Increased stromal collagen deposition in human breast tumours correlates with metastases. We show that activation of the collagen I receptor DDR2 (discoidin domain receptor 2) regulates SNAIL1 stability by stimulating ERK2 activity, in a Src-dependent manner. Activated ERK2 directly phosphorylates SNAIL1, leading to SNAIL1 nuclear accumulation, reduced ubiquitylation and increased protein half-life. DDR2-mediated stabilization of SNAIL1 promotes breast cancer cell invasion and migration in vitro, and metastasis in vivo. DDR2 expression was observed in most human invasive ductal breast carcinomas studied, and was associated with nuclear SNAIL1 and absence of E-cadherin expression. We propose that DDR2 maintains SNAIL1 level and activity in tumour cells that have undergone epithelial–mesenchymal transition (EMT), thereby facilitating continued tumour cell invasion through collagen-I-rich extracellular matrices by sustaining the EMT phenotype. As such, DDR2 could be an RTK (receptor tyrosine kinase) target for the treatment of breast cancer metastasis.
Tumour–microenvironment interactions are critical for tumour development and metastasis. In addition to alterations in cell types and their function, the stroma associated with primary tumours also exhibits increased deposition of matrix proteins, including fibrillar collagens. In human breast tumours and transgenic mouse models, increased collagen in the tumour stroma correlates with metastases1,2, and can increase stromal stiffness or tension, influencing breast tumour invasion through tumour cell integrin-dependent mechanotransduction signalling3. Breast tumour cells also preferentially migrate along linear collagen fibrils near the tumour/extracellular matrix (ECM) boundary4. Although integrins are important collagen receptors for tumour cell invasion and migration, the role of other collagen receptors, such as the DDRs, in cancer development and metastasis has not been extensively explored5.
DDRs are distinctive RTKs in that their ligands are fibrillar collagens, their activation kinetics is slow, and they may exist as preformed dimers in the absence of ligand6. DDR1 is present in epithelial but not mesenchymal cells, whereas DDR2 is expressed in mesenchymal cells and not epithelia. DDR2 exhibits a more selective ligand specificity and is activated by collagens I and III, which are abundant in the ECM. DDR1 has a broader collagen specificity that includes collagen IV, a major component of the basement membrane.
EMT contributes to breast cancer metastasis7. EMT facilitates primary tumour invasion through the basement membrane and migration through the tumour-associated stroma or ECM by suppressing tumour cell–cell adhesion and stimulating tumour cell invasion and migration8. Various signals generated within the tumour microenvironment activate tumour cell intrinsic transcription factors considered to be inducers of EMT, such as SNAIL1 (also known as SNAI1), SNAIL2 (SLUG), TWIST1 and ZEB (refs 9, 10, 11).
Human and mouse cancer cell line experiments and human pathology studies implicate SNAIL1 in breast cancer metastasis9,12,13. SNAIL1, which has a very short half-life is degraded by the proteasome, is predominantly present in tumour cells at the tumour/ECM boundary and in mesenchymal cells within the tumour-associated stroma13, whereas SNAIL1 messenger RNA is more broadly expressed throughout the primary tumour. Although EMT-stimulating signals (for example, TGF-β) induce SNAIL1 transcription, post-transcriptional and post-translational regulation influence SNAIL1 stability and subcellular localization, which is critical for the capacity of SNAIL1 to induce or sustain cancer EMT (refs 14, 15, 16). Given the distinct pattern of SNAIL1 mRNA and protein levels in tumours, the environmental signals regulating SNAIL1 stability probably originate at the tumour/ECM boundary. The identity, mechanism of SNAIL1 stabilization, and functional consequences of these signals on metastasis are only beginning to be determined15,16.
Here we show that the collagen receptor DDR2 stabilizes SNAIL1 and is critical for breast cancer invasion and migration in vitro and for metastasis in vivo. We propose that DDR2 contributes to breast cancer metastasis by sustaining SNAIL1 stability and activity in tumour cells that have undergone EMT, to promote their invasion and migration through collagen-I-enriched tumour-associated matrices.
DDR2 stabilizes cellular SNAIL1 level, post-transcriptionally
We previously identified DDR2 in a human kinome/phosphatome short interfering RNA (siRNA) screen for regulators of the total cellular SNAIL1 level16. RNA interference (RNAi)-mediated depletion of DDR2, but not DDR1, in HEK293 cells that stably expressed a bioluminescent SNAIL1 fusion protein, SNAIL1–clic beetle green (SNAIL1–CBG; ref. 16), reduced SNAIL1–CBG bioluminescence and protein levels without affecting the transcription of SNAIL1 or SNAIL1–CBG (Supplementary Fig. S1a,b).
DDR2 depletion in human MDA-MB-231 and mouse 4T1 cells that have undergone EMT resulted in a decreased level of SNAIL1 without affecting SNAIL1 transcription (Fig. 1a,b), demonstrating that DDR2 also affected endogenous SNAIL1 levels. DDR2 depletion also affected other EMT inducers: the SNAIL2 level was slightly diminished and the ZEB1 level increased slightly, whereas the TWIST1 level was unchanged (Fig. 1a). Depletion of DDR2 or SNAIL1 in MDA-MB-231 cells that stably expressed bioluminescent SNAIL1–CBG reduced SNAIL1–CBG bioluminescence, SNAIL1–CBG protein and endogenous SNAIL1 protein levels (Fig. 1c,d). Thus, SNAIL1–CBG bioluminescence could serve as an approximation of the cellular SNAIL1 level to allow live-cell analyses.
Overexpression of DDR2 in MDA-MB-231 or 4T1 cells increased the SNAIL1 level (Fig. 1e,f) without affecting mRNA levels (Supplementary Fig. S1c). This depended on DDR2 kinase activity, as overexpression of a kinase-dead DDR2 form did not affect the SNAIL1 level (Fig. 1e). Overexpression of DDR1 may slightly increase SNAIL1 levels (Fig. 1f). These experiments indicated that DDR2 stabilized SNAIL1 without affecting SNAIL1 transcription.
Collagen-I-induced stabilization of SNAIL1 requires DDR2 and can occur independently of integrin or TGF-βR
Addition of HEK293 cells, co-transfected with myc–DDR2 and SNAIL1–Flag, or MDA-MB-231 cells to collagen I resulted in tyrosine phosphorylation and activation of DDR2 and an increased SNAIL1 level (Fig. 2a,b), in contrast to exposure of MDA-MB-231 cells to collagen IV, fibronectin or gelatin, which had little effect on SNAIL1 levels (Fig. 2b). DDR2 depletion in MDA-MB-231 cells abrogated the increase in the SNAIL1 level following collagen I stimulation (Fig. 2b). SNAIL1 was stabilized by exposing MDA-MB-231 cells to collagen I in a two-dimensional (2D) or 3D context (Supplementary Fig. S1d), and this was also the case for MDA-MB-231 cells containing SNAIL1–CBG, which led to increased SNAIL1–CBG bioluminescence and protein level (Supplementary Fig. S1e,f).
When MDA-MB-231 cells were mixed with neutralizing antibodies against β1 integrin, before and during their exposure to collagen I, the SNAIL1 cellular protein level increased despite inhibition of integrin signalling (for example, pFAK; Fig. 2c). Inhibition of TGF-β signalling in MDA-MB-231 cells also had no effect (Fig. 2d), despite TGF-β increasing the SNAIL1–CBG protein level in the absence of collagen I stimulation (Supplementary Fig. S1g), indicating that collagen-I-induced stabilization of the SNAIL1 level required DDR2 and could occur independently of integrin and TGF-β.
DDR2 expression is induced during EMT but is not required for EMT induction
Normal breast epithelial cells (MCF-10A) do not express SNAIL1 (ref. 12). When exposed to TGF-β they undergo EMT and SNAIL1 expression is induced, the level of DDR2 expression is increased and the DDR1 protein level is decreased (Fig. 2e). DDR2 was not required for TGF-β-dependent EMT induction in MCF-10A cells, as DDR2 depletion minimally affected EMT (Fig. 2e). Consistent with a role for DDR2 in stabilizing SNAIL1, the levels of SNAIL1 were reduced in DDR2-depleted cells as compared with control cells following EMT (Fig. 2e). Constitutive overexpression of SNAIL1 in MCF-10A cells induces TGF-β-independent EMT (Fig. 2f)12; however, DDR2 overexpression did not induce morphological or biochemical EMT (Fig. 2f). These results indicated that although DDR2 expression was induced during EMT, the presence or absence of DDR2 did not influence TGF-β-induced EMT.
DDR2 influences breast cancer cell invasion and migration ex vivo and metastases in vivo
A major function for SNAIL1 during cancer metastasis is induction of invasion through the basement membrane and migration through the ECM (ref. 8). Depletion of SNAIL1 and DDR2 in 4T1 and MDA-MB-231 cells inhibited their migration in 3D collagen I matrices (Fig. 3a and Supplementary Fig. S2a), without influencing their proliferation (Fig. 3b and Supplementary Fig. S2b), and also inhibited cell invasion through Matrigel (Supplementary Fig. S3c).
To determine whether DDR2 influenced breast cancer cell metastases in vivo SNAIL1 or DDR2 was stably depleted in 4T1-Luc–GFP cells and equal numbers of cells were surgically implanted into the breast tissue of syngeneic BALB/cJ mice. After correcting for the small differences observed in the growth of the SNAIL1- and DDR2-depleted primary tumours in vivo (Fig. 3c,d), reduced lung metastases were observed in DDR2- and SNAIL1-depleted tumour-bearing mice (Fig. 3e–g). Both the number of mice that developed lung metastases (Fig. 3f) and the extent of metastasis present (Fig. 3g) were reduced in mice injected with DDR2-depleted tumour cells. Fluorescent GFP imaging of isolated lungs (Supplementary Fig. S2d) and histologic analysis of lung slices (Supplementary Fig. S2e) confirmed the bioluminescent results, demonstrating that DDR2 promoted breast to lung metastases in vivo.
Collagen-I-activated DDR2 induces collagen I transcription and remodelling of collagen fibres at the tumour/ECM boundary
During breast cancer progression, increased deposition of ECM proteins, particularly collagens, can occur within and near the primary tumour17,18. This is associated with poor outcome due to increased local invasion and distant metastasis19,20. The collagen fibres deposited near tumours can also thicken and realign, and tumours with fibres aligned perpendicularly to the tumour edge have an increased propensity to invade and metastasize. This observation has led to a classification of tumour-associated collagen signatures (TACS) in primary breast tumours that correlates with breast tumour progression and patient outcome2,4,21. The DDRs have been implicated in both collagen synthesis by cells and remodelling of the ECM (refs 22, 23, 24).
Collagen fibre alignment was measured and TACS was calculated using second-harmonic generation imaging of collagen fibres at the tumour/ECM boundary. Tumours depleted of SNAIL1 or DDR2 showed an increase in TACS-2, the non-invasive collagen signature, and a decrease in TACS-3, the invasive collagen signature, compared with control tumours (Fig. 4a,b and Supplementary Fig. S3a,b), suggesting that breast cancer cell-expressed DDR2 might influence collagen remodelling by invasive tumours.
Addition of MDA-MB-231 cells to collagen-I-coated plates, induced collagen I mRNA transcription by 2–2.5-fold, and depletion of DDR2 or SNAIL1 reduced this transcriptional response (Fig. 4c). Similar results were observed with 4T1 cells (Supplementary Fig. S3c). Despite this effect on collagen I transcription there was no discernible change in the collagen content of DDR2-depleted 4T1 tumours when compared to control tumours, as determined from trichrome staining (Supplementary Fig. S4a).
DDR2-activated ERK2 associates with and phosphorylates SNAIL1
Stabilization of SNAIL1 by collagen-I-induced DDR2 activation was unlikely to be the result of direct phosphorylation by DDR2, as tyrosine phosphorylation of SNAIL1 has not been observed in phosphoproteomic analyses16,25.
Addition of MDA-MB-231 cells to collagen I led to activation of ERK1/2, but not p38, JNK or Akt, which was shown by pharmacological inhibitor studies to be required for stabilization of SNAIL1 (Fig. 5a), but had no effect on SNAIL1 mRNA levels (Supplementary Fig. S5a). RNAi depletion revealed that ERK2, but not ERK1, was required for SNAIL1 stabilization following collagen I activation (Supplementary Fig. S5b). ERK2 was recently shown to be critical for EMT induction of RasV12-transfected MCF-10A cells26. Sequence analysis of SNAIL1 revealed the presence of potential D-domain and DEF-domain ERK-docking sites (Fig. 5b)27. In MDA-MB-231 cells a small amount of endogenous ERK2 associated with endogenous SNAIL1, and this association increased when cells were added to collagen I, yet was blocked by pretreating cells with the MEKK inhibitor PD98059 (Fig. 5b). SNAIL1 and ERK2 readily co-immunoprecipitated from HEK293 cells co-transfected with both (Supplementary Figs S5c,d and S6a). Mapping experiments indicated that the SNAIL1 D-domain docking site, but not the DEF-domain, was important for ERK2 association (Supplementary Fig. S5c). Mutation of the D-domain ERK-docking site in SNAIL1 (RK-MM) reduced its association with ERK2 (Supplementary Fig. S5d). Mutation of the D-domain recognition region of ERK2 (D319N), but not the DEF-domain recognition region (Y261A; ref. 26), reduced ERK2–SNAIL1 association (Supplementary Fig. S6a). In vitro kinase assays with purified ERK2 and purified GST–SNAIL1 proteins containing S–A mutations in three potential ERK phosphorylation sites identified, revealed that Ser 82 and Ser 104 were phosphorylated (Fig. 5c).
To determine if phosphorylation of SNAIL1 by ERK2 was functionally relevant we investigated whether mutation of Ser 82 and Ser 104 affected SNAIL1 stability in response to collagen I. When expressed in MDA-MB-231 cells, the protein half-life of an 82/104A SNAIL1–CBG mutant was half that of WT SNAIL1 (Fig. 5d). When cells were added to collagen I, the WT SNAIL1–CBG half-life increased whereas the half-life of 82/104A SNAIL1–CBG remained unaltered (Fig. 5d). In contrast to WT SNAIL1–CBG-expressing cells, which required DDR2 for SNAIL1 stabilization in response to collagen I, the stability of 82/104A SNAIL1–CBG in collagen-I-stimulated cells was unaffected by the presence or absence of DDR2 (Supplementary Fig. S6b).
Depletion of endogenous SNAIL1 in MDA-MB-231 cells and concurrent rescue with RNAi-resistant WT SNAIL1, using dual-copy lentiviruses28, completely restored 3D cell migration to SNAIL1-depleted cells, whereas rescue with the 82/104A mutant (2A) or ERK D domain mutant of SNAIL1 (RK-MM) only partially rescued the migration defect of SNAIL1-depleted cells (Fig. 5e and Supplementary Fig. S7a). Control western blots indicated that endogenous SNAIL1 was effectively depleted and that RNAi-resistant rescue mutants were expressed at levels equivalent to endogenous SNAIL1 (Supplementary Fig. S7b). Endogenous DDR2 was not affected by these manipulations (Supplementary Fig. S7b).
Both WT SNAIL1–CBG and 82/104A.SNAIL1–CBG were stabilized by proteasome inhibition, but only WT SNAIL1–CBG was stabilized by exposure to collagen I (Fig. 6a). SNAIL1 is degraded by proteasomes in the cytosol and when nuclear SNAIL1 is more stable14,16. When transfected in MCF-10A cells, WT SNAIL1 was predominantly nuclear whereas 82/104A SNAIL1 was predominantly cytosolic (Fig. 6b); however, leptomycin B treatment, to inhibit nuclear export, resulted in accumulation of both forms in the nucleus (Fig. 6b). This suggested that the 82/104A mutant was not retained in the nucleus and that ERK2-mediated phosphorylation of SNAIL1 leads to its nuclear accumulation and protection from ubiquitylation and subsequent proteosomal degradation. When MDA-MB-231 cells were treated with collagen I, the increase in WT SNAIL1 was exclusively nuclear (Fig. 6c). In HEK293 cells transfected with DDR2, SNAIL1 and HA-tagged ubiquitin and then stimulated with collagen I, in the absence of proteosomal inhibition, immunoprecipitated 82/104A SNAIL1 was more ubiquitylated than WT SNAIL1 (Fig. 6d). These results indicated that phosphorylation of SNAIL1 on Ser 82 and Ser 104 by collagen I–DDR2-stimulated ERK2 led to its accumulation in the nucleus and decreased ubiquitylation. Thus, SNAIL1 was stabilized, its level increased and its capacity to mediate tumour cell invasion and migration was maintained.
Delineation of the intracellular signalling pathway whereby DDR2 leads to SNAIL1 stabilization and tumour cell invasion/migration
To evaluate the contribution of more proximal DDR2-activated signalling enzymes to ERK2 activation and SNAIL1 stabilization we examined whether FAK or Src was required for collagen-I-induced SNAIL1 stabilization5,29. FAK depletion in MDA-MB-231 cells did not affect ERK activation or SNAIL1 stabilization (Supplementary Fig. S8a). Src, however, was required for collagen-I-induced stabilization of SNAIL1 (Supplementary Fig. S8b).
Glycogen synthase kinase 3β (GSK3β) is a well-established regulator of SNAIL1 stability14. ERK can associate with and inactivate GSK3β by acting as a scaffold for p90RSK and GSK3β (ref. 30), and this ERK–p90RSK–GSK3β pathway can be stimulated by RTKs such as HER2 (refs 30, 31). Thus, we investigated whether DDR2 activation could modulate GSK3β activity downstream of ERK2. pS9-GSK3β (inhibited) was present in unstimulated MDA-MB-231 cells, and its levels were increased when cells were added to collagen I (Fig. 7a). DDR2 depletion in unstimulated cells decreased the level of inhibited GSK3β; however, when these cells were added to collagen I the level of pS9-GSK3β increased to the same extent as in control cells (Fig. 7a). This suggested that when breast cancer cells interacted with collagen I, GSK3β activity was inhibited independently of DDR2. This DDR2-independent pathway could also contribute to stabilization of SNAIL1 in response to collagen. We tested this possibility by assessing whether depleting GSK3β affected SNAIL1 stabilization following exposure to collagen I. The level of SNAIL1 increased 2.5–3.5-fold in unstimulated cells, indicating that GSK3β contributed to basal SNAIL1 levels (Fig. 7b). However, when GSK3β-depleted cells were added to collagen I, SNAIL1 increased to the maximal level seen in control cells (Fig. 7b), indicating that although GSK3β contributed to SNAIL1 stabilization, it did so in unstimulated cells and was not necessary for collagen-I-induced stabilization of SNAIL1.
To determine whether ERK2 activation in response to collagen I was required for GSK3β inhibition and SNAIL1 stabilization, ERK2 was depleted in MDA-MB-231 cells. In unstimulated cells, depletion of ERK2 resulted in decreased inhibition of GSK3β (Fig. 7c); however, when these cells were added to collagen I, pS9-GSK3β levels were equal to those in control or ERK1-depleted cells (Fig. 7c). As expected, depletion of ERK2 resulted in a decreased level of SNAIL1 in unstimulated cells (Fig. 7c), which was equivalent to SNAIL1 levels in DDR2-depleted cells (Fig. 7a). Therefore, ERK2 was required to inhibit GSK3β in unstimulated cells. When ERK2-depleted cells were added to collagen I, the SNAIL1 levels increased twofold, whereas those in control cells increased threefold (Fig. 7c). The level of SNAIL1 in ERK2-depleted cells stimulated with collagen I did not approach that observed in control or ERK1-depleted cells, despite equivalent levels of pS9-GSK3β (Fig. 7c). Taken together, these results indicated that although ERK2 also stabilized SNAIL1 by inhibiting GSK3β this was a minor contribution.
These signalling experiments suggested a pathway whereby collagen-I-mediated activation of DDR2 stabilized SNAIL1 (Fig. 7d). Collagen-I-activated DDR2 stimulated ERK2 activity, in a Src-dependent manner. ERK2 then interacted with and phosphorylated SNAIL1 on Ser 82 and Ser 104 and inhibited GSK3β activity, with both effects contributing to stabilization of SNAIL1. Elevated SNAIL1 levels sustained both MT1-MMP production and activity32,33 and collagen synthesis (Fig. 4c), both of which could contribute to the remodelling of collagen fibres seen at the breast tumour/ECM interface (Fig. 4a,b).
To test this proposed pathway, we depleted each component individually, and determined the activity and level of other proteins in the pathway. DDR2 depletion affected the activity or protein level of all downstream components (Fig. 7e). ERK2 depletion influenced the activity or protein level of GSK3β, SNAIL1 and MT1-MMP, but not DDR2 (Fig. 7e). GSK3β depletion affected only the SNAIL1 levels in unstimulated cells and had no effect on other pathway components, or SNAIL1 in collagen-I-stimulated cells (Fig. 7e). SNAIL1 depletion affected only the level of MT1-MMP (Fig. 7e). Finally, depletion of MT1-MMP did not affect SNAIL1 levels (Supplementary Fig. S8c), but was required for MDA-MB-231 cell migration in 3D collagen I gels (Supplementary Fig. S8d).
DDR2 is expressed in human invasive ductal breast cancers and correlates with nuclear SNAIL1, loss of E-cadherin and increased collagen
We performed immunohistochemistry on human invasive ductal cancer (IDC) and invasive lobular breast cancer samples compared with tumour-associated normal breast tissue. SNAIL1 and DDR2 were not expressed in normal breast ducts or acini (Fig. 8a,b and Supplementary Fig. S8e); however, 71% of IDC cells expressed DDR2 (Fig. 8a). There was no difference in DDR2 expression and lymph node involvement (Supplementary Fig. S8e). Only 21% of invasive lobular cancers expressed low levels of DDR2 (Fig. 8a). Ongoing TCGA RNA sequence analysis of gene copy number in breast cancers34 found that 5% of invasive breast tumours exhibited amplified copy number of the DDR2 gene and these patients had worse survival rates (Fig. 8b). Thirty-eight per cent of IDCs were SNAIL1 positive and of these 91% were DDR2 positive and 81% E-cadherin negative (Fig. 8c). However, in SNAIL1-negative tumours 47% were DDR2 positive (Fig. 8c). Thus, DDR2 was expressed in most invasive ductal breast carcinomas studied, and most of those positive for nuclear SNAIL1 expression also expressed DDR2 and lacked E-cadherin.
Trichrome staining was used to assess collagen content in IDCs. Of IDCs with increased collagen (70%), 86% were positive for DDR2, whereas 55% of collagen-negative tumours (30%) were DDR2 positive (Fig. 8d). Starting with DDR2-positive tumours (77%), 78% exhibited increased collagen (Fig. 8e). In DDR2-negative tumours (23%), 42% exhibited increased collagen (Fig. 8e). This suggested that in tumours with increased collagen content there was a strong association with the presence of DDR2; however, this was not specific as significant numbers of collagen-negative tumours expressed DDR2 and a significant number of DDR2-negative tumours had increased collagen content.
We have identified the collagen receptor DDR2 as contributing to invasion, migration and metastasis of breast cancer cells by stabilizing the SNAIL1 level. Our working model posits that breast tumour environmental signals such as TGF-β or Wnt induce SNAIL1 transcription14,15 and EMT (ref. 12). As a result tumour cells become mesenchymal and DDR2 is expressed as part of the EMT process. Following invasion through the basement membrane, tumour cells are exposed to the collagen-I-rich ECM associated with breast tumours and this activates DDR2 and serves as a positive signal to maintain SNAIL1 level and activity, thereby sustaining tumour cell migration leading to metastasis.
Most human invasive ductal breast cancer expressed DDR2. Although genetic events leading to DDR2 expression cannot be excluded at present, it is likely that many invasive breast cancers have undergone EMT and DDR2 expression could reflect this process. Although we observed an association between nuclear SNAIL1, lack of E-cadherin and the presence of DDR2, a significant proportion of SNAIL1-negative tumours expressed DDR2, suggesting that in addition to stabilizing SNAIL1, DDR2 could contribute to cancer metastasis by other means. RNA sequence analysis found that 5% of invasive breast tumours have amplified DDR2 gene expression34, with this single variable revealing that these patients exhibited decreased survival. Whether these tumours also express SNAIL1 is not known, and it would be interesting to study whether they represent the aggressive claudin-low subtype of triple-negative breast cancers that exhibit mesenchymal features, including the presence of EMT inducers such as SNAIL1 (ref. 7).
The presence of collagen fibres perpendicularly aligned to the tumour boundary predicts for increased relapse and decreased survival4. An increase in collagen in vitro drives expression of EMT markers35, providing support for a link between increased collagen and EMT. Our studies suggest that DDR2 could provide a connection between collagen and sustained EMT. However, DDR2 is not the only collagen I receptor expressed by invasive breast cancer cells. Invasive breast cancers can exhibit increased tissue stiffness with tumour cells responding to increased ECM tension through integrin-mediated mechanotransduction3. Thus, integrin activity is important for breast cancer invasion3,36. Whether DDR2 contributes to integrin-mediated mechanotransduction responses during tumour invasion is not known, but other RTKs have been shown to influence integrin function37,38. In addition, the increase in ERK activity in tumour cells in a stiff environment was ascribed to integrin-dependent focal adhesion activation and EGFR signalling3. Collagen-induced DDR2 activation also led to an increase in ERK activity that contributed to SNAIL1 stabilization and tumour cell invasion/migration, raising the possibility that, in addition to integrin/focal adhesion-Rho-ROCK-mediated mechanotransduction3, SNAIL1 stabilization and EMT maintenance could also contribute to tumour cell invasion in stiff tissues. Regardless, DDR2 depletion from aggressive, metastatic breast cancer cell lines that have undergone EMT and express SNAIL1 led to reduced in vivo metastases, implying that DDR2, like integrin, is an important collagen receptor during breast cancer metastasis.
The collagen-I-induced, DDR2–ERK2 signalling axis that maintains SNAIL1 level and activity in breast tumour cells that have undergone EMT could contribute to the increased aggressiveness of breast cancer in women with dense breasts that are due, in part, to increased fibrillar collagen deposition39, tumours in an involuting breast microenvironment that are associated with increased collagen deposition40 and in breast cancer patients with extensive fibrotic reactions around the primary tumours1. Thus, DDR2 could be a potential target to treat breast cancer metastasis.
Cell culture and viral infections.
HEK293 and 293T cells were from ATCC. 4T1-Luc and MDA-MB-231-Luc cells were provided by K. Weilbaecher (Washington University in St Louis, USA). MCF-10A cells were provided by L. Michel (Washington University in St Louis, USA). HEK293 clones stably expressing SNAIL1–CBG were described previously16. 4T1-Luc–GFP and MDA-MB-231-Luc–GFP cells were generated by infecting 4T1-Luc or MDA-MB-231-Luc cells with lentiviruses expressing EGFP and selecting for stable clones. Before use, cells were FACS sorted for equivalent GFP expression. Production of lentiviruses and pBabe containing-amphotropic viruses, and infection of target cells were described previously16. To make stable cell lines of MDA-MB-231 and 4T1 cells selection was carried out in 4 μg ml−1 puromycin, and MCF-10A cells were selected in 1.5 μg ml−1 puromycin.
Antibodies and chemical inhibitors.
For antibodies and dilutions used for immunoprecipitations, western blot and immunohistochemistry, see Supplementary Table S1. PD98059, SB203580, wortmannin and SP600125 were from Calbiochem. Cycloheximide, leptomycin B, MG132, gelatin, the TGF- β inhibitor SB431542 and the Src inhibitor PP2 were from Sigma. Type I collagen, fibronectin and collagen IV were from BD Biosciences.
Immunoprecipitation, western blots and in vitro kinase assay.
Cells were lysed with a non-denaturing lysis buffer as described previously16. Five hundred micrograms of whole-cell lysate, as determined by BCA analysis (Pierce), was used for immunoprecipitation. Integrated relative densities of individual bands were quantified using NIH ImageJ software. In vitro kinase assays were performed as previously described16 with minor modifications. Briefly, GST–SNAIL1 and its mutants were produced in bacteria and purified with glutathione Sepharose 4B (GE Healthcare Life Sciences). Purified, active P42 MAP kinase (NEB, Cat. no. P6080) was used in kinase reactions carried out in p42 MAP Kinase Reaction Buffer (NEB) with 100 units of MAP Kinase, 2 μg of GST–SNAIL1, 50 μM ATP and 10 μCi [γ-32P]ATP. Reactions were incubated at 30 °C for 45 min, terminated and resolved by SDS–PAGE.
Bioluminescence imaging of live cells.
Bioluminescence measurements on MDA-MB-231 (SNAIL1–CBG) cells were acquired in an IVIS 100 imaging system (Caliper Life Sciences) at 37 °C under 5% CO2 flow. Bioluminescence photon flux (photons per second) data were analysed by region of interest measurements in Living Image 3.2 (Caliper Life Sciences). This raw data were imported into Excel (Microsoft) and normalized to controls.
In vitro cell proliferation analysis.
Cells were seeded at 2×104cells per well in triplicate 24-well plates in 500 μl of growth medium on day 0. Cells were trypsinized, resuspended in a total volume of 500 μl of medium, and counted with a haemacytometer at the intervals shown in the figure legends.
Migration and invasion assays.
For 3D cell migration assays, 105 cells were embedded in 20 μl of type I collagen gel (2.0 mg ml−1) extracted from rat tail (BD Biosciences). After gelling, the plug was embedded in a cell-free collagen gel (2.0 mg ml−1) within a 24-well plate. After allowing the surrounding collagen matrix to gel (1 h at 37 °C), 0.5 ml of culture medium was added on the top of the gel and cultured for another 2 days. Invasion distance from the inner collagen plug into the outer collagen gel was quantified. For invasion assays, Transwell cell invasion assays were performed using a 24-well FluoroBlok Transwell insert (BD) with 8-μm pore size. Inserts were prepared by coating the upper surface with 1 mg ml−1 of Matrigel (BD Biosciences) for 4–6 h at 370 C in a 5% CO2 incubator. 5×104 MDA-MB-231-Luc or 4T1-Luc cells in DMEM containing 1% FBS were seeded into the upper chamber of the insert. The bottom chamber contained DMEM with 10% FBS. After 24 h, luminescence intensity was measured using a FluoStar Optima microplate reader (BMG Labtech) for 10 individual fields on the bottom of each insert.
Breast implant in vivo metastasis assay.
Eight-week old female BALB/cJ mice (Jackson Labs) were anaesthetized with a ketamine/xylazine cocktail (90 mg kg−1 ketamine and 13 mg kg−1 xylazine, intraperitoneal injection) and the abdomen was sterilized using povidone–iodine (Betadine) solution and ethanol. A small Y-shaped incision was made in the lower abdominal skin to expose the fourth mammary gland using surgical scissors and bleeding vessels were cauterized. 4T1-Luc–GFP cells (1×106) in 50 μl DMEM were injected into the fourth mammary gland using a 29-gauge needle. The skin flaps were replaced and closed using 9 mm wound clips, and the surgical site was swabbed with triple-antibiotic cream. The mice were imaged 24 h post-surgery and weekly thereafter using the IVIS 100 bioluminescent imaging system (Perkin Elmer) following an intraperitoneal injection of D-luciferin (150 mg kg−1). After 5 weeks, mice were injected with luciferin, euthanized and lungs removed and immediately imaged ex vivo. The lungs and primary tumours were then fixed in 10% formalin for 24 h, cryopreserved in 30% sucrose overnight, and finally embedded in OCT and frozen in a dry ice/ethanol bath. Frozen specimens were sectioned with a cryostat (6 μm) and analysed by fluorescence microscopy for GFP expression or stained with haematoxylin and eosin for histological analysis. All mouse work was approved by Washington University Institutional Animal Care and Use Committee.
Second-harmonic generation imaging of collagen fibres at the tumour boundary.
Excised 4T1 tumours were fixed, embedded in 3% agarose, and then sectioned into 100 mm slices using a vibrotome. The collagen matrix images were observed using multiphoton laser scanning microscopy and second-harmonic generation on a custom-built multiphoton microscope platform with an excitation source produced by a Spectra Physics Mai Tai DeepSee laser (Newport), tuned to a wavelength of 890 nm, mounted around a Nikon Eclipse TE300 inverted microscope (Nikon Instruments). Images were focused onto a Nikon ×20 Plan Apo multi-immersion lens (numerical aperture 1.4), and second-harmonic generation emission was observed at 445 nm and discriminated from fluorescence using a 445 nm 20 nm narrow band-pass emission filter (Semrock). Three tumours were sectioned for each tumour type and three z-stacks were collected with a 3 μmstep size for each tumour. The z-stacks were compressed into one image and collagen fibre analysis was completed using CurvAlign software. All collagen fibres with angles between 0°–30° from the tumour boundary were considered to be TACS-2 and fibres with angles between 60° and 90° were classified as TACS-3 (ref. 21).
Cytoplasmic and nuclear fractions were prepared by resuspending cells in lysis buffer A (10 mM HEPES (pH 7.9), 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 0.1% NP40, 1 mM dithiothreitol and 1.0 mM phenylmethyl sulphonyl fluoride with complete protein inhibitors). Nuclei were collected by centrifugation (10,000g) for 10 min at 4 °C, and washed 3 times in buffer A. Nuclei were lysed by shaking vigorously in buffer B (20 mM HEPES (pH 7.9), 0.4 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol and 1 mM phenylmethylsulphonyl fluoride) at 4 °C for 20 min, and the resulting nuclear extracts were cleared by microcentrifugation (10,000g) at 4 °C for 10 min. All volumes were kept constant starting with 106 cells per 100 μl of hypotonic buffer. Fraction purity was determined by western blotting for the cytoplasmic marker β-tubulin and the nuclear marker lamin A/C.
Formalin-fixed paraffin-embedded microarrays of breast cancer tissues and normal breast tissues were from US Biomax (Rockville). Tissue samples were stained with mouse monoclonal SNAIL1 antibody (17EC3, 1:200), mouse monoclonal E-cadherin antibody (BD Bioscience, 1:500) and mouse monoclonal DDR2 antibody (R&D systems, 1:50). Staining was scored by an Hscore method that combines the values of immunoreaction intensity and the percentage of tumour cell staining as described previously41.
Results are expressed as the means ± s.d. Comparisons among groups were done with P values calculated using unpaired, two-sided Student’s t-tests. *P<0.05; **P<0.01. For each experiment point n = 3. Each experiment was repeated at least two times.
Real-time PCR with reverse transcription.
Real-time PCR reactions were done using the SYBR Green PCR Master Mix (Applied Biosystems) in the ABI detection system (Applied Biosystems). The thermal cycling conditions were composed of 50 °C for 2 min followed by an initial denaturation step at 95 °C for 20 s, 40 cycles at 95 °C for 3 s, and 60 °C for 30 s. The experiments were carried out in triplicate for each data point. The relative quantification in gene expression was determined using the 2−ΔΔCt method as described previously42. Primers used for PCR with reverse transcription and real-time PCR with reverse transcription are listed in Supplementary Table S2.
Plasmids, siRNA oligonucleotides and shRNAi lentiviruses.
pcDNA 3.1/zeo–DDR1–myc and pcDNA 3.1/zeo–DDR2–myc plasmids were provided by B. Leitinger (Imperial College, London, UK)43. pCDNA3–HA–ERK2 WT, Y261A and D319N plasmids were provided by J. Blenis (Harvard University, Boston, USA)26. Double-copy lentiviruses expressing an shRNAi that specifically targets SNAIL1 and an RNAi-resistant SNAIL1 isoform or mutant isoforms of SNAIL1 were generated as described previously28. shRNAis were from the Broad collection (Washington University Genome Center). There were 5–7 shRNAis for each target. These were all screened and 2–3 used for all experiments. All shRNAis were subcloned into the pFLRu vector28. shRNAi target sequences used are listed in Supplementary Table S3.
Kauppila, S., Stenback, F., Risteli, J., Jukkola, A. & Risteli, L. Aberrant type I and type III collagen gene expression in human breast cancer in vivo. J. Pathol. 186, 262–268 (1998).
Provenzano, P. P. et al. Collagen density promotes mammary tumor initiation and progression. BMC Med. 6, 11–26 (2008).
Paszek, M. J. et al. Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 241–254 (2005).
Conklin, M. W. et al. Aligned collagen is a prognostic signature for survival in human breast carcinoma. Am. J. Pathol. 178, 1221–1232 (2011).
Valiathan, R. R., Marco, M., Leitinger, B., Kleer, C. G. & Fridman, R. Discoidin domain receptor tyrosine kinases: new players in cancer progression. Cancer Metastasis Rev. 31, 295–321 (2012).
Leitinger, B. Transmembrane collagen receptors. Annu. Rev. Cell Dev. Biol. 27, 265–290 (2011).
Perou, C. M. Molecular stratification of triple-negative breast cancers. Oncologist 15 (Suppl. 5), 39–48 (2010).
Thiery, J. P., Acloque, H., Huang, R. Y. & Nieto, M. A. Epithelial-mesenchymal transitions in development and disease. Cell 139, 871–890 (2009).
Moody, S. E. et al. The transcriptional repressor Snail promotes mammary tumor recurrence. Cancer Cell 8, 197–209 (2005).
Yang, J. et al. Twist, a master regulator of morphogenesis, plays an essential role in tumor metastasis. Cell 117, 927–939 (2004).
Drasin, D. J., Robin, T. P. & Ford, H. L. Breast cancer epithelial-to-mesenchymal transition: examining the functional consequences of plasticity. Breast Cancer Res. 13, 226–239 (2011).
Tran, D. D., Corsa, C. A., Biswas, H., Aft, R. L. & Longmore, G. D. Temporal and spatial cooperation of Snail1 and Twist1 during epithelial-mesenchymal transition predicts for human breast cancer recurrence. Mol. Cancer Res. 9, 1644–1657 (2011).
Franci, C. et al. Expression of Snail protein in tumor–stroma interface. Oncogene 25, 5134–5144 (2006).
Zhou, B. P. et al. Dual regulation of Snail by GSK- 3β-mediated phosphorylation in control of epithelial-mesenchymal transition. Nat. Cell Biol. 6, 931–940 (2004).
Yook, J. I. et al. A Wnt-Axin2-GSK3β cascade regulates Snail1 activity in breast cancer cells. Nat. Cell Biol. 8, 1398–1406 (2006).
Zhang, K. et al. Lats2 kinase potentiates Snail1 activity by promoting nuclear retention upon phosphorylation. EMBO J. 31, 29–43 (2011).
Egeblad, M., Rasch, M. G. & Weaver, V. M. Dynamic interplay between the collagen scaffold and tumor evolution. Curr. Opin. Cell Biol. 22, 697–706 (2010).
Schedin, P. & Keely, P. J. Mammary gland ECM remodeling, stiffness, and mechanosignaling in normal development and tumor progression. Cold Spring Harbor Perspect. Biol. 3, a003228 (2011).
Walker, R. A. The complexities of breast cancer desmoplasia. Breast Cancer Res. 3, 143–145 (2001).
Kreike, B. et al. Gene expression profiling and histopathological characterization of triple-negative/basal-like breast carcinomas. Breast Cancer Res. 9, R65 (2007).
Provenzano, P. P. et al. Collagen reorganization at the tumor–stromal interface facilitates local invasion. BMC Med. 4, 38–54 (2006).
Ferri, N., Carragher, N. O. & Raines, E. W. Role of discoidin domain receptors 1 and 2 in human smooth muscle cell-mediated collagen remodeling: potential implications in atherosclerosis and lymphangioleiomyomatosis. Am. J. Pathol. 164, 1575–1585 (2004).
Flynn, L. A., Blissett, A. R., Calomeni, E. P. & Agarwal, G. Inhibition of collagen fibrillogenesis by cells expressing soluble extracellular domains of DDR1 and DDR2. J. Mol. Biol. 395, 533–543 (2010).
Sivakumar, L. & Agarwal, G. The influence of discoidin domain receptor 2 on the persistence length of collagen type I fibres. Biomaterials 31, 4802–4808 (2010).
MacPherson, M. R. et al. Phosphorylation of serine 11 and serine 92 as new positive regulators of human Snail1 function: potential involvement of casein kinase-2 and the cAMP-activated kinase protein kinase A. Mol. Biol. Cell 21, 244–253 (2010).
Shin, S., Dimitri, C. A., Yoon, S. O., Dowdle, W. & Blenis, J. ERK2 but not ERK1 induces epithelial-to-mesenchymal transformation via DEF motif-dependent signaling events. Mol. Cell 38, 114–127 (2010).
Jacobs, D., Glossip, D., Xing, H., Muslin, A. J. & Kornfeld, K. Multiple docking sites on substrate proteins form a modular system that mediates recognition by ERK MAP kinase. Genes Dev. 13, 163–175 (1999).
Feng, Y. et al. A multifunctional lentiviral-based gene knockdown with concurrent rescue that controls for off-target effects of RNAi. Genom. Proteom. Bioinform. 8, 238–245 (2010).
Ikeda, K. et al. Discoidin domain receptor 2 interacts with Src and Shc following its activation by type I collagen. J. Biol. Chem. 277, 19206–19212 (2002).
Ding, Q. et al. Erk associates with and primes GSK- 3β for its inactivation resulting in upregulation of β-catenin. Mol. Cell 19, 159–170 (2005).
Desbois-Mouthon, C. et al. Insulin and IGF-1 stimulate the β-catenin pathway through two signalling cascades involving GSK- 3β inhibition and Ras activation. Oncogene 20, 252–259 (2001).
Ota, I., Li, X. Y., Hu, Y. & Weiss, S. J. Induction of a MT1-MMP and MT2-MMP-dependent basement membrane transmigration program in cancer cells by Snail1. Proc. Natl Acad. Sci. USA 106, 20318–20323 (2009).
Rowe, R. G. et al. Mesenchymal cells reactivate Snail1 expression to drive three-dimensional invasion programs. J. Cell Biol. 184, 399–408 (2009).
Cerami, E. et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2, 401–404 (2012).
Provenzano, P. P., Inman, D. R., Eliceiri, K. W. & Keely, P. J. Matrix density-induced mechanoregulation of breast cell phenotype, signaling and gene expression through a FAK-ERK linkage. Oncogene 28, 4326–4343 (2009).
White, D. E. et al. Targeted disruption of β1 integrin in a transgenic mouse model of human breast cancer reveals an essential role in mammary tumor induction. Cancer Cell 6, 159–170 (2004).
Mariotti, A. et al. EGF-R signaling through Fyn kinase disrupts the function of integrin α6β4 at hemidesmosomes: role in epithelial cell migration and carcinoma invasion. J. Cell Biol. 155, 447–458 (2001).
Soung, Y. H., Clifford, J. L. & Chung, J. Crosstalk between integrin and receptor tyrosine kinase signaling in breast carcinoma progression. BMB Rep. 43, 311–318 (2010).
Aiello, E. J., Buist, D. S., White, E. & Porter, P. L. Association between mammographic breast density and breast cancer tumor characteristics. Cancer Epid. Biom. Prev. 14, 662–668 (2005).
McDaniel, S. M. et al. Remodeling of the mammary microenvironment after lactation promotes breast tumor cell metastasis. Am. J. Pathol. 168, 608–620 (2006).
Sharp, T. V. et al. The chromosome 3p21.3-encoded gene, LIMD1, is a critical tumor suppressor involved in human lung cancer development. Proc. Natl Acad. Sci. USA 105, 19932–19937 (2008).
Livak, K. J. & Schmittgen, T. D. Analysis of relative gene expression data using real-time quantitative PCR and the 2(−ΔΔC(T)) method. Methods 25, 402–408 (2001).
Noordeen, N. A., Carafoli, F., Hohenester, E., Horton, M. A. & Leitinger, B. A transmembrane leucine zipper is required for activation of the dimeric receptor tyrosine kinase DDR1. J. Biol. Chem. 281, 22744–22751 (2006).
This work was supported by NIH grant P50CA94056 to the Imaging Core of the Siteman Cancer Center at Washington University, and NIH grants GM080673 and CA143868, and Susan G. Komen for the Cure KG110889 to G.D.L. C.A.C. was supported by NIH grant F31CA165729.
The authors declare no competing financial interests.
Supplementary Information (PDF 4260 kb)
Supplementary Information (XLSX 12 kb)
Supplementary Information (XLSX 12 kb)
Supplementary Information (XLSX 26 kb)
Supplementary Information (XLSX 19 kb)
About this article
Cite this article
Zhang, K., Corsa, C., Ponik, S. et al. The collagen receptor discoidin domain receptor 2 stabilizes SNAIL1 to facilitate breast cancer metastasis. Nat Cell Biol 15, 677–687 (2013). https://doi.org/10.1038/ncb2743
Medicinal Chemistry Research (2021)
Comprehensive Proteomic Characterization Reveals Subclass-Specific Molecular Aberrations within Triple-negative Breast Cancer
Discoidin Domain Receptors in Melanoma: Potential Therapeutic Targets to Overcome MAPK Inhibitor Resistance
Frontiers in Oncology (2020)
Cancer cells grown in 3D under fluid flow exhibit an aggressive phenotype and reduced responsiveness to the anti-cancer treatment doxorubicin
Scientific Reports (2020)
Matrix Biology (2020)