1 Introduction

Solid tumors usually contain multiple cellular components, collectively forming the tumor microenvironment [1]. Different cell subtypes in the tumor microenvironment exhibit distinct characteristics, which is illustrated by their diverse gene expression patterns. The latter include genes encoding cell surface receptors, which define differential sensitivities of the different cell subtypes to external stimuli [2]. Cancer associated fibroblasts (CAFs) display a spindle-shaped morphology and express specific markers such as alpha smooth muscle actin (αSMA) and fibroblast activation protein alpha (FAPα). CAFs are highly capable of secreting signaling molecules and, by doing so, contribute to key features of tumor biology such as growth, invasion and metastasis [3, 4]. Unlike epithelial tumor cells (ETCs), CAFs are considered to be highly responsive to growth factor stimuli [2]. Given their distinctive features, together with their high abundance within tumor masses, CAFs have emerged as prominent targets in the design of novel therapeutic strategies [5, 6].

Transforming growth factor β (TGFβ) is a well-established inducer of epithelial-to-mesenchymal transition (EMT) [7]. Its action is exerted through binding to one of its receptors: TGFBR1, TGFBR2 or TGFBR3 [8], the presence of which determines the cellular responsiveness to TGFβ-induced signaling. Interactions between stromal and tumor cells may also yield indirect, downstream effects on other cells within the tumor microenvironment.

Inspired by recent interest in the differential action of TGFβ on CAFs and ETCs [2], we set out to investigate the effects of TGFβ on CAFs and ETCs at the proteomic level. To this end, we established a co-culture system combined with TGFβ stimulation, as well as FACS separation of the respective cell types (Fig. 1). Through independent analysis of the co-cultured cells we show that TGFβ treatment has a strong impact on the protein composition of the CAFs, but not the ETCs.

Fig. 1
figure 1

Proteome profiling workflow of CAFs and CECs grown in co-culture. CAFs and ETCs were stably transfected with GFP and mCherry, respectively. Labeled cells were grown in co-culture and treated for 48 h with TGFβ. Untreated cells were used as controls. Cells were FACS-sorted according to expression of the respective fluorescent labels. After lysis and tryptic digestion the peptides were labeled with light or heavy FA. Samples of treated and untreated cells of the same cell type were mixed at a 1:1 ratio. Peptides were purified, pre-fractionated on HPLC and analyzed by LC-MS/MS.

Full size image

2 Materials and methods

2.1 Cell lines and culture

The generation of CT5.3 cancer associated fibroblasts (CAFs) with shRNA-silenced FAPα expression (CT5.3shFAP) or scrambled shRNA expression (CT5.3shctr) has been reported before [9,10,11]. P-56-HM epithelial cancer cells were derived from a distal bile duct adenocarcinoma and isolated by Bachem’s outgrowth method [12] into Quantum 333 medium from a human ampullary adenocarcinoma in 2012, expanded and immortalized through SV40 transduction. HCT116 epithelial colon tumor-derived cells were purchased from the ATCC. The 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 (Gibco/Invitrogen, Paisley, UK) at 37 °C in a humidified atmosphere containing 5% CO2.

2.2 Cell line-specific fluorescent protein expression

CT5.3 cells were stably transfected with a pAcGFP1-Hyg-N plasmid (Genecopoeia) using SuperFect Transfection Reagent (Qiagen). After transfection cells were selected for 2 weeks using 300 μg/ml hygromycin B (EMD Millipore Corp., USA.). HCT116 cells were transfected with pmCherry-N1 as described above and selected for 2 weeks using 750 μg/ml Geniticin (Invitrogen). Both cell lines expressing fluorescent proteins (FPs) were subjected to FACS sorting to enrich for GFP or mCherry Red expressing cells at equal levels.

2.3 Direct co-culture and mono-culture approaches to TGFβ treatment

For the co-culture approach, CT5.3shFAPGFP cells (CAFs) and HCT116mCherry cells (ETCs) were mixed in a 1:1.2 ratio. At 70% confluence the cells were treated for 48 h with10 ng/ml recombinant human TGFβ1 (R&D Systems), whereas control cells remained untreated. Next, the cells were trypsinized and sorted on a FACS Aria III machine (BD Biosciences). The wavelengths used for GFP were 488 nm for excitation and 515–545 nm for emission, while those for mCherry were 561 nm for excitation and 600–620 nm for emission. In mono-culture approaches CAFs and ETCs (HCT116, SW480 or SW620) were grown until 70% confluency and treated with 10 ng/ml recombinant human TGFβ1 for 48 h. The cells were lysed with RIPA buffer (150 mM NaCl, 50 mM Tris, pH 7.5, 1% NP-40) supplemented with protease inhibitors (5 mM EDTA, 0.01 mM trans-epoxysuccinyl-L-leucylamido(4-guanidino)butane (E64) and 1 mM phenylmethanesulfonyl fluoride (PMSF)).

2.4 Quantitative proteome profiling

For proteomic analyses, samples were prepared as reported previously, using differential dimethylation [9]. For mass spectrometry analyses, a Q-Exactive plus system (Thermo Scientific, Bremen, Germany) was used again as reported before [9]. Likewise, LC-MS/MS data analysis was performed as reported before [9]. STRING [13] was used to annotate functional interactions of proteins and non-redundant Gene Ontology (GO) term enrichment analysis was performed using the TopGO R package.

2.5 Western blot analysis

Western blot analysis was performed as reported before [9] using the following primary antibodies: ant-FAPα (R&D, AF3715), anti-N-Cadherin (Cell Signaling 4061), anti-Phospho-Smad3 (S423/S425; R&D, AB3226) and anti-α-Tubulin (Sigma-Aldrich, T 6199), and the following secondary antibodies: anti-mouse (Dianova, 115-035-003), anti-rabbit (BioRad, 172-1019) and anti-sheep (Dianova, 713-035-147). Tubulin or GAPDH served as internal loading control standards for the Western blots. In case of CCM samples, Coomassie staining of the gels was used as an equal loading control.

3 Results and discussion

TGFβ is a powerful signal transducer, involved in the regulation of various hallmark cancer processes. In addition, TGFβ signaling has been associated with a poor prognosis in cancer patients [14, 15]. Previously, we have shown that TGFβ treatment of FAPα-deprived CAFs can rescue their fibroblastic phenotype [9]. A strong responsiveness of CAFs to TGFβ is in line with previous reports on differential cellular responses to external signals. Calon et al. [2] showed that TGFβ-stimulated ETCs were not tumorigenic upon subcutaneous injection into mice, whereas co-injection of fibroblasts did lead to tumor formation, which highlights a unique role of CAFs in tumorigenesis. To investigate the effect of TGFβ on individual tumor cellular components, we performed a proteomic analysis of co-cultured CAF (CT5.3shFAPGFP) and ETC (HCT116mCherry) cells treated with TGFβ. The applied workflow is schematically depicted in Fig. 1. FACS separation of cells expressing different fluorescent proteins (FPs) allowed for a quantitative proteome profiling of individual cell types.

Single-cell suspensions of CAFs and ETCs expressing different FPs were seeded in dishes to form 2D co-cultures. By doing so, we found that TGFβ treated co-cultures exhibited differential morphologies compared to control cells (Fig. 2a) and a shift in CAF/ETC ratio towards a higher CAF abundance. The decreased ETC abundance in TGFβ-treated co-cultures might suggest a TGFβ-driven growth suppression of epithelial cells [16]. We found, however, that TGFβ stimulation had no detectable effect on HCT116 cell proliferation in mono-culture (data not shown).

Fig. 2
figure 2

a Co-culture of CAFs and ETCs after treatment with TGFβ compared to the untreated control. The picture was created by overlaying the green and red channels; red and green colors were added manually. b FACS separation of cells expressing GFP and mCherry (representative bivariate plots are shown). SSC – side scatter, FSC – forward scatter, pos – positive cells. (i) Gating on forward and side scatter to remove cell debris, (ii) DAPI staining for dead cell exclusion, (iii) and (iv) gating for single cells using FSC and SSC, respectively, in order to eliminate possible doublets and cell clusters, (v) Separation of two distinct cell populations based on the expression of either GFP or mCherry

Full size image

To further elucidate the differential effect of TGFβ on CAFs and ETCs in a direct co-culture system, we used FPs as unique CAF (GFP) and ETC (mCherry) markers for FACS sorting. In doing so, TGFβ treated and control cultures were harvested and subjected to FACS sorting, after which 5 × 104 to 2 × 106 separated cells (both CAFs and ETCs) were isolated (Fig. 2b). These cells were used for quantitative global proteomic profiling, in which TGFβ treated CAFs and ETCs were compared to their untreated counterparts, respectively (Fig. 1), in two biological replicates. Peptides were isotopically labeled using either light or heavy formaldehyde. To exclude label bias, we applied a label swap for each replicate. Through LC-MS/MS analysis 2909 and 3366 proteins (false discovery rate (FDR) < 1%) were identified and quantified for the CAFs in replicates 1 and 2, respectively, with 2349 proteins consistently identified and quantified in both replicates. For the ETCs, 3253 proteins were identified in replicate 1 and 4018 proteins in replicate 2 (FDR < 1%), with 2748 proteins consistently identified and quantified in both replicates. In all experiments we found an overlap of 1881 identified and quantified proteins (Fig. 3a). Relative protein abundances were measured for TGFβ treated and untreated samples and protein ratios were expressed as fold change (Fc) values (log2 of TGFβ treated/untreated). The fold change distribution in all experiments was close to normal (Fig. 3b), with a narrower range in ETCs than in CAFs, indicating less alterations in protein abundances in the ETCs.

Fig. 3
figure 3

a Overlap of proteins consistently identified and quantified. 1881 proteins were found in all four experiments. The Venn Diagam (created by software Venny 2.1 http://bioinfogp.cnb.csic.es/tools/venny/) shows higher numbers of shared proteins between replicates of ETCs or CAFs compared to the overlap between different cell lines. b Distribution of fold-change values (log2 of light/heavy or heavy/light ratios) in four independent experiments. Histograms show a broader distribution and, thus, a higher number of altered proteins (Fc > 0.58 or Fc < −0.58) in CAFs than in ETCs. c Biological themes emerging from proteins up-regulated in CAFs co-cultured with ETCs upon TGFβ treatment, according to GO annotations performed using a R studio-based script. GO-Biological Processes annotations indicate enrichment in proteins clustering around extracellular processes. The full output data are listed in Supplementary Table 2. d STRING v10 (Search Tool for the Retrieval of Interacting Genes/Proteins) analysis of proteins altered in ETCs upon TGFβ treatment. Connections show confidence (line thickness indicates the summed-up strength of data support). Connections are based on the categories “Neighborhood”, “Gene Fusion”, “Co-occurrence”, “Co-expression Experiments”, “Databases”, and “Textmining” at a medium confidence level (0.4)

Full size image

To determine the impact of TGFβ on the proteome composition of CAFs and ETCs, we defined the following criteria: (a) the protein was identified and quantified in both experiments, (b) protein abundance was consistently increased or decreased > 50% in both experiments, (c) a favorable manual inspection of the extracted ion chromatograms. Similar criteria have successfully been used in previous studies to quantitatively delineate proteins affected by treatment [17]. By using these criteria, 118 up-regulated and 71 down-regulated proteins were highlighted in CAFs upon TGFβ treatment, whereas only 27 proteins showed increased and 25 proteins decreased abundances in TGFβ treated ETCs (Tables 1 and 2, Supplementary Table 1 a-d). This finding suggests that exposure to TGFβ causes a substantial perturbation in proteome composition in CAFs but not in ETCs, underscoring a higher responsiveness of CAFs to TGFβ compared to ETCs. Lack of overlap between proteins affected in ETCs and CAFs indicates a differential effect of TGFβ on cells of epithelial versus mesenchymal origin.

Table 1 List of proteins significantly and consistently affected in both replicates 1 and 2 (r1 and r2) of CAFs by TGFβ treatment - Uniprot ID and recommended name according to Uniprot database (Uniprot Consortium 2013). Fc- values (log2 of TGFβ treated / control cells ratios for replicates 1 and 2) indicate the fold change of protein in each experiment as well as average Fc
Full size table
Table 2 List of proteins significantly and consistently affected in both replicates 1 and 2 (r1 and r2) of CECs by TGFβ treatment - Uniprot ID and recommended name according to Uniprot database (Uniprot Consortium 2013). Fc- values (log2 of TGFβ treated / control cells ratios for replicates 1 and 2) indicate the fold change of protein in each experiment as well as average Fc
Full size table

To elucidate biological “themes” within the proteins affected by TGFβ treatment in CAFs, we performed a non-redundant GO-enrichment analysis using TopGO. By doing so, a notable enrichment was noted for proteins involved in extracellular matrix (ECM) and cell-matrix adhesion, such as fibronectin and different collagens (Fig. 3c and Supplementary Tables 2 a and b). The impact of TGFβ on mechanical properties of the ECM and on cellular interactions with extracellular proteins is well-documented [18] and several proteins that we found to be up-regulated in CAFs upon TGFβ stimulation in co-culture have been previously reported to be TGFβ targets [19]. TGFβ exhibits a dual role in cancer, i.e., as an anti-proliferative factor and as an inducer of apoptosis [20, 21]. In the proteome analysis of co-cultured ETCs upon TGFβ treatment, we observed a down-regulation of proteins involved in nucleic acid synthesis and in ribosome biosynthesis (Fig. 3d). This finding is indicative of an impaired cellular growth and proliferation and is in agreement with the decreased number of ETCs observed in the TGFβ treated co-culture compared to the corresponding untreated control (Fig. 2a).

One of the most prominent modes of action of TGFβ is the induction of an EMT-like cellular behavior [7, 21]. We found that the EMT marker N-cadherin [22] was consistently up-regulated in CAFs (Fc values in replicates 1 and 2: 1.32 and 2.81, respectively), whereas in ETCs its expression was not detectable (Table 2, Supplementary Table 1 c and d). Next, we independently validated the impact of TGFβ on N-Cadherin expression in mono-cultured CAFs. In both CT5.3 CAFs (derived from colorectal cancer) and P-56-HM CAFs (derived from distal bile duct adenocarcinoma) we found that TGFβ treatment resulted in increased N-cadherin levels (Fig. 4a). As expected, N-cadherin expression was not detected in ETCs, nor in any of the three colon cancer-derived cell lines tested: HCT116, SW480 and SW620. We also validated the impact of TGFβ on the CAF marker protein FAPα in both CAF lines (CT5.3 and P-56-HM) and found increased levels upon TGFβ treatment, which is of interest since earlier studies have indicated that FAPα depletion results in reduced TGFβ levels, thus pointing at a mutual feedback system. Lastly, we assessed the phosphorylated Smad3 (pSmad3) levels upon TGFβ treatment in monocultured CAFs and ETCs. In agreement with the strong effect of TGFβ on co-cultured CAFs, and a limited effect on co-cultured ETCs, we found that TGFβ increased the phosphorylation of Smad3 in monocultured CAFs, but not in monocultured ETCs (Fig. 4b). This finding further strengthens a differential responsiveness of CAFs and ETCs to TGFβ.

Fig. 4
figure 4

a Expression of EMT markers in TGFβ treated CAFs. CT5.3 and P-56-HM cells were treated with TGFβ after which cell lysates were analyzed by Western blotting for the expression of N-cadherin and FAPα. None of these proteins was detected in ETCs, either in the presence or absence of TGFβ (data not shown). b Smad3 phosphorylation analysis in TGFβ treated CAFs and ETCs. CT5.3 and HCT116 cells were treated with TGFβ for 48 h in monocultures. Smad3 phosphorylation was detected by Western blotting. Smad3 phosphorylation was stimulated in CAFs, but not in ETCs

Full size image

Previously, it has been shown that different cellular components within tumors may exhibit differential responses to external stimuli [2]. Our current study nicely illustrates how cell type-specific effects of growth factors can be assessed. We performed proteome profiling of two cell types, CAFs and ETCs, grown in a 2D co-culture and stimulated with TGFβ. We observed different architectures in TGFβ-treated co-cultures compared to untreated control cultures. Quantitative proteomic profiling of the two different cell types indicated a strong response to TGFβ by CAFs but not by ETCs. Our findings are in agreement with previous results indicating that TGFβ may lead to up-regulation of structural ECM proteins in CAFs and may stimulate N-cadherin expression. The differential responsiveness of stromal and tumor cells may be exploited for novel therapeutic avenues targeting stromal cells within the tumor microenvironment.