Hypo-phosphorylated CD147 promotes migration and invasion of hepatocellular carcinoma cells and predicts a poor prognosis

TRIzol was purchased from Sigma (St Louis, Missouri, USA). Antibodies directed against phospho-serine (ab9392) and α-tubulin (ab80779) were purchased from Abcam (Cambridge, MA, USA); Antibodies directed against HA tag (0906–1), STAT3 (ET1605–45), p-STAT3 (ET1603–40), Akt (ET1609–47), p-Akt (ET1607–73), c-Jun (ET1608–3), p-c-Jun (ET1608–4) and p38 (ET1602–26) were obtain from HuaAn Biotechnology (Hangzhou, China); goat anti-mouse IgG antibody (31430), goat anti‐rabbit IgG antibody (31460), anti‐p‐p38 antibody (44‐684G) and anti‐NEK6 antibody (MA5–24947) were purchased from Thermo Fisher Scientifc (Waltham, MA, USA); CD147-specific antibody was produced by our lab.

Human hepatocellular carcinoma (HCC) cell line Huh-7 was obtained from the Japanese Collection of Research Bioresources (JCRB, Osaka, Japan). HepG2 and SMMC-7721 cells were obtained from the Chinese Academy of Medical Sciences (Shanghai, China). A Huh-7 CD147-KO (Huh-7 CD147^−/−) cell line was generated using a CRISPR/Cas9 system as previously reported [14]. All cells were cultured at 37 °C, 5% CO₂, in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS). All cells have been authenticated using short tandem repeat profiling.

76 HCC tissue specimens were collected from the Department of Pathology, Eastern Hepatobiliary Surgery Hospital, which is affiliated with the Second Military Medical University, from 2008 to 2012 and were histological confirmed by staining with hematoxylin and eosin (HE). All patients provided written informed consent, and the study was approved by the hospital Ethics Committee

Immunohistochemical (IHC) staining was performed on 5 μm tissue sections. To this end, paraffin sections were dewaxed, followed by antigen retrieval with 10 μM citrate buffer at pH 6.0. The deparaffinized sections were treated with methanol containing 3% hydrogen peroxide for 15 min. After washing with PBS, the sections were incubated with blocking serum for 30 min. Then, the sections were incubated with anti-CD147 antibody at 4 °C overnight. Following incubation, immunoperoxidase staining was conducted using a streptavidin-peroxidase kit (Zhongshan Jinqiao Co., Beijing, China) and the sections were treated with 3,3′-diaminobenzidine (Zhongshan Jinqiao Co., Beijing, China) to detect the target proteins. Hematoxylin was used to counterstain the nuclei. The expression levels were independently evaluated by two senior pathologists according to the proportion and intensity of positive cells. The following criteria were used to score each specimen: 0 (no staining), 1 (any percentage with weak intensity or < 30% with intermediate intensity), 2 (> 30% with intermediate intensity or < 50% with strong intensity) or 3 (> 50% with strong intensity).

Immunofluorescence was performed as described previously [13]. Briefly, cells were harvested and allowed to attach for 24 h to fibronectin pre-coated cell culture dishes with glass bottoms (NEST Biotechnology Co., LTD.). After washing twice with PBS, the cells were fixed with paraformaldehyde in PBS, permeabilized with 0.1% Triton X-100, and blocked with 1% BSA in PBS for 1 h. The resulting cells were first incubated with the indicated antibodies for 1 h, washed twice with PBS, and then incubated with Alexa 488-phalloidin solution and the corresponding FITC-conjugated secondary antibodies for 30 min in the dark. Cell nuclei were stained with DAPI (Vector Labs). After washing, the cells were visualized using an A1R-A1 confocal laser microscope (Nikon, Japan).

In situ proximity ligation assay (PLA) experiments were performed using reagents and instructions provided by a commercially available kit (Duolink In Situ Detection Reagents Red) from Sigma-Aldrich (St Louis, MI, USA). Briefly, cells were seeded into dishes with glass bottoms (NEST Biotechnology Co., LTD.). After washing twice with PBS, the cells were fixed in paraformaldehyde for 15 min at room temperature and blocked with Blocking Solution for 1 h at 37 °C. Next, the cells were rinsed twice with PBS/0.1% Tween 20 (PBST) after which primary mouse anti-CD147 (HAb18, prepared by our laboratory, 1:1000) and rabbit anti-phospho-serine (Abcam, ab9392, 1:100) antibodies in Duolink In Situ Antibody Diluent were applied and incubated overnight at 4 °C. Next, the cells were rinsed three times with Wash Buffer A. Secondary probes (anti-mouse-PLUS and anti-rabbit-MINUS, conjugated to oligonucleotides) were diluted to final concentrations of 1:5 in antibody diluent. The secondary probe mix was added to each sample, incubated for 1 h at 37 °C and washed with Wash Buffer A, after which 40 μl ligation solution was added. The dishes were incubated for 30 min at 37 °C. After washing with Wash Buffer A, 40 μl amplification solution was added and incubated for 100 min at 37 °C. Next, the cells were rinsed three times with Wash Buffer B. Before detection, 50 μl Duolink In Situ Mounting Medium with DAPI was added to each sample. Images were captured using a fluorescence microscope, and PLA signals were analyzed using the Duolink ImageTool software (Sigma-Aldrich). Formalin-fixed paraffin-embedded (FFPE) HCC tissue samples were prepared as described under 2.3 prior to the blocking step. Next, the assay was performed according to the manufacturer’s protocol.

One day prior to transfection, 4 × 10⁵ cells were seeded per well in a 12-well plate in complete medium. Subsequent transfection was carried out using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. After transfection, the cells were subjected to selection in 100 μg/ml G418 for 2 weeks. Antibiotic resistant colonies were subsequently picked, pooled, and expanded for further analysis under selective conditions.

Co-immunoprecipitation (co-IP) was performed using a Pierce™ Co-IP Kit (Thermo Fisher Scientifc, MA, USA) according to the manufacturer’s protocol. For each co-IP assay 10 μg affinity-purified antibody was used. Cell lysates were incubated with gentle rocking overnight at 4 °C. The eluted samples were analyzed by Western blotting using antibodies as indicated.

Western blotting was performed as described previously [13]. Briefly, equal amounts of protein were separated by denaturing SDS-PAGE and transfered to polyvinylidene fluoride (PVDF) microporous membranes (Millipore, Boston, MA). Next, the resulting blots were blocked with 5% nonfat milk in TBS/0.5‰ Tween (TBS-T). The primary antibodies were diluted in TBS-T, and the blots were incubated with these antibodies overnight at 4 °C followed by washing in TBS-T and incubation with HRP-conjugated secondary antibodies for 1 h at room temperature. Signal detection was conducted using a ChemiDoc™ Touch Imaging System and analyzed using Image Lab™ Software (Bio-Rad, CA, USA).

In vitro scratch wound healing assays were performed as described previously [13]. Briefly, 24 h after treatment, the cells were harvested, seeded in 12-well plates and grown until confluence. Next, a pipette was used to scratch (‘wound’) the monolayer after which the remaining cells were washed with serum-free medium. Subsequently, photomicrographs were taken at various time points.

Chambers with polycarbonate filters with a 8 μm nominal pore size (Millipore, Boston, MA) coated on the upper side with Matrigel (BD Bioscience, San Jose, CA) were used to assess cell invasiveness. The chambers were placed into a 24-well plate. Cells were trypsinized, resuspended in serum-free medium, and seeded at a density of 1 × 10⁵ cells per well in the upper chambers. The lower chambers were filled with 500 μl RPMI-1640 containing 10% FBS. After 24 h, the chambers were moved to a fresh 24-well plate and stained with 0.2% crystal violet for 20 min. The number of cells that had attached to the lower surface was counted under a light microscope and statistically analyzed.

Cells were transfected with siRNAs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. siRNAs targeting NEK6 were designed and synthesized by Shanghai GenePharma Co. (Shanghai, China). The siRNA sequences used are depicted in Table S1.

The apoptosis and cell cycle rates of Huh-7 CD147-KO cells transfected with wildtype CD147 or S246A/S252A mutants were assessed using an Annexin V-FITC/propidium iodide (PI) apoptosis detection kit and a cell cycle detection kit, respectively (KeyGEN Biotech, Nanjing, China). Quantification of PI and FITC signals was performed using a fluorescence activated cell sorter FACSAria (BD Bioscience, San Jose, CA) system.

Total RNA was extracted using TRIzol reagents (OMEGA Bio-Tek, Norcross, GA, USA). Reverse transcription was performed using a PrimeScript RT reagent kit (TaKaRa Biotechnology, Japan). All primers were synthesized by BGI (BGI, Shenzhen, China) and their sequences are listed in Table S1. Quantitative real-time PCR (qRT-PCR) was performed using a SYBR Premix Ex Taq II Kit (TaKaRa Biotechnology, Japan).

Total RNA was extracted using Trizol (Tiangen, Beijing) and assessed for quantity and quality using an Agilent 2100 BioAnalyzer (Agilent Technologies, Santa Clara, CA, USA) and a Qubit Fluorometer (Invitrogen, Carlsbad, CA, USA), respectively. Total RNA samples that met the following requirements were used in subsequent experiments: RNA integrity number (RIN) > 7.0 and a 28S:18S ratio > 1.8. Sequence libraries were generated and sequenced by CapitalBio Technology (Beijing, China). A NEB Next Ultra RNA Library Prep Kit for Illumina (NEB, Ipswich, MA, USA) was used to construct the libraries for sequencing. A NEB Next Poly(A) mRNA Magnetic Isolation Module (NEB, Ipswich, MA, USA) kit was used to enrich poly(A) tailed mRNA molecules from 1 μg total RNA. The mRNA was fragmented into ∼200 base pair pieces. First-strand cDNA was synthesized from the mRNA fragments using reverse transcriptase and random hexamer primers, after which second-strand cDNA was synthesized using DNA polymerase I and RNase H. The ends of the cDNA fragments were subjected to an end repair process including the addition of a single “A” base, followed by ligation of adapters. The resulting products were purified and enriched by PCR to amplify the library DNA. The final libraries were quantified using a KAPA Library Quantification kit (KAPA Biosystems, South Africa) and an Agilent 2100 Bioanalyzer. After qRT-PCR validation, the libraries were subjected to paired-end sequencing with a pair-end 150-base pair reading length on an Illumina HiSeq sequencer (Illumina) [30].

The sequencing quality was assessed using FastQC (Version 0.11.5) after which low quality data were filtered using NGSQC (v0.4). The clean reads were subsequently aligned to the reference genome using HISAT2 (Johns Hopkins University, USA) with default parameters [31]. The processed reads from each sample were aligned using HISAT (Johns Hopkins University, USA) against the corresponding human reference genome. Gene expression analyses were performed using Cuffquant and Cuffnorm (Cufflinks 2.2.1). Cuffdiff was used to analyze the differentially expressed genes (DEGs) between samples. The standardization method of Cuffdiff is geometric, with per-condition and pooled as discrete model [32]. Thousands independent statistical hypothesis-driven tests were conducted on DEGs, separately, after which a p value was obtained that was corrected using a FDR method. This p value was used to perform significance analysis. The parameters for classifying significantly DEGs was ≥ 2-fold difference (|log2FC| ≥ 1, FC: fold change of expression) in transcript abundance. By searching the ENSEMBL, NCBI, UniProt, GO and KEGG databases, BLAST (Basic Local Alignment Search Tool) alignment was performed to determine the functional annotation of the DEGs. The best matches were selected to annotate the DEGs. Finally, a KEGG pathway enrichment analysis was performed for the DEGs using KOBAS 3.0 software (Available online: http://​kobas.​cbi.​pku.​edu.​cn).

All experiments were performed in triplicate, and the results were expressed as mean ± SD. Statistics were evaluated using GraphPad Prism V7.0 software (GraphPad Software, La Jolla, CA). The statistical analyses were carried out using one-way ANOVA (multiple comparisons) and Student’s t test (two comparisons). Differences were deemed significant when p < 0.05. *** indicates p < 0.001, ** indicates p < 0.01, * indicates p < 0.05, and # indicates p > 0.05.

As mass spectrometry-based global phosphoproteome analysis has revealed that CD147 can be phosphorylated at Ser246 and/or Ser252 in various human tissues and cell lines, we first examined whether CD147 is also phosphorylated in HCC tissues and cell lines. Total CD147 was purified from HCC tissue and cell lysates by immunoprecipitation after which phosphorylation modification was detected using an anti-phospho-serine antibody. We found that CD147 was phosphorylated in HCC tissues (Fig. 1a) and Huh-7, HepG2 and SMMC-7721 cells (Fig. 1b), whereas no positive result was obtained using the anti-phospho-serine antibody in a mouse IgG (mIgG) isotype control group. To characterize CD147 phosphorylation in situ, we applied a PLA assay, which detects post-translation modifications (PTMs) including phosphorylation in situ in fixed biological samples employing two paired antibodies, i.e., an anti-phospho-serine antibody and an anti-CD147 antibody (Fig. 1c) [33, 34]. The validity of this approach was established by monitoring CD147 phosphorylation in FFPE HCC specimens. As a negative control, we analyzed a HCC tissue without CD147 expression as verified by IHC staining. Little background staining was observed, whereas distinct red spots were detected in a HCC tissue with positive CD147 expression (Fig. 1d). We also detected a similar pattern of red spots in Huh-7, HepG2 and SMMC-7721 cells, whereas little staining was observed in Huh-7 CD147-KO cells (Fig. 1e). Together, these results indicate that CD147 is serine phosphorylated in HCC tissues and cell lines.

Based on the observation that CD147 is serine phosphorylated in HCC tissues, we next set out to identify the precise phosphorylation site of CD147 in HCC tissue. Native CD147 was isolated from HCC tissue using an anti-CD147 antibody (Fig. 2a). The isolated CD147 protein was subsequently subjected to mass spectrometry. By doing so, an approximately 80 Da mass shift was observed on serine-246, suggesting that CD147 is phosphorylated at this serine residue in HCC tissue (Fig. 2b). Previous studies showed that CD147 may be phosphorylated at serine-246 and/or serine-252 in different tissues, and sequence analysis has revealed that serine-246 and serine-252 are evolutionarily conserved in CD147 in species ranging from human, mouse, rat, chicken, rabbit and Chinese hamster (Fig. 2c). Next, we set out to evaluate the phosphorylation level of each of these serines in Huh-7 cells. To this end, we constructed S246A, S252A and S246A/S252A mutants of CD147 and assessed their steady-state expression in Huh-7 CD147 KO cells (Fig. 2d). We found that wild type CD147 and S246A/S252A mutants exhibited similar turnover rates at the same time (Fig. 2e) and that the mutations did not alter the membrane distribution pattern of CD147 (Fig. 2f). Next, we analyzed the phosphorylation level of each mutant by immunoprecipitation using an anti-phospho-serine antibody. We fount that wildtype CD147 exhibited a higher level of phosphorylation modification than the mutants and that no phosphorylation of S252A and the S246A/S252A mutants could be detected (Fig. 2g). To further confirm the above results, In situ Duolink-PLA analysis was performed. As a negative control, we found that the PLA signal of Huh-7 CD147-KO cells was below the detection threshold. Compared to wild type, the PLA signal of S246A phosphorylation was found to be reduced, while the PLA signals of S252A and S246A/S252A were found to be low (Fig. 2h). Taken together, these results indicate that the major phosphorylation sites of CD147 are serine-246 and serine-252 in HCC tissues and cells.

To determine the biological significance of CD147 phosphorylation, we transfected Huh-7 CD147-KO (Fig. 2d) and HepG2 shCD147 (Fig. 3a) cells with S246A/S252A mutant or wildtype CD147. Subsequent cell migration abilities were determined using scratch wound healing assays. We found that CD147 expression rescue of the S246A/S252A mutants significantly increased the migration capacity of both HepG2 shCD147 (Fig. 3b) and Huh-7 CD147-KO (Fig. 3c) cells compared to the respective wildtype controls. Consistently, transwell invasion assays showed that S246A/S252A mutants enhanced cell invasion compared to the wildtype control in both HepG2 shCD147 (Fig. 3d) and Huh-7 CD147-KO cells (Fig. 3e). These results indicate that CD147-facilitated migration and invasion of HCC cells are incompatible with phosphorylation events. As CD147 was reported to be able to regulate cytoskeleton rearrangements, we assessed the effects of CD147 phosphorylation on the cytoskeleton in HCC cells. We found that CD147 mutants as well as wildtype CD147 were mainly located in the cell membrane, and no significant alterations in the cytoskeleton were observed (Fig. 2f). We also determined the effects of CD147 phosphorylation on the occurrence of cell death and cell cycle progression. We found that HCC cells expressing S246A/S252A mutants showed similar apoptotic cell proportions (Fig. 3f) and cell cycle progression parameters (Fig. 3g) compared to the wildtype controls, suggesting that phosphorylation modification of CD147 is not involved in the regulation of CD147-mediated cytoskeleton rearrangements, apoptosis and/or cell cycle progression.

In order to explore the mechanisms by which non-phosphorylated CD147 promotes HCC cell migration and invasion, we employed gene expression profiling in Huh-7 KO cells transfected with S246A/S252A mutant CD147 or wildtype CD147. Of the 632 differentially expressed genes identified, 358 genes were up-regulated and 274 genes were down-regulated (Fig. 4a). Fifteen of the differentially expressed genes (DEGs) were validated by qRT-PCR (Fig. 4b). Subsequent KEGG pathway analysis showed that ECM organization and ECM degradation were the top two pathways involved (Fig. 4c), which corroborates the importance of CD147 in ECM remodeling and subsequent invasion during tumor progression. Since CD147 has been reported to regulate several signaling pathways, such as the STAT3 [35, 36], Akt [37, 38] and c-Jun [39] pathways, we wondered whether this regulation may be dependent on the phosphorylation state of CD147. We found that, in contrast to wildtype CD147, expression of S246A/S252A mutant CD147 significantly increased the STAT3 and Akt phosphorylation levels (Fig. 4d), suggesting that CD147 promotes STAT3 and Akt signaling mainly through its non-phosphorylated form. To ascertain whether the altered gene expression profiles observed were due to activated STAT3/Akt signaling, Huh-7 CD147-KO cells transfected with S246A/S252A mutant CD147 or wildtype CD147 were treated with STAT3 inhibitor Niclocide or Akt inhibitor LY294002. We found that the S246A/S252A mutant-induced upregulation of LABM3 and COL12A1 could be partially reversed by inhibition of STAT3 in a dose‐dependent manner (Fig. 4e). Consistently, we found that S246A/S252A mutant-induced downregulation of SPP1 could be reversed by inhibition of Akt in a dose-dependent manner (Fig. 4f). These results suggest that other pathways besides the STAT3/Akt signaling pathways may be responsible for non-phosphorylated CD147-induced alterations in gene expression in HCC cells.

As we found that S246A/S252A can promote HCC cell migration and invasion and increase STAT3 and Akt activation, we wondered whether STAT3 and Akt signaling are involved in hypo-phosphorylated CD147-induced cell migration and invasion. When Huh-7 CD147-KO cells expressing S246A/S252A mutant CD147 were treated with the STAT3 inhibitor Niclocide or the Akt inhibitor LY294002, we found that the S246A/S252A mutants promoted cell migration compared to the wild type group and that this effect could be abolished by both the STAT3 and Akt inhibitors (Fig. 5a and Fig. 5b). Consistently, we found that Huh-7 CD147-KO cells expressing S246A/S252A mutants showed enhanced invasive abilities compared to the wild type group and that these enhanced abilities could be abolished by both the STAT3 and Akt inhibitors (Fig. 5c and Fig. 5d). Together, these results suggest that S246A/S252A mutant CD147 can promote cell migration and invasion, at least partially through STAT3 and Akt signaling.

Next, we set out to assess whether CD147 phosphorylation is regulated by and involved in signal transduction. Serine-252 is located in a conserved sequence APLKSSGQ within human CD147, or APLKGSGT within murine CD147. The peptide matches the typical consensus motif (L-X-X-S/T) (Fig. 6a) [40], which is also found in TPP1 that is known to be phosphorylated by NEK6 [41]. Based on this information, we focused on the involvement of NEK6 as a possible kinase responsible for phosphorylating CD147. To this end, we transfected Huh-7 cells with siRNAs targeting NEK6 (Fig. 6b). We found that siRNA-mediated silencing of NEK6 decreased CD147 phosphorylation (Fig. 6c). To subsequently test a direct interaction between CD147 and NEK6, reciprocal co-IP was performed in Huh-7 cells. By doing so, we found that NEK6 was specifically immunopurified with CD147 (Fig. 6d). These results indicate that NEK6 directly interacts with CD147 and phosphorylates the protein at serine-252 in Huh-7 cells.

Since CD147 has been found to act as a cell surface signaling receptor for extracellular CyPA [42, 43] and to interact with the ectodomain of CD147 [44], we assessed a possible role of the intracellular domain of CD147 in signal transduction initiated by CyPA. To this end, we probed CD147 phosphorylation of cell lysates from Huh-7 cells stimulated with CyPA for different periods of time. After stimulation by CyPA, the phospho-CD147 band showed a specific down-shift after 10 min (Fig. 6e). In addition, phosphorylation of p38, known to be regulated by CyPA [45], was detected during the same time period. Based on these results, we conclude that serine-252 of CD147 is dephosphorylated after CyPA stimulation and that reversible phosphorylation of CD147 may be involved in signal transduction upon CyPA stimulation.

Since we found that hypo-phosphorylated CD147 increases cell migration and invasion, we sought to determine whether CD147 phosphorylation is associated with the prognosis and clinical characteristics of HCC patients. To this end, we randomly sampled 76 tumor tissues from HCC patients with complete clinical and follow-up data (Table 1). Expression of CD147 was determined by IHC staining and CD147 phosphorylation was assessed by in situ PLA analysis (Fig. 7a). We found that the metastasis group had a lower PLA signal than the non-metastasis group (Fig. 7b), whereas no significant differences in the expression of CD147 between the two groups was noted (Table 2). The group with lower serum AFP levels (AFP ≤ 87 μg/l) exhibited higher PLA signals (Fig. 7c) and the recurrence group lower PLA signals compared to the group without recurrence (Fig. 7d). Pearson correlation analysis showed that high levels of phosphorylated CD47 expression were negatively correlated with serum AFP levels (p = 0.0016), distant metastasis (p = 0.0004) and recurrence (p = 0.0461), but were not correlated with age and/or any other factors tested (Table 1). Kaplan-Meier analysis showed that low PLA signals of CD147 phosphorylation were correlated with a poor overall survival (Fig. 7e). These results indicate that low CD147 phosphorylation levels correlate with the occurrence of distant metastasis and an unfavorable prognosis in HCC.