AF3715, 1:700). After an overnight incubation at 56 °C, a heat‐induced antigen retrieval was performed at 97–99 °C for a period of 40 min in pH 6.1 using Dako antigen retrieval buffer S1699. Primary antibody was incubated at room temperature for 30 min by using nVision™ FLEX+ (Dako, Autostainer Plus). As the second antibody Dako rabbit‐anti‐goat (E0466) was used. All slides were counterstained with hematoxylin, dehydrated in ascending alcohol concentrations and covered.
CT5.3 cancer associated fibroblasts (CAFs) were described previously (Van Hoorde, Braet et al., 1999). Briefly, fibroblasts were derived from a human colon cancer specimen by the outgrowth method, characterized by αSMA stress fibers and presence of vimentin and prolyl 4‐hydroxylase and absence of cytokeratin. Primary cells were immortalized by transduction with the human telomerase gene. The P‐48GB cells were isolated by Bachem's outgrowth method in Quantum 333 medium (attached) from a human ampullary adenocarcinoma in 2012, expanded to passage 2 in 75 cm2 standard culture flasks and cryopreserved in standard cryomedium containing DMSO and immortalized by transduction with SV40 gene.
Cells were cultured in Dulbecco's modified Eagle's medium (DMEM, PAN, Aidenbach, Germany) supplemented with 10% fetal calf serum (PAN) and 1% penicillin/streptomycin solution (Gibco/Invitrogen, Paisley, UK) at 37 °C in a humidified atmosphere containing 5% CO2. In assays with TGFβ treatment, recombinant human TGFβ1 protein (R&D Systems) at 10 nM concentration was used.
The MISSION shRNA system (Sigma–Aldrich) was used to generate a stable knock‐down of FAPα expression. The puromycin resistance of the pLKO.1‐puro vector was replaced with a gentamicin resistance cassette. The FAPα shRNA sequence is CCGGGCATTGTCTTACGCCCTTCAACTCGAGTTGAAGGGCGTAAGACAATGCTTTT, the scrambled control sequence is CCGGCAACAAGATGAAGAGCACCAATTTTT. Retroviral transduction was used for transfection of CT5.3 and P‐48GB cells (Soneoka et al., 1995). 24 h after transfection, cells were selected with medium containing 500 μg/ml gentamicine (Gibco/Invitrogen, Paisley, UK) for two weeks.
The bicistronic pMIG system (Tholen et al., 2013) was used for stable overexpression of active FAPα or the enzymatically inactive S624A mutant (Cheng et al., 2005). FAPα cDNA was cloned into a pMIG expression vector, yielding a construct that encodes for both FAPα and green fluorescent protein (GFP). Both are controlled by the same promoter and are transcribed as a combined mRNA. An internal ribosomal entry site between both coding sequences enables expression of untagged FAPα in direct correlation to GFP expression, which was used to fractionate the cell population by preparative fluorescence‐assisted cell sorting. Individual fractions were probed for FAPα expression by immunoblotting. For the S624A mutation, site directed mutagenesis was performed, modifying the triplet TCC (coding for S624) to GCC (coding for A624). The complete insert was sequenced to verify that no unintentional mutations occurred.
CT5.3 cells were cultivated in DMEM containing 10% FCS until 90% confluence, washed three times with phosphate buffered saline (PBS), and incubated for 24 h (37 °C, 5% CO2) in serum‐free DMEM without phenol red. After 24 h, cell‐conditioned medium (CCM) was collected, supplemented with protease inhibitors (5 mM ethylenediaminetetraacetic acid (EDTA), 0.01 mM trans‐epoxysuccinyl‐l‐leucylamido(4‐guanidino)butane (E64), 1 mM phenylmethanesulfonyl fluoride (PMSF)), centrifuged (5 min, 1000 × g, 4 °C), and filtered using a 0.2 μm filter to remove debris. Samples were stored in −80 °C until processing. Samples for comparative proteomic analysis were prepared as described previously (Tholen et al., 2014). Briefly, proteins were precipitated with trichloracetic acid (TCA), solubilized, trypsinized, reduced, and alkylated. Samples were then labeled with 20 mM either “light” 12CH2O formaldehyde (CT5.3shctr and CT5.3FAPact) or “heavy” 13CD2O formaldehyde (CT5.3shFAP and CT5.3FAPasm) in the presence of 20 mM sodium cyanoborohydride. After quenching the reaction with glycine, samples were combined in pairs (CT5.3shctr & CT5.3shFAP; CT5.3FAPasm & CT5.3FAPact) in 1:1 ratio. Following desalting by C18 solid phase extraction (Sep‐Pak C18 Plus Light Cartridge, Waters, Frankfurt, Germany), samples (ca. 300 μg) were fractionated by strong cation exchange chromatography as described previously (Tholen et al., 2013; Shahinian et al., 2014; Tholen et al., 2014) and analyzed by liquid chromatography–tandem mass spectrometry (LC–MS/MS).
For mass spectrometry analysis, an Orbitrap XL system (Thermo Scientific, Bremen, Germany) or Q‐Exactive plus system (Thermo Scientific, Bremen, Germany) was used. The Orbitrap XL system was coupled to an Ultimate mate3000 micro pump (Thermo Scientific) with a flow rate of 300 nl/min. For peptide separation, 0.5% acetic acid and 0.5% acetic acid in 80% acetonitrile (both water and acetonitrile were at least HPLC gradient grade quality) with a gradient of increasing acetonitrile concentration was used. Column tips with 75 μm inner diameter and 11 cm length were self‐packed (Olsen et al., 2004) with Reprosil‐Pur 120 ODS‐3 (Dr. Maisch, Ammerbuch, Germany). The mass spectrometer was operated in the data‐dependent mode and switched automatically between MS and MS/MS. The Q‐Exactive system was coupled to an Easy nanoLC 1000 (Thermo Scientific) with a flow rate of 300 nl/min each. Buffer A was 0.5% formic acid, and buffer B was 0.5% formic acid in 100% acetonitrile (water and acetonitrile were at least HPLC gradient grade quality). A gradient of increasing organic proportion was used for peptide separation (5–40% acetonitrile in 80 min). The analytical column was an Acclaim PepMap column (Thermo Scientific), 2 μm particle sizes, 100 Å pore sizes, length 150 mm, I.D. 50 μM. The mass spectrometer operated in data dependent mode with a top 10 method at a mass range of 300–2000.
LC–MS/MS data in raw format was converted to the mzXML format (Pedrioli et al., 2004) using msconvert (Kessner et al., 2008) with centroiding of MS1 and MS2 data, and deisotoping of MS2 data. Peptide sequences were identified by X!Tandem (Version 2013.09.01) (Craig and Beavis, 2004) using the human reviewed canonical Uniprot sequences (without isoforms) together with an equal number of randomized decoy sequences, generated by DBtoolkit (Martens et al., 2005). Tryptic cleavage specificity with no missed cleavage sites was applied. Mass tolerance was 10 ppm for parent ions and 0.3 Da (Orbi XL data) or 20 ppm (Q‐Exactive data), respectively, for fragment ions. Static modifications were cysteine carboxyamidomethylation (+57.02 Da) and lysine and N‐terminal dimethylation (light formaldehyde 28.03 Da; heavy formaldehyde 34.06 Da). X!Tandem results were further validated by PeptideProphet at a confidence level of >95% (MPT = 0.05). Corresponding protein identifications are based on the ProteinProphet algorithm (Nesvizhskii et al., 2003) with a false discovery rate <1.0%. The relative quantitation for each protein was calculated from the relative areas of the extracted ion chromatograms of the precursor ions and their isotopically distinct equivalents using the ASAPratio algorithm (Li et al., 2003) together with XPRESS algorithm (Han et al., 2001) as described previously (Shahinian et al., 2014; Tholen et al., 2014). Ingenuity Pathway Analysis (QIAGEN) was used for enrichment analysis; STRING (Franceschini et al., 2013) was used to annotate functional interaction of proteins.
Cell‐conditioned media were prepared as described above and concentrated using ultrafiltration (Microcon Millipore, 10 kDa molecular weight cutoff (MWCO)). Protein content was quantified using the Bradford assay (Bio‐Rad). Total cell lysates were prepared by on‐plate lysis. Cells were washed twice with cold PBS and lysed with lysis buffer containing 20 mM Tris–HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 1% Triton X‐100, and protease inhibitors (0.01 mM E64, 1 mM PMSF). Protein samples were denatured for 5 min at 95 °C in reducing sample buffer (50 mM Tris, pH 6.8, 1% SDS, 10% glycerol, 10 mM DTT, bromphenol blue). 15–25 μg of protein were loaded on 10 or 12% SDS–polyacrylamide gels. After electrophoretic separation (1 h in 160 mV), proteins were transferred on polyvinylidene fluoride (PVDF) membranes using a semi‐dry blotting system (BioRad). After blocking in 4% (wt/vol) milk powder in PBS supplemented with 0.1% Tween 20 (PBS‐T), membranes were incubated with primary antibodies (FAPα, polyclonal, sheep, 1:1000, R&D, AF3715; α‐SMA, polyclonal, rabbit, 1:500, Abcam, ab15734; PEDF, goat, polyclonal, 1:1000, R&D, AF1177; CCL2, goat, polyclonal, 1:2000, R&D, AF279NA; VEGFC, polyclonal, goat, 1:2000, R&D, AF752; LAP‐TGFbeta, polyclonal, goat, 1:500, R&D, AF246NA; TGFbeta, polyclonal, chicken, 1:250, R&D, AF101NA; OPN, polyclonal, rabbit, 1:1000, Abcam, ab63856; MMP1, monoclonal, mouse, 1:250, R&D, MAB901; TPA, polyclonal, sheep, 1:1000, R&D, AF7449; tubulin, monoclonal, mouse, 1:1000, Sigma–Aldrich, T 6199) overnight at 4 °C on a rotator. After washing 3 times in PBS‐T, membranes were incubated for 1.5 h with the secondary antibody (anti‐goat, Sigma–Aldrich, A5420; anti‐mouse, Dianova, 115‐035‐003; anti‐rabbit, BioRad, 172‐1019; anti‐sheep, Dianova, 713‐035‐147; anti‐chicken, Sigma–Aldrich, A9046‐1ML), dissolved in PBS‐T containing 4% milk powder. Membranes were washed three times in PBS‐T buffer and developed using the West Pico Chemiluminescent substrate (Pierce). Horseradish peroxidase activity was detected with a LumiImager device (Roche, Mannheim, Germany). Tubulin or GAPDH served as internal standards for total cell extract immunoblots; for CCM samples, coomassie staining of gels was used as an equal loading control.
1 × 106 CT5.3 cells were embedded in 500 μl of rat tail collagen type I (3 mg/ml; BD Biosciences) solution, in DMEM supplemented with 10% FCS. The mixture was loaded on wells of 24‐well plate and incubated in humidified atmosphere in 37 °C until solidified. After 2 h, 500 μl DMEM supplemented with 10% FCS was added on top of the collagen layer. The plate was further incubated (37 °C, 5% CO2) and pictures of the collagen gel plugs were taken after 48 h. Quantification of collagen areas was performed using ImageJ (Schneider et al., 2012).
Preparation of endothelial cell spheroids and sprouting assay were performed as described previously (Buehler et al., 2013). Briefly, human umbilical vein endothelial cells (HUVEC) (PromoCell) until fifth passage were cultured as monolayers at 37 °C and 5% CO2 in a humidified atmosphere in endothelial cell growth medium ECGM (Cell Systems). HUVECs were harvested from subconfluent monolayers and suspended in ECGM containing 10% FCS and 0.25% carboxy‐methylcellulose (Sigma–Aldrich). Five hundred cells were seeded in one hanging drop to assemble into a single spheroid. After 24 h, spheroids were harvested and used for sprouting analysis in a matrix of rat type I collagen (Corning). 0.5 ml of matrix containing 30 endothelial cell spheroids were used per well. As negative control, spheroid containing collagen matrix was seeded into non‐adherent plastic wells of a 24‐well plate. In positive controls, sprouting was induced through 25 ng/ml recombinant human VEGF165 (R&D Systems 293‐VE‐010). To test angiogenic properties of CAFs, spheroid containing matrix was seeded on top of a 90% confluent CAF monolayer. 25,000 CAFs had been seeded per well in 24‐well plates 24 h in advance. Freshly prepared collagen gels were incubated in a humidified incubator (37 °C, 5% CO2), until solidified. 100 μl of serum‐free medium (Cell Systems) per well was added on top and incubation continued for 24 h. After 24 h, gels were photomicrographed and spheroid sprouting assessed quantitatively. Results are expressed as mean number of sprouts per spheroid ± S.E.M. The relative sprouting rate of each condition was normalized to the sample containing CT5.3 cells with higher FAPα level set to 100%.
FAPα specificity was probed using tryptic proteome‐derived peptide libraries from Escherichia coli as described previously (Schilling and Overall, 2008; Biniossek et al., 2011; Schilling et al., 2011b). Briefly, E. coli were grown in Luria‐Bertani medium, harvested, and lysed as described previously, followed by tryptic digestion and chemical protection of all primary amines (Schilling et al., 2011b). Trypsin was inactivated by heat and addition of 1 mM PMSF. The peptide library was purified by reverse‐phase solid phase extraction. The purified peptide library was then incubated with recombinant human FAPα (R&D systems). Protease:library ratio was 1:300 in 50 mM HEPES‐HCl, pH 7.5, 150 mM NaCl with 18 h incubation at 37 °C. Prime‐side cleavage products present newly generated, free amino termini, which were tagged by amine‐reactive sulfoNHS‐SS‐biotin, followed by streptavidin‐based isolation of biotinylated cleavage products and their analysis by LC–MS/MS. The corresponding non‐prime cleavage products were derived bioinformatically, as described previously (Schilling and Overall, 2008). WebPICS was used for visualization (Schilling et al., 2011a).
TAILS was performed as described previously (Kleifeld et al., 2010; Shahinian et al., 2014; Tholen et al., 2014). Briefly, 2.5 mg of CCM (prepared as previously described) per condition was used and the N‐termini were blocked with either heavy or light formaldehyde, and then mixed in a 1:1 ratio. After tryptic digest samples were desalted using a reversed phase C18 column, internal peptides were captured by the HPG‐ALD polymer. Non‐bound, N‐terminal peptides were collected by ultrafiltration, prefractionated by SCX as described above, and analyzed by LC–MS/MS as described above. TAILS data analysis followed the procedures outlined above with the following differences. Semi Arg‐C specificity with up to three missed cleavage sites was applied. Static modifications were (+57.02 Da), lysine and N‐terminal dimethylation (light formaldehyde + 28.03 Da; heavy formaldehyde + 34.06 Da). The relative quantification (fold changes) for each peptide was calculated using XPRESS algorithm (Han et al., 2001).
Formalin‐fixed and paraffin‐embedded tissue specimens of 20 patients with colorectal cancer were examined for FAPα expression (Figure 1). FAPα expression was predominantly found in stromal fibroblast adjacent to tumor glands in 14 of 19 cases, with 13/14 colorectal carcinomas of NOS histology (Figure 1A–D) and 1/14 colorectal cancers of mucinous histology (Figure 1G). Slight or negligible FAPα staining was seen in 5 of 19 cases, with 3/5 colorectal cancers with NOS histology (Figure 1E), 2/5 mucinous colorectal carcinomas (Figure 1F, H). Interestingly in one case with mixed differentiation (mucinous and NOS), there was an increased staining for FAPα in fibroblasts only adjacent to parts of tubular gland formation (Figure 1D), but not in fibroblasts close to mucinous tumor cells (Figure 1H). In one case, a tubular adenoma with high grade intraepithelial neoplasia (dysplasia) also showed slight FAPα staining in fibroblasts adjacent to neoplastic (dysplastic) epithelium.
FAPα is expressed in fibroblasts within the stroma of colorectal cancers. Immunohistochemical stainings were performed for tissue specimens of colorectal cancer patients, showing strong FAPα staining in 14/19 cases of NOS (A–D*) ...
These finding corroborate stromal fibroblast expression of FAPα in colorectal cancers and are in line with previous reports, including several tumor entities (Scanlan et al., 1994; Cheng et al., 2002; Huber et al., 2003; Henry et al., 2007; Shan et al., 2012; Yang et al., 2013a; Kim et al., 2014a; Mhawech‐Fauceglia et al., 2015).
To study the role of FAPα in regulating CAFs biology, we chose CT5.3 fibroblasts isolated from colorectal cancer CT5.3 fibroblasts are characterized by prototypic, spindle‐like shape and substantial expression of the CAF markers αSMA and FAPα (Figure 2a).
(a) Western blot detection of CAFs markers αSMA and FAPα in the CT5.3 cell models. The expression of both markers confirms the cancer associated fibroblast origin of CT5.3 cell line. Human recombinant FAPα was loaded as a positive ...
To create the loss‐of‐function system, FAPα expression in CT5.3 cells was repressed by stable transfection with a suitable shRNA (referred to as CT5.3shFAP), yielding >90% reduction of FAPα abundance, compared to the corresponding control of scrambled shRNA (referred to as CT5.3shctr) (Figure 2b). In order to focus on effects mediated by FAPα enzymatic activity, we generated a corresponding gain‐of‐function system, consisting of stable overexpression of either wild‐type FAPα or the enzymatically inactive S624A mutant. The S624A mutation has been established previously to specifically abolish FAPα activity (Cheng et al., 2005). Based on the bicistronic pMIG technology, similar overexpression levels of FAPαwt and FAPαS624A were achieved (Figure 2b). We aim to highlight FAPα‐dependent effects that are apparent in both the loss‐ and gain‐of‐function system. This approach allows focusing specifically on FAPα enzymatic activity.
Due to its emerging role in mediating and triggering key features of cancer, great interest in the analysis of the cancer secretome composition has recently developed (Patel et al., 2014). Secretome profiling has previously yielded useful insight into the role of secreted or cell surface proteases, e.g. highlighting a link of kallikrein proteases to TGFβ‐signaling in ovarian cancer cells (Shahinian et al., 2014). The remarkable alterations in the secretome were previously reported upon cathepsin B and L double deficiency (Tholen et al., 2014). Since CAFs are an important source of secreted proteins in the tumor microenvironment, the impact of FAPα on their secretome composition is of particular interest. We performed quantitative proteomic analysis of CCM derived from the FAPα loss‐of‐function setting (comparing CT5.3shctr vs CT5.3shFAP) as well as from the FAPα gain‐of‐function setting (comparing CT5.3FAPact vs CT5.3asm). The quantitative proteomic approach included stable isotope labeling with either light (d(0)12C) or heavy (d(2)13C) formaldehyde, followed by LC–MS/MS. A total of 2420 (CT5.3shctr vs CT5.3shFAP) and 2140 proteins (CT5.3FAPact vs CT5.3FAPasm), respectively, were identified and quantified with considerable overlap between both experiments The average number of peptides per protein was 20.0 and 20.8 and the average protein sequence coverage was 16.6% and 17.9% for the loss‐ and gain‐of‐function experiment respectively (Figure 3a, Suppl. Tab. S1a and b).
Quantitative secretome profiling. (a) Overlap of 1713 proteins (around 80%) which were consistently identified and quantified in both experiments. (b) Distribution of fold‐change values (log2 of light/heavy ratios) for the secretome comparison ...
Relative protein abundances were normalized to increased FAPα activity vs decreased FAPα activity; yielding the comparative settings “CT5.3shctr:CT5.3shFAP” and “CT5.3FAPact:CT5.3FAPasm”, respectively. Protein ratios are expressed as fold‐change (Fc) values (log2 of light to heavy ratios, adjusted as previously described). Both proteome comparisons displayed a normal distribution (Figure 3b). The software ASAPRatio was used to denote the statistical significance of a protein being quantitatively changed (Li et al., 2003). We further defined the following criteria to identify proteins with affected abundance upon altered FAPα activity: (a) protein identified and quantified in both experiments; (b) protein annotated as secreted or localized at the cell surface; (c) protein abundance was consistently increased or decreased >50% in both experiments; (d) joint ASAPRatio p value for the quantitative alteration <0.1 (Shahinian et al., 2014; Tholen et al., 2014); (e) favorable manual inspection of the extracted ion chromatograms. These strict criteria yielded 47 proteins with significantly affected abundance upon altered FAPα activity (12 proteins with decreased abundance; 35 proteins with increased abundance; Table 1, Suppl. Tab. S1a and b).
List of proteins significantly and consistently affected by FAPα expression – in both knock‐down and overexpression. Uniprot ID and recommended name according to Uniprot database (Uniprot Consortium, 2013). p‐value denotes ...
We corroborated the quantitative alteration for several proteins by Western blotting including tissue plasminogen activator (tPA), pigment epithelium‐derived factor (PEDF), vascular endothelial growth factor C (VEGFC), TGFβ – mature for and LAP, chemokine (C–C motif) ligand 2 (CCL2), interstitial collagenase (MMP1), and osteopontin (OPN). Hereby we show a good agreement of proteomic results with Western blot analysis in the secretome of CT5.3 cells (Figure 4a). To define which domain of the TGFβ molecule was identified, we additionally performed peptide mapping (Suppl. Figure S1).
Western blot detection of several proteins identified as altered in the LC–MS/MS quantitative secretome profiling in CT5.3 cell conditioned medium: tPA, PEDF, VEGFC, LAP, mature TGFβ dimer, CCL2, MMP1, and OPN. The electrophoresis was ...
To further investigate if the observed changes can be considered as a generalizable phenomenon, we created a FAPα depleted variant of P‐48GB CAF cells (Suppl. Figure S2). We derived cell conditioned medium as described for the CT5.3 cells and investigated the level of several proteins by Western blotting. As shown in Suppl. Figure S3, the panel of probed proteins corroborates the changes observed in CT5.3 cells.
Several of the FAPα‐affected proteins cluster into biological themes that are considered as key functionalities of cancer associated fibroblasts (Han et al., 2015). These include regulation of angiogenesis (PEDF, angiopoietin‐1, VEGFC) and processing of extracellular matrix (a disintegrin and metalloproteinase with thrombospondin motifs 8 (ADAMTS8), MMP1).
STRING analysis of the 47 affected proteins highlighted functional connectivity for >50% of them, linking to the ECM organization, further emphasized an important role of TGFβ (Figure 5a, Suppl. Tab. S2a). In line with this observation, enrichment analysis with Ingenuity Pathway Analysis (IPA) highlighted the affected TGFβ system in both loss‐ and gain‐of‐function systems (Figure 5b, Suppl. Tab. S2b).
(a) STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) analysis of proteins affected by altered FAPα activity. The cluster represents overall connectivity of around 50% of significantly and consistently altered proteins in the ...
Alterations of FAPα levels result in perturbations of the CAF secretome composition, which mediate pivotal biological functionalities, such as TGFβ signaling, ECM structure, and angiogenesis signaling. The enrichment of collagens and other secreted structural proteins can be due to the FAPα gelatinase/collagenase activity, which brings deposited ECM proteins into the soluble form. We consider the observed, FAPα‐dependent perturbations of protein levels within the CAF secretome to be the combined result of the persistently altered FAPα activity and accumulated downstream effects.
When cultivated in low confluence, CT5.3 cells do not exhibit major morphological differences dependent of FAPα level (Figure 6a). Interestingly, when highly confluent, depletion of FAPα (CT5.3shFAP) resulted in a less spindle‐shaped and more epithelial cell morphology in comparison to the corresponding control cells (CT5.3shctr) (Figure 6b). Such changes in morphology are reminiscent of those that are induced by exogenous TGFβ, which yields a more spindle‐shaped, fibroblastic morphology from initially rather epithelial cells (Kern et al., 2015). Since FAPα depletion leads to reduced TGFβ levels (Table 1 and Figure 4), we hypothesized that the partial loss of fibroblastic morphology can be rescued by exogenous application of TGFβ. Indeed, addition of 10 nM TGFβ for 96 h restored the original spindle‐shaped morphology (Figure 6b). To further corroborate this phenomenon, we investigated αSMA levels upon TGFβ treatment. We observed an increase of αSMA in FAPα depleted CAFs treated with recombinant TGFβ, which was not cell line specific (Figure 6c and Suppl. Figure 4). These findings further underline that FAPα affects CAF biology in part through TGFβ signaling.
(a) Light microscopic image of low density CT5.3 cells expressing differential FAPα levels. FAPα knock‐down cells (CT5.3shFAP) and the control cells (CT5.3shctr) do not display any morphological differences. (b) Light microscopic ...
Collagen contraction, based on fibroblasts embedded in collagen type I gels is considered an in vitro model for fibrotic processes (Fang et al., 2004). Collagen contraction has been originally investigated in the context of wound healing (Bell et al., 1979). Due to numerous highly similar biological processes, tumors are sometime referred to as “wounds that never heal” (Kalluri and Zeisberg, 2006); a notion that emphasizes the importance of the stroma and the microenvironment in tumor biology (Kalluri and Zeisberg, 2006). Collagen is the predominant component of the extracellular matrix (ECM), determining the physico‐mechanical nature of the ECM, e.g. its stiffness and rigidity, which further affects tumor malignancy (Kumar and Weaver, 2009). In murine breast cancer models, increased collagen density leads to more invasive tumor phenotypes (Provenzano et al., 2008).
FAPα has been described to process collagens and gelatins (Christiansen et al., 2007; Brokopp et al., 2011). Due to its proteolytic activity, FAPα is believed to affect structure and composition of deposited ECM (Lee et al., 2011). Our proteomic data show that FAPα increases the abundance of several proteins that are known to promote collagen contraction. These include TGFβ (Montesano and Orci, 1988), lumican (Montesano and Orci, 1988), and matrix metalloproteases (Daniels et al., 2003; Bildt et al., 2009).
Based on these findings we aimed to elucidate whether FAPα affects the ability of CAFs for collagen contraction. CT5.3 cells with gain‐ or loss of FAPα function, as well as the appropriate control cells, were embedded in type I rat tail collagen. After 48 h of culturing, contraction of the collagen patches was observed and their area was determined to quantitatively measure this effect. All CT5.3 cell variants (CT5.3shctr, CT5.3shFAP, CT5.3FAPact, and CT5.3FAPasm) induced noticeable collagen contraction. In both gain‐ and loss of FAPα settings, FAPα activity positively correlated with increased collagen contraction (Figure 7), thereby functionally corroborating the secretome data and linking FAPα to a key fibroblast functionality.
Collagen contraction assay shows differential ability of CT5.3 cells s to process collagen depending on FAPα activity. After 48 h, CAFs embedded in collagen matrix processed collagen which was observed as shrinkage. The difference in ...
Vascularization is a key step in the formation of highly malignant tumors and is a prominent therapeutic target (Cao, 2009). CAFs are a major source of secreted pro‐angiogenic proteins, which stimulate the emergence of novel blood vessels (Kalluri and Zeisberg, 2006). Our secretome profiling revealed a strong FAPα “footprint” concerning the abundance of proteins with roles in the regulation of angiogenesis. Increased FAPα activity resulted in decreased levels of the anti‐angiogenic protein PEDF as well as elevated levels of the pro‐angiogenic proteins angiopoietin‐1 and VEGFC in the secretome of CAFs (Table 1, Figure 4). These findings suggest a rather pro‐angiogenic effect of FAPα activity.
To test this hypothesis, sprout formation of HUVEC cells in co‐culture with CT5.3 cells was investigated. As outlined above, we used both the FAPα gain‐ and loss‐of‐function systems. HUVEC spheroids were seeded in a collagen matrix on top of a confluent monolayer of CT5.3 cells with differential FAPα expression profiles. In this setting sprout formation was influenced by the CAF secretome. The co‐culture was incubated for 24 h and afterwards the sprouting rate was calculated as an total number of sprouts and sprout length. All CT5.3 cell variants (CT5.3shctr, CT5.3shFAP, CT5.3FAPact, and CT5.3FAPasm) induced noticeable HUVEC sprout formation. In both gain‐ and loss‐of‐FAPα settings, increased FAPα activity led to mildly elevated levels of sprout formation. These results are in line with the proteomic data, which originally suggested a pro‐angiogenic function of FAPα by affecting the balance of pro‐ and anti‐angiogenic mediators (Figure 8).
The effect of CAFs co‐culture on sprouting from endothelial cell spheroids (HUVEC). The pro‐angiogenic effect was determined by number and length of sprouts. The graphs represent the relative sprouting rate of each condition normalized ...
A pro‐angiogenic effect of FAPα is further substantiated by in vivo studies of mouse models deficient for FAPα (Santos et al., 2009). Furthermore, in human pancreatic adenocarcinoma, high levels of VEGF correlate with elevated FAPα levels (Patsouras et al., 2015). Collectively, these findings suggest a decisive role of FAPα in regulating CAF‐driven angiogenesis.
FAPα is an amino‐dipeptidase preferentially cleaving after proline in the P1 position, additionally favoring glycine in the P2 position. In some cases, e.g. gelatins and collagens, FAPα also exerts endopeptidase activity with increased selectivity for Gly‐Pro sequence motifs (P2GlyP1Pro). FAPα specificity has been investigated by multiple methods (Park et al., 2005, 2008, 2006, 2006, 2011, 2012, 2011, 2009, 1999).
Protease specificity profiling with proteome‐derived peptide libraries has emerged as a powerful technique that harnesses natural sequence diversity for protease characterization and simultaneously investigates prime‐ and non‐prime specificity (Schilling and Overall, 2008; Schilling et al., 2011b). We investigated FAPα specificity using tryptic peptide library (Figure 9a) and corroborated the prototypical specificity for Gly‐Pro sequence motifs, with P1Pro being the major selectivity determinant, followed by a less strict preference for P2 glycine.
Identification of FAPα substrates. (a) FAPα specificity PICS profile performed using recombinant FAPα and E. coli peptide library. (b) Amino acid enrichment in nonprime peptides with proline in P1 position, corresponding to the ...
In order to investigate cell‐contextual proteolysis that is mediated by FAPα, we used Terminal Amine Isotopic Labeling of Substrates (TAILS) (Kleifeld et al., 2010), to identify and quantify proteolytically generated protein N‐termini in cell conditioned medium of CT5.3 cells. Through stable isotopic labeling, we compared N‐termini in either the gain‐ or the loss‐of‐function setting (CT5.3shctr vs CT5.3shFAP and CT5.3FAPact vs CT5.3FAPasm, respectively, Suppl. Tab. S3a and b).
We identified 2987 protein N‐termini for the loss‐of‐function setting and 3807 N‐termini for the gain‐of‐function setting. In both cases, quantitation followed near normal distribution (Figure 10a). Since FAPα specifically cleaves at P1Pro, we focused on cleavage sites with a P1 proline residue. In both gain‐ and loss‐of‐function settings, the quantitative distribution of these sites indicates a mild enrichment in correlation to elevated FAPα activity (Figure 10b).
(a) Distribution of fold‐change values (log2 of light/heavy ratios) of CT5.3shFAP/CT5.3shctr, and CT5.3FAPact/CT5.3FAPasm in N‐terminome profiling of secretome using TAILS. In total, respectively, 2982 and 3968 free N‐termini ...
We chose the following criteria to determine putative FAPα cleavage events: (a) proteolytically generated N‐terminus with P1Pro, (b) the corresponding protein is annotated as secreted or cell surface localized, (c) the N‐terminus was enriched >50% in CT5.3shctr vs CT5.3shFAP or CT5.3FAPact vs CT5.3FAPasm, without contradictory findings in other setup, (d) favorable manual inspection of the underlying extracted ion chromatograms. These criteria yielded 76 N‐termini, representing 42 cleavage sites in collagen proteins and 34 cleavage sites in “other”, non‐collagenous proteins. The distinction between collagen and “other” proteins revealed further differences: positional profiling of the FAPα‐dependent cleavage events for collagen proteins showed a rather random positional profile (Figure 11a and c). It cannot be, however, determined, whether these sites represent true endopeptidase activity of FAPα or whether they represent amino‐dipeptidase activity that follows initial endoproteolysis by a different collagenase. Nevertheless, FAPα has been reported to show endoproteolytic activity for denatured collagens (Aggarwal et al., 2008; Keane et al., 2011). Moreover, previous reports show FAPα to cleave antiplasmin (Lee et al., 2006; Keane et al., 2011). In our studies we observed that FAPα‐dependent cleavage events for non‐collagenous proteins showed strong N‐terminal enrichment, corresponding to the amino‐dipeptidase activity of FAPα (Figure 11b). N‐terminal clustering of FAPα‐dependent, P1Pro cleavage sites is highlighted for number of proteins, including ADAM15, interleukin 6, fibrillin‐2, matrillin‐3, serine protease 23, testican‐1, and TGFβ‐induced protein, as depicted in Figure 11d. These cases represent non‐collagenous FAPα substrate candidates.
(a) Positional clustering of all P1Pro N termini with Fc ≥ 0.58 found in collagens. The random distribution of identified peptides in the collagen sequences showed endopeptidase activity of FAPα. (b) Positional clustering ...
As an example, two cleavage events were identified in IL‐6 in the vicinity of the signal peptide removal site. IL‐6 is a pleiotropic cytokine, expressed by several cell types, such us immune cells and fibroblasts. The two processed N‐termini found in our analysis have been previously identified by Edman sequencing, initially distinguishing different IL‐6 proteins (Van Damme et al., 1988; Ming et al., 1989). The role of IL‐6 role in cancer is widely investigated, mainly towards its pro‐inflammatory properties (Mauer et al., 2015). Generally, the enrichment of post P1Pro N‐termini in the secretome of CAFs with higher FAPα activity indicates its proteolytic activity exerted over proteins secreted by CAFs. Thus, our results underline cell‐contextual FAPα proteolysis with substrate candidates that play important roles in tumor biology and cell–cell communication.
Due to the complex biological context and interconnectedness of proteolytic activities (Fortelny et al., 2014), the N‐TAILS method, as any other N‐terminomic strategy, does not fully allow to distinguish between direct‐ and indirect‐FAPα cleavage events. The observed both collagenous and non‐collagenous cleavage events display the typical FAPα cleavage specificity and extent of cleavage quantitatively depends on FAPα activity. Therefore, these proteins can be postulated as putative FAPα substrates.
FAPα is a proteolytic enzyme that is predominantly expressed by CAFs in vicinity of the tumor–stroma interface. We successfully generated FAPα loss‐ and gain‐of‐function systems in a CAF cell line. Secretome profiling showed that FAPα regulates the levels of CAF‐secreted proteins that play important roles in tumor–stroma interaction and in tumor‐associated processes such as neo‐angiogenesis. Functional assays strengthened a pivotal role of FAPα in CAF biology; highlighting an impact on CAF functionality in collagen contraction as well as on sprout formation by endothelial cells. A strong link to TGFβ signaling was also evident. Our degradomic analysis corroborated the preference of FAPα for P1 proline. Substrate profiling identified cell contextual FAPα proteolysis in collagens and non‐collagenous proteins, thereby highlighting multiple FAPα substrate candidates. Generally, the present study emphasizes a central role of FAPα in shaping hallmark features of CAF biology through proteome and degradome alterations.
The following are the supplementary data related to this article:
Supplementary Figure S1. Peptide mapping for TGFβ. Tryptic TGFβ peptides that were identified in the global secretome comparison are mapped on both LAP and TGFβ region, both in FAPα loss‐of‐function and gain‐of‐function experiments (green bands), the corresponding numbers above indicate the Fc‐values of the individual peptide, where identified peptides were enriched at least 50% in samples with higher FAPα activity.
Supplementary Figure S2. FAPα expression in P‐48GBshctr and P‐48GBshFAP cells. The P‐48GBshFAP cells show massive reduction of FAPα level.
Supplementary Figure S3. Western blot detection of several proteins identified as altered in the LC‐MS/MS quantitative secretome profiling in P‐48GBshctr and P‐48GBshFAP cell conditioned medium: tPA, PEDF, VEGFC, LAP‐TGFβ, TGFβ, CCL2, MMP1, and OPN. The electrophoresis was run on 12.5% PAA gel and 15–25 μg of CCM was loaded.
Supplementary Figure S4. In TGFβ treated P‐48GBshFAP cells the αSMA level is elevated, comparing to the untreated control.
O.S. is supported by grants of the Deutsche Forschungsgemeinschaft (DFG, SCHI 871/2 and SCHI 871/5, SCHI 871/6, GR 1748/6, and INST 39/900‐1), the SFB850 (Project B8), a starting grant of the European Research Council (Programme “Ideas” – Call identifier: ERC‐2011‐StG 282111‐ProteaSys), the Excellence Initiative of the German Federal and State Governments (EXC 294, BIOSS), and a DKTK project on breast cancer. S.L. is supported by the DFG with SFB850 (projects C5 and Z1), SFB993 project C3 and a DKTK project on colorectal cancer. The authors thank Franz Jehle for excellent technical assistance with mass spectrometry analysis, Asli Aras Taskin for the FAPα specificity analysis, Julia Knopf for cell growth assays, Nicola Bittermann for excellent immunohistochemistry, and Manuel Schlimpert, Charlotte Friedemann, and Lennart Enders for general assistance.
Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.molonc.2015.08.001.
Koczorowska M.M., Tholen S., Bucher F., Lutz L., Kizhakkedathu J.N., De Wever O., Wellner U.F., Biniossek M.L., Stahl A., Lassmann S., Schilling O., (2016), Fibroblast activation protein‐α, a stromal cell surface protease, shapes key features of cancer associated fibroblasts through proteome and degradome alterations, Molecular Oncology, 10, doi: 10.1016/j.molonc.2015.08.001.