Carcinoma-associated fibroblasts (CAFs) are now widely appreciated for their contributions to tumor progression. However, the ability of CAFs to regulate anoikis, detachment-induced cell death, has yet to be investigated. Here, a new role for CAFs in blocking anoikis in multiple cell lines, facilitating luminal filling in three-dimensional cell culture, and promoting anchorage-independent growth is defined. In addition, a novel mechanism underlying anoikis inhibition is discovered. Importantly, it was demonstrated that CAFs secrete elevated quantities of insulin-like growth factor–binding proteins (IGFBPs) that are both necessary for CAF-mediated anoikis inhibition and sufficient to block anoikis in the absence of CAFs. Furthermore, these data reveal a unique antiapoptotic mechanism for IGFBPs: the stabilization of the antiapoptotic protein Mcl-1. In aggregate, these data delineate a novel role for CAFs in promoting cell survival during detachment and unveil an additional mechanism by which the tumor microenvironment contributes to cancer progression. These results also identify IGFBPs as potential targets for the development of novel chemotherapeutics designed to eliminate detached cancer cells.
Implications: The ability of CAF-secreted IGFBPs to block anoikis in breast cancer represents a novel target for the development of therapeutics aimed at specifically eliminating extracellular matrix–detached breast cancer cells. Mol Cancer Res; 12(6); 855–66. ©2014 AACR.
This article is featured in Highlights of This Issue, p. 813
The stromal cells located in the microenvironment of various tumors are now widely appreciated to directly influence the malignant characteristics of adjacent tumor cells (1). A better understanding of the precise molecular mechanisms used by these stromal cells has the potential to lead to knowledge that may be the target of novel and more effective chemotherapeutic approaches. In particular, a number of recent studies have revealed carcinoma-associated fibroblasts (CAFs) to be a major contributor to tumor progression through multiple mechanisms (2, 3). Despite this information, the precise role CAFs play in augmenting tumorigenesis is only beginning to be understood.
One facet of CAF-mediated malignant progression that remains unexamined is the ability of CAFs to inhibit apoptosis. More specifically, in malignant lesions of the breast, CAFs and tumor cells are known to intermingle after cancer cells break through the basement membrane (4). Once the basement membrane has been breached, cancer cells will be subject to a foreign matrix environment and the induction of anoikis (5). Anoikis is defined as cell death induced by detachment from normal extracellular matrix (ECM) and can also be induced by attachment to atypical or unfamiliar ECM (6). It is now accepted that cancer cells must acquire resistance to anoikis to survive during cancer progression (7). Given that the interaction between CAFs and cancer cells occurs during a time when cancer cells need to acquire anoikis resistance, we hypothesized that CAFs may contribute to the survival of ECM-detached cancer cells.
An abundance of evidence as to the molecular mechanisms used by breast cancer cells to evade anoikis has been discovered using MCF-10A cells, a nontumorigenic mammary epithelial cell line (8). Studies using this cell line have revealed that ECM-detached cells accumulate high levels of Bim (9). In addition, metabolic changes can compromise ECM-detached cells (10, 11). Oncogenic signaling from ErbB2 and EGFR can promote the survival of ECM-detached cells by inhibiting Bim and rectifying metabolic alterations, which suggests that breast cancer cells use similar signaling pathways to block the induction of anoikis (10, 12).
Interestingly, previous studies have discovered that CAFs secrete a variety of proteins used as survival factors. However, an examination of the role CAFs play in modulating anoikis has yet to be completed. In this study, we have discovered that CAFs inhibit the induction of anoikis in breast cancer cells. This inhibition is dependent on the secretion of insulin-like growth factor–binding proteins (IGFBPs; -2/-5) by CAFs, which ultimately leads to the stabilization of Mcl-1. Collectively, these data establish CAFs as an antagonist of anoikis in cancer cells and reveal potential targets for the development of novel therapeutics aimed at inducing cell death in ECM-detached cancer cells.
Materials and Methods
MCF-10A cells were grown as described previously (10). WT, Cav−/−, MDA-MB-231, EG (Whitehead Institute, Cambridge, MA), and PD-CAFs were cultured in Dulbecco's Modified Eagle's Media (DMEM) plus 10% FBS. CCD 1074SK cells were cultured in Iscove's medium plus 10% FBS. BPH-1 (University of Michigan, Ann Arbor, MI) and BT549 cells were cultured in RPMI-1640 media plus 10% FBS. SKBR3 cells were cultured in McCoy media plus 10% FBS. 48L cells (LBNL) were cultured in DMEM/F12 plus 10% FBS. Antibiotics were added to all cell cultures except those cultured in DMEM/F12 plus 10% FBS.
Collection of conditioned media
Upon reaching 70% to 80% confluence, the indicated cultures were changed to appropriate epithelial cell media and were cultured continuously for 24 hours. The conditioned media were collected, filtered through a 0.22-μm filter, and used fresh or stored at −20°C until use.
Caspase/cell death assays
Caspase activation was measured using the Caspase-Glo 3/7 Assay Kit (Promega) as described previously (13). Cytochrome c release was assayed as described previously (14). Error bars represent SD and P values were determined using a 2-tailed t test.
Three-dimensional cell culture
To generate acini, cells were grown in reconstituted basement membrane (Matrigel) and imaged as described previously (15). For examination of luminal filling, acini were imaged as described previously (10). The figures, including data from these assays, show representative experiments from at least 3 independent replicates.
Soft agar assays
Soft agar assays were conducted as described previously (10). Conditioned media were added at the time of plating and at feedings as indicated throughout the duration of the experiment. The figures, including data from these assays, show representative experiments from at least 3 independent replicates. Error bars represent SD and P values were determined using a 2-tailed t test.
Cells were fixed in paraformaldehyde and immunofluorescence was conducted as described previously (10).
Cell lysates and subsequent immunoblotting were completed as described previously (10). The following antibodies were used for immunoblotting: cytochrome c (BD Biosciences), β-actin (Sigma-Aldrich), Bcl-2 (Millipore), Bcl-XL (Cell Signaling), Mcl-1, (Millipore, RC13), MCL1 (phospho S159; Abcam), IGFBP-2 (Lifespan Biosciences), ERK1&2 [pTpY185/187] (Invitrogen), ILK1 (4G9; Cell Signaling), Bim (C34C5; Cell Signaling), actin, α-smooth muscle (Sigma-Aldrich 1A4), FSP1/S100A4 (Millipore), α-tubulin (Sigma-Aldrich), and vimentin (Abcam).
Upon reaching 70% to 80% confluence, indicated cultures were changed to standard 10A serum-free media and were continuously cultured for 24 hours. The conditioned media were collected, filtered, concentrated, and stored at −80°C until use. Protein concentration and trypsin digestion were carried out similar to that described previously (16, 17). Mass spectrometry was performed as described in Supplementary Fig. S1 (18).
Short hairpin RNA
Short hairpin RNA (shRNA) constructs were obtained from Sigma (Mission shRNAs). Viruses were produced and selection was carried out as described previously (10).
siRNA SMARTpools (Dharmacon) against ILK and IGFBP-5 were transfected as described previously (10). To note, ILK experiments were plated 48 hours after siRNA transfection and IGFBP-5 experiments were plated 72 hours after siRNA transfection. Cell death was measured using a caspase assay and cell lysates were subjected to immunoblotting as described above.
Female athymic mice (Foxn1-nu), 6-week-old, were obtained from Harlan Laboratories and/or Taconic. A total of 1 × 106 MDA-MB-231 cells and 1 × 106 Cav−/− (WT or IGFBP-2 deficient) fibroblast cells were injected into the lateral tail vein of 8-week-old female nude mice. About 5.5 to 6.5 weeks postinjection, the mice were sacrificed and lungs were removed. Lungs were fixed in Bouin solution for 48 hours and stored in 70% ethanol before analysis. Procedures were evaluated and approved by the University of Notre Dame IACUC (Protocol number 16–027).
Ex vivo lung fluorescence
To confirm that coinjection of MDA-MB-231 cells and fibroblasts resulted in both cell types traveling to the lungs, mice were injected with labeled MDA-MB-231 (DiR) and fibroblasts (DiD) as reported previously (13). In short, cultured MDA-MB-231 or fibroblasts were incubated with a 5 μg/mL final concentration of DiR [1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide ('DiR';DiIC18); ref. 7; Invitrogen] or DiD (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine perchlorate, Invitrogen) before tail vein injection. Fluorescence images of lungs were acquired on the Xtreme 4 MP image station (Bruker Molecular Imaging) with exposure time of 60 seconds, bin 4 × 4, f-stop 1.1, and 150 mm field of view (FOV). For DiR, an excitation/emission filter pair of 750/830 nm was chosen, and for DiD, 630/700 nm was used. Images were subsequently processed, overlaid, and montaged using ImageJ v1.44 p.
Hematoxylin and eosin staining
Lung tissues were fixed with Bouin fixative for 48 hours. The tissues were then transferred into 70% EtOH. After processing, the tissues were embedded and cut in 4-μm sections. The slides were stained with hematoxylin and eosin.
Conditioned media from caveolin−/− NIH-3T3 cells inhibits the induction of anoikis in epithelial cells
To address the possibility that CAFs secrete factors that contribute to the inhibition of the anoikis program, we used NIH-3T3 cells devoid of caveolin-1 (hereafter referred to as Cav−/− cells), an established model cell line for the behavior of CAFs (19). We cultured MCF-10A (10A) cells on nonadherent plates, exposed these cells to conditioned media from Cav−/− cells or wild-type NIH-3T3 cells (hereafter referred to as WT cells), and then assayed for the induction of anoikis. Interestingly, we discovered that exposure of 10A cells to Cav−/− conditioned media led to a striking inhibition of caspase activation during ECM detachment (Fig. 1A). Similar experiments with other cell lines revealed that the ability of Cav−/− conditioned media to inhibit anoikis extended to other cell types (HMEC, BPH) and invasive breast cancer cells (MDA-MB-231, BT-549, SKBR3 cells; Fig. 1B). In addition, we examined the ability of Cav−/− conditioned media to mediate luminal filling in a 3-dimensional cell culture model of mammary acinus formation (8). We observed a prominent escalation in the number of structures scoring as either “mostly filled” or “filled” in those exposed to Cav−/− conditioned media (Fig. 1C). Finally, we investigated the ability of Cav−/− conditioned media to modulate anchorage-independent growth in soft agar assays. Using both 10A cells (engineered to permit minimal colony growth by HPV-E7 and Bcl-2 expression; ref. 10) and MDA-MB-231 cells, we found that treatment with Cav−/− conditioned media could substantially enhance colony formation in both cell lines (Fig. 1D).
Conditioned media from patient-derived CAFs inhibits anoikis in epithelial cells
We next sought to expand our studies to include CAFs derived from patient tumors. To do this, we obtained diagnostic needle biopsy samples from women who had developed breast tumors and isolated pure CAF populations (20). To confirm the isolation of fibroblasts, we stained our cells for cytokeratin and vimentin. As expected, we found that patient-derived CAFs (PD-CAFs) had negligible cytokeratin staining and intense staining for vimentin (Fig. 2A). Additional confirmation using immunoblotting of whole-cell lysates revealed that these cells expressed the well-established CAF markers α-smooth muscle actin (α-SMA) and fibroblast-specific protein-1 (FSP1; Fig. 2A; ref. 3). We then examined whether conditioned media from PD-CAFs could inhibit anoikis. Using multiple PD-CAFs, we discovered that in all cases there was significant inhibition of anoikis comparable to the level of protection conveyed by Cav−/− cells (Fig. 2B and C). Given that these data do not rule out the possibility that fibroblasts of human origin are inherently better able to obstruct the anoikis program, we obtained CCD-1074Sk (CCDs) cells as a control. CCDs are normal human fibroblasts derived from the non-cancerous region of mastectomy tissue (21). Using CCDs as a control, we found that PD-CAF conditioned media could significantly inhibit anoikis in 10A cells as well as a number of invasive breast cancer lines (Fig. 2D). In addition, 10A acini that were exposed to PD-CAF conditioned media exhibited a significant increase in the number of structures scoring as “filled” or “mostly filled” (Fig. 2E). Finally, 10A-E7-Bcl-2 cells plated in soft agar and treated with PD-CAF media enhanced colony formation as compared with treatment with CCD media (Supplementary Fig. S2).
Exposure of epithelial cells to conditioned media from CAFs leads to Mcl-1 stabilization
We next began to assess the molecular mechanism by which CAF conditioned media could impede the activation of the anoikis program. 10A cells grown in the absence of ECM-attachment have been shown to release cytochrome c from the mitochondria (9, 12). Indeed, in all cases, we found that cytosolic cytochrome c was not present when cells were exposed to media from CAFs, suggesting that factors secreted by CAFs can block mitochondrial cytochrome c release (Fig. 3A). In addition, cytochrome c release during anoikis has previously been linked to a stabilization of Bim protein (9, 12). However, Bim levels were not diminished in the presence of CAF conditioned media (Supplementary Fig. S3). Given the well-documented role that Bcl-2 family members play in modulating cytochrome c release from the mitochondria (22), we hypothesized that Bcl-2 family members were regulated in some fashion by CAF-secreted factors. To address this possibility, we treated 10A cells that were exposed to conditioned media with obatoclax, a drug that mimics the actions of pro-apoptotic BH3 only proteins by interfering with antiapoptotic Bcl-2 family members (23). As expected, addition of obatoclax to 10A cells treated with CAF conditioned media reversed the protection from caspase activation (Fig. 3B).
We next assessed the protein levels of antiapoptotic Bcl-2 family members in ECM-detached cells exposed to CAF conditioned media. While we detected no changes in Bcl-2 or Bcl-XL in the presence of CAF conditioned media, we did find a striking increase in the protein levels of Mcl-1 (Fig. 3C). To examine the contribution of this stabilized Mcl-1 to CAF-mediated protection from anoikis, we eliminated Mcl-1 expression in 10A cells by siRNA and found that the protection from anoikis conferred by CAF conditioned media was completely abrogated (Fig. 3D). Mcl-1 is well-known to have a short half-life due to ubiquitination and subsequent degradation by the proteasome (24, 25). Thus, we examined whether the increase in Mcl-1 levels in the presence of CAF conditioned media was due to the prevention of proteasome-mediated degradation. Treatment with the proteasome inhibitor MG132 resulted in a substantial increase in Mcl-1 in 10A cells treated with WT conditioned media while leaving Mcl-1 levels unchanged in cells treated with Cav−/− or PD-CAF media (Fig. 3E). In addition, ECM-detached cells that had been exposed to CAF conditioned media had dramatically decreased levels of phospho-Mcl-1 (serine 159) (Fig. 3F), a site that has previously been shown to be important in the degradation of Mcl-1 (26).
CAFs secrete elevated amounts of IGFBPs that function to inhibit anoikis
To identify proteins secreted by CAFs involved in the inhibition of anoikis, we conducted mass spectrometric analysis of CAF conditioned media. Using nano-UHPLC/MS/MS, we identified numerous proteins that appeared to have differential abundance compared with control. Of particular interest was IGFBP-2 as the LC/MS/MS data suggested it was substantially elevated in Cav−/− conditioned media (Fig. 4A, Supplementary Fig. S4). Although IGFBP-2 typically acts to inhibit the actions of IGF, it has more recently been appreciated to function in alternate roles to promote cancer progression (27, 28). To confirm these observations, we next used differential quantitative nano-UHPLC/MS/MS Multiple Reaction Monitoring (MRM) to determine the precise fold change in IGFBP-2. Targeted MRM transitions to IGFBP-2 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; internal standard) were developed from the empirical LC/MS/MS data and analyzed in triplicate (Supplementary Figs. S5–S7). We observed an approximately 10-fold change in secreted IGFBP-2 levels in Cav−/− conditioned media as normalized to GAPDH levels (Fig. 4B).
To better assess whether these elevated amounts of IGFBP-2 contribute to the inhibition of anoikis by CAFs, we generated Cav−/− cells deficient in IGFBP-2. Interestingly, conditioned media from IGFBP-2-deficient Cav−/− cells was incapable of inhibiting anoikis in ECM-detached 10A cells (Fig. 4C). In addition, this loss of anoikis inhibition correlated with a loss of Mcl-1 stability, suggesting that IGFBP-2 was involved in the stabilization of Mcl-1 and the inhibition of anoikis (Fig. 4C). Furthermore, depletion of IGFBP-2 in Cav−/− cells resulted in a loss of anoikis inhibition in ECM-detached MDA-MB-231 cells (Fig. 4D) and compromised growth of MDA-MB-231 cells in soft agar (Supplementary Fig. S8), suggesting that IGFBP-2 mediated anoikis inhibition is a critical barrier to long-term anchorage-independent growth. Moreover, the addition of recombinant IGFBP-2 to media from Cav−/− cells deficient in IGFBP-2 rescued the anoikis inhibition (Fig. 4E), suggesting that off-target effects were not responsible for the loss of anoikis inhibition. Finally, the addition of recombinant IGFBP-2 to WT or CCD (normal fibroblasts) conditioned media was sufficient to provide caspase protection in 10A cells in the absence of CAFs (Fig. 4F). These data suggest that IGFBP-2 is both necessary and sufficient for CAF-mediated inhibition of anoikis.
Interestingly, when conducting similar LC/MS/MS analysis in PD-CAFs, we did not detect any significant changes in IGFBP-2 when compared with CCD cells (Fig. 4A). However, we did observe an appreciable increase in the secretion of IGFBP-5 via LC/MS/MS (Fig. 4A). To confirm these findings, we conducted MRM analysis and compared the secreted quantities of IGFBP-5 in CCD cells with fibroblast cell lines derived from nonmalignant reduction mammoplasty tissue (EG, 48L; refs. 29–32) and to multiple PD-CAF cell lines. Enhanced IGBFP-5 was observed in all PD-CAFs when compared with the normal mammary fibroblasts (Fig. 4G, Supplementary Fig. S9). IGFBP-5 has been found to activate many of the same noncanonical downstream signaling pathways as IGFBP-2 (33), and thus we examined whether IGFBP-5 was critical for anoikis inhibition by PD-CAFs. Indeed, when we eliminated IGFBP-5 from PD-CAFs by siRNA, anoikis inhibition and stabilization of Mcl-1 were lost (Fig. 4H). These data suggest that enhanced IGFBP-5 secretion is responsible for the anoikis inhibition seen in PD-CAFs and that multiple IGFBPs (−2, −5) can antagonize the induction of anoikis through the activation of similar signaling pathways (i.e., stabilization of Mcl-1).
IGFBP-2 regulates Mcl-1 stability through activation of MAPK and inhibition of GSK-3β
Given that we had previously seen a decrease in phosphorylation of Mcl-1 at serine 159 in the presence of CAF conditioned media (Fig. 3F), we posited that IGFBPs may inhibit anoikis (and stabilize Mcl-1 levels) by inhibiting GSK-3β. To address this possibility, we first treated 10A cells with the GSK-3β inhibitor TDZD-8. As expected, TDZD-8 treatment stabilized Mcl-1 levels and blocked anoikis in 10A cells exposed to WT conditioned media (Fig. 5A). We next treated 10A cells with TDZD-8 and exposed them to media from IGFBP-2 deficient Cav−/− cells to determine whether IGFBP-2 was important for the inhibition of GSK-3β–mediated Mcl-1 degradation. Indeed, we observed the recovery of Mcl-1 stability and restoration of anoikis protection (Fig. 5B).
Given the fact that mitogen-activated protein kinase (MAPK) signaling was upregulated in cells exposed to CAF conditioned media (Fig. 5C) and that MAPK signaling has been shown to inactivate GSK-3β (34, 35), we investigated the role of ERK/MAPK in the stabilization of Mcl-1 and the inhibition of anoikis by IGFBP-2. Addition of MEK1/2 inhibitor, U0126, to 10A cells treated with CAF conditioned media revealed that ERK/MAPK activity was required for the ability of CAF conditioned media to stabilize Mcl-1 and to block anoikis (Fig. 5D and E).
IGFBP-mediated activation of integrin-linked kinase is necessary for anoikis inhibition
While our data suggest that IGFBPs can modulate ERK/MAPK activation to block anoikis, it was unclear how extracellular IGFBPs could stimulate the ERK/MAPK pathway. Interestingly, both IGFBP-2 and IGFBP-5 have previously been revealed to activate integrin-linked kinase (ILK; refs. 28, 33), and ILK can promote phosphorylation and activation of p-ERK (36). To assess the importance of ILK to anoikis inhibition, we treated ECM-detached cells that had been exposed to CAF conditioned media with Compound 22 (Cpd22), a recently described inhibitor of ILK (37). Treatment of 10A cells with Cpd22 resulted in a complete loss of Mcl-1 stability in cells exposed to conditioned media from Cav−/− (Fig. 6A) or PD-CAF cells (Fig. 6B). In addition, Cpd22 treatment inhibited the activation of ERK/MAPK, induced the phosphorylation of Mcl-1, and stimulated anoikis in 10A cells exposed to CAF conditioned media (Fig. 6A and B). We confirmed these data by reducing ILK expression via siRNA and found that a reduction in ILK expression abrogates Mcl-1 expression in 10A cells exposed to CAF media (Fig. 6C). Furthermore, we found that the addition of recombinant IGFBP-2 can overcome the loss of anoikis protection conferred by ILK siRNA suggesting that enhanced quantities of IGFBPs can signal through residual ILK to inhibit anoikis (Fig. 6D). Finally, the loss of anoikis inhibition, decrease in Mcl-1 stability, and diminished activation of ERK/MAPK in response to ILK inhibition was observed in a number of breast cancer lines (Fig. 6E–G).
CAF-secreted IGFBP-2 is critical for tumor formation in vivo
Given that our data demonstrate that CAFs secrete IGFBPs in elevated quantities to block the induction of anoikis, we reasoned that this anoikis inhibition may be critically important for tumor formation in vivo. To examine this, we coinjected MDA-MB-231 cells and either normal or IGFBP-2-deficient CAFs into the tail vein of immunocompromised mice. We chose to inject cells into the tail vein to force the breast cancer cells to grow in the absence of ECM attachment (in the circulation) or in a foreign matrix environment (the lungs). Labeling of the cell populations with dyes that fluoresce at distinct wavelengths (i.e., fibroblasts with DiD and cancer cells with DiR) confirmed that both cell populations did localize to the lungs following coinjection (Fig. 7A). Interestingly, when lungs were excised from tumor-laden mice, the mice that had received the injection of IGFBP-2–deficient CAFs with MDA-MB-231s had significantly lower tumor burden when compared with the injection of normal CAFs with MDA-MB-231s (Fig. 7B). Histologic analysis of the lungs demonstrates robust tumor formation in the lungs of animals coinjected with MDA-MB-231 and normal CAFs that is substantially compromised in the mice that were coinjected with MDA-MB-231 and IGFBP-2–deficient CAFs (Fig. 7C). We quantified this effect by determining the percentage of the lungs that were covered with tumor and found a statistically significant decrease in tumor burden in animals that were injected with MDA-MB-231s and IGBFP-2–deficient CAFs (Fig. 7D). These data suggest that a deficiency in IGFBP secretion by CAFs can substantially compromise tumor formation in vivo.
Despite the heightened awareness of the importance of CAFs in tumor biology, the ability of CAFs to modulate cell death is poorly understood. There have been studies linking CAFs to chemoresistance in prostate cancer cells (38) and to epithelial–mesenchymal transition (39), which has been linked to the inhibition of anoikis through the activity of the protein NRAGE (40). However, a direct connection between CAFs and anoikis protection has yet to be conclusively demonstrated. Here, we describe a critical role for CAFs in blocking anoikis through the secretion of IGFBPs (see model in Fig. 7E). We have elucidated a molecular mechanism that implicates IGFBPs (−2 or −5) in the stabilization of Mcl-1 and subsequent inhibition of anoikis. This connection between CAF-secreted IGFBPs from the tumor microenvironment and enhanced intracellular Mcl-1 levels suggests that targeting IGFBPs with novel chemotherapeutics may be an effective strategy to induce anoikis in ECM-detached cancer cells.
Our data also suggest that CAF-mediated inhibition of anoikis may be particularly important for metastasizing cancer cells. However, the precise point at which cancer cells would benefit from the IGFBP-mediated anoikis inhibition described here is unclear. Given that media from both Cav−/− cells and PD-CAFs can promote luminal filling in 10A acini (Figs. 1C and 2E), it seems likely that CAFs could facilitate the induction of premalignant conditions like ductal carcinoma in situ (DCIS). Whether CAFs could inhibit anoikis at later stages of the metastatic cascade is less clear. That being said, previous studies have revealed that metastasizing cancer cells can carry CAFs from the site of the primary tumor with them through the circulatory system and to the secondary site (41). In fact, our in vivo data (Fig. 7) demonstrate that in this precise circumstance, CAF-mediated IGBFP-2 secretion can significantly impact tumor formation. Thus, it may be possible that CAFs travel with cancer cells during the metastatic cascade and can assist in the inhibition of anoikis by paracrine signaling. This model suggests that CAFs themselves would need to be resistant to anoikis and indeed previously published data indicate that this is the case (42).
Furthermore, the results presented in this study identify IGFBP-2/IGFBP-5 as a potentially important target for the design of novel therapeutics aimed at inducing anoikis in cancer cells. Interestingly, previous research has revealed that both IGFBP-2 and IGFBP-5 levels are high in breast cancers (43, 44). Our data reveal a potential source for these high levels of IGFBPs and suggest that the reason these IGFBPs are highly expressed is their ability to inhibit the anoikis program. Detection of IGFBP-5 in our PD-CAFs but an inability to identify IGFBP-2 could lead one to conclude that IGFBP-5 is only secreted by CAFs in human cells whereas IGFBP-2 is only secreted in mouse CAFs. However, additional analysis of IGFBP-2 and IGFBP-5 in CAFs derived from a large patient cohort will be necessary to make any definitive conclusions. In addition, it will be important to conduct future studies aimed at better assessing IGFBP levels at various stages of the metastatic cascade to determine when antagonizing IGFBPs would be most effective and how an IGFBP inhibitor could be optimally delivered. Nonetheless, the data presented here represent a significant advance in our understanding of how CAFs modulate tumorigenesis and uncover a novel molecular mechanism used by ECM-detached cancer cells to survive. We believe these data pave the way for future studies aimed at eliminating ECM-detached cancer cells through the inhibition of CAF-mediated paracrine signaling and the neutralization of IGFBP-2/5.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: K.J. Weigel, W.M. Leevy, Z.T. Schafer
Development of methodology: K.J. Weigel, A. Jakimenko, W.M. Leevy
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.J. Weigel, A. Jakimenko, B.A. Conti, S.E. Chapman, W.J. Kaliney, W.M. Leevy, M.M. Champion
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.J. Weigel, A. Jakimenko, B.A. Conti, M.M. Champion, Z.T. Schafer
Writing, review, and/or revision of the manuscript: K.J. Weigel, A. Jakimenko, B.A. Conti, S.E. Chapman, M.M. Champion, Z.T. Schafer
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.J. Weigel, A. Jakimenko, Z.T. Schafer
Study supervision: Z.T. Schafer
This work was supported by a V Scholar Award from the V Foundation for Cancer Research to Z.T. Schafer, by funds from the Walther Cancer Research Foundation to A. Jakimenko, by funds from the Coleman Foundation to Z.T. Schafer, by funds from the Notre Dame Glynn Family Honors Program to B.A. Conti, and by a Notre Dame Departmental Graduate Fellowship to K.J. Weigel.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The authors thank Calli Versagli, Cassandra Buchheit, Raju Rayavarapu, Joshua Mason, Crislyn D'Souza-Schorey, Patricia Champion, Shaun Lee, Reginald Hill, Veronica Schafer, and the entire Schafer Lab for helpful comments, experimental assistance, and valuable discussion. They also thank Amy Leliaert for technical and organizational assistance, Jill Macoska at the University of Michigan for the BPH-1 cells, Robert Weinberg at the Whitehead Institute for the EG cells, and Martha Stampfer at Lawrence Berkeley National Laboratory for the 48 L cells. They also thank the physicians at South Bend Medical Foundation and Memorial Hospital of South Bend for assisting with the acquisition of PD-CAFs.
Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).
- Received February 17, 2014.
- Accepted February 18, 2014.
- ©2014 American Association for Cancer Research.