Human tissue specimens and HCC tissue microarray
Human HCC tissues and matched adjacent non-tumour liver tissues were obtained from patients who received curative surgery at the Sun Yat-sen Cancer Center (SYSUCC; Guangzhou, China). Two independent cohorts, including the training cohort (94 HCC specimens collected from December 2003 to July 2010) and the validation cohort (378 HCC specimens collected from January 2010 to May 2015) were used for survival analysis. The related clinicopathological features of the enrolled patients are presented in Table S1. All samples were obtained with the informed consent of the patients. This study complied with the standards of the 1975 Declaration of Helsinki and the experiments were approved by the Ethics Committee of Sun Yat-Sen University Cancer Center.
Immunohistochemical staining (IHC)
After deparaffinized and blocked off the nonspecific antigen with goat serum (Zsbio, Beijing, China), HCC tissue microarrays were probed with anti-Elafin antibody, anti-Sp1 antibody, anti-Vimentin antibody, and anti-p-AKT antibody, as described in Table S4, overnight at 4 °C. Then, the microarrays were incubated with HRP anti-rabbit/mouse antibodies (Dako, Copenhagen, Denmark) for half an hour at 37 °C, and then the diaminobenzidine chromogen (Dako, Copenhagen, Denmark) was applied for reaction, followed by counterstaining with hematoxylin (Leagene, Beijing, China).
For IHC scoring, two experienced pathologists evaluated the staining intensity of specific markers, independently. The immunoreactivity for Elafin protein was scored using a semi-quantitative method by evaluating the number of positive tumor cells over the total number of tumor cells, and the IHC scores were assigned by using 5% increments (0, 5, 10%...100%), as described in previous studies [21, 22]. Based on the IHC scores, we then dichotomized these patients as negative group, 0–25%; weak group, 25–50%; moderate group, 50–75%; strong group, 75–100%. The high Elafin expression group is defined as patients with tumors of moderate or strong intensities, while the low Elafin expression group is defined as patients with tumors of negative or weak intensities [23, 24].
Cell lines and cell culture
Cell lines and cell culture were performed as described previously . Human HCC cell lines (PLC-8024, Huh7, Hep3B, MHCC-97H, and HepG2) and MIHA (the normal liver cell line) were purchased from the Shanghai Cell Bank of the Chinese Academy of Sciences (Shanghai, China) with STR (short tandem repeat) appraisal certificates. Cells were maintained in Dulbecco’s Modified Eagle medium (DMEM; ThermoFisher, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Califonia, USA) at 37 °C in 5% CO2.
Co-immunoprecipitation (Co-IP) assay and mass spectrometry
The cells were lysed with RIPA (Fdbio Science, Hangzhou, China) lysis buffer supplemented with protease inhibitors, and the lysates were mixed with the corresponding conditioned medium of the cells. Fifty microlitres of the mixture was removed and used as input, and the remaining mixture was incubated with primary anti-EGFR or anti-Elafin or with anti-IgG as a negative control overnight, as described in Table S4, at 4 °C. Thereafter, an appropriate volume of Protein A/G PLUS-Agarose beads (sc-2003, Santa Cruz Biotechnology) was added to the mixture for conjugation for another 4 h at 4 °C according to the manufacturer’s protocol. After that, the beads were washed with RIPA buffer, and the precipitated proteins binding on the beads were collected. Finally, the beads were resuspended in electrophoresis loading buffer and boiled for western blotting assays with anti-EGFR or anti-Elafin antibodies. Mass spectrometry was conducted by Saizhe Science (Guangzhou, China) using PLC-8024 cells.
Dimerization and internalization of EGFR
After starvation for 12 h, wild-type HCC cells were stimulated with 10 μg/ml recombinant Elafin (rElafin, R&D Systems, MN, USA) or 20 ng/ml EGF (236-EG, R&D Systems, MN, USA) or with BSA (Biofroxx, Germany) as the control for 30 min. After that, the cells were treated with the crosslinking reagent BS3 reagent (Thermo Fisher, MI, USA) at a concentration of 3 mM for 30 min at room temperature. Then, 40 μL 1 M Tris-HCl pH 7.5 was added for an additional 15 min to terminate the reaction. Finally, the cells were lysed and subjected to western blotting to assess EGFR dimerization. The dimers were represented by bands at approximately 300 KD, while EGFR monomers were represented by bands at 170 KD.
To assess the internalization of EGFR, after starvation for 12 h, the cells were stimulated with 10 μg/ml recombinant Elafin or 20 ng/ml EGF or BSA for 30 min. Then, the cells were subjected to immunofluorescence (IF) to assess the cell surface expression of EGFR.
To further investigate the detailed binding mode of Elafin and EGFR, we docked Elafin to EGFR via the ClusPro online server (https://cluspro.bu.edu) with the default settings. Considering the similarity between Elafin and EGF, both of which are composed of ~ 50 residues and bind to the extracellular domain (residues 25–645) of EGFR, we selected the crystal structure of the EGFR/EGF complex (PDB ID: 1IVO) as a receptor for the docking procedure. The crystal structure of Elafin was derived from its complexed form (PDB ID: 1FLE). Among the resulting top 10 best binding modes, we chose the top one based on the analysis of interaction effects. To verify the predicted binding mode, we also repeated the docking through the Zdock online server (http://zdock.umassmed.edu), which yielded a result consistent with that of ClusPro.
Luciferase reporter assay
Luciferase reporter plasmids, including full-length, different truncated, and mutant Elafin promoters, were constructed by GeneCopoeia (Rockville, USA). For the assay, PLC-8024 and Huh7 cells (5 × 104) were seeded in 24-well plates under regular conditions and then co-transfected with Sp1 overexpression plasmids, or Sp1 siRNA or negative controls, together with Elafin promoters. After 48 h, the conditioned medium was collected, and the Gaussia luciferase (Gluc) and secreted alkaline phosphatase (SEAP) luciferase activities were measured consecutively using the Secrete-Pair™ Dual Luminescence Assay Kit (GeneCopoeia, Rockville, USA) according to the manufacturer’s instructions. The gluc activity was normalized to the SEAP activity, and each group was analysed in triplicate experiments.
Chromatin Immunoprecipitation (ChIP) assay
ChIP assays were carried out in PLC-8024 and Huh7 cells with a ChIP kit (Cell Signaling Technology, Boston, USA) according to the manufacturer’s instruments. Briefly, 1% formaldehyde (Sigma-Aldrich, Germany) solution was added to induce PLC-8024 or Huh7 cell crosslinking followed by glycine solution to quench the reaction. Afterwards, the cells were lysed, and the nucleoprotein complexes were sonicated for 10 cycles of 10 s power-on and 20 s interval with an intensity of 200 W with the sonicate conductor (Qsonica, USA). Then, anti-Sp1 antibody or IgG, as described in Table S4, was added and incubated with the complexes overnight at 4 °C. The next day, Protein A/G magnetic beads were added to precipitate the indicated fragments for an addition 4 h at 4 °C. After extraction and purification of the indicated DNA, semiquantitative PCR was performed to identify the region interacting with the Elafin specific primers. The indicated primers were present in Table S5. The experiments were performed in triplicate and the amount of immunoprecipitated DNA was normalized to the input.
In vivo lung metastasis model
To establish the lung metastasis model, 5- to 6-week old male nude mice (n = 12 per group) were injected with stable PLC-8024 or Huh7 cells (2 × 106 cells suspended in 200 μL PBS) through the tail veins. Fifty days after injection, the mice were sacrificed, and the lungs were harvested for haematoxylin and eosin (H&E) staining, which was used to count the number of metastatic nodes in lung sections.
To assess the effects of erlotinib treatment, after injection with Huh7 stable Elafin overexpressing or vector cells (2 × 106) via the tail veins of the nude mice, the mice were treated with erlotinib (S1023, Selleck Chemicals) in PBS with 10% Captisol (HY-17031, MedChemExpress) at 80 mg/kg orally  or with vehicle (PBS with 10% Captisol) daily from the 4th to 7th week after injection. On day 50, the mice were sacrificed, and the lungs were collected. H&E staining was performed on paraffin-embedded lungs to identify the metastatic nodes in the lungs.
All BALB/c nude mice were purchased from the Medical Experimental Animal Center of Guangdong Province (China) and fed at the Animal Experimental Center of Sun Yat-Sen University. All animal research procedures were performed according to the Animal Care and Use Ethics Committee of Sun Yat-Sen University Cancer Center.
Analysis of the Cancer genome atlas (TCGA) and gene expression omnibus (GEO) databases
The clinicopathological features, as well as data on the relative expression levels of Elafin, Sp1, and vimentin in the 371 HCC tissues, were downloaded from the TCGA-Liver Hepatocellular Carcinoma RNA sequencing dataset. The survival and expression data from another 221 HCC patients were previously deposited in the National Center for Biotechnology Information (NCBI) GEO dataset with accession number: GSE 14520.
The experiments were repeated at least three times independently, and the measured data were represented as the mean ± SD. Binary variables were compared by using the Chi-squared test, and ordinal categorical variables were compared by the Kruskal-Wallis test. In the TCGA and GEO datasets, the optimal cutoff points for the expression of Elafin was determined by the maximally selected rank statistic from the “maxstat” R package for survival analysis. Survival curves were constructed using the Kaplan-Meier method and analysed by the log-rank test. Significant prognostic factors found by univariate analysis were entered into a multivariate analysis using the Cox proportional hazards model. The correlations between Elafin and vimentin or Sp1 or pAKT were assessed by Spearman correlation analysis. All analyses were two-sided, and differences with P values less than 0.05 were considered significant. Statistical analyses were performed using the R program (R version 3.5.0; R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism 7.0 software (GraphPad, Inc., La Jolla, CA, USA).
Other experiments are described in the ‘Supplementary Data’.