Up-regulation of PCOLCE by TWIST1 promotes metastasis in Osteosarcoma.

Procollagen C-proteinase enhancer protein (PCOLCE) was originally identified as an enhancer to facilitate the catalysis of procollagens by BMP1. PCOLCE participates in the reconstitution of extracellular and corneal repair. The elevation of PCOLCE in blood indicates that breast cancer has metastasized into the bones. However, direct research on PCOLCE has not been reported.
Methods: ECM candidates were identified by RNA-seq analysis from 4 normal and 16 osteosarcoma tissues. The in vitro migration and invasion abilities of osteosarcoma cells were determined by a Transwell assay. A spontaneous metastatic osteosarcoma model was established to assess osteosarcoma metastasis in vivo. The N-linked glycosylated amino acids were identified by PNGase F treatment combined with Western blotting. The mechanism of TWIST1 regulating PCOLCE transcription was elucidated by luciferase, qPCR and ChIP assays.
Results: PCOLCE was markedly up-regulated in human osteosarcoma tissues compared to its expression in noncancerous adjacent tissues; high PCOLCE expression in tissues correlated with a poor patient prognosis, and the knockdown of PCOLCE by shRNAs impaired the migration, invasion and lung metastasis of osteosarcoma cells. The overexpression of wild-type PCOLCE, but not its N29Q mutant, promoted migration, invasion and metastasis, indicating that the glycosylation of PCOLCE at Asn29 is necessary for its functions in osteosarcoma. TWIST1, a key transcription factor in metastasis, was also overexpressed in osteosarcoma tissues and positively correlated with either PCOLCE or its potential procollagen substrates, such as COL1A1, COL1A2, COL5A1, COL8A2 and COL10A1.
Conclusion: Our findings are the first to provide evidence that PCOLCE plays a critical role in promoting the lung metastasis of osteosarcoma, and this up-regulation of PCOLCE by TWIST1 may lead to a new therapeutic strategy to treat patients with metastatic osteosarcoma.

Introduction

Osteosarcoma, the most common primary malignant n class="Disease">bone tumor, is composed of malignant osteoblasts that produce immature bone or osteoid tissues 1. It arises primarily in adolescence and childhood and has a second peak incidence in those over the age of 50 2. Despite the improvement of treatment strategies, such as surgery or chemotherapy, the overall 5-year survival is approximately 75% for nonmetastatic osteosarcoma but has remained at 20% for metastatic cases over the past 30 years 3, 4. Thus, it is urgent to understand the molecular mechanism of osteosarcoma metastasis.

The extracellular matrix (ECM) is composed of numerous biochemically distinct proteins that participate in various biological processes, such as embryonic development, tissue homeostasis and cell differentiation 5, 6. Abnormal ECM dynamics also play a critical role in cancer progression and metastasis. For instance, matrix metalloproteinases (MMPs), the main enzymes involved in ECM degradation, are highly expressed in n class="Disease">cancer and promote cancer metastasis 7, 8. Lysyl oxidase (LOX) crosslinks collagens and elastins in the extracellular matrix and promotes bone metastasis in hypoxic cancers by inducing premetastatic bone lesions 9. Collagens, the most abundant components of the ECM, are excessively deposited in the extracellular matrix to form the tumor microenvironment and to modulate cancer cell polarity and migration 9. The down-regulation of BMP-1 (bone morphogenetic protein-1), which processes fibrillar collagen types I-III, leads to the suppression of TGFβ in non-small cell lung cancer cells 10.

Procollagen C-proteinase enhancer protein (n class="Gene">PCOLCE) is a secreted glycoprotein that enhances the activity of procollagen C-proteinases to participate in ECM reconstruction 11-13. PCOLCE binds to the C-propeptide of procollagen III and heparin sulfate through its CUB and NTR domains, respectively, leading to the enhancement of BMP-1 activity and collagen precursor maturation 14, 15. It has been reported that the dysregulation of PCOLCE is involved in numerous diseases. For instance, the expression of PCOLCE positively correlates with muscle and liver fibrosis 16. A PCOLCE deficiency contributes to deficient corneal repair 17. Mutated PABPN1 binds to PCOLCE and entraps it within the nuclear compartment, leading to oculopharyngeal muscular dystrophy 18. However, the role of PCOLCE in cancer remains to be explored. In the present study, we found that PCOLCE was highly expressed in osteosarcoma and may play an important role in promoting the lung metastasis of osteosarcoma.

Materials and methods

RNA preparation, RNA-seq and analysis

The total RNA from n class="Disease">osteosarcoma tissues was extracted using the TRIzol Reagent (Invitrogen). The cDNA was synthesized with mRNA enriched by Oligo(dT) beads. Libraries for next-generation sequencing were constructed by blunting the ends of the cDNA; this was followed by A-tailing, ligation to sequencing adapters, and PCR with indexed sequencing primers. RNA-seq was carried out using Illumina HiSeqTM 2500 by Gene Denovo Biotechnology Co. (Guangzhou, China), and 10 Gb of the RNA-seq data of every sample are available from the GEO database (GEO: GSE00000). The low-quality reads were trimmed by Trimmomatic (Version 0.32)19, and the clean reads were aligned to the human genome (GRCh38) using Tophat2 (Version 2.0.13)20. The expression levels of the genes were quantified by Cuffquant in the Cufflinks (Version 2.2.1)21 suite. All the processes were processed by Gene Denovo Co. (Guangzhou, China).

Clinical samples

Paracancerous tissues and n class="Disease">tumor samples were obtained from recruited osteosarcoma patients. The patients were diagnosed according to their clinicopathologic characteristics at the Department of Musculoskeletal Oncology at The First Affiliated Hospital of Sun Yat-Sen University (Guangzhou, China). No patient had received radiotherapy and/or chemotherapy prior to surgery. The samples were immediately frozen in liquid nitrogen for RNA extraction and qRT-PCR. The clinical information of 20 patients, whose samples were collected for RNA-seq, are listed as Supplementary Table S1 (sheet 3), whereas the clinical information of 68 patients, whose samples were collected for qRT-PCR, are listed as Supplementary Table S2.

Cell culture and transfection

U2OS and n class="CellLine">HOS were obtained from American Type Culture Collection. U2OS/MTX300 is methotrexate-resistant and is derived from U2OS. All cell lines were cultured in 10% fetal bovine serum containing DMEM (Dulbecco's Modified Eagle Medium) at 37°C and 5% CO2. Stable cell lines were cultured in complete medium supplemented with 0.5 μg/mL puromycin. The transfection experiments were performed according to the manufacturer's instructions included with Lipofectamine 2000TM (Invitrogen).

His-tagged protein purification

Stable cell lines expressing PCOLCE-His or n class="Mutation">N29Q-His were lysed by cell lysis buffer (phosphate 20 mM, NaCl 0.5 M, imidazole 20 mM, Triton X-100 2%, pH 7.4) at 4°C for 1 h, and the lysate was centrifuged to remove the cell debris before it was transferred to HiTrap (GE, 17040501). After washing with 5 ml binding buffer (phosphate 20 mM, NaCl 0.5 M, and imidazole 20 mM, pH 7.4), the protein was eluted with 1 ml elution buffer (phosphate 20 mM, NaCl 0.5 M, and imidazole 500 mM, pH 7.4). The elution product was diluted by adding 14 ml of PBS before it was concentrated to a 0.5 ml product by 10 KD Amicon Ultra-4 (Millipore). All buffer used in this process was supplemented with a protease inhibitor cocktail (Sigma).

Construct and plasmids

PCOLCE and n class="Mutation">N29Q CDSs were amplified in a thermal cycler using PrimeSTAR (TaKaRa, Japan) and were then infused with psin-EF1, which is used for producing lentiviruses, using the Clone Express II One Step Cloning Kit (Vazyme Biotech). The primers used for cloning PCOLCE and the N29Q mutant are as follows:

psin-PCOLCE-F: 5'-ATT GGG ATC CCC GGA CGG CCA CCA TGC TGC CTG CAG CCA CA-3';

psin-PCOLCE-R: 5'-GCA TGC GGA n class="Chemical">TCA CTA GTG TCA GTC CTG GGA CGC AGC A-3';

N29QF: 5'-AGA CCC CCC AAT ACA CCA GAC CCG n class="Gene">TGT T-3';

N29QR: 5'-n class="Gene">TGT ATT GGG GGG TCT GGC CCT GGG CAA-3'.

The PCOLCE n class="Chemical">CDS was amplified from cDNA derived from U2OS, and the N29Q mutant was generated by an overlap-extension strategy.

pLKO.1-puro was inserted with double-stranded oligonucleotides, which genen class="Species">rates shRNA in the cell line, and the sequences are as follows:

PCOLCE-sh#1-s: 5'-CCG GTG AAG AAA GGA GTC n class="Gene">AGT TAT CCT CGA GGA TAA CTG ACT CCT TTC TTC ATT TTT-3';

PCOLCE-sh#1-as: 5'-AAT TAA AAA n class="Gene">TGA AGA AAG GAG TCA GTT ATC CTC GAG GAT AAC TGA CTC CTT TCT TCA-3';

PCOLCE-sh#2-s: 5'-CCG GCG CTG ACC TTC GAG AAG TTT GCT n class="Gene">CGA GCA AAC TTC TCG AAG GTC AGC GTT TTT-3';

PCOLCE-sh#2-as: 5'-AAT TAA AAA CGC n class="Gene">TGA CCT TCG AGA AGT TTG CTC GAG CAA ACT TCT CGA AGG TCA GCG-3'.

RNA extraction and qRT-PCR

The total RNA was extracted using the Rn class="Chemical">NAprep Pure Cell/Bacteria Kit (Tiangen). The total RNA (500 ng) was reverse-transcribed using a HiScript II 1st Strand cDNA Synthesis Kit (Vazyme). qRT-PCR was performed using the ChamQ Universal SYBR qPCR Master Mix (Vazyme) on LightCycler 480 (Roche). All qRT-PCR samples were repeated at least 3 times. The primer sequences are as follows:

qGAPDH-F: 5'-GGA GCG AGA TCC CTC CAA n class="Gene">AAT-3';

qGAPDH-R: 5'-GGC n class="Gene">TGT TGT CAT ACT TCT CAT GG-3';

qPCOLCE-F: 5'-GTG CGG AGG GGA n class="Gene">TGT GAA G-3';

qPCOLCE-R: 5'-n class="Gene">CGA AGA CTC GGA ATG AGA GGG-3';

qTWIST1-F: 5'-GTC CGC n class="Gene">AGT CTT ACG AGG AG-3';

qTWIST1-R: 5'-GCT n class="Gene">TGA GGG TCT GAA TCT TGC T-3';

qCOL1A1-F: 5'-GAG GGC CAA GAC GAA GAC ATC-3';

qCOL1A1-R: 5'-CAG ATC ACG n class="Chemical">TCA TCG CAC AAC-3';

qCOL1A2-F: 5'-GGC n class="Gene">CCT CAA GGT TTC CAA GG-3';

qCOL1A2-R: 5'-CAC n class="Gene">CCT GTG GTC CAA CAA CTC-3';

qRT-col5A1-F: 5'-TAC AAC GAG CAG GGT ATC CAG-3';

qRT-col5A1-R: 5'-ACT TGC CAT CTG ACA GGT n class="Gene">TGA-3';

qRT-col8A2-F: 5'-GCT GGC TTA GGC AAA n class="Gene">CCT G-3';

qRT-col8A2-R: 5'-GGA CTC CCA CAC n class="Gene">CGT CTA CT-3';

qRT-col10A1-F: 5'-ATG CTG CCA CAA ATA CCC TTT-3';

qRT-col10A1-R: 5'-GGT n class="Gene">AGT GGG CCT TTT ATG CCT-3'.

Dual-Luciferase assay

293T cells were plated in 24-well plates and were transiently transfected with 200 ng of n class="Gene">PCOLCE promoter-containing pGL-3 plasmids together with pTK-cLuc as the normalization control. 48 h later, the luciferase activity was measured for 3 independent experiments using the Dual-Luciferase Assay Kit (Promega).

Assessment of cell migration and invasion

Cells were seeded into Boyden chambers containing 24-well Transwell plates with a pore size of 8 μm (BD Bioscience). The upper chamber was either left uncoated for the migration assay or precoated with 50 μl 1:8 diluted Matrigel (BD Bioscience) for the invasion assay. n class="CellLine">U2OS (5×104), U2OS/MTX300 (5×104) and HOS (3×104) cells were seeded into the upper chamber. The upper chamber was filled with 200 μL serum-free specified medium, whereas the lower chamber was filled with complete medium containing 10% FBS. Following incubation for 12 h (U2OS, HOS) and 24 h (U2OS/MTX300), the cells that had invaded into the lower chamber were fixed with 4% paraformaldehyde and were stained with crystal violet for 1 h at room temperature. The cells in the upper chamber were removed by a cotton swab, and the remaining cells were counted in 5 randomly selected microscopic fields. All experiments were performed in triplicate.

Chromatin immunoprecipitation assay

The chromatin immunoprecipitation assay was performed using a ChIP Kit according to the manufacturer's instructions 22. Briefly, each cell line was seeded into a 15 cm plate. The cells were fixed by adding 10% of the volume of the growth medium. The fixation was stopped by adding 5% of the volume of the stop solution to the existing culture medium. The cells were scratched, collected and lysed to obtain the nuclear pellet. The nuclear pellet was resuspended and sonicated. Five micrograms of antibodies together with protein A/G n class="Chemical">agarose beads was incubated overnight at 4°C. The bound DNA-protein complexes were eluted after a series of washes. Purified DNA was resuspended in TE buffer for qRT-PCR.

The PCR primer sequences are as follows:

CHIP-PCOLCE-EBX1-F1: 5'-AAC AAG GAC ACT CCC TTC AT-3';

CHIP-PCOLCE-EBX2-R1: 5'-TAG AGG CAG GGT TTC AGC AT-3'.

Western Blotting

Western blotting was performed as previously described 22. The cells were collected and lysed in RIPA buffer (150 mM n class="Chemical">NaCl, 0.5% EDTA, 50 mM Tris, 0.5% NP40) and centrifuged for 20 min at 14,000 x g and 4°C. Before being loaded onto the polyacrylamide gel, the protein lysate or concentrated. cell culture supernatant was denatured at 105°C for 10 min in loading buffer (Tris-HCl pH 6.8 60 mM; SDS 2%; bromophenol blue 0.1%; glycerol 25%) supplemented with or without 14.4 mM beta-mercaptoethanol for a reduced or nonreduced reaction, respectively. A total of 20 μg of harvested protein was loaded and separated on a 10% SDS-polyacrylamide gradient gel. The proteins were then transferred onto polyvinylidene difluoride membranes and were blocked with 5% fat-free milk for 2 h at room temperature. The membranes were incubated with the primary antibody and the horseradish peroxidase-conjugated secondary antibody; then, the proteins were detected using an ECL chemiluminescence system (Pierce). The primary antibodies used were as follows: the antibody against GAPDH (13937-1-AP) was from Proteintech, and the antibody against HA (3724) was obtained from Cell Signaling Technology.

PNGase F Treatment

In this study, PNGase F was purchased from NEB corpon class="Species">ration and was used to remove almost all N-linked oligosaccharides from the glycoproteins. The cell lysates were prepared as described in the Western blotting process. According to the PNGase F denaturing protocol, 5 μg of cell lysate was mixed with 1 μl of glycoprotein denaturing buffer (10X) and H2O to make a 10 μl total reaction volume. The mixture was then incubated at 100°C for 10 min and was then chilled on ice for 10 sec. Then, 2 μl GlycoBuffer 2 (10X), 2 μl 10% NP-40, 6 μl H2O, and 1 μl PNGase F was added into the mixture and incubated at 37°C for 1 h.

Supernatant concentration

The cells were cultured in 15 cm dishes with complete media. The media were removed and replaced by 25 ml of DMEM without antibiotics or n class="Disease">FBS. After 48 h, the cell supernatant was loaded onto an Ultracel-10 centrifugal Filter Unit for centrifugation at 4000 rpm at 4°C for 1 h. The concentrated supernatants were lysed in reduced and nonreduced reactions for subsequent Western blotting.

Animal experiments

All animal experiments were performed according to the protocols approved by the Research Animal Resource Center of Sun Yat-Sen University and complied with the guidelines. The animal ethics number of this project is GZR2015-137. All animals were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. For the spontaneous n class="Disease">metastasis osteosarcoma model in vivo, U2OS/MTX300 cells stably expressing luciferase, including the vector, PCOLCE, N29Q, shNC, PCOLCE-sh#1, and PCOLCE-sh#2, were used as described. Six mice were injected separately with 1x106 cells in each group. When the tumor reached 1.5 cm in diameter, the mice were injected with a luciferase substrate, and the IVIS Lumina Imaging System (Xenogen) was used to capture the luciferase signal by taking photographs. The mice were then sacrificed, and the lungs were fixed in formalin (4 μmol/L) and stained with hematoxylin and eosin (H&E).

Statistical analysis

The SPSS software was used for statistical analysis. The statistical significance was tested by Student's t-test or a chi-square test. The curve of correlation between PCOLCE expression level and the overall survival was plotted by the Kaplan-Meier method. Differences were considered statistically significant when p values were <0.05.

Results

PCOLCE is up-regulated in osteosarcoma tissues

By analyzing our RNA-seq data from 4 noncancerous adjacent tissues and 16 n class="Disease">osteosarcoma tissues (Fig. ), we found that 169 genes encoding ECM proteins were overexpressed in osteosarcoma tissues compared to the expression in noncancerous adjacent tissues (sheets 1-2 of Table ). Notably, PCOLCE, a procollagen C-proteinase enhancer protein for stimulating procollagen maturation, was one of the most up-regulated genes in osteosarcoma (Fig. ). Interestingly, data from CCLE (https://portals.broadinstitute.org/ccle/page?gene=PCOLCE) also indicated that the PCOLCE mRNA was dramatically elevated in osteosarcoma cell lines, such as U2OS and HOS cells (Fig. ). The up-regulation of PCOLCE was further confirmed by qRT-PCR using 68 pairs of human osteosarcoma tissues and their adjacent nontumor tissues (Fig. ), whose clinical information are listed as Supplementary Table More importantly, the osteosarcoma patients with high levels of PCOLCE had poorer total survivals compare those with low levels of PCOLCE (Fig. ). The results indicate that PCOLCE is up-regulated in osteosarcoma tissues and may be involved in osteosarcoma progression.

Knockdown of PCOLCE by shRNA inhibits the lung metastasis of osteosarcoma

To determine the function of PCOLCE in n class="Disease">osteosarcoma, we generated two independent short hairpin RNAs against PCOLCE, which could efficiently reduce the endogenous expression of PCOLCE in U2OS, U2OS/MTX300 and HOS cells (Fig. ). The knockdown of PCOLCE significantly inhibited cell migration and invasion in vitro (Fig. ), but did not affect cell viability (Fig. ). More importantly, in the orthotopic osteosarcoma metastasis mouse model using U2OS/MTX300-Luc cells with the stable knockdown of PCOLCE, only one or two out of five mice in the PCOLCE-shRNAs groups, while four out of five mice in the control group had luminescent signals within the lungs (Fig. ). Consistently, the mice bearing PCOLCE-sh#1 or PCOLCE-sh#2 cells had significant decreases in lung metastatic burden compared to those bearing control cells (Fig. ).

PCOLCE promotes osteosarcoma cell migration and invasion

The stable overexpression of PCOLCE in n class="CellLine">U2OS, U2OS/MTX300 and HOS cells was also generated (Fig. ). Cell migration and invasion, but not cell viability, were dramatically promoted in these stable cells compared to those in cells stably overexpressing the empty vector (Fig. ). Furthermore, we purified His-tagged PCOLCE recombinant protein from HEK293T cells transiently expressing His-tagged PCOLCE (Fig. ) to test whether this recombinant protein could promote the migration of osteosarcoma cells. As shown in Fig. , cell migration was significantly promoted in U2OS cells treated with the purified His-tagged PCOLCE recombinant protein in a concentration-dependent manner compared to the cell migration of the control cells. The results suggest that PCOLCE may play a critical role in promoting osteosarcoma metastasis.

The N-glycosylation of PCOLCE is essential for its enhancement of metastasis

Glycosylation modifications are crucial for the ECM-mediated malignant phenotype during cancer 23, and n class="Gene">PCOLCE may have N-linked oligosaccharides modified with sialic residues in human serum 24; we explored which residue of PCOLCE could be glycosylated and the significance of this event. Two putative N-glycosylation sites of PCOLCE were predicted using the N-X-S/T consensus sequence 25 (Fig. ). Therefore, we generated two PCOLCE mutants by individually changing each Asn (N) into Gln (Q). As shown in Fig. , the molecular weight of the N29Q mutant, but not that of the N431Q mutant, was slightly lower than its wild-type counterpart, whereas their molecular weights became identical under PNGase F treatment. As PNGase F specifically removes the N-linked glycosylation, the band shifts of both wild-type and its N431Q mutant, but not of its N29Q mutant, were observed after PNGase F treatment, indicating that PCOLCE has the N-linked glycosylation at N29. Notably, the protein level of the PCOLCE-N29Q mutant was markedly reduced compared to that of its wild-type protein (Fig. ), indicating that the N-linked glycosylation of PCOLCE at Asn29 may impact its protein stability. Indeed, as shown in Fig. , the PCOLCE-N29Q mutant protein had a shorter half-life than that of its wild-type protein. In addition, as shown in Fig. wild-type PCOLCE was detected in both cell lysate and supernatant, while the N29Q mutant was hardly detected in supernatant, indicating that the secretion of the PCOLCE-N29Q mutant into media was much less compared to wild-type PCOLCE. The results suggest that PCOLCE is N-glycosylated at Asn29, which positively regulates its protein stability and its secretion into the ECM.

Next, we asked whether the n class="Chemical">Asn29 glycosylation of PCOLCE is required for its functions in osteosarcoma. As shown in Fig. and Fig. , the enhancement of cell migration and metastasis induced by wild-type PCOLCE was not observed in the N29Q mutant both in vitro and in vivo. This finding indicates that the glycosylation of PCOLCE at Asn29 is necessary for its promotion of osteosarcoma metastasis.

PCOLCE is transcriptionally up-regulated by TWIST1

Since PCOLCE has been reported to be directly regulated by n class="Gene">TWIST1 in mouse embryonic tissues26, and two putative E-box binding sites for TWIST1, -542~-533 and -408~-399, were predicted in the PCOLCE promoter using JASPAR 27 as shown in Fig. . In addition, the expression levels were positively correlated between TWIST1 and PCOLCE in osteosarcoma tissues (Fig. ). Therefore, we speculated that TWIST1 could transcriptionally up-regulate PCOLCE in osteosarcoma cells. This was the case, as the mRNA levels of PCOLCE were significantly increased in both U2OS and HOS cells stably overexpressing TWIST1 compared to those stably expressing the empty vector (Fig. ). To test whether PCOLCE is a direct target of TWIST1, the depletion of a single or double E-box in the dual-luciferase reporter of the PCOLCE promoter was generated. As shown in Fig. , E-box1, but not E-box2, was crucial for the up-regulation of the PCOLCE promoter by TWIST1. Indeed, the association of TWIST1 with the PCOLCE promoter was further verified by a chromatin immunoprecipitation (ChIP) assay (Fig. ). Taken together, our results indicate that TWIST1 transcriptionally up-regulates PCOLCE by directly binding to E-box1 within the PCOLCE promoter.

TWIST1 may also up-regulate some procollagens in osteosarcoma tissues

Since PCOLCE promotes the matun class="Species">ration of procollagens, we determined the status of several procollagens in osteosarcoma tissues. As shown in Fig. , the mRNA levels of several procollagens, such as COL1A1, COL1A2, COL5A1, COL8A2 and COL10A, were positively correlated with TWIST1 in our RNA-seq data (Fig. ). Indeed, the mRNA levels of these procollagens were significantly increased in U2OS cells stably expressing TWIST1 compared to those in cells stably expressing the vector (Fig. ). These results indicate that TWIST1 may also up-regulate some procollagens in osteosarcoma tissues.

Discussion

In this report, we revealed that PCOLCE and some procollagens were up-regulated by n class="Gene">TWIST1 in osteosarcoma cell lines and tissues and that the N-glycosylation of PCOLCE at Asn29 was crucial for promoting osteosarcoma metastasis.

PCOLCE specifically enhances the activity of n class="Gene">BMP-1, a zinc metalloproteinase that modulates collagen deposition in the extracellular matrix by removing C-propeptides from procollagen I, II and III 11. Increased collagen deposition is the most well-recognized ECM alteration during cancer progression 28, suggesting that PCOLCE may be involved in cancer metastasis. Recently, Salza et al. reported that PCOLCE might be involved in tumor growth based on the interaction network of PCOLCE 29. Our results showed that PCOLCE plays a critical role in promoting the metastasis of osteosarcoma, as the overexpression of PCOLCE promotes cell migration, invasion and lung metastasis (Figs. ) whereas the knockdown of PCOLCE had the opposite effects (Fig. ) in osteosarcoma.

A previous study has indicated that PCOLCE may have n class="Chemical">N-linked oligosaccharides decorated with sialic residues in human serum 24. Here, we have demonstrated that PCOLCE is N-glycosylated at Asn29, which is essential for its protein stability, secretion and the enhancement of metastasis in osteosarcoma (Fig. ). Notably, serum PCOLCE may act as a biomarker of diseases, such as muscle 16, liver 30 and cardiac 31 fibrosis. PCOLCE could be detected in the supernatant of osteosarcoma cells (Fig. ), and the purified PCOLCE recombinant protein could promote cell migration in U2OS cells (Figs. ); this indicates that PCOLCE may be a potential noninvasive biomarker to predict metastasis in osteosarcoma patients.

We proved that TWIST1 could up-regulate n class="Gene">PCOLCE by directly binding to the E-box1 of the PCOLCE promoter (Fig. ). In addition, TWIST1 also increased the mRNA levels of procollagens, such as COL1A1, COL1A2, COL5A1, COL8A2 and COL10A (Fig. ), which are potential substrates for PCOLCE. This may provide a new explanation that TWIST1 is a key regulator in cancer metastasis in addition to being a key regulator of EMT32 33 because the correlations between TWIST1 mRNA levels and PCOLCE mRNA levels were also observed in 18 different cancer types from the ChIPBase v2.0 database 34 (Supplementary Table ).

In summary, we propose that TWIST1 may up-regulate both n class="Gene">PCOLCE and procollagens and that PCOLCE helps BMP1 to accelerate the maturation of collagen, which in turn, promotes osteosarcoma metastasis; this is illustrated in Fig. .

Supplementary figures and table 1.

Click here for additional data file.

Supplementary table S1.

Click here for additional data file.

Supplementary table S2.

Click here for additional data file.