Transforming growth factor β (TGFβ) is a pleiotropic cytokine whose effects are both cell type and context dependent. The complexity of TGFβ signaling begins with its synthesis as a pro-peptide requiring cleavage by endogenous enzymes such as furins before it can be secreted as a latent complex.1–3 Active TGFβ dimers must then be released from the latency-associated peptide (LAP) before acquiring signaling competency. Mature TGFβ signals via an oligomeric complex consisting of the type I (TGFβRI) and type II (TGFβRII) TGFβ receptors.4–7 Binding of cytokine to the TGFβ receptor complex leads to phosphorylation of Smad2 and/or Smad3 which then bind Smad4 and translocate to the nucleus to modulate gene transcription. While Smad signaling is controlled exclusively by members of the TGFβ superfamily, several other pathways, including MAPK and PI3K/Akt, are susceptible to modulation by TGFβ. Furthermore, crosstalk exists between intermediates from each pathway and different thresholds of activation have been reported for the activation of different signaling cascades.4, 8
In normal epithelia, including that of the gut, the main effects of TGFβ are the induction of differentiation and inhibition of proliferation.9, 10 However, the growth and migration of many late stage neoplasms are enhanced by the direct actions of TGFβ and increased expression of this cytokine is a feature common to many advanced tumors.10, 11 TGFβ is also a potent effector within the tumor microenvironment. It exerts a predominantly immunosuppressive effect on lymphocytes and has been shown to be an active player in tumor immune evasion.11–13 Previous work has documented the presence of a cell surface-associated form of TGFβ on a subset of CD4+CD25+ regulatory T cells known to have potent immunosuppressive effects.14–16 Importantly, cell–cell contact between the regulatory T cells and CD8+ cytotoxic T cells is required to mediate this immunosuppression. Experiments using blocking antibodies and inhibition of TGFβ processing demonstrated that surface-bound TGFβ was the critical molecule in this process.
In light of these findings as well as of the immunosuppressive nature of the tumor microenvironment and the association of advanced tumors with increased TGFβ production, we investigated the possibility of a colorectal cancer (CRC) cell surface-bound form of TGFβ as a potential contributor to cancer mediated immunosuppression.
Material and methods
Human colorectal cancer cell lines RKO, LoVo and SK-CO-1 were purchased from ATCC (Manassas, VA). Moser was a generous gift from Dr. Michael Brattain (Roswell Park Cancer Institute, New York, NY) and Vaco5 was kindly provided by Dr. James K.V. Willson (University of Texas Southwestern Medical Center at Dallas, Dallas, TX). All cells were maintained in DMEM (HyClone, Logan, UT) supplemented with 10% FBS (HyClone), 200 mM L-glutamine (HyClone), 100 μg/ml streptomycin and 100 U/ml penicillin (Gibco, Burlington, ON, Canada).
Following trypsinization, adherent CRC cells were resuspended in starvation medium (DMEM containing 0.2% FBS) and incubated on a rocker for 18 hr at 37°C and 5% CO2. This allowed for recovery of membrane-associated proteins and also minimized cell contact with extracellular matrix components to which TGFβ can be bound. Unpermeabilized cells were then stained with antibodies against TGFβ that detect either both the mature and latent forms (sc-146, Santa Cruz Biotechnology, Santa Cruz, CA) or only the mature form (MAB240, R&D Systems, Minneapolis, MN) of TGFβ or an IgG control (R&D Systems). In some experiments, cells were also stained using an anti-LAP antibody (MAB2463, R&D Systems). Primary antibodies were detected using an Alexa-Fluor 488 conjugated secondary antibody (Invitrogen, Burlington, ON, Canada). Fluorescence was read using a FACSCalibur machine (BD Biosciences, Mississauga, ON, Canada) and data was analyzed using CellQuest software.
In an attempt to dissociate the surface-associated TGFβ, in some experiments cells were treated with a low pH buffer, as described previously, before staining.17 Briefly, cells were pulsed for 60 sec in an acidic buffer consisting of 140 mM NaCl/10 mM Citrate (pH 4.0). The solution was then neutralized with DMEM and cells were washed extensively before staining for membranous TGFβ. The integrity of E-cadherin was also monitored to ensure that this treatment did not destroy membrane receptors (data not shown).
Paracrine cell surface-associated TGFβ assay
The functionality of the cell surface-bound TGFβ was evaluated using two adherent CRC cell lines as indicators and a suspension CRC cell line as a stimulator. The protocol is illustrated in Figure 2a. All cell lines were incubated for 18 hr in starvation medium prior to the assay. To account for the effects of soluble TGFβ secreted by the stimulator cell line (Vaco5), starvation medium was first conditioned by these cells for either 30 or 60 min. Cells were then separated from the conditioned medium via centrifugation and the conditioned medium was added to 1 well each of the indicator cells (Moser and SK-CO-1). The pellet of cells used to condition the medium was then resuspended in an equal volume of starvation medium and the stimulator cell suspension was added to 1 well each of the indicator cells. Responder cells were incubated in the presence of the conditioned medium or of the stimulator cell suspension for a time equivalent to the medium conditioning period. The indicator cells were then rinsed with ice cold PBS to remove any remaining stimulator cells, lysed and analyzed for phospho-Smad2 (Cell Signaling Technology, Danvers, MA) expression by standard Western blotting techniques. To ensure that endogenously produced cell surface-bound TGFβ was responsible for the effects seen, stimulator cells were first incubated for 24 hr in 50 μM Furin Inhibitor I (Calbiochem, San Diego, CA) in some experiments. Densitometry was performed on the blots using GeneTools software (Syngene, Frederick, MD).
A modified version of this coculture assay was used to evaluate the effects of CRC surface-bound TGFβ on a CD8+ intraepithelial lymphocyte (IEL) cell line, 32891, described previously and generously donated by Dr. Asma Nusrat (Emory University, GA).18 Since both populations of cells were suspension cultures, evaluation of phospho-Smad2 had to be carried out in such a way as to exclude the stimulator cell population in the cell–cell contact condition. This was done by examining phospho-Smad2 levels using flow cytometry as opposed to Western blotting. The setup of the coculture was as described earlier, using a 1:1 ratio of V5:IEL in the cell–cell contact condition. Cells from each sample were then stained first with PE conjugated anti-CD3 (BD Pharmingen), permeabilized using the Cytofix/Cytoperm kit (BD Pharmingen) and stained with anti-phospho-Smad2 and an Alexa-Fluor 488 conjugated secondary antibody. During acquisition, cells were gated on CD3-PE positivity and phospho-Smad2 was evaluated on cells only within this gated population to exclude any changes in the stimulator Vaco5 cells. Fluorescence was read using a FACSCalibur machine (BD Biosciences) and data was analyzed using CellQuest software.
A student's t test was used to evaluate the densitometric values from 3 independent blots of phospho-Smad2. The analysis was performed using MiniTab 13.1 (MiniTab, State College, PA).
Colorectal cancer cells express a surface-bound form of TGFβ
As illustrated in Figure 1a, all of the CRC cells tested expressed membranous TGFβ. Since cells had been incubated overnight in serum free medium, it is unlikely that this TGFβ was derived from the extracellular environment, leaving endogenous production as the most probable source. Notably, only TGFβ-specific antibodies capable of recognizing both the active and latent forms of the cytokine were able to stain membrane-bound cytokine. A significant overlap in the staining pattern was also found between membrane-bound TGFβ and LAP (Fig. 1b). A further indication that the detected cell surface TGFβ was in latent form was the significant increase seen in membrane-bound ligand after a brief pulse of the cells in a low pH buffer (Fig. 1c). This treatment had previously been shown to remove exogenously added TGFβ from regulatory lymphocytes.17 However, acid pulsing is a well recognized means of activating latent TGFβ and activation of a latent form of membrane-bound cytokine is the most plausible explanation for the phenomenon we observed.3 Together, these data indicated that the surface-bound TGFβ on our CRC cells was in a latent form and thus not necessarily signaling competent.
Surface-bound TGFβ on CRC cells can initiate paracrine signaling
The signaling capability of the surface-bound TGFβ was verified as outlined in Figure 2a using a coculture system of suspended Vaco5 cells as stimulators and adherent CRC cells as responders. Figure 2b demonstrates that levels of phospho-Smad2 in both responder cell lines are clearly increased when incubated with the cell suspension as opposed to the medium conditioned by the same cells. While the adherent cells do respond to soluble TGFβ, as indicated by the phospho-Smad2 increase in the conditioned versus the control medium, there is an additional increase in phospo-Smad2 above that of the conditioned medium when responder cells are incubated directly with stimulator cells. This is reflected in the desitometric graph where density has been normalized to that of the conditioned medium to clearly indicate the portion of increased phosphor-Smad2 that is attributable to surface-bound TGFβ. These results indicate that the membranous form of TGFβ expressed on Vaco5 is signaling competent or can be made so by cell–cell contact.
No blocking antibodies exist for the latent form of TGFβ. Since the enzyme furin is required for processing of the TGFβ pro-peptide before its release in latent form, Vaco5 stimulator cells were incubated with a furin inhibitor for 24 hr prior to conditioning of medium and exposure to responder cells.3, 16 Flow cytometry confirmed that this treatment decreased by 36% the mean fluorescent intensity of surface-bound TGFβ on Vaco5 cells (Supplementary Fig. 1). The previously seen increase in phospho-Smad2 in responder cells incubated with stimulator cells was abrogated following this Vaco5 treatment, implicating cell surface TGFβ expressed on Vaco5 in the activation of Smad signaling by responder cells (Fig. 2c). It is worth noting that the variation in induction of phospho-Smad2 in the cell–cell culture condition between Figures 2b and 2c represent the lower and upper limits of the signaling activation we observed in multiple trials. While the magnitude of activation of Smad2 varied in different trials, phosphorylation of Smad2 was consistently observed in all repeats. Since the values were determined from densitometric analysis of a blot, it is likely that background variables may account for some of the observed discrepancy between trials.
Since membranous-TGFβ on regulatory T cells has been reported to mediate immunosuppression, we investigated whether an intraepithelial derived lymphocyte cell line (IELs) could respond to the membrane-bound TGFβ on our CRC cells. A modified version of our coculture assay using flow cytometry and CD3 gating to separate out the cell populations in the cell–cell contact condition (Fig. 3a) revealed increased phospho-Smad2 in IELs incubated in direct contact with Vaco5 stimulator cells over and above that seen in IELs incubated in medium conditioned by these same stimulator cells (Fig. 3b). Preincubation of the Vaco5 stimulator cells with a furin inhibitor for 24 hr decreased the activation of phospho-Smad2 in the cell contact condition, again implicating membrane-bound TGFβ in Smad activation (Fig. 3c). Membrane-associated TGFβ also activated Smad2 in IELs when they were cocultured with two other CRC cell lines (Fig. 3d). The paracrine effects of cell surface TGFβ on our CRC cells are, therefore, not limited to adjacent cancer cells or to a single membrane-associated TGFβ expressing CRC cell line.
Signaling by TGFβ is characterized by many layers of complexity and its effects vary not only with different cell types but also across different contexts within the same cell type. Within advanced adenocarcinomas, TGFβ works to promote tumor progression by encouraging tumor cell growth and migration while simultaneously inhibiting anti-tumor immune responses. A complete understanding of the nature of TGFβ signaling within the tumor microenvironment is thus an essential component of the arsenal in the fight against cancer. The present study reports for the first time the existence of a cell-surface bound, endogenously derived form of TGFβ expressed by CRC cells which is capable of signaling both to adjacent tumor cells and to IELs.
All of the five CRC cell lines tested were found to express detectable cell surface-bound TGFβ when analyzed in a nonpermeabilized state via flow cytometry. Since cells had been incubated overnight in serum free medium, the observed TGFβ is unlikely to have been derived from the cell culture medium. The extended incubation of the CRC cells on a rocker kept the cells in suspension, thereby minimizing contact with extracellular matrix molecules to which TGFβ can be bound. Thus, endogenous production is the most probable source for the cell surface associated TGFβ we observed. Notably, only TGFβ-specific antibodies capable of recognizing both the active and latent forms of the cytokine were able to stain surface-bound TGFβ. This finding is consistent with those of previous studies on regulatory T cells in which all antibodies used detect both latent and active cytokine.15, 19, 20 Probing the cells with an anti-LAP antibody demonstrated a significant overlap in the staining patterns of cell-bound TGFβ and LAP in most of the CRC cell lines. These initial findings suggested that the cell surface-bound TGFβ detected on the CRC cells was in latent form.
A commonly used method for activating soluble latent TGFβ in vitro is treatment with a low pH buffer.3 Broderick et al. previously demonstrated that pulsing lymphocytes in an acidic NaCl/citrate buffer (pH 4.0) was capable of removing exogenously added TGFβ from the cell surface.17 This same method was used in order to investigate whether acid treatment could alter the pattern of surface-associated TGFβ on the CRC cells. In contrast to the results obtained by Broderick et al. using lymphocytes, briefly pulsing our CRC cells in a low pH solution resulted in an increase in the amount of detectable surface-bound TGFβ. Since this acid stripping method had previously been shown capable of removing exogenously added TGFβ from cell membranes, these results support the notion that the surface-bound TGFβ on our CRC cells is endogenously derived. Furthermore, the rise in observable surface-bound TGFβ following treatment with a well-known activating method supports the notion that the detectable cytokine is present in a latent form on the CRC cell membranes.
Latent TGFβ is not known to be signaling competent and so it was important to determine if the surface TGFβ detected on our CRC cell lines could initiate signaling. A coculture assay capable of differentiating the effects of both soluble and surface-associated TGFβ was thus derived utilizing both suspended and adherent CRCs. Phosphorylation of Smad2 was chosen as the endpoint of this assay since it is specifically activated by members of the TGFβ superfamily and is thus a reliable indicator of activation of a predominant TGFβ signaling cascade. The assay demonstrated that levels of phospho-Smad2 in both responder CRC cell lines are clearly increased when incubated with the stimulator cell suspension as opposed to the medium conditioned by the same cells. Soluble TGFβ is present in both the conditioned medium and cell–cell contact conditions and so the greater magnitude of the increase in Smad2 phosphorylation seen in the latter case can be attributed to the surface-associated TGFβ on the stimulator cells. While densitometric analysis fell short of statistical significance, it is reasonable to assume that changes of 20–50% in the phosophorylation of a signalling mediator are of biological significance. These findings thus indicate that the membranous form of TGFβ is either signaling competent or can be made so by direct cell–cell contact.
Since no blocking antibodies exist for the latent form of TGFβ, we used a furin inhibitor to decrease production of surface-bound TGFβ by the stimulator cells, as has been successfully done before.16 The enzyme furin is required for processing of the TGFβ pro-peptide before its release in latent form and inhibition of this enzyme decreases production of TGFβ.3, 16 Preincubation of the stimulator cells for 24 hr decreased, although did not eliminate, surface-bound TGFβ (Supplementary Fig. 1). Nevertheless, the previous increase in phospho-Smad2 seen in responder cells incubated with stimulator cells was abrogated following treatment of the latter with furin inhibitor. While furin is involved in the processing of many pro-peptides, amongst them TGFβ is the primary activator of the Smad signaling cascade even though some involvement of activin cannot entirely be ruled out.21 These results thus strongly implicate surface-bound TGFβ expressed on the stimulator CRC cells in activation of Smad signaling in the responder cells and supports the notion of paracrine cell type specific signaling. Indeed, the present findings may help to explain the noted discrepancy between sensitivity of cells to autocrine versus paracrine TGFβ.22, 23 The presence of cell surface associated TGFβ offers an alternative mechanism for cytokine signaling and, while not demonstrated in the present study, may lower the threshold required for autocrine ligand because of its proximity with receptors on the same cell. Such a mechanism could explain the phenomenon of retained sensitivity to endogenously produced TGFβ which we have documented in microsatellite unstable CRC that lack functional TGFβRII and are refractory to exogenously derived TGFβ.24 Further work is required to clarify whether or not surface associated TGFβ can mediate autocrine signaling.
The previous findings highlighting a role for membrane-bound TGFβ in immunosuppression led us to develop a modified version of our coculture assay to verify if the cell-associated TGFβ expressed by our CRC cells could activate TGFβ signaling in lymphocytes. To do this, we used an IEL derived cell line that had been isolated from the human small intestine and possessed a phenotype similar to that of CRC-associated intraepithelial tumor infiltrating lymphocytes (TILs).18 These IELs were cocultured in the presence of suspended surface-bound TGFβ expressing responder cells and then examined for changes in phospho-Smad2 status via flow cytometry using CD3 gating to focus on the correct cell population. In comparison with IELs incubated in the conditioned medium, IELs cocultured directly with the stimulator cells demonstrated an important increase in phospho-Smad2 expression. This increase in activated Smad2 was significantly smaller when the stimulator cells were first incubated with the furin inhibitor. Similar cell–cell contact dependent Smad2 activation was seen when IELs were incubated with two other CRC cell lines, RKO and Moser. While further experimentation will be required to link this surface-TGFβ mediated phenomenon directly to immunosuppression, several previous studies have already implicated Smad signaling in inhibition of T cell activation.25, 26
Several mechanisms of surface-associated TGFβ activation may be proposed. While acid activation was observed here, this is unlikely to be a physiologically relevant process. In contrast, the concept of αVβ6 integrin-mediated “activation by traction” is particularly appealing.1, 2 This process involves the activation of LAP-bound TGFβ via a direct interaction of the αVβ6 integrin with LAP. The noncovalent bond between LAP and TGFβ is disrupted by the binding of the αVβ6 integrin, thereby exposing the active site of the cytokine. Importantly, activation by traction does not release TGFβ into the surrounding medium and cell–cell contact is required for signaling via the activated cytokine.2 Other integrin molecules, such as αVβ8, have also been implicated in activation of latent TGFβ and may play a role in the activation of surface-bound TGFβ on CRC.27, 28
Additional research is clearly required to elucidate the detailed mechanism of activation of surface associated TGFβ on CRC cells. Nevertheless, the preliminary finding of signaling competent surface-bound TGFβ on CRC cells is likely to have important implications for host-tumor dynamics. Besides providing an autocrine and paracrine source of TGFβ for adjacent tumor cells, active cell surface-associated TGFβ on CRC cells may contribute to immunosuppression within the tumor microenvironment that may be relevant to cancer treatment by immune-mediated therapies.
The authors state that they have no conflicts of interest. The IEL cell line 32891 was generously donated by Dr. Asma Nusrat (Emory University, GA). The authors thank Dr. Inti Zlobec for helpful discussions and manuscript revision.