Heat shock protein 47 (HSP47) binds to discoidin domain–containing receptor 2 (DDR2) and regulates its protein stability

Cell–collagen interactions are crucial for cell migration and invasion during cancer development and progression. Heat shock protein 47 (HSP47) is an endoplasmic reticulum–resident molecular chaperone that facilitates collagen maturation and deposition. It has been previously shown that HSP47 expression in cancer cells is crucial for cancer invasiveness. However, exogenous collagen cannot rescue cell invasion in HSP47-silenced cancer cells, suggesting that other HSP47 targets contribute to cancer cell invasion. Here, we show that HSP47 expression is required for the stability and cell-surface expression of discoidin domain–containing receptor 2 (DDR2) in breast cancer tissues. HSP47 silencing reduced DDR2 protein stability, accompanied by suppressed cell migration and invasion. Co-immunoprecipitation results revealed that HSP47 binds to the DDR2 ectodomain. Using a photoconvertible technique and total internal reflection fluorescence microscopy, we further demonstrate that HSP47 expression significantly sustains the membrane localization of the DDR2 protein. These results suggest that binding of HSP47 to DDR2 increases DDR2 stability and regulates its membrane dynamics and thereby enhances cancer cell migration and invasion. Given that DDR2 has a crucial role in the epithelial-to-mesenchymal transition and cancer progression, targeting the HSP47–DDR2 interaction might be a potential strategy for inhibiting DDR2-dependent cancer progression.

Adhesion of cells to the extracellular matrix (ECM) 4 is essential for cell migration, proliferation, and differentiation (1). Collagen is the most abundant ECM protein, and 28 different types of collagen have been characterized in mammals. The basic unit of collagen is the triple-helical structure, which is formed by the repeating motif Gly-Pro-Xaa (2). HSP47 was identified as a collagen-specific chaperone that regulates maturation of collagen molecules. Binding of Hsp47 to the triple helix region inhibits the aggregation of collagen in the endoplasmic reticulum (ER), thereby facilitating its secretion and deposition (3).
Fibroblasts are considered the major source of collagen in stroma. Consistently, HSP47, encoded by serpin family H member 1 (SERPINH1) gene, is highly expressed in fibroblasts (4 -6). Interestingly, we and others show that HSP47 expression in cancer cells is up-regulated during cancer development and progression (7)(8)(9)(10). HSP47 silencing suppresses cancer cell migration and invasion in vitro and in vivo. However, the addition of collagen cannot fully rescue cell migration and invasion in HSP47-silenced cells (7). HSP47 is a heat shock protein that is produced when cells are exposed to stressful conditions (11)(12)(13). These results suggest that HSP47 has additional targets during cancer progression and in response to stressful conditions. Identification of HSP47 targets may advance our understanding how HSP47 expression promotes cell migration and provide new insights into function of HSP47 in cancer progression.
Cell-collagen interactions are mainly mediated by integrins and discoidin domain-containing receptors (DDRs), and dysregulation of the collagen-receptor interaction contributes to cancer development and progression (14,15). DDRs, including DDR1 and DDR2, are receptor tyrosine kinases. DDRs consist of an extracellular discoidin domain, a transmembrane region, a cytoplasmic juxtamembrane region, and a catalytic domain (16). Binding of collagen to the ectodomain induces DDR phosphorylation in the cytoplasmic domain and activation of downstream signaling (17). A 3-5% incidence of DDR2 point mutations has been detected in lung squamous cell carcinoma, suggesting that the DDR2 gene is a potential oncogenic target (14). Although DDR2 mutation has not been detected in breast cancer, roles of DDR2 in breast cancer metastasis are well-established. It has been shown that DDR2 expression is induced during epithelial-mesenchymal transition (EMT) (15,18,19). Enhanced DDR2 activities in the EMT cells stabilize Snail protein by stimulating ERK2 activity and subsequently promote cancer cell invasion and colonization at distant organs (15). Therefore, determining how DDR2 protein expression is regu-lated may identify potential strategies to target this oncogene and suppress cancer progression.
In the present study, we have identified DDR2 as a new target of HSP47. Binding of HSP47 is crucial for the stability and cellsurface localization of DDR2 protein. We also show that introducing exogenous DDR2 rescues cell migration and invasion in HSP47-silenced cells. These results suggest that the function of HSP47 in cancer progression is at least partially mediated by DDR2.

HSP47 regulates DDR2 expression at the protein level
Our previous study showed that HSP47 expression is induced during breast cancer development and progression. Silencing HSP47 in breast cancer cells suppressed cancer cell migration and invasion (7). However, the molecular mechanism by which HSP47 regulates cell migration and invasion remains to be determined. Using gene co-expression analysis, we showed that mRNA levels of HSP47 and DDR2 were significantly correlated in human breast cancer tissues (Fig. 1A).
The correlated gene expression suggests a functional connection between HSP47 and DDR2; therefore, we examined DDR2 expression in HSP47-silenced breast cancer cells. We found that DDR2 protein levels were reduced in HSP47-silenced MDA-MB-231 cells (Fig. 1, B and C), which is associated with reduced cell migration and invasion (7). In addition, knockdown of HSP47 had little effect on protein expression of ␤1-integrin and epidermal growth factor receptor. To assess whether DDR2 activity is regulated by HSP47, we analyzed DDR2 phosphorylation in control and HSP47-silenced cells using immunoprecipitation experiments. Total protein DDR2 levels were reduced in HSP47-silenced cells; however, silence of HSP47 had little effect on the relative levels of tyrosine phosphorylated DDR2 (Fig. 1D). It has been shown that DDR2 activation in breast cancer cells promotes cell migration and invasion (18,20,21). These results suggest that DDR2 is the potential HSP47 target that mediates its function in regulating cell migration and invasion.

HSP47 binds to DDR2 and enhances its protein stability
To understand how HSP47 regulates DDR2 expression, we first determined whether DDR2 is regulated by HSP47 at the transcription level. Quantitative RT-PCR data showed that silencing HSP47 had little effect on DDR2 mRNA levels ( Fig.  2A). Next, we examined whether the protein stability of DDR2 is regulated by HSP47. Control and HSP47-silenced MDA-MB-231 cells were treated with cycloheximide, and DDR2 protein levels were assessed by Western blotting at different time points. We found that DDR2 protein levels decreased much faster in HSP47-silenced cells compared with the levels in control cells (Fig. 2B). These results indicate that silencing HSP47 reduced DDR2 protein stability, which may lead to the reduction in DDR2 protein levels.
HSP47 is a molecular chaperone that facilitates protein maturation through protein-protein interaction (22,23). We asked whether HSP47 binds to DDR2 and enhances its stability. An expression construct containing FLAG-tagged HSP47 gene was introduced in HEK 293 cells, and protein complexes were immunoprecipitated with anti-FLAG M2 beads. The DDR2-HSP47 protein complex was detected in the IP products (Fig.  3A). In another experiment, the protein complexes were immunoprecipitated from cells expressing FLAG-DDR2, and binding of HSP47 to DDR2 was also detected in the complexes (Fig.  3B). To confirm the DDR2 and HSP47 interaction in breast cancer cells, we performed co-IP experiments using endoge-

HSP47 regulates DDR2 stability
nous protein extracted from HS578 cells. DDR2 was detected in the protein complexes immunoprecipitated by the antibody against HSP47, but not in the IgG control group (Fig. 3C). In addition, confocal microscope imaging analysis showed that DDR2 protein was mainly localized at cell surface and in cytoplasm. The co-localization of HSP47 and DDR2 was mainly detected in the cytoplasm (Fig. 3D).
DDR2 is composed of extracellular discoidin homology (DS) and DS-like domain and an intracellular domain (Fig. 3E). To determine which domain is involved in the DDR2-HSP47 interaction, we generated constructs containing the DDR2 extracellular region, DS domain, DS-like domain, or the intracellular domain. Results from co-IP experiments showed that HSP47 bound to DS domain and DS-like domain but had no detectable interaction with the intracellular region (Fig. 3F). These results suggest that binding of HSP47 to DS domain and DS-like domain contributes to the DDR2 protein maturation and stabilization.

Dynamics of membrane DDR2 are regulated by HSP47
DDR2 is synthesized in the ER and traffics to the cell surface as a collagen receptor (24). We asked whether binding of HSP47 regulates DDR2 translocation and dynamic at cell surface. The photoconvertible fluorescent protein, Dendra2, has recently been used to track protein translocation and dynamics (25). Dendra2 is synthesized as a fluorescent protein, with excitation at 488 nm. Exposure to 405 nm light can photoconvert Dendra2 from green to red emission. Combining this technique with total internal reflection fluorescence microscopy (TIRF), we can monitor the translocation and dynamics of Dendra2 fusion proteins located on the plasma membrane (Fig. 4A).
An expression vector containing DDR2-Dendra2 fusion gene was generated, and the photoconversion of the fused protein was tested in HEK 293 cells. Exposure of the fusion protein-expressing HEK293 cells to 405-nm light successfully converted the fluorescence from green to red (Fig. 4B). To A, co-IP analysis was performed in HEK293 cells transfected with DDR2 and/or FLAG-HSP47 expression constructs. Protein complexes were pulled down with M2 beads and analyzed by Western blotting. B, co-IP analysis was performed in HEK293 expressing FLAG-DDR2 and/or HSP47. C, endogenous co-IP was performed in HS578 cells. Protein complexes were pulled down with HSP47 antibody and protein A/G beads. D, confocal microscopy imaging analyzing localization of HSP47 (green) and DDR2 (red) in HEK 293 cells. Nuclei was staining with (DAPI; blue). E, the cartoon showing the DDR2 protein structure. F, co-IP analysis was performed in HEK 293 cells expressing FLAG-HSP47 and truncated DDR2 (EX-domain and In-Domain). IB, immunoblotting.

HSP47 regulates DDR2 stability
determine whether HSP47 regulates the translocation of DDR2 to cell surface, the DDR2-Dendra2 construct was transfected in control and HSP47-silenced epithelial cells. DDR2-Dendra2 on the cell surface was selectively converted using a 405-nm laser in a TIRF configuration. Red emission was monitored overtime in TIRF. A decay in fluorescence corresponds to endocytosis of DDR2 moving it out of the TIRF focal volume. TIRF imaging analysis showed that silencing HSP47 did not block translocation of DDR2 to the cell surface (Fig. 4C). Next, we monitored the dynamics or stability of DDR2-Dendra2 fusion protein on cell surface after photoconversion by quantifying the red florescence intensity with TIRF. We found that the half-life of the fusion protein at the cell surface was significantly shorter in HSP47-silenced cells compared with that in control cells (Fig. 4D). These results suggest that HSP47 regulates cell membrane dynamics of DDR2 protein.

DDR2 mediates function of HSP47 in enhancing cancer cell invasion
It has been shown that DDR2 expression promotes EMT and enhances cell migration and invasion (15). To determine whether HSP47 promotes cancer cell migration and invasive phenotypes through DDR2, we have introduced exogenous DDR2 in HSP47-silenced cells. By tracking single-cell movement with live cell imaging, we showed that silencing HSP47 in MDA-MB-231 cells significantly inhibited cell migration as previously described (Fig. 5, A and B). Overexpression of DDR2 in HSP47-silenced cells at least partially rescued cell migration (Fig. 5, A and B).
3D tissue culture assay provides a physiologically relevant microenvironment to investigate cell-ECM interaction and cancer progression. MDA-MB-231 cells forms invasive stellate structures in 3D culture (26,27). Knockdown of HSP47 significantly reduced the invasive branching, and expression of DDR2 partially restored the invasive phenotype of MDA-MB-231 cells in 3D culture (Fig. 6, A and B). We further confirmed that DDR2 expression rescued invasion in HSP47-silenced MDA-MB-231 cells using the Transwell assay (Fig. 6, C and D). These results indicate that HSP47 promotes cancer cell migration and invasion at least partially through enhancing DDR2 protein stability.

Discussion
HSP47 was considered a collagen-specific chaperone; however, IRE1a, decorin, and lumican have recently been identified as HSP47 targets (28,29). Binding of HSP47 to decorin and lumican is crucial for protein maturation and secretion (28). HSP47 directly binds to the ER luminal domain of IRE1a, and this interaction is crucial for the negative regulation of ER stress (29). Here, we identified DDR2 as a new HSP47 target that mediates HSP47 function in cell migration and invasion.
We found that the presence of HSP47 enhanced the DDR2 protein stability. The co-IP data showed that HSP47 bound to the extracellular region of DDR2. Photoconversion and TIRF imaging analyses showed that binding of HSP47 was not required for the cell surface translocation of DDR2 protein; however, HSP47 expression significantly enhanced the half-life of DDR2 at cell surface. Interestingly, confocal images showed that HSP47 and DDR2 were mainly co-localized in cytoplasm but not at the cell surface. DDR2 is synthesized and matures in the ER, and the ectodomain locates in the ER lumen during synthesis. HSP47 mainly localizes in the ER; therefore, we predict that HSP47 binds to DDR2 in the ER during DDR2 synthesis and maturation.

HSP47 regulates DDR2 stability
Protein stability is regulated by post-translational modification, such as glycosylation, hydroxylation, and phosphorylation (30 -32). Silencing HSP47 induced a band shift of DDR2 in the Western blotting analysis (Fig. S1), suggesting that HSP47 regulates DDR2 post-translational modification. N-Glycosylation has been characterized in DDR2 (33). We have some data showing that knockdown of HSP47 reduced DDR2 glycosylation (Fig. S2). Therefore, HSP47 may regulate DDR2 stability by facilitating its glycosylation.
It has been shown that expression of DDR2 is induced during EMT (34,35). EMT is characterized by enhanced cell migration and invasion. Aberrant activation of EMT promotes metastatic spread in many solid tumors (36). The collagen/DDR2 axis plays crucial roles in regulating EMT and promoting breast cancer metastasis (15,37,38). Activation of DDR2 increases the level and activity of Snail in breast cancer cells and subsequently promotes cell invasion and cancer metastasis (15). We also detected increased expression of HSP47 during EMT. HSP47 promotes EMT phenotypes in breast cancer cells, such as cell migration and invasion (7). DDR2 overexpression at least partially rescued cell migration and invasion in HSP47-silenced cancer cells, suggesting that HSP47 contributes to the EMT process by enhancing collagen/DDR2 signaling. DDR2 has been identified as an oncogene (14); however, the regulation of DDR2 during breast cancer progression largely remains to be determined. We found that expression of HSP47 and DDR2 are associated during breast cancer development and progression. Importantly, HSP47 binds to DDR2 and enhances its protein stability. Given the crucial function of DDR2 in EMT and cancer progression, targeting the HSP47-DDR2 interaction may be a potential strategy to inhibit DDR2dependent cancer progression.

Antibodies and reagents
Dulbecco's modified Eagle's medium (DMEM) and DMEM/ Ham's F-12 medium were obtained from Sigma-Aldrich. Fetal

Immunoprecipitation assay
293FT cells were transfected with DDR2 and HSP47 expression plasmids using FuGENE HD transfection reagent (Promega). Following 48 h of transfection, the cells were washed with PBS then lysed with ice-cold hypotonic gentle lysis buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 2 mM EDTA, 0.5% Triton X-100, Protease Inhibitor Mixture Set 1 (Calbiochem)) and centrifuged at 4°C for 15min. The cell lysates were incubated with 20 l of anti-FLAG M2 affinity gel (Sigma) for 4 h at 4°C. The protein complexes were eluted with 3ϫ FLAG peptide and analyzed by Western blotting. Each IP experiment was repeated at least twice.

Western blotting analysis
The cells were lysed with 2% SDS in PBS buffer containing phosphatase and protease inhibitor cocktails. Protein concentration was measured using DC TM protein assay (Bio-Rad). Equal amounts of protein lysates were subjected to SDS gel electrophoresis, immunoblotted, and detected with an ECL system (Pierce). Exposures were acquired and quantified using a FluroChem HD2 (Alpha Innotech).

Quantitative RT-PCR
Total RNA was extracted from cells using TRIzol reagent (Invitrogen). cDNA synthesis was performed with SuperScript III first-strand synthesis system (Invitrogen) according to the protocol. Quantitative RT-PCRs were carried out using SYBR Green PCR master mix reagents on an ABI 7500 Fast real-time PCR system (Applied Biosystems). Thermal cycling was conducted at 95°C for 30 s, followed by 40 cycles of amplification at 95°C for 5 s, 55°C for 30 s, and 72°C for 15 s. The following primers were used to amplify: DDR2 forward, 5Ј-CCAGTCA-GTGGTCAGAGTCCA-3Ј; and DDR2 reverse, 5Ј-GGGTCCC-CACCAGAGTGATAA-3Ј. The quantitative RT-PCR experiments were performed three times. DDR2 Ct value was subtracted by corresponding 18s value, which produced the ⌬Ct of DDR2. The ⌬Ct value of each group was subtracted by ⌬Ct value of shctrl to generate the ⌬⌬Ct value.

Single-cell migration assay
Control and HSP47-silenced MDA-MB-231 cells (0.04 ϫ 10 6 ) with or without DDR2 overexpression were plated on type I collagen precoated 35-mm dishes in DMEM/Ham's F-12 medium containing 2% fetal bovine serum and 4 ng/ml epider-mal growth factor. After 2 h of incubation at 37°C, the images were taken by live cell/incubator imaging system (Nikon Biostation IMQ) every 10 min for 8 h. The cell migration distance and velocity were quantified as previously described (27).

Invasion assay
The Transwell (Corning) were coated with 60 l of 1 mg/ml Matrigel and incubated for 30 min at 37°C. The cells were plated in 200 l on top of the Transwell filter and incubated in 37°C 5% CO 2 for 24 h. The invaded cells on the bottom face of the filter were fixed by methanol and stained with 8% crystal violet.

Stability assay
Control and HSP47-silenced MDA-MB-231 cells were plated with a density of 2 ϫ 10 6 cells. The cells were treated with cycloheximide (100 g/ml) for 0, 2, 4, and 8 h. The protein levels of DDR2 were analyzed with Western blotting (39).

Total internal reflection fluorescence microscopy
Control and HSP47-silenced HEK293 cells were transfected with the DDR2-Dendra2 expression plasmid and then were plated on a 35-mm Matrigel-coated glass-bottomed dish for 24 h. Before imaging, the growth medium was replaced with Leibovitz's L-15 imaging medium. TIRF images of selected cells were taken before Dendra2 photoconversion using both 561-nm and 488-nm excitation. This verified the absence of any fluorescence in the red emission channel prior to photoconversion. These cells were then exposed to TIRF-oriented 405-nm laser (ϳ16.5 milliwatts at the objective) excitation for 3 s. Realtime TIRF images were then acquired for both 561-nm and 488-nm excitation to collect emission in both the red and green channels at 10 min intervals for a total of 7 h. Image collection was initiated 20 min after photoconversion to allow for the clearance of endosomes. The cells were kept at 37°C for the duration of the experiment using a stage top incubator to avoid any temperature induced changes in endocytosis rates.

Statistical analysis
The correlation expression of HSP47 and DDR2 was assessed with Spearman correlation analysis using the microarray data set generated from human breast cancer tissues (TCGA). All experiments were repeated at least twice. Results were reported as means Ϯ S.E.; significance of difference was assessed by an independent Student t test or one-way ANOVA test. A p value of 0.05 represented statistical significance, and a p value of 0.01 represented high statistical significance.