SFRP1 and PTGS2 are potential biomarkers of tumorigenic properties in human placenta-derived mesenchymal stem cells

Mesenchymal stem cells (MSCs) display tumour tropism and have been explored as cellular vehicles to deliver anti-cancer agents. As cellular components of the tumour microenvironment, MSCs also inuence tumour progression. However, the tumour transformation-related genes of MSCs are not well understood since either oncogenic or tumour suppressor effects within these cells have been reported. Here, we aimed to identify potential biomarkers capable of tumorigenic risk by RNA-seq analysis of human placenta tissue-derived MSCs (hPTMSCs) exposed to the carcinogenic agent 3-methylcholanthrene (3-MC). Polycyclic aromatic hydrocarbons (PAHs), including 3-MC, are ubiquitous environmental pollutants that have many untoward effects, including carcinogenesis. In the present study, we aimed to identify potential biomarkers of tumorigenic risk by RNA-seq analysis of hPTMSCs exposed to 3-MC. In total, 268 tumour transformation-related genes of MC-induced MSCs were identied by RNA-seq, including 17 upregulated (C1QTNF1, CD14, CHI3L1, CXCL8, CXCL16, GATA3, GCH1, IRAK3, MN1, PID1, PLXDC2, PTGS2, PTK2B, TNFAIP3, TNFAIP6, TNFAIP8 and TNFSF13B) and 12 downregulated (IL18, KRT18, MTUS1, MTUS2, PRAME, PTTG1, SFRP1, TACSTD2, TLR4, TNFAIP8L1, TNFRSF19 and TXNIP) genes. In addition, 66 pluripotency-related genes were identied, including 3 downregulated Notch signalling pathway genes: including JAG1, LFNG and MAML3. We then characterized the expression of these DEGs by qPCR and found that the expression trends of 27 tumour transformation-related DEGs as well as 3 pluripotency-related DEGs in the Notch signalling pathway were consistent with the RNA sequencing results; the other 2 tumour transformation-related DEGs did not change. Frizzled-related which is located within the 8p11-12 region, is the 1 µg Sequencing libraries RNA following and each The index-coded performed on a cBot Generation System using PE Cluster Kit v4-cBot-HS (Illumia) according After generation, library preparations sequenced on Hiseq 2500 platform and paired-end were generated. TopHat2 tools soft were used to map with human genome sequence. Gene expression levels were estimated by FPKM (fragments per kilobase of transcript per million fragments mapped) using Cuffquant and Cuffnorm[27]. Prior to differential gene expression analysis, for each sequenced library, the read counts were adjusted by edgeR program package through one scaling normalized factor. Differential expression analysis of two samples was performed using the DEGseq R package. Pvalue was adjusted using q value. qvalue < 0.005 & |log2(foldchange)| ≥ 1 was set as the threshold for signicantly differential expression.


Results
Twenty-nine tumour transformation-related genes and three pluripotency-related genes were identi ed as differentially expressed genes (DEGs) in hPTMSCs. Importantly, SFRP1 and PTGS2 were further identi ed as tumour suppressors and oncogenes in hPTMSCs, respectively. The overexpression of SFRP1 and PTGS2 was positively correlated with the expression of the pluripotent markers OCT4, CMYC and NANOG in the human lung adenocarcinoma cell line A549, and the low expression of SFRP1 and PTGS2 was inversely correlated with the expression of the pluripotent markers SOX2, OCT4 and NANOG in hPTMSCs. Interestingly, overexpression of SFRP1 led to reduced cell viability, colony formation and migration of A549 cells. In contrast, the ectopic expression of PTGS2 exerted the opposite effect.

Conclusions
These results indicate that hPTMSCs with high PTGS2 expression but low SFRP1 expression may have more potential tumorigenic capacity. Taken together, this study suggests that PTGS2 and SFRP1 may be valuable biomarkers for quality and safety control of hPTMSC preparations in clinical applications, which warrants further study.

Background
Mesenchymal stem cells (MSCs) are excellent candidates for a diverse range of clinical applications in tissue regeneration, cell therapy and immunomodulation [1]. Their characteristics include plastic adherence; trilineage differentiation capacity into osteoblasts, chondrocytes and adipocytes in vitro; and a surface phenotype de ned by the absence of haematopoietic and antigen-presenting surface markers and the presence of MSCcharacteristic antigens CD73, CD90, and CD105 [2]. MSCs can be found in various tissues, including bone marrow, cord blood, adipose tissue and placentas [3]. Among these sites, human placental tissue is accessible and abundant for the isolation of adult stem cells. Therefore, human placenta tissue-derived MSCs (hPTMSCs) have become an attractive source for cell therapy. Ra JC et al reported the safety of intravenous infusion of human adipose tissue-derived MSCs (hATMSCs) in animals and humans [4]. MacIsaac et al also showed that no tumorigenic properties were detected in mice treated with adipose stromal/stem cells [5]. However, in tumour models, MSCs have been reported to play a role in the process of carcinogenesis, tumour growth and metastasis, although numerous studies have reported contradictory results; some investigators found that MSCs promote tumour growth, and others reported that MSCs inhibit tumour growth [6][7][8]. Zhang T et al reported that bone marrow-derived MSCs (BM-MSCs) promote the growth and angiogenesis of breast and prostate tumours [9]. Rodini et al. also found that MSCs enhanced the tumorigenic properties of human glioblastoma through independent cell-cell communication mechanisms [10]. Nevertheless, Qiao L et al. reported that Dkk-1 secreted by MSCs inhibits the growth of breast cancer cells via the depression of Wnt signalling [11]. Yin J also found that MSC-mediated concurrent delivery of endostatin and carboxylesterase to mouse xenografts suppresses glioma initiation and recurrence [12]. Some mechanisms of MSC-related tumour supporting effects have been reported, including vascular [9,13,14] and tumour microenvironment support, immunosuppressive effects and metastatic support [15][16][17]. On the other hand, the tumour inhibitory mechanisms of MSCs could be due to cell cycle arrest [18,19], angiogenesis inhibition [20] and regulation of certain soluble factors, such as the Wnt inhibitor DKK-1, that may block tumour-related cell signalling pathways [21]. Despite these explanations regarding the action of MSCs, the tumour modulation mechanisms of MSCs are extremely complicated and di cult to demonstrate. Thus, the potential tumorigenicity of MSCs should be further explored and monitored to elucidate the risk of potential tumorigenicity related to MSC-based therapies [22].
The accumulation of mutations, genetic predisposition of some factors, exposure to environmental agents, and lifestyle all play important roles in the carcinogenic process. 3-Methylcholanthrene (3-MC), a potent polycyclic aromatic hydrocarbon (PAH), is a ubiquitous environmental pollutant that can be obtained as a by-product of combustion of various organic substances, such as coal, garbage, gas and oil [23]. MC and other PAHs have many untoward effects, including immune suppression, endocrine disruption, wasting syndrome, birth defects and carcinogenesis [24,25]. These compounds suppress the proliferation of human umbilical vascular endothelial cells and induces tumours in mice and rats [26].
Here, we aimed to identify potential biomarkers capable of tumorigenic risk by RNA-seq analysis of hPTMSCs exposed to the carcinogenic agent 3-MC and provide markers for detecting tumorigenicity related to quality and safety control of clinical-grade MSCs in cell therapy.

Results
Sequencing data and quality control RNA samples were extracted from DMSO control and 3-MC-treated hPTMSCs using TRIzol. The Illumina HiSeq 2500 RNA sequencing platform was used to sequence the whole transcriptome of hPTMSCs. Raw data (raw reads) in FASTQ format were rst processed. In this step, clean data (clean reads) were obtained by removing reads containing adapters and low-quality reads from raw data. In addition, the Q20, Q30, GC content and sequence duplication level of the clean data were calculated. All downstream analyses were based on clean data with high quality. After stringent data ltering and quality checks, approximately 62 million high-quality clean reads were obtained from both samples with 87.11% and 86.44% Q30 bases (base quality was more than 30) for DMSO control and 3-MC treated hPTMSCs, respectively. In total, 29,251,305 and 32,739,136 clean paired-end reads as well as 8,730,066,236 and 9,771,496,192 clean bases were generated from DMSO control and 3-MC treated hPTMSCs, respectively (Table 1). These clean reads were then mapped to the reference genome sequence. Only reads with a perfect match or mismatch were further analysed and annotated based on the reference genome. Tophat2 tools software was used to map these reads based on the reference genome. Differentially expressed unigene analysis Unigene expression was calculated by FPKM (fragments per kilobase of transcript per million fragments mapped) [27] using Cuffquant and Cuffnorm. Differential expression analysis was determined by EBsEq. The threshold of the false discovery rate (FDR) was set to less than 0.01 and with a fold change greater than or equal to 2. Through this calculation, the upregulation and downregulation of both DMSO control-and 3-MC-treated transcripts were determined. Ultimately, 1081 differentially expressed genes (DEGs) were identi ed, including 498 upregulated and 583 downregulated genes ( Fig. 1A and 1B). Hierarchical cluster analysis of DEGs is presented in downregulated (IL18, KRT18, MTUS1, MTUS2, PRAME, PTTG1, SFRP1, TACSTD2, TLR4, TNFAIP8L1, TNFRSF19 and TXNIP) ( Fig. 2A, 2C). In addition, 66 pluripotency-related genes were identi ed, including JAG1, LFNG and MAML3, which were downregulated in the Notch signalling pathway (Fig. 2B, 2C).

Functional annotation, gene ontology classi cation and pathway analysis
Functional annotation provided information on protein function annotation, pathway annotation, COG annotation (Clusters of Orthologous Groups of proteins) and GO annotation (Gene Ontology). Unigenes were subjected to sequence alignment with protein databases, such as NR (non-redundant protein sequence database), Swiss-Prot (a manually annotated, non-redundant protein sequence database), GO, KOG (Clusters of Protein homology), Pfam (Homologous protein family) and KEGG (Kyoto Encyclopedia of Genes and Genomes) using BLAST software, which is based on sequence similarities to public databases. In total, 1077 DEGs were successfully annotated. All DEGs were annotated using the NR (1077) followed by eggNOG (1063) GO database is an international standard biological annotation system with three ontologies: biological process, cellular component and molecular function [28]. The GO classi cation of DEGs is presented in Fig. 3A. Pathwaybased analysis provides information and further understanding on how hPTMSCs regulate their biological functions in response to 3-MC at the molecular level. KEGG is a database resource for understanding the highlevel functions and utilities of biological systems, such as cells, organisms and ecosystems, from molecular-level information, especially large-scale molecular datasets generated by genome sequencing and other highthroughput experimental technologies. We used KOBAS software to test the statistical enrichment of DEGs in KEGG pathways. DEGs in the same pathways typically cooperate with each other to exercise their biological functions. In total, 865 DEGs were mapped to the KEGG database. These DEGs were classi ed into 50 KEGG pathways of four main KEGG categories, including cellular processes (129 DEGs), environmental information processing (294 DEGs), human diseases (343 DEGs) and organismal systems (99 DEGs) (Fig. 3B). Out of 50 pathways, the rst 20 pathways (Q value ≤ 1.00) are presented in Fig. 3C.

Veri cation of tumour transformation-related and pluripotencyrelated DEGs
To verify the results of tumour transformation-related and pluripotency-related DEGs in RNA sequencing, we rst characterized the expression of these DEGs by qPCR. The expression trend of 27 tumour transformation-related DEGs as well as 3 pluripotency-related DEGs in the Notch signalling pathway exhibited consistency with the RNA sequencing, whereas the other 2 tumour transformation-related DEGs did not change (Fig. 4A). Among these genes, it has been reported that the downregulation of SFRP1 can lead to the overactivation of Wnt signalling pathways, which promote tumorigenesis in human mammary tissues [29,30]. In addition, NF-kB-mediated in ammation-driven carcinogenesis involves transactivation of prostaglandin-endoperoxide synthase 2 (PTGS2) [31]. PTGS2 is upregulated in various premalignant and malignant tissues [31]. Abnormally high PTGS2 levels contribute to tumorigenesis in various animal models of cancer [32][33][34]. More recently, genomic studies performed on MSCs have provided evidence that these cells express the pluripotent markers Oct-4, Nanog, and Sox2, which clari ed their undifferentiated state [35]. To elucidate the possible roles of SFRP1 and PTGS2 in the 3-MC-induced tumorigenesis of hPTMSCs, gain-and loss-of-function studies were performed in hPTMSCs and the human lung adenocarcinoma cell line A549. Stable SFRP1 and PTGS2 overexpression or silencing in hPTMSCs and A549 cells were achieved using the respective lentiviruses. QPCR and western blotting veri ed that SFRP1 overexpression and silencing in hPTMSCs was accompanied by a signi cant decrease in CMYC and an increase in SOX2 and NANOG. PTGS2 overexpression and silencing was accompanied by a signi cant decrease in CMYC and an increase in SOX2 ( Fig. 4B and 4C). Moreover, SFRP1 overexpression and silencing in A549 cells were accompanied by a signi cant decrease in SOX2 and an increase in OCT4 and NANOG, and PTGS2 overexpression and silencing were accompanied by a signi cant decrease in SOX2 and an increase in OCT4 and NANOG ( Fig. 4D and 4E).
Additionally, CCK-8, colony-formation and wound healing assays showed that overexpression of SFRP1 led to reduced cell viability, colony formation and migration of A549 cells. In contrast, the ectopic expression of PTGS2 resulted in signi cantly increased cell viability, colony formation and migration of A549 cells. These results indicate that hPTMSCs with high PTGS2 expression but low SFRP1 expression may have more potential tumorigenic capacity; cells with low PTGS2 expression and high SFRP1 expression may exhibit less tumorigenic capacity. Taken together, these results suggest that PTGS2 and SFRP1 may represent valuable biomarkers for the quality and safety control of hPTMSC preparations in clinical applications, which warrants further study (Fig. 5).

Discussion
MSCs are multipotent, self-renewing cells with the capacity to differentiate into cells of mesenchymal origin, including osteoblasts, chondrocytes, and adipocytes in vitro and give rise to bone, cartilage, fat and muscle tissues in vivo. Their differentiation potential together with their ease of isolation and cultivation has made them a potentially important source of clinical application in tissue engineering and regenerative medicine.
Clinically, MSCs may aid in the reconstitution of tissues as well as the maintenance of tissue homeostasis, which brings about some concern regarding the potential tumorigenicity of MSC-based therapies [36]. Although some studies suggest that MSCs have inhibitory effects on tumour development, an overwhelming number of studies demonstrate that MSCs exert stimulatory effects on tumour pathogenesis [22]. Djouad et al. showed that MSCs favour earlier onset and growth of syngenic tumours in allogenic animals given their immunosuppressive effects [37,38].
Secreted Frizzled-related protein 1 (SFRP1), which is located within the 8p11-12 region, is homologous to the extracellular cysteine-rich domain of the wingless and integration site growth factor (WNT) receptor Frizzled but lacks the intracellular and transmembrane domains [39]. It has been reported that the downregulation of SFRP1 can lead to the overactivation of Wnt signalling pathways, which promotes tumorigenesis in human mammary tissues [29,30]. A recent focus in cancer prevention involves the modulation of molecular cross-talk between in ammation and cancer growth signalling networks converging on the redox-sensitive transcription factor nuclear factor-kB (NF-kB). NF-kB is directly associated with in ammatory responses with further links to the promotion of carcinogenesis [40][41][42][43]. NF-kB-mediated in ammation-driven carcinogenesis involves transactivation of a major molecular target, prostaglandin-endoperoxide synthase 2 (PTGS2) [31]. PTGS2 is involved in prostaglandin (PG) biosynthesis and in ammation and is upregulated in various premalignant and malignant tissues [31]. Abnormally high levels of PTGS2 contribute to tumorigenesis in various animal models of cancer, and downregulation of PTGS2 is spontaneously preventive [32][33][34]. Various stimuli and mediators are involved in transient induction of NF-kB and transactivation of PTGS2, thus eliciting in ammation and consequent tumorigenesis [31][32][33][34]41]. Thus, targeted inhibition of NF-kB and PTGS2 has become a promising and practical tool in cancer prevention. Reports suggest that diminishing NF-kB-PTGS2 crosstalk may prevent tumorigenesis. More recently, genomic studies performed on MSCs demonstrate that these cells express the pluripotent markers Oct-4, Nanog, and Sox2, clarifying their undifferentiated state [35].
Given this nding, gain-and loss-of-function studies were performed in hPTMSCs and the human lung adenocarcinoma cell line A549. QPCR and western blotting veri ed that SFRP1 overexpression and silencing in hPTMSCs was accompanied by a signi cant decrease in CMYC and an increase in SOX2 and NANOG expression. PTGS2 overexpression and silencing in hPTMSCs was accompanied by a signi cant decrease in CMYC and an increase in SOX2 expression. Moreover, SFRP1 overexpression and silencing in A549 cells was accompanied by a signi cant decrease in SOX2 and an increase in OCT4 and NANOG expression, and PTGS2 overexpression and silencing in A549 cells was accompanied by a signi cant decrease in SOX2 and an increase in OCT4 and NANOG expression. Additionally, CCK-8, colony-formation and wound healing assays showed that overexpression of SFRP1 led to reduced cell viability, colony formation and migration of A549 cells. In contrast, the ectopic expression of PTGS2 exerted the opposite effect.

Conclusions
In our study, 29 tumour transformation-related genes and 3 pluripotency-related genes were identi ed as differentially expressed genes (DEGs) in hPTMSCs exposed to 3-MC based on RNA-seq transcriptional pro le analysis using functional annotation, Gene Ontology classi cation and pathway analysis. Importantly, SFRP1 and PTGS2 were further identi ed as tumour suppressors and oncogenes in hPTMSCs, respectively. Interestingly, overexpression of SFRP1 led to a reduced cell viability, colony formation and migration of A549 cells. In contrast, the ectopic expression of PTGS2 exerted the opposite effect. These results indicate that hPTMSCs with high PTGS2 expression but low SFRP1 expression may have more potential tumorigenic capacity. Cells with low PTGS2 expression and high SFRP1 expression may have less tumorigenic capacity. Taken together, this study suggests that PTGS2 and SFRP1 may be valuable biomarkers for quality and safety control of clinical-grade hPTMSC preparations, which warrants further study.

Cell Culture and treatment
The collection of human placentas was approved by the Ethics Committee of Human Research at Ningxia Medical University. Written consent was obtained from each mother prior to delivery. Three human full-term placentas were obtained from healthy mothers after elective caesarean section in the General Hospital of Ningxia Medical University. Human PTMSCs were isolated as described previously [44]. Cells were cultured in serum-free medium containing MesenCult-XF Basal Medium and MesenCult-XF Supplement (STEMCELL Technologies Inc., Grenoble, France). At about 90% con uence, cells were dissociated using the MesenCult-ACF Dissociation Kit (STEMCELL Technologies Inc., Grenoble, France) and passaged. The hPTMSCs at passage 3 were treated with 3-MC (SIGMA-ALDRICH, Saint Louis, USA)) at a concentration of 5 µg/ml for serial 7days to induce carcinogenesis.
All the studies were conducted with ve passages of the isolated hPTMSCs. Human lung adenocarcinoma cell line A549 were purchased from the Cell Resource Center of the Chinese Academy of Sciences (Shanghai, China). A549 cells were cultured in RPMI-1640 medium containing 10% fetal bovine serum (Gibco, Grand Island, NY, USA). Cells were harvested to conduct assays while con uence reached about 90% using TrypLE™ Express (Invitrogen, USA). All cultures were maintained at 37 °C in a sterile humidi ed incubator with 5% CO 2 .

Flow cytometry
The immunological characterization of hPTMSCs was analyzed by ow cytometry on a FACS Calibur ow cytometer (BD Biosciences, San Diego, CA, USA). All monoclonal antibodies used for ow cytometry were

RNA extraction and sequencing
Total RNA was extracted from DMSO control and 3-MC treated hPTMSCs using TRIzol Reagent (Ambion, America). RNA degradation and contamination were monitored on 1% agarose gels. RNA purity was checked using the NanoPhotometer spectrophotometer (IMPLEN, CA, USA). RNA concentration was measured using Qubit RNA Assay Kit in Qubit 2.0 Flurometer (Life Technologies, CA, USA). RNA integrity was assessed using the RNA Nano 6000 Assay Kit of the Agilent Bioanalyzer 2100 system (Agilent Technologies, CA, USA). A total amount of 1 µg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using NEBNext UltraTM RNA Library Prep Kit for Illumina (NEB, USA) following manufacturer's recommendations and index codes were added to attribute sequences to each sample. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v4-cBot-HS (Illumia) according to the manufacturer's instructions. After cluster generation, the library preparations were sequenced on an Illumina Hiseq 2500 platform and paired-end reads were generated. TopHat2 tools soft were used to map with human genome sequence. Gene expression levels were estimated by FPKM (fragments per kilobase of transcript per million fragments mapped) using Cuffquant and Cuffnorm [27]. Prior to differential gene expression analysis, for each sequenced library, the read counts were adjusted by edgeR program package through one scaling normalized factor. Differential expression analysis of two samples was performed using the DEGseq R package. Pvalue was adjusted using q value. qvalue < 0.005 & |log2(foldchange)| ≥1 was set as the threshold for signi cantly differential expression.

qRT-PCR
Total RNA was isolated using TRIzol Reagent (Ambion, USA). RNA degradation and contamination were monitored on 1% agarose gels. RNA purity and concentration were checked using a Nanodrop™ spectrophotometer (Thermo Fisher Scienti c Inc., San Diego, CA, USA). Then, 5 µg of total RNA was converted to cDNA with TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen Biotech, Beijing, China). Afterwards, real-time PCR was carried out using TransStart Tip Green qPCR SuperMix (TransGen Biotech, Beijing, China) on the Bio-rad IQ5 Real Time PCR System (Bio-Rad Inc., USA). The expression levels of gene transcripts were estimated from the respective standard curves, and normalized to the amount of ACTB or GAPDH endogenous control transcripts by the comparative Ct value. Melting curves for each PCR reaction were generated to ensure the purity of the ampli cation products. Speci c primers sets were synthesized by shenggong (Shanghai, China) and illustrated in Table 2.     in triplicate and the data are presented as the mean ± SD. *P 0.05, **P 0.01 denote signi cant differences compared with the MOCK or sh-NC group.