Calreticulin promotes EGF-induced EMT in pancreatic cancer cells via Integrin/EGFR-ERK/MAPK signaling pathway

Our previous study showed that Calreticulin (CRT) promoted the development of pancreatic cancer (PC) through ERK/MAPK pathway. We next investigate whether CRT promotes EGF-induced epithelial–mesenchymal transition (EMT) in PC via Integrin/EGFR-ERK/MAPK signaling, which has not been reported yet to our knowledge. EGF simultaneously induced EMT and activated Integrin/EGFR–ERK/MAPK signaling pathway in 3 PC cells. However, CRT silencing significantly inhibited EGF function, including inhibiting EGF-induced EMT-like cell morphology, EGF-enhanced cell invasion and migration, and EGF induced the decrease of E-cadherin, ZO-1, and β-catenin and the increase of the key proteins in Integrin/EGFR-ERK/MAPK signaling (pEGFR-tyr1173, Fibronectin, Integrinβ1, c-Myc and pERK). Conversely, CRT overexpression rescued the change of EMT-related proteins induced by EGF in CRT silencing PC cells. Additionally, CRT was co-stained with pEGFR1173 (with EGF), Fibronectin and Integrinβ1 by IF under confocal microscopy and was co-immunoprecipitated with Fibronectin, Integrinβ1 and c-Myc in both PC cells, all of which indicating a close interaction of CRT with Integrin/EGFR–ERK/MAPK signaling pathway in PC. In vivo, CRT silencing inhibited subcutaneous tumor growth and liver metastasis of pancreatic tumor. A positive relationship of CRT with Fibronectin, Integrinβ1, c-Myc and pERK and a negative association of CRT with E-cad was also observed in vivo and clinical samples. Meanwhile, overexpression of the above proteins was closely associated with multiple aggressive clinicopathological characteristics and the poor prognosis of PC patients. CRT promotes EGF-induced EMT in PC cells via Integrin/EGFR-ERK/MAPK signaling pathway, which would be a promising therapy target for PC.

predominant cytoplasmic CRT expression was also observed in clinical tissues by immunohistochemistry (IHC; Figure 10). All of the above indicated predominantly intracellular functions of CRT in PC development. Our previous study showed EGF was much more reliable to induce EMT in AsPC-1, BxPC-3 and Capan-2 cell lines. 3 Thus above three PC cell lines with relative CRT high expression were used to construct CRT-silencing stable cells via CRISPR/Cas9 system. Western blotting (WB) verified that CRT protein level in Capan-2, AsPC-1 and BxPC-3 cells in the sg1-CRT and sg2-CRT groups were significantly lower than that in the corresponding scramble groups (Figures 1b-d).
CRT silencing inhibited EGF-induced EMT-like cell morphology, EGF-enhanced cell invasion and migration and EGF-induced the change of EMT-related proteins in two PC cell lines. EGF-induced EMT is well studied in our and other previous studies. 3,23 Thus we intended to investigate the relationship of CRT with EGF-induced EMT in PC. Upon EGF induction, both Capan-2 and AsPC-1 cells exhibited EMT-like cell morphology: most cells lost their epithelial characteristics and presented a spindle-shaped and fibroblast-like morphology ( Figure 2). However, CRT silencing significantly reversed EGF-induced EMT-like cell morphology. Most EGF-treated Capan-2 and AsPC-1 cells restored original cell morphology in the sg1-CRT and sg2-CRT groups. No significant spindle-shaped and fibroblast-like morphology was observed in the sg1-CRT and sg2-CRT groups compared with the corresponding scramble groups ( Figure 2). The same results were also observed in BxPC-1 cells (Supplementary Figure 1a).
Meanwhile, EGF significantly induced the decrease of E-cad, ZO-1 and β-catenin and the increase of Fibronectin, MMP9, Vimentin and a-SMA protein expression (except for N-cad) in both cell lines. However, CRT silencing partially inhibited EGF-induced change of EMT-related proteins, including inhibiting EGF-induced decrease of E-cad, ZO-1 and β-catenin and increase of Fibronectin ( Figure 3). The same results were also observed in BxPC-1 cells (Supplementary Figure 1b).
It is well known that EMT has a significant role in cell invasion and metastasis of PC. 24 In current study, EGF significantly enhanced cell invasion and migration in both Capan-2 and AsPC-1 cells ( Figure 4). However, CRT silencing inhibited EGF-enhanced cell invasion and migration ( Figure 4). Upon EGF induction, a significant increase of cell invasion and migration were found in the scramble groups compared with the corresponding sg1-CRT1 and sg2-CRT groups. The difference of cell invasion and migration in the scramble groups with or without EGF treatment was much more significant than that in the sg1-CRT and sg2-CRT groups ( Figure 4). The same results were also observed in BxPC-1 cells (Supplementary Figure 1c).
All together, CRT silencing inhibited EGF-induced EMT in PC cells.
Overexpression of CRT rescued the change of EMTrelated proteins induced by EGF in CRT-silencing Capan-2 cells. sg1-CRT, sg2-CRT and scramble-infected Capan-2 and AsPC-1 cell lines were transiently transfected with CRT-GFP and GFP plasmids, respectively. Because Capan-2 cells showed much high transfection efficiency compared with AsPC-1 cells, we only used Capan-2 cells for further rescue experiment (Figures 5a and b).
We found that the expression of EMT biomarkers (Fibronectin, E-cad, ZO-1 and β-catenin) was rescued by CRT overexpression in sg1-CRT and sg2-CRT-infected Capan-2 cells under EGF treatment (Figure 5c). In detail, the downregulation of E-cad, ZO-1 and β-catenin and the upregulation of Fibronectin all induced by EGF in the sg1-CRT+GFP and Black bars: CRT protein expression in the sg1-CRT and sg2-CRT groups. **Po0.01 compared with the control sg2-CRT+GFP groups was rescued by CRToverexpression in the sg1-CRT+CRT-GFP and sg2-CRT+CRT-GFP groups ( Figure 5c). It was well known that membrane-bound CRT is a danger/eat-me signal and important for immunogenic cell death. 25,26 However, in the current study, AsPC-1 and Capan-2 cells with CRT-GFP exhibited predominantly cytoplasmic green fluorescence in a perinuclear pattern (but not only on cell surface), which is consistent with the studies by Bibi et al. 27 and by Ihara et al. 28 That might be why high expression of predominantly intracellular CRTexecuted a pro-tumor role in the development of PC.
CRT silencing inhibited EGF-activated Integrin/EGFR-ERK/MAPK signaling in two PC cell lines. EGF significantly activated Integrin/EGFR-ERK/MAPK signaling in both Capan-2 and AsPC-1 cells followed by the increase of pEGFR1173, pEGFR1068, pEGFR845, Integrinβ1, Integ-rinα5, pERK and c-Myc protein expression (Figures 3 and  6). Without EGF, CRT silencing alone did not influence all phosphorylation EGFR expression. However, its silencing significantly inhibited EGF-induced increase of pEGFR1173 but had no effect on EGF-induced pEGFR1068 and pEGFR845 protein levels. Meanwhile, pEGFR1173 was hard to detect and showed no co-staining with CRT by IF without EGF treatment, whereas pEGFR1173 expression was enhanced in cytoplasm and was co-stained with CRT in predominant cytoplasm of Capan-2 and AsPC-1 cells induced by 50 ng/ml of EGF (Figures 6c and d). All the above results implicated a specific interaction of CRT and EGF/EGFR signal pathway.
In addition, CRT silencing alone partially downregulated the key proteins of Integrin/EGFR-ERK/MAPK signaling, including Fibronectin, Integrinβ1, c-Myc and pERK (except for Integrinα5 and caveolin-1) ( Figure 3). Meanwhile, both CRT/ Fibronectin and CRT/Integrinβ1 (except for CRT/E-cad) were co-stained in predominant cytoplasm in normal Capan-2 and AsPC-1 cells (Figure 7). Moreover, CRT was coimmunoprecipitated with Fibronectin, Integrinβ1 and c-Myc in the lysates of normal Capan-2 and AsPC-1 cells (Figure 8). Upon EGF induction, a significant decrease of Fibronectin, Integrinβ1, pERK and c-Myc were found in the sgCRT1 and sgCRT2 groups compared with the corresponding scramble   Figure 3). Taking together, a close interaction between CRT and EGF-induced Integrin/EGFR-ERK/MAPK signaling pathway was observed in PC cells.
Alteration of intracellular Ca 2+ mediated by CRT might be involved in regulating EGF-induced EMT and Integrin/ EGFR-ERK/MAPK signaling pathway. It is well known that the intracellular Ca 2+ acting as a second messenger is essential for appropriate cellular functioning, including regulating gene transcription, cell proliferation and migration. Using Fluo-3 to detect intracellular Ca 2+ under conforcal microscopy, we showed that EGF (50 ng/ml) significantly enhanced intracellular Ca 2+ just as the function of ionomycin with 100 nmol (a Ca 2+ ionophore that increased intracellular Ca 2+ concentrations) in Capan-2 cells (Supplementary Figure 2). Without any stimulus, intracellular Ca 2+ was partially decreased by CRT silencing, and this trend was much significant under EGF or ionomycin treatment (Supplementary Figure 2). It indicates that the increase of intracellular Ca 2+ by EGF and ionomycin is finally regulated by CRT. We further showed that Integrinβ1, Fibronectin and c-Myc proteins were significantly activated by ionomycin in PC cells just as the function of EGF (Supplementary Figure 3). Taking together, we infer that alteration of intracellular Ca 2+ mediated by CRT has a significant role in EGF-induced EMT and Integrin/EGFR-ERK/MAPK signaling pathway, which will be further investigated in our future study.
CRT silencing inhibited subcutaneous tumor formation and distant liver metastasis of pancreatic tumor in nude mice. Capan-2 (derived from primary PC tissue) and AsPC-1 (derived from metastatic ascites) cells were used to construct subcutaneous tumor and distal liver metastasis models, respectively. Tumor volumes in nude mice transplanted with sg-CRT-infected Capan-2 cells were much smaller than that in paired corresponding scramble groups (P = 0.043) ( Figure 9a). The primary tumors were diagnosed under hematoxylin and eosin (HE) staining ( Figure 9b). Meanwhile, IHC further verified that CRT, Fibronectin, Integrinβ1, c-Myc and pERK expression in the sg-CRT groups were much lower than that in the scramble groups (Po0.001; Po0.01; Po0.01; Po0.05 and Po0.05, respectively), whereas E-cad expression in the sg-CRT groups were much higher than that in the scramble groups using paired sample non-parametric test ( Figure 9e). However, Vimentin expression showed no significant difference between the sg-CRT and scramble groups. All above results were consistent with the results in vitro.
The average number of liver metastases in nude mice transplanted with sgCRT-infected AsPC-1 cells was much lower than that in the scramble groups (P = 0.014) (Figure 9c). The distant liver metastases were diagnosed by HE staining (Figure 9d).
Clinical significance and relationship of CRT with Fibronectin, Integrinβ1, c-Myc, pERK, E-cad and Vimentin expression levels in PC. Finally, we investigate the close relationship between CRT, Fibronectin, Integrinβ1, c-Myc and pERK and two classic EMT markers (E-cad and Vimentin) with the clinical significance of PC patients.
The location of CRT in cytoplasm, Fibronectin and Integrinβ1 in membrane and cytoplasm and c-Myc and pERK in cytoplasm and nuclei were considered for scoring, while the membrane location of E-cad was identified as normal expression (E-cad-negative and cytoplasmic expression were considered as abnormal expression; Figure 10). In 68 cases of PC tissues, proteins were overexpressed in 63.2% ( (Table 1).
Spearman correlation tests showed that CRT was positively associated with Fibronectin, Integrinβ1, c-Myc and pERK expression (P = 0.006; P = 0.020; P = 0.008 and P = 0.036, respectively) and negatively associated with normal E-cad expression (P = 0.022), but had no association with Vimentin. Using serial sections, PC tissues with high CRT expression was associated with positive Fibronectin, Integrinβ1, c-Myc, pERK and abnormal E-cad expression levels ( Figure 10b) and vice versa ( Figure 10c). However, no relationship was found between CRT and Vimentin, although their co-expression was shown in most of PC tissues. All above results were consistent with our observation in vitro and in vivo.
The clinical significant of c-Myc, pERK, E-cad and Vimentin expression in PC are well studied in other and our previous study. [29][30][31][32] In the current study, CRT and Fibronectin overexpression had positive association with tumor differentiation, UICC stage and lymph nodes metastasis, respectively, while positive Integrinβ1 expression was closely associated with UICC stage of PC patients (Table 2).
In Kaplan-Meier analysis, overexpression of CRT, Fibronectin or Integrinβ1 contributed to the poor prognosis of PC patients (P = 0.026; P = 0.043: P = 0.021, respectively; Figures 11a-c). Moreover, patients with high CRTand positive Fibronectin expression had a much worse prognosis than in patients with both low/weak expression (P = 0.006; Figure 11d). This trend was also observed in patients with the co-expression of CRT and Integrinβ1 (P = 0.014; Figure 11e). All together, the close interaction of the above proteins coordinately contributed to clinical stage and poor prognosis of PC patients.

Discussion
Our study first found that CRT silencing inhibited EGF-induced EMT in three PC cell lines. The relationship between CRT and EMT in cancers is rarely reported currently. In gastric cancer cells, TGF-β1-induced cell migration and invasion and reciprocal downregulation of E-cad could be abrogated by CRT knockdown. 7 In MDCK cells, CRT regulates the EMT-like change of cellular phenotype by modulating the Slug/E-cad pathway through altered Ca 2+ homeostasis. 28 Our study supplies a new sight to investigate the close interaction of CRT with EGF-induced EMT in PC, which has not been reported to our knowledge. It is well known that EMT (the initial transformation from benign to invasive carcinoma following E-cad decrease) and mesenchymal-epithelial transition (MET, the reverse of EMT in the later metastatic stage with the rescue of E-cad) are recognized as critical events for cancer metastasis. 33,34 Overexpression of CRT rescued the change of EMT-related proteins (E-cad, ZO-1, β-catenin and Fibronectin) induced by EGF in CRT-silencing PC cells, which further indicates that CRT also has a significant role in EGFinduced MET in PC.
Integrin/EGFR-ERK/MAPK signaling pathway also has a critical role in EMT of various cancers. Inhibition of Fibronectin fibrillogenesis blocks activation of the TGF-β signaling pathway via Smad2, decreases cell migration and ultimately leads to inhibition of EMT in colorectal cancer (CRC). 35 Integrinβ1 enhances the EMT in association with gefitinib resistance of non-small cell lung cancer. 36 The activation of integrinβ1 by type I collagen coupling with the hedgehog pathway promotes EMT in PC. 37 Both c-Myc expression and MEK1-induced ERK2 nuclear localization are required for TGF-β-induced EMT in prostate cancer. 38 In current study, EGF activated Integrin/EGFR-ERK/MAPK signaling pathway in PC cells. The intreaction between Integrin/EGFR-ERK/MAPK signaling in response to EGF remains controversial. EGF promotes human keratinocyte locomotion on collagen by increasing the integrinα5. 39 EGF has a synergistic effect on EGFR phosphorylation and ERK/MAPK activation in non-small cell lung cancer induced by Integrinβ1-dependent adhesion. 40 EGFinduced MAPK signaling inhibits hemidesmosome formation through phosphorylation of the integrinβ4 in keratinocytes. 41 Conversely, EGF-mediated Integrinβ1 inactivation in squamous and colon cancer cell lines was dependent on an EGFR kinase/ERK/p90RSK/filamin A pathway. 42 Meanwhile, inactivation of integrin a5β1 by EGF is correlated with both the level of EGFR expression and p90RSK phosphorylation in squamous carcinoma cells. 43 Based on our current study, Integrin signaling coordinates with EGF/EGFR-ERK/MAPK pathway in the co-operative control of cell biology as Cabodi et al. 44 and Moro et al. 45 suggest. However, CRT silencing significantly inhibited EGF-induced pEGFR1173, Fibronectin, Integrinβ1, c-Myc and pERK in PC cells. Moreover, co-staining of CRT with pEGFR1173 (with EGF), Fibronectin and Integrinβ1 and co-immunoprecipitation of CRT with the key proteins of Integrin/EGFR-ERK/MAPK signaling (Fibronectin, Integrinβ1 and c-Myc) were detected in PC cells, which is partially observed in Jurkat cells (CRT/integrin α2β1) 46 and atria tissues of patients (CRT/integrinα5 complex). 47 In addition, Integrin α1β1 expression is controlled by c-Myc in CRC cells detected by CHIP assays. 48 All the above indicated a close interaction of CRT and Integrin/EGFR-ERK/MAPK signal pathway in EGF-induced EMT of PC. We conclude that CRT promotes EGF-induced EMT in PC cells via Integrin/ EGFR-ERK/MAPK signaling pathway.
In vivo, CRT silencing inhibited subcutaneous tumor growth and liver metastasis of pancreatic tumor, which was consistent with the studies in gastric and bladder cancers. 7,49 Moreover, consistent with the results in vitro, CRT silencing significantly inhibited Fibronectin, Integrinβ1, c-Myc and pERK and The corresponding molecular mechanism in the current study is further investigated. With EGF stimulus, CRT colocalized with (bound to) pEGFR1173, directly activated pEGFR and corresponding downstream of ERK/MAPK signaling. Meanwhile, CRT enhanced integrins' signaling via interacting with Integrinβ1/Fibronectin. Finally, these two signal pathways coordinately regulated EGF-induced EMT in PC cells as mentioned above.
In addition, CRT might regulate the above signaling via mediating Ca 2+ homeostasis just as the research in MDCK cells. 28 Increasing evidences have demonstrated that intracellular alteration of Ca 2+ homeostasis in various cancer cells is closely involved in tumor initiation, angiogenesis, progression and metastasis. 50 In the current study, the increase of both intracellular Ca 2+ and Integrinβ1, Fibronectin and c-Myc expression induced by EGF and ionomycin is finally regulated by CRT. CRT also couples Ca 2+ release and Ca 2+ influx in integrin-mediated Ca 2+ signaling. 51,52 A positive modulation of ERK/MAPK signaling pathways by an increase in intracellular Ca 2+ is also well reported in neurons and PC12 cells after stimulation with neurotrophins or growth factors. 53,54 The Ca 2+ sensor calmodulin enhance Myc transcriptional and oncogenic activities in Ca 2+ -dependent manner. 55 Additionally, the interaction of Integrin/Fibronectin and ERK/MAPK pathway is always mediated by Ca 2+ signaling. In immortalized thyroid TAD-2 cells, activation of the integrin receptors by Fibronectin binding stimulates the Ras/ERK pathway, which is controlled by calcium calmodulin-dependent kinase IImediated Ca 2+ signaling. 56 In normal thyroid cells, both induction of pERK and increase of intracellular Ca 2+ were mediated by Fibronectin binding to integrin αvβ3. 57 Taking together, CRT promotes EGF-induced EMT in PC cells via Integrin/EGFR-ERK/MAPK signaling pathway, which might be regulated by CRT-mediated alteration of intracellular Ca 2+ . Recently, it remains unclear how ER or non-ER CRT functions are stimulated in different cancers. All of the above will be further investigated in our future study.

Materials and Methods
Tissue samples. All patient-derived specimens were collected and archived under protocols approved by the institutional review board of China Medical University. Sixty-eight cases of paraffin-embedded pancreatic ductal adenocarcinoma were obtained from the patients at the First Hospital of China Medical University between 2011 and 2016. All diagnoses were confirmed pathologically. Patients with endocrine carcinoma, acinar cell carcinoma and invasive intraductal papillary mucinous carcinoma were excluded from this study. Patients' characteristics were summarized in Supplementary Table 1.   Immunohistochemistry. IHC was performed as described previously. 13,58 Staining intensity was scored as 0 (negative), 1 (weak), 2 (medium) and 3 (strong). Extent of staining was scored as 0 (0%), 1 (1-25%), 2 (26-50%), 3 (51-75%) and 4 (76-100%) according to the percentage of the carcinoma involved area that was positively stained. The final IHC staining scores were determined by three professional pathologists. The sum of the extent and intensity score was used to arrive at the final staining scores (0-7). For CRT, tumors with a final staining score  Immunoprecipitation. For immunoprecipitation, as described previously 58 , whole protein lysates (prepared from Capan-2 and AsPC-1 cells) were extracted in a lysis buffer (20 mM Tris/HCl, pH7.4, 1.0% NP-40, 1 mM EDTA, 150 mM NaCl, 50 μg/ml PMSF, 10 μg/ml leupeptin). Briefly, CRT, c-Myc and IgG (Santa Cruz) antibodies were preincubated with magnetic beads (Bio-Rad, Hercules, CA, USA) for 4 h at 4°C. The antibody-beads complex was washed three times with the lysis buffer and incubated with the soluble supernatants of protein lysates overnight at 4°C . Next, the immunocomplexes were washed three times with lysis buffer, eluted by boiling in sample loading buffer for SDS-PAGE and then subjected to WB analysis with CRT, c-Myc, Integrinβ1, Fibronectin and GAPDH antibodies. Each experiment was repeated three times.
Construction of CRT-silencing stable cell lines using CRISPR/ Cas9 and transient transfection for rescue experiment. Lenti-cas9 and Lenti-sgRNA were synthesized from Genechem (GenePharma Co, Ltd, Shanghai, China). Cells were firstly infected with lenti-cas9 in AsPC-1 and Capan-2 cell lines and next selected by puromycin (Sigma). The stable sublines were then infected with CRT-sgRNA (sg1-CRT/sg2-CRT) and sgRNA control (scramble) to specifically silence target genes. The target sequences of sg1-CRT, sg2-CRT and scramble are shown in Supplementary Table 2. CRT-silencing effect was successfully verified by WB shown in 'Results section' (Figure 1). The PCMV2-CRT plasmid (CRT-GFP) and corresponding empty plasmid (GFP) were gifts from Dr. Hassan Dihazi as described previously. 27 sg1-CRT, sg2-CRT and scrambleinfected PC cells with EGF treatment were transiently transfected with CRT-GFP and GFP plasmids for 72 h using oligofectamine3000 (Invitrogen, Carlsbad, CA, USA) as described by the manufacturer. Transfection efficiency was detected by confocal fluorescence microscopy.
EMT construction. According to our previous study, 3 sg1-CRT, sg2-CRT and scramble infected three PC cell lines (Capan-2, AsPC-1 and BxPC-3 cells) were  A total of 15 cases of nude mice (BALB/c-nu) were acclimatized for a week. sg-CRT and scramble-infected Capan-2 cells (5 × 10 6 /ml) suspended in 200 μl of FBS-free 1640 were subcutaneously transplanted into bilateral axillas of five mice, respectively. A cotton swab was used to avoid possible bleeding and leakage of tumor cells from the injection site. The mice were killed 4 weeks later. Tumors were excised and documented by measurements using vernier calipers. Tumor volumes were calculated by the following formula: length × width × height × 0.52 in millimeters. Finally, samples were extracted and fixed for late HE and IHC staining.
sg-CRT and scramble-infected AsPC-1 cells (1 × 10 7 /ml) suspended in 100 μl of FBS-free 1640 were injected into the tail vein of 10 cases of nude mice, respectively. A cotton swab was held over the injection site for 1 min to prevent leakage from tail vein. The mice were killed 8 weeks later. The number of liver metastases was investigated immediately and then fixed for HE and staining. Statistical analysis. Statistical analyses were performed using the SPSS software 13.0 (SPSS, Chicago, IL, USA). The differences of orthotopic tumor volumes were compared by paired sample t-test. The differential expression of target proteins by IHC in orthotopic tumor were compared through paired sample non-parametric test. The differences of WB analysis, cell migration and invasion assays and of the number of liver metastases were expressed as mean ± S.E. and compared by Student's t-test. The clinicopathological significance of and relationship between target proteins were analyzed by Chi-squared and Spearman correlation tests, respectively. Kaplan-Meier curve was used to estimate survival, and differences were analyzed by log-rank test. A value of Po0.05 indicated statistical significant.

Conflict of Interest
The authors declare no conflict of interest.