FSTL1 Promotes Metastasis and Chemoresistance in Esophageal Squamous Cell Carcinoma Through Nfκb-BMP Signaling Crosstalk

Author Manuscript Published OnlineFirst on September 7, 2017; DOI: 10.1158/0008-5472.CAN-17-1411 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

FSTL1 promotes metastasis and chemoresistance in esophageal squamous cell carcinoma through NFκB-BMP signaling crosstalk

Marco Chi-Chung Lau1#, Kai Yu Ng1#, Tin Lok Wong1, Man Tong1, Terence K. Lee5, Xiao-Yan Ming2, Simon Law3, Nikki P. Lee3, Annie L. Cheung1, Yan-Ru Qin6, Kwok Wah Chan4, Wen Ning7, Xin-Yuan Guan2, Stephanie Ma1

1School of Biomedical Sciences, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong; Departments of 2Clinical Oncology, 3Surgery and 4Pathology, Li Ka Shing Faculty of Medicine, Queen Mary Hospital, The University of Hong Kong, Hong Kong; 5Department of Applied Biology and Chemical Technology, Hong Kong Polytechnic University, Hong Kong; 6Department of Clinical Oncology, First Affiliated Hospital, Zhengzhou University, Zhengzhou, China and 7Department of Genetics and Cell Biology, College of Life Sciences, Nankai University, Tianjin, China

#Equal contribution

Corresponding author: Dr. Stephanie Ma, PhD, School of Biomedical Sciences, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Room 47, 1/F, Laboratory Block, Faculty of Medicine Building, 21 Sassoon Road, Pok Fu Lam, Hong Kong. Phone: (852) 3917 9238; Fax: (852) 2817 0857; E-mail: [email protected].

Grant support: This study was supported by funding from the Research Grants Council of Hong Kong - Collaborative Research Fund (C7038-14G) to XY Guan and S Ma.

Short title: FSTL1-mediated NFκB and BMP pathway deregulation in ESCC

Keywords: FSTL1, NFκB, SMAD, BMP, chemoresistance, metastasis, ESCC

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2017 American Association for Cancer Research. Author Manuscript Published OnlineFirst on September 7, 2017; DOI: 10.1158/0008-5472.CAN-17-1411 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Abstract

Esophageal squamous cell carcinoma (ESCC) has a generally poor prognosis and molecular markers to improve early detection and predict outcomes are greatly needed. Here we report that the BMP-binding follistatin-like protein FSTL1 is overexpressed in ESCCs where it correlates with poor overall survival. Genetic amplification of FSTL1 or chromosome 3q where it is located occurred frequently in ESCC, where FSTL1 copy number correlated positively with higher FSTL1 protein expression. Elevating FSTL1 levels by various means was sufficient to drive ESCC cell proliferation, clonogenicity, migration, invasion, self-renewal and cisplatin resistance in vitro and tumorigenicity and distant metastasis in vivo. Conversely, FSTL1 attenution by shRNA or neutralizing antibody elicited the opposite effects in ESCC cells. mRNA profiling analyses suggested that FSTL1 drives ESCC oncogenesis and metastasis through various pathways with deregulation of NFκB and BMP signaling figuring prominently. Crosstalk between the NFκB and BMP pathways was evidenced by functional rescue experiments using inhibitors of NFκB and TLR4. Our results establish the significance of FSTL1 in driving oncogenesis and metastasis in ESCC by coordinate NFκB and BMP pathway control, with implications for its potential use as a diagnostic or prognostic biomarker and a candidate therapeutic target in this disease setting.

Introduction

Esophageal squamous cell carcinoma (ESCC) is the most common histologic subtype of esophageal cancer. The disease is ranked as the eighth most common cancer and the sixth leading cause of cancer mortalities worldwide, with a dismal prognosis (1). The overall prognosis for ESCC is poor due to late presentation, high incidences of tumor recurrence and metastasis, as well as the ability of the tumor to acquire chemoresistance. A better understanding of the recurrently altered genomic and molecular profiles involved in ESCC development, progression and therapy resistance should aid in the identification of novel targets and the development of new therapies for the more effective clinical management of this disease. To date, there is no effective biomarker for the diagnosis or prognosis of ESCC, nor a good targeted therapeutic drug for the adjuvant treatment of the disease.

To identify better molecular markers for early detection and therapeutic targeting, our lab has previously performed transcriptome sequencing on three pairs of patient-derived ESCC samples and their adjacent non-tumor tissue counterparts (2) and successfully identified a number of commonly and differentially expressed genes at a global level. Based on this profiling data, we have identified PTK6 and RAB25 as novel tumor and metastasis suppressors of ESCC (2-3), CD90 to represent a novel ESCC cancer stem cell subpopulation (4), and NRP2 as a novel oncogene in ESCC (5). However, our knowledge of the exact cellular and molecular mechanisms leading to ESCC is incomplete.

Through data mining of our transcriptome sequencing data, we found the transmembrane glycoprotein follistatin-like 1 (FSTL1), belonging to the BM-40/SPARC/osteonectin family (6), to be commonly overexpressed in all three ESCC samples compared with its corresponding non- tumor counterparts (p<0.005). FSTL1 became an obvious candidate gene of interest as it is located on chromosome 3q13.33, which has previously been reported to be a chromosomal amplification hotspot in ESCC (7-9). FSTL1 is involved in the development of different organogenesis including early development of the lung, ureter, central nervous system and skeleton (10-14). FSTL1 has also been implicated to act as an auto-antigen associated with rheumatoid arthritis and to elicit a cardioprotective role in various cardiovascular diseases (15-18). The link between FSTL1 and rheumatoid arthritis is particularly strong, where Tanaka et al. first cloned the gene from synovial 3

tissues of rheumatoid arthritis patients in 1998 and found FSTL1 to be detected in the sera and synovial fluid of patients suffering from the disease (19). Mechanistically, FSTL1 have been shown to regulate the TGF-β/BMP pathway and as a result leading to blockage of erythroid cell differentiation and subsequent apoptosis (20), as well as abnormal skeletal and lung organogenesis in mice (11-12). In addition to TGF-β/BMP, FSTL1 has also been shown to activate NFκB, via both canonical and non-canonical means, to trigger inflammation in osteoarthritis (21) and obesity (22). Only a small number of studies have reported on the role of FSTL1 in cancer, and many of them with contradictory organ-specific roles. FSTL1 has been shown to negatively regulate the migratory and invasive abilities of ovarian, endometrial, renal, lung and nasopharyngeal cancers (23-26). However, FSTL1 was found to be overexpressed in astrocytic brain tumors (27) and also to enhance the metastasis of cancer cells in breast prostate, skin and pancreatic cancers (28-29). A recent study have also found knockdown of FSTL1 to induce mitotic cell death and BIM up-regulation in non-small cell lung carcinoma cells (30). However, to date, the clinical significance and functional role of endogenous and secretory FSTL1 and the molecular mechanism by which it drives ESCC pathogenesis has not been reported. The link between FSTL1 and the crosstalk between BMP and NFκB pathways have also not been implicated in cancer previously.

In this present study, we found both endogenous and secretory FSTL1 to be markedly up- regulated in ESCC tissue and serum samples and that their overexpression was significantly correlated with worst survival of ESCC patients. FSTL1 copy number positively correlated with high FSTL1 expression in both ESCC cell lines and clinical samples, suggesting that FSTL1 gene and/or chromosome 3q amplification contributes, at least in part, to the preferential up- regulation of FSTL1 in ESCC. Overexpression of FSTL1 provoked, whereas silencing FSTL1 abrogated proliferation, clonogenicity, migration, invasion, self-renewal and cisplatin resistance in ESCC in vitro as well as tumorigenicity and distant metastasis in vivo. Co-culture of FSTL1- containing conditioned medium or recombinant FSTL1 in low FSTL1-expressing ESCC cells enhanced the cells’ ability to migrate and invade in vitro as well as form tumors in vivo; whereas treatment of high FSTL1 expressing ESCC cells with a FSTL1 neutralizing antibody resulted in decreased ESCC metastasis. Mechanistically, FSTL1 modulated ESCC tumorigenicity and metastasis through de-regulation of both NFκB and BMP signaling pathways crosstalk, as evidenced by mRNA profiling of ESCC cells with or without FSTL1 stably overexpressed, coupled 4

with functional rescue experiments involving inhibitors for NFκB and TLR4 (21, 31-35). Collectively, our findings provide evidence that FSTL1 functions as an important oncogene in ESCC development and metastasis through canonical NFκB pathway activation and BMP pathway attenuation. This is also the first study to report an NFκB and BMP pathways crosstalk, where they both work in concert to promote ESCC. Clinically, FSTL1 represents a potential diagnostic and prognostic biomarker as well as a therapeutic target in ESCC.

Materials and Methods

Clinical samples For RNA-Seq, 3 pairs of fresh frozen ESCC and their adjacent non-tumor tissue specimens (patients 16, 18, and 19) were randomly selected from Linzhou Cancer Hospital (Henan, China). Patient 16 is a 43-year-old man with a TNM grade of T2N0M0 (moderate differentiation and no sign of lymph node metastasis). Patient 18 is a 60-year-old woman with a TNM grade of T2N0M0 (moderate differentiation and no lymph node metastasis). Patient 19 is a 54-year-old male with a TNM grade of T3N1M0 (moderate differentiation and lymph node metastasis). A total of 73 primary human ESCC and adjacent non-tumor esophageal tissue samples used for FSTL1 expression studies by RT-qPCR were collected from Queen Mary Hospital (Hong Kong). 394 ESCC and 225 non-tumor esophageal formalin-fixed, paraffin-embedded tissue specimens were obtained from a total of 619 patients undergoing esophagectomy at Linzhou Cancer Hospital and Queen Mary Hospital and used for tissue microarray construction and immunohistochemical analysis. Serum samples from a total of 104 ESCC patients (75% male, 25% female, age range 22 to 89) and 30 normal healthy individuals (66.7% male, 33.3% female, age range 21 to 58) were collected from Queen Mary Hospital and used for ELISA analysis for secretory FSTL1 expression. Above patients had received no previous local or systemic treatment prior to operation. An additional 37 ESCC clinical samples, of which 22 received preoperative chemotherapy treatment and 15 did not were also collected from Queen Mary Hospital for immunohistochemistry work. All samples used in this study were approved by the committee for ethical review of research involving human subjects at Zhengzhou University and The University of Hong Kong / Hospital Authority Hong Kong West Cluster, with informed consent from the subjects.

Cell lines ESCC cell line EC109 was kindly provided by Professor George Tsao (School of Biomedical Sciences, The University of Hong Kong). ESCC cell lines HKESC1, EC9706 and KYSE520 were kindly provided by Professor Gopesh Srivastava (Department of Pathology, The University of Hong Kong). ESCC cell lines KYSE150 and KYSE510 were obtained from DSMZ, the German Resource Centre for Biological Material, Germany (36). 293FT and 293FN cells used for lentiviral transduction and human umbilical vein endothelial cells (HUVEC) were purchased from Invitrogen. All cell lines used in this study were obtained between 2013 to 2016, regularly 6

authenticated by morphologic observation and AuthentiFiler STR (Invitrogen) and tested for absence of mycoplasma contamination (MycoAlert, Lonza). Cells were used within 20 passages after thawing.

Reagents Normal goat IgG control and FSTL1 neutralizing antibody (clone 22B6) used for functional studies was purchased from R&D Systems and obtained from Professor Wen Ning (Nankai University, China), respectively. Human recombinant FSTL1 protein used for functional studies was purchased from R&D Systems (1694-FN-050). NFκB inhibitor (IMD-0354) and TLR4 inhibitor (C34) were purchased from Tocris. Mitomycin C was purchased from Calbiochem.

Quantitative real-time qPCR Total RNA was extracted using RNA-IsoPlus (Takara) and cDNA was synthesized by PrimeScript RT Master Mix (Takara). RT-qPCR was performed with EvaGreen qPCR MasterMix (Applied Biosystems) and the following primers FSTL1: F (5’- GCCATGACCTGTGACGGAAA -3’); R (5’- CAGCGCTGAAGTGGAGAAGA -3’) and β-actin: F (5’- CATCCACGAAACTACCTTCAACTCC-3’); R (5’- GAGCCGCCGATCCACACG-3’) on an ABI Prism 7900 System with data analyzed using the ABI SDS v2.3 software (Applied Biosystems). Relative expression differences were calculated using the 2- ΔΔCt method.

Western blot Protein lysates were quantified and resolved on a SDS-PAGE gel, transferred onto a PVDF membrane (Millipore) and immunoblotted with a primary antibody, followed by incubation with a secondary antibody. Antibody signal was detected using an enhanced chemiluminescence system (GE Healthcare). The following antibodies were used: FSTL1 (1:2000; AF1694, R&D Systems), p- IκBα (1:1000, 2859, Cell Signaling Technology), total IκBα (1:1000, 4812, Cell Signaling Technology), p-SMAD1 (1:1000, 5753, Cell Signaling Technology), total SMAD1 (1:1000, 6944, Cell Signaling Technology), p-p65 (1:1000, sc101749, Santa Cruz), total p65 (1:1000, sc372, Santa Cruz), β-actin (1:10,000; Sigma-Aldrich) and histone H3 (1:1000, ab24834, Abcam). For collection of conditioned medium from ESCC cell lines, cells were first serum-starved for 24 hrs and conditioned medium was subsequently collected. Protein in concentrated conditioned medium was resolved on SDS-PAGE and Silver Staining (PlusOne Silver Staining Kit, GE Healthcare) was 7

performed to normalize the amount of protein. ImageJ software was used for densitometric analyses of Western blot bands, and the quantification results were normalized to the loading control.

Immunohistochemistry Slides were heated for antigen retrieval in 10 mM sodium citrate (pH 6.0). Endogenous peroxidase activity was inhibited with 3% hydrogen peroxide. Sections were subsequently incubated with monoclonal FSTL1 antibody (clone 22B6) (Prof. Wen Ning, Nankai University, China). Reaction was developed with LSAB+ System-HRP (Dako). Slides were counterstained with Mayer’s hematoxylin. Evaluation for FSTL1 expression was performed by a pathologist who had no prior knowledge of patient data. Staining intensity was divided into three scores: no/low, medium and high. Note only expression in the squamous epithelium lining of the non-tumor esophageal samples were taken into account.

Enzyme-linked immunosorbent assay (ELISA) Serum FSTL1 and secretory FSTL1 in conditioned media was quantified using the human FSTL1 ELISA Kit (Cloud-Clone Corp.).

Fluorescence in situ hybridization (FISH) Dual-color FISH was undertaken using BAC clone at 3q13.33 covering the FSTL1 gene (RP11- 343B4) labeled in Spectrum Green (Vysis) and a reference BAC clone localized to the centromere of chromosome 3 (RP11-301H7), labeled in Spectrum Red (Vysis). BACs were obtained from the BACPAC Resource Center at the Children’s Hospital Oakland Research Institute, USA. FISH reactions were performed as previously described (37). Stained sections were counterstained with DAPI anti-fade solution and were examined under a Zeiss Axiophot microscope equipped with a triple-band pass filter. Samples were scored as normal (2 copies) and abnormal (> 2 copies).

Lentiviral transduction FSTL1 shRNA expression vector (NM_007085) and the scrambled shRNA non-target control (NTC) were purchased from Sigma-Aldrich. Sequence of the shRNAs directed against FSTL1 are (clone ID 352) 5’- CCGGCCAGGTTGATTACGATGGACACTCGAGTGTCCATCGTAATCAACCTGGTTTTTG -3’ and (clone ID 1185) 5’ 8

CCGGGTCGCCAAATCACCAGTATTTCTCGAGAAATACTGGTGATTTGGCGACTTTTTG – 3’; sequence of NTC is 5’- CCGGCAACAAGATGAAGAGCACAACTCGAGTTGGTGCTCTTCATCTTGTTGTTTTT -3’. Sequences were transfected into 293FT cells, packaged using MISSION Lentiviral Packaging Mix (Sigma-Aldrich) and used to infect EC109 ESCC cells to establish cells constitutively repressing FSTL1. FSTL1 lentiviral overexpression (EX-A1145-Lv105) or empty vector control plasmids were purchased from GeneCopoeia. Sequences were transfected into 293FN cells, packaged using Lenti- Pac HIV expression packaging mix (GeneCopoeia) and used to infect KYSE150-Luc ESCC cells to establish cells constitutively expressing FSTL1. Stable clones were selected with puromycin.

Conditioned medium preparation for functional studies Stable FSTL1 expressing cells and empty vector transfected cells were cultured in DMEM medium with 10% FBS until 70% confluent. After complete removal of the culture medium, these cells were then continually cultured in DMEM medium supplemented with 3% FBS for 24h before medium collection. Culture medium was then centrifuged at 1000g for 30 min and the supernatant was collected as conditioned medium for further studies.

XTT cell proliferation assay XTT cell proliferation assay (Roche) was performed according to manufacturer’s instructions. Absorbency was measured with a spectrophotometer (Victor3 1420 Multilabel Plate, Perkim Elmer) at 450nm with a reference wavelength of 750nm.

Colony formation assay Cells were seeded in a 6-well plate at a low density and cultured in complete medium for 2 weeks with medium change every 2 days. Colonies were counted and stained with 2% crystal violet (Sigma-Aldrich).

Cell motility and invasion assays Migration and invasion assays were conducted in 24-well Millicell hanging inserts (Millipore) and 24-well BioCoat Matrigel Invasion Chambers (BD Biosciences), respectively. Cells re-suspended in serum free medium were added to the top chamber and medium supplemented with 10% FBS was added to the bottom chamber as a chemoattractant. After 48 hrs of incubation at 37oC, cells that migrated or invaded through the membrane (migration) or Matrigel (invasion) were fixed 9

and stained with 2% crystal violet (Sigma-Aldrich). The number of cells was counted in 3 random fields under 40x objective lens.

Spheroid formation assay Single cells were cultured in 300 µl of serum-free DMEM/F12 medium (Invitrogen) supplemented with 20 ng/ml human recombinant epidermal growth factor (Sigma-Aldrich), 10 ng/ml human recombinant basic fibroblast growth factor (Sigma-Aldrich), 4 µg/ml insulin (Sigma-Aldrich), B27 (1:50; Invitrogen), 500 U/ml penicillin, 500 µg/ml streptomycin (Invitrogen) and 1% methylcellulose (Sigma-Aldrich). Cells were cultured in suspension in poly-HEMA-coated 24-well plates. Cells were replenished with 30 µl of supplemented medium every second day. To propagate spheres in vitro, spheres were collected by gentle centrifugation and dissociated to single cells using TrypLE Express (Invitrogen). Following dissociation, trypsin inhibitor (Invitrogen) was used to neutralize the reaction, and the cells were cultured to generate the next generation of spheres.

Annexin V apoptosis assay Cells were treated with cisplatin (1µg/mL) for 48 hrs. Following treatment, cells were harvested and stained with propidium iodide (PI) and FITC-conjugated Annexin V as provided by the Annexin V-FLUOS Staining Kit (Roche). Samples were analyzed on BD FACSCanto II (BD Biosciences) with data analyzed by FlowJo (Tree Star Inc.).

In vivo subcutaneous tumorigenicity and tail vein metastasis models The study protocol was approved by and performed in accordance with the Committee of the Use of Live Animals in Teaching and Research at The University of Hong Kong. Tumorigenicity was determined by subcutaneous injection into the flank of 4-5 weeks old female Balb/c nude mice (n = 9). For conditioned medium treatment, ESCC cells were first pretreated with conditioned medium for 24 hrs and then co-injected subcutaneously into the flank of 4-5 weeks old female Balb/c nude mice (n = 5), with conditioned media replenished subcutaneously directly around the injection site every 3 days. For neutralizing antibody treatment, ESCC cells were subcutaneously injected into the flanks of 4-5 weeks old female Balb/c nude mice. When tumor volume reaches ~100 mm3, 25 µg of FSTL1 neutralizing antibody or IgG control were injected intratumorally every other day. Mice were sacrificed two days after the sixth treatment (n = 5). Tumor incidence 10

and tumor size were recorded. Tumor volumes were calculated as volume (cm3) – L x W2 x 0.5 with L and W representing the largest and smallest diameters, respectively. Tumors formed were harvested for histological analysis. For tail vein metastasis, luciferase-labeled cells were intravenously injected into 6-7 weeks old Balb/c nude mice (n = 10) through the lateral tail vein. Seven weeks after implantation, mice were administered with 100 mg/kg D-luciferin via peritoneal injection 5 mins before bioluminescent imaging (IVISTM 100 Imaging System, Xenogen). Lungs of the mice were collected and fixed for paraffin sectioning. Sections were stained with H&E for histological examination for sign of tumor growth and metastasis.

Agilent cDNA microarray and analysis Total RNA was harvested and purified using RNeasy Plus Mini Kit (Qiagen). Quality of the RNA was analyzed on Agilent 2100 Bioanalyzer at Centre for Genomic Sciences at the University of Hong Kong. SurePrint G3 Human Gene Expression v3 8x60K Microarray Kit (Agilent Technologies) covering 26,083 Entrez genes was used for gene expression analysis. Microarray profiling was performed as a service at Macrogen Inc. (Seoul). Statistical significance of the expression data was determined using fold change. Hierarchical cluster analysis was performed using complete linkage and Euclidean distance as a measure of similarity. Data analysis and visualization of differentially expressed genes was conducted using R 3.1.2. Pathway analyses were performed using Gene Set Enrichment Analysis (GSEA) and Ingenuity Pathway Analysis (IPA) softwares. Statistical significance of the expression data was determined using z-score and p-value.

Statistical analysis Statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software, Inc.) and SPSS version 21.0 (IBM). Independent Student’s t-test was used for the analysis of all in vitro and in vivo assays. Clinico-pathological significance in clinical samples was evaluated by χ2 test or Fisher’s exact test for categorical data and independent Student’s t-test for continuous data. Kaplan-Meier analysis and log-rank test were used for survival analysis. Statistical significance was defined as p≤0.05 (* stands for p≤0.05, ** stands for p≤0.01 and *** stands for p≤0.001). Error bars represent SEM of the data.

Results

Transcriptome sequencing profiling identifies FSTL1 to be frequently overexpressed in ESCC. Transcriptome sequencing was previously performed in our lab on three pairs of matched tumor and adjacent non-tumor tissues from patients of Chinese origin with ESCC (patients 16, 18 and 19) using the Illumina Genome Analyzer IIx platform (Solexa) (GSE29968) (2). FSTL1 was a candidate gene of interest due to increased expression in all three ESCC samples compared with its corresponding non-tumor counterparts (Fig. 1A left; p<0.005) and that the chromosomal location (3q) on which FSTL1 is localized, has previously been reported to be frequently amplified in ESCC (7-9). RT-qPCR was performed to validate our RNA-Seq finding, and the results were concordant with the transcriptomic results (Fig. 1A right). To determine whether the up- regulation of FSTL1 was a common event, we extended our analysis to an additional 73 paired non-tumor and primary ESCC samples. By RT-qPCR, FSTL1 was found to be up-regulated (defined by a fold change increase of greater than 2) in over 57% of primary ESCCs when compared with their corresponding non-tumor esophageal tissue samples. On average, ESCC displayed an average of 2.57-fold higher FSTL1 expression than non-tumor esophageal tissue (Fig. 1B, Supplementary Table 1). Proteomic FSTL1 levels was also examined by immunohistochemistry using a tissue microarray comprising of 225 non-tumor and 394 ESCC cases in which FSTL1 was consistently detected in the majority of the ESCC sectors but at reduced levels in the normal esophageal epithelia (Fig. 1C, Supplementary Fig. 1). Note only expression in the squamous epithelium lining of the non-tumor esophageal samples (highlighted in red dotted lines in Fig. 1C) was taken into account. 84% of all non-tumor samples displayed low or medium FSTL1 expression levels, whereas only 56.1% of the tumor samples showed low or medium FSTL1 expression level. In contrast, only 16% of the non-tumor samples showed high FSTL1 levels as compared to 43.9% in ESCC samples. High FSTL1 expression correlated with “older age”, as defined by age over the median in our cohort sampled (>59), but did not correlate with other clinic-pathological parameters like pathological stage, tumor size, lymph node metastasis nor differentiation (Supplementary Table 2). Importantly, medium to high FSTL1 expression significantly correlated with a worst overall survival rate. Patients with low FSTL1 expression had, on average, a longer survival time (47 months, 95% confidence interval 10.28) and better prognosis as compared to patients with medium and high FSTL1 expression (27 months, 95% confidence

interval 6.70 and 25 months, 95% confidence interval 4.64, respectively) (Fig. 1D). Thus, endogenous FSTL1 overexpression is tightly associated with adverse ESCC pathogenesis.

High serum FSTL1 expression is tightly associated with older age and represents a sensitive and specific biomarker for ESCC diagnosis. An old study by Tanaka et al. have suggested FSTL1 to exist in a soluble form, where it can be detected in the sera and synovial fluid of patients suffering from rheumatoid arthritis (19). To investigate whether serum FSTL1 levels was altered in ESCC patients, serum from 30 healthy individuals and 104 ESCC patients were collected for ELISA analysis. A marked difference between the two groups was observed (Fig. 1E, Supplementary Table 3).

Gene amplification as a mechanism of FSTL1 overexpression in ESCC. In addition to clinical samples, endogenous and secretory FSTL1 expression was also examined in a panel of esophageal cancer cell lines by Western blot. 5 of the 6 ESCC cell lines examined showed a range of endogenous FSTL1 expression, which we later used for selection of which cells to use for overexpression and knockdown studies (Fig. 2A). Overexpression of oncogenes in cancers is often associated with gene-copy number or chromosomal amplification (38). Because the locus of FSTL1 at 3q has previously been reported to be frequently amplified in ESCC (7-9), we performed FISH analysis to examine whether gene and/or chromosomal amplification represents a mechanism to FSTL1 overexpression in ESCC (Fig. 2B). FISH analysis was carried out on both ESCC cell lines as well as formalin-fixed paraffin-embedded sections from 45 non-tumor and 51 ESCC tissue samples. Chromosomal region encompassing FSTL1 was not found to be amplified in the PBMC from healthy individual control nor the 2 ESCC cell lines with absent or low FSTL1 expression levels (luciferase labeled KYSE150 (KYSE150-Luc) and HKESC1). Most cells only displayed 2 signals representing chromosome region of FSTL1 in green and 2 control signals in red (Fig. 2C). In contrast, amplification was evident in the 2 ESCC cell lines that express high levels of FSTL1 (EC109 and EC9706), where most cells displayed an abnormal FSTL1 copy number (Fig. 2C). Of the 45 non-tumor and 51 ESCC tissue samples tested, FSTL1 copy number abnormality (ranging from 3 to 6) was found in ~7% (3 of 45) non-tumor epithelial tissue, while >60% (31 of 51) ESCC cases displayed FSTL1 gene amplification (Figs. 2D, E). More importantly, FSTL1 copy number abnormality positively correlated with higher proteomic FSTL1 expression (Fig. 2F).

These observations provide evidence in support of the notion that gene or chromosome 3q amplification is, at least in part, implicated in FSTL1 overexpression in human ESCC.

FSTL1 contributes to augmented tumorigenic, metastatic and chemoresistance potential in ESCC. To ascertain whether there is causative relationship between FSTL1 overexpression and altered cancer and stem cell-like phenotype in ESCC, we generated stably overexpressed FSTL1 KYSE150-Luc and stably repressed FSTL1 EC109 cells by lentiviral-based transduction. The successful overexpression and knockdown of FSTL1 was successfully validated at proteomic levels (Fig. 3A). Overexpressing FSTL1 in KYSE150-Luc resulted in a significant increase in the efficiency of cells to proliferate (Fig. 3B) and form foci (Fig. 3C) when compared with empty vector (EV) controls. Conversely, repressing FSTL1 expression in EC109 cells yielded an opposing effect (Figs. 3B, C). Next, we also investigated whether overexpression of FSTL1 has a promoting effect on ESCC cell migration and invasion. Stable FSTL1 overexpression led to a significant increase in the ability of cells to migrate and invade through an extracellular matrix coating. On the contrary, stable knockdown of FSTL1 resulted in opposing effects (Figs. 3D, E). Similar results were also obtained when the same experiment was performed in the presence of mitomycin C, where cells were inhibited to proliferate, suggesting that FSTL1-mediated migration and invasion is not a misinterpretation of the cells’ altered ability to proliferate (Supplementary Fig. 2). Stem cell-like properties of cancer cells are thought to contribute to cancer relapse due to their ability to self-renew and resist chemotherapy (39). Thus, we also evaluated whether overexpression of FSTL1 can promote self-renewal and confer resistance to chemotherapy. FSTL1 overexpression enhanced the self-renewing ability of ESCC cells as demonstrated by spheroid formation assays, while FSTL1 knockdown resulted in the contrary (Fig. 3F). To examine whether FSTL1 confers chemoresistance advantage to ESCC cells, Annexin V apoptosis flow cytometry analysis was performed on cisplatin-treated ESCC cells with or without FSTL1 expression modulated. The apoptotic index of FSTL1-expressing ESCC cells was significantly lower than that of empty vector only control cells (26.98% vs. 73.33%) after a 48 hour exposure to cisplatin; while the apoptotic index of FSTL1-repressed ESCC cells was significantly higher than that of knockdown control cells (56.80% and 66.10% vs. 35.98%) (Fig. 3G). To further obtain evidence in support of this observation, we analyzed FSTL1 expression on tissue sections obtained from our recently established cisplatin-resistant KYSE520 ESCC xenograft model. KYSE520 ESCC xenografts in NOD- SCID mice were treated with varying doses of cisplatin (0, 1, 2.5 and 5mg/kg), with treatment 14

resulting in variable tumor inhibition among the xenografts in a dose dependent trend (40). Interestingly, IHC results showed a stepwise enrichment in the frequency of FSTL1(+) cells in xenograft tumors treated with increasing doses of cisplatin (Fig. 3H). To further extrapolate this to a clinical setting, we next studied whether chemotherapy would enrich FSTL1(+) cells in ESCC clinical samples. To this end, 22 and 15 ESCC clinical samples were collected from patients who had or had not undergone preoperative chemotherapy treatment prior to esophagectomy, respectively. Although not statistically significant (p=0.0767), possibly due to a small sample size as samples of this sort is hard to find, a trend towards increased FSTL1(+) cells in ESCC patients who had received chemotherapy prior to surgery was indeed observed (Supplementary Fig. 3). To further substantiate the role of FSTL1 in driving ESCC tumorigenicity and metastasis, subcutaneous tumorigenicity and experimental tail vein metastasis models were also conducted in vivo. KYSE150-Luc cells with FSTL1 stably overexpressed formed larger tumors as compared to EV control cells when injected subcutaneously into immunodeficient mice (Fig. 4A top; n=9). Histologic analysis of the xenografts revealed tumor cells to form packed nests and invasive cords surrounded by a desmoplastic stroma, suggestive of a poorly differentiated carcinoma (Fig. 4A bottom). Likewise, KYSE150-Luc cells with FSTL1 stably overexpressed displayed a superior ability to metastasize to the lung following cell injection through the tail vein as compared to EV controls, as supported by stronger bioluminescence signal in the lung (Figs. 4B, C; n=10). Histological analysis revealed significantly more and larger metastatic foci in the harvested lung tissues of mice injected with ESCC cells with FSTL1 overexpressed (Fig. 4D). These observations suggest a role for FSTL1 in the regulation of proliferation, metastasis and chemoresistance in ESCC.

Secretory FSTL1 expression promotes while suppression of FSTL1 by neutralizing antibody attenuates aggressive features of ESCC cells. In light of the critical roles of endogenous FSTL1 in conferring multiple cancer and stem cell-like properties in ESCC and the clinical relevance of both endogenous and secretory FSTL1 in ESCC, we hypothesize that secretory FSTL1 may also be crucial in driving these oncogenic properties. To this goal, in vitro and in vivo experiments were carried out in three different settings – (1) FSTL1 overexpressing conditioned medium, (2) recombinant FSTL1 and (3) FSTL1 neutralizing antibody (41) – all aimed at to examine the functional roles of secretory FSTL1 in ESCC (Figs. 5A-C top). Treatment of low FSTL1 expressing KYSE150-Luc ESCC cells with FSTL1 overexpressing conditioned medium or human recombinant FSTL1 led to a significant increase in the ability of cells to migrate and invade (Figs. 5A, B). In 15

addition, treatment of KYSE150-Luc cells with FSTL1 conditioned medium also promoted tumor growth potential in vivo, as demonstrated by increased proliferative rate, end-point tumor volume and tumor weight (Fig. 5D). Next, we then treated high FSTL1 expressing EC109 ESCC cells with a homemade monoclonal FSTL1 neutralizing antibody (41) to examine its therapeutic potential in suppressing the oncogenic and metastatic properties of ESCC. EC109 displayed a diminished ability to migrate and invade when treated with FSTL1 neutralizing antibody, as compared to IgG control (Fig. 5C). In vivo, treatment of ESCC cells with the FSTL1 neutralizing antibody resulted in a marked diminished ability of the cells to grow and even led to a regression in the tumor size, as compared to IgG control (Fig. 5E).

FSTL1 drives ESCC through crosstalk between NFκB and SMAD signaling pathways. To better understand the molecular mechanism underlying FSTL1-driven ESCC tumor progression, we compared the transcriptome profiles of KYSE150-Luc ESCC cells with or without FSTL1 stably overexpressed by Agilent microarray profiling. Using a fold change cutoff of ≥1.5, 1399 probes corresponding to 723 protein-coding genes (471 up and 252 down) were found deregulated. Differential expression pattern between EV control and FSTL1 OE cells was visualized using a hierarchical clustering heat map (Fig. 6A). Pathway analysis was also carried out using Gene Set Enrichment Analysis (GSEA), where 10 gene sets were found to be significantly enriched with a FDR≤0.05 (Supplementary Table 4). In addition to TGFβ signaling (NES=1.61; FDR=0.036), which FSTL1 has previously been shown to be involved via antagonizing BMP (11-13, 20, 41), the other two most significantly enriched gene sets identified were TNFα signaling via NFκB (NES=1.99; FDR=0.0) and epithelial mesenchymal transition (EMT) (NES=1.87; FDR=0.002) (Fig. 6B and Supplementary Table 4). Similar results were also obtained through IPA analysis where both the NFκB complex and TNF family were identified to be significantly activated downstream effectors in FSTL1 overexpressing ESCC cells (Fig. 6C and Supplementary Table 5). IPA analysis predicted 33 deregulated genes identified in our microarray to be regulated by NFκB and of these, 10 of them were also found in the list predicted by GSEA (Fig. 6D), including BIRC3, CCL2, CCL3, CD69, GFPT2, IL1A, IL1B, IL23A, IL6 and TNFRSF9, and for this reason, were selected for RT-qPCR validation. Correlation between RT-qPCR and microarray data indicate a high level of concordance between the differential expression measurements from both platforms (Fig. 6D). More importantly, phosphorylation of key members of the NFκB pathway, namely IκBα and p65, were also found to be significantly altered at the proteomic level, not only in FSTL1 overexpressing 16

KYSE150-Luc cells, but also in FSTL1 repressed EC109 cells, KYSE150-Luc cells treated with EV or FSTL1 OE conditioned medium, KYSE150-Luc cells with control or recombinant FSTL1 and EC109 cells treated with IgG control or FSTL1 neutralizing antibody, suggesting NFκB to be a critical downstream effector of FSTL1-mediated ESCC (Fig. 6E). A number of studies in the past have reported FSTL1 to act as a BMP signaling antagonist in development (11-13, 20, 41). Here, we also found altered phosphorylated SMAD1 expression in ESCC cells with or without FSTL1 modulated. More specifically, p-SMAD1 was found to be down-regulated in ESCC cells with FSTL1 stably overexpressed or treated with FSTL1 OE conditioned medium or recombinant FSTL1. Conversely, p-SMAD1 was up-regulated in ESCC cells with FSTL1 stably repressed or treated with FSTL1 neutralizing antibody (Fig. 6E). This observation coincides with our previous pathway enrichment analysis where we found TGF-β signaling to be positively enriched in ESCC cells with FSTL1 overexpressed (Fig. 6B) and together, suggests that TGF-β/BMP/SMAD is also a critical effector of FSTL1-mediated ESCC.

To further validate the role of NFκB activation in FSTL1-induced tumorigenicity, metastasis and stemenss, we analyzed the impact of introducing IMD-0354, an inhibitor of NFκB, into ESCC cells with FSTL1 stably overexpressed on these altered phenotypes. Successful suppression of the NFκB pathway following addition of IMD-0354 in ESCC cells with FSTL1 stably overexpressed was confirmed by a diminished p-IκBα and p-p65 protein expression (Fig. 7A). IMD-0354 suppressed the cancer and stem cell-like properties conferred by FSTL1 overexpression, as evidenced by the diminished abilities of ESCC cells to migrate, invade, form spheroids and proliferate (Fig. 7B). Similar functional observations were also noted when IMD-0354 was used against ESCC cells treated with FSTL1 OE conditioned medium (Supplementary Fig. 4A) and ESCC cells treated with recombinant FSTL1 (Supplementary Fig. 4B). Previous study have demonstrated the existence of a crosstalk between TLR4/MyD88/NFκB and BMP/SMAD signaling in osteoblastic differentiation (42). Given that we observed both NFκB and SMAD to be deregulated in FSTL1-mediated ESCC, we went on to examine the link between these two pathways and noted that addition of IMD-0354 led to a rescue in p-SMAD1 signaling (Fig. 7A), suggesting that NFκB acts upstream of the BMP pathway and crosstalk between the two pathways do exist and act in concert to promote ESCC.

Previous studies have demonstrated effects of FSTL1 on TLR4/MyD88/NFκB pathway (21, 31-35). To examine the role of TLR4 in mediating FSTL1 induced NFκB activation in ESCC, we performed 17

rescue experiments utilizing an inhibitor of TLR4 (C34). Inhibition of TLR4 attenuated expression of p-IκBα and p-p65, promoted p-SMAD1 expression and a concomitant diminished abilities of ESCC cells to migrate, invade, form spheroids and proliferate (Figs. 7C, D). Again, similar functional observations were also noted when C34 was used against ESCC cells treated with FSTL1 OE conditioned medium (Supplementary Fig. 5A) and ESCC cells treated with recombinant FSTL1 (Supplementary Fig. 5B). Taken together, our results suggest that FSTL1 promotes ESCC through stimulating NFκB activation, attenuating BMP signaling and modulating the crosstalk between these two pathways (Fig. 7E).

Discussion

FSTL1 has previously been reported to play both oncogenic and tumor suppressive roles in various cancer types (23-30). Yet, to the best of our knowledge, this is the first study to report its significance in ESCC. FSTL1 was first systematically identified to be frequently up-regulated in ESCC as compared to adjacent non-tumor esophageal tissues by RNA-Seq profiling. Amongst the other genes that were also found to be de-regulated in ESCC, we chose FSTL1 to focus on as first, the gene is located on chromosome 3q, which is a known chromosomal amplification hotspot in ESCC (7-9) and second, because the gene encodes for a secretory protein. The late clinical presentation of ESCC often limits the success of treatment at the stage of discovery. Thus far, no serum screening methods have been employed in a clinical setting for neither ESCC diagnosis nor prognosis. Thus, identification of a functionally important secretory protein as a biomarker is urgently needed. Data from our study found FSTL1 to have obvious potential to be developed into a diagnostic and prognostic biomarker, especially for non-invasive screening in high-risk areas, for instance China. This is the first study to identify the significance of secretory FSTL1 in ESCC and any tumor type. The markedly lower FSTL1 levels in healthy individual patient sera in our study allows for a good prediction of ESCC by FSTL1. However, since the number of healthy individuals samples included in our current study is limited (n=30), a study with a larger cohort of samples is required for further validation. Since our functional studies identified a role of FSTL1 in conferring resistance of ESCC cells to the chemo-drug cisplatin, it would also be worthwhile to explore whether FSTL1 can be used to predict response to cisplatin treatment and/or tumor recurrence following cisplatin therapy.

Gene amplification is one of the most common mechanisms of gene up-regulation in cancer. Our results here suggest gene amplification of FSTL1 and/or chromosome 3q to be tightly correlated with high FSTL1 protein expression, suggesting gene amplification may at least in part be implicated in FSTL1 activation in human ESCC. Yet with that said, it is unlikely that gene amplification represents the sole mechanism leading to FSTL1 up-regulation in ESCC as we also found a small portion of clinical samples which did not have chromosome 3q amplified but had FSTL1 overexpressed. A recent study in prostate cancer have suggested the involvement of androgen receptor transcriptional activity in driving FSTL1 up-regulation (29). FSTL1 is also identified as a TGF-β inducible gene (43). Studies in rheumatoid arthritis, scleroderma endothelial 19

cells and nasopharyngeal carcinoma have also identified other potential regulatory mechanisms leading to FSTL1 inactivation, including miR-27a (44), histone deacetylase 5 (45) de-regulations and hypermethylation in the FSTL1 promoter (25).

Through functional studies using overexpression and suppression lentiviral-based expression systems as well as recombinant FSTL1, FSTL1 expressing conditioned medium and a homemade FSTL1 neutralizing antibody (kindly provided by Prof. Wen Ning) (41), we found the up- regulation of FSTL1 to promote tumorigenicity, metastasis, self-renewal and resistance to cisplatin. Note that since FSTL1 can also lead to increased cell proliferation, as demonstrated by foci formation and XTT cell proliferation assays, we must take caution when we interpret our metastasis findings, such that we must ensure that the metastasis effect is not a by-product of the cells’ altered proliferation potential. To address this, we repeated our migration and invasion assays again, in the presence of mitomycin C, a drug used to inhibit cell proliferation. Previously, our collaborator Ning et al. have demonstrated blockade of FSTL1 with this neutralizing antibody in mice to reduce bleomycin-induced fibrosis in vivo suggesting that FSTL1 may serve as a novel therapeutic target for treatment of progressive lung fibrosis (41). However, use of this neutralizing antibody for treatment of cancer in general has not been applied nor tested. FSTL1 encodes a 308 amino acid secretory protein with a 20 amino acid signaling peptide located at the N-terminus. In this work, we have demonstrated FSTL1 to be secreted from ESCC cells and that it will act, via an autocrine manner, to promote aggressive cancer features in ESCC development and progression. Past literature have found, via immunohistochemical staining of murine joint sections, expression of FSTL1 in all cell types of the mesenchymal lineage including osteocytes, chondrocytes, adipocytes and fibroblasts. More specifically, FSTL1 could be induced in osteoblasts, adipocytes and human fibroblast-like synoviocytes (46). In colorectal cancer, FSTL1 was also found to be significantly increased in cancer-associated fibroblasts, without significant expression in the cancer epithelial cells (47-48). Further studies on whether FSTL1 is also expressed in the cancer-associated fibroblasts in ESCC stroma would be worthwhile to further delineate origin of FSTL1 in addition to ESCC cells and whether the secretory protein functions in both autocrine and paracrine manners.

For a systematic identification of the underlying molecular mechanism mediating FSTL1 induced ESCC, we compared the expression profiles of KYSE510 ESCC cells with or without FSTL1 20

overexpressed using Agilent microarray profiling. By pathway enrichment analyses, we found the NFκB pathway to be activated and the BMP pathway to be silenced in FSTL1-overexpressing ESCC cells. De-regulation in both of these pathways have previously been implicated in FSTL1-mediated ESCC (11-13, 20-21, 31-35, 41). The two pathways have also been found to interlink (41), yet our study is the first to demonstrate the crosstalk between NFκB and BMP signaling where they both work in concert to promote ESCC. More specifically, there is data to show TLR signaling pathways to culminate in activation of NFκB that goes on to control the expression of an array of inflammatory cytokine genes (34). Recently, FSTL1 was found to bind to TLR4 (32) and its co- receptor CD14 (36). Recombinant FSTL1 was also found to induce IL-6 and IL-8 production from 293 cells in a CD14- and TLR4-dependent manner (32). In culture fibroblast-like synoviocytes, FSTL1 was found to activate NFκB signaling, as evident by Western Blot and ChIP assays at p65 binding sites (35). Further, two recent reports have also demonstrated a link between FSTL1, TLR4 and NFκB (31, 33). However, none of the reports stated are related to cancer and ESCC. Taking into account the literature and in combination with the fact that our profiling data, comparing empty vector and FSTL1 overexpressing ESCC cells, enriched for an altered NFκB and TGFβ signaling pathway, we hypothesized that FSTL1 will drive ESCC via modulating TLR4/NFκB/BMP signaling.

In addition, there is a handful of studies where FSTL1 has been found to modulate tumor immune response. For instance, Murakami et al. have demonstrated FSTL1 to evoke an innate immune response via CD14 and TLR4, thereby modulating organ development (32). Kudo-Saito et al. have shown FSTL1 to promote bone metastasis by causing immune dysfunction (28); and a separate study have also found epigenetic inactivation of FSTL1 to mediate tumor immune evasion in NPC (25). Thus, it would also be of interest to further explore whether FSTL1 secreted from ESCC cells would mediate alterations in immune cells like macrophages, T cells, etc. to modulate tumor immune response.

Findings from our current study suggest the novel oncogenic, metastatic and cisplatin resistance role of FSTL1 in ESCC and its potential use as a diagnostic and prognostic ESCC biomarker. Both endogenous and secretory FSTL1 plays important roles in mediating aggressive cancer properties in ESCC via canonical NFκB pathway activation and BMP pathway attenuation. BMP and NFκB

pathways crosstalk to drive these aggressive cancer features in FSTL1-mediated ESCC. With more studies, targeting FSTL1 could hold potential as a novel therapy for the treatment of ESCC.

Acknowledgements

We thank the Faculty Core Facility at the Faculty of Medicine, HKU for providing and maintaining the equipment needed for flow cytometry, confocal microscopy and animal imaging work.

References

1. Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011; 61: 69-90. 2. Ma S, Bao JY, Kwan PS, Chan YP, Tong CM, Fu L, et al. Identification of PTK6, via RNA sequencing analysis, as a suppressor of esophageal squamous cell carcinoma. Gastroenterology 2012; 143: 675-86. 3. Tong M, Chan KW, Bao JY, Wong KY, Chen JN, Kwan PS, et al. Rab25 is a tumor suppressor gene with antiangiogenic and anti-invasive activities in esophageal squamous cell carcinoma. Cancer Res 2012; 72: 6024-35. 4. Tang KH, Dai YD, Tong M, Chan YP, Kwan PS, Fu L, et al. A CD90(+) tumor-initiating cell population with an aggressive signature and metastatic capacity in esophageal cancer. Cancer Res 2013; 73: 2322-32. 5. Fung TM, Ng KY, Tong M, Chen JN, Chai S, Chan KT, et al. Neuropilin-2 promotes tumorigenicity and metastasis in oesophageal squamous cell carcinoma through ERK-MAPK- ETV4-MMP-E-cadherin deregulation. J Pathol 2016; 239: 309-19. 6. Miyamae T, Marinov AD, Sowders D, Wilson DC, Devlin J, Boudrea R, et al. Follistatin-like protein-1 is a novel proinflammatory molecule. J Immunol 2006; 177: 4758-62. 7. Bandla S, Pennathur A, Luketich JD, Beer DG, Lin L, Bass AJ, et al. Comparative genomics of esophageal adenocarcinoma and squamous cell carcinoma. Ann Thorac Surg 2012; 93: 1101- 6. 8. Bellinin MF, Silva AE, Varella-Garcia M. Genomic imbalances in esophageal squamous cell carcinoma identified by molecular cytogenetic techniques. Genet Mol Biol 2010; 33: 205-13. 9. Song Y, Li L, Ou Y, Gao Z, Li E, Li X, et al. Identification of genomic alterations in oesophageal squamous cell cancer. Nature 2014; 509: 91-5. 10. Adams D, Larman B, Oxburgh L. Developmental expression of mouse follistatin-like 1 (Fstl1): dynamic regulation during organogenesis of the kidney and lung. Gene Expr Patterns 2007; 7: 491-500. 11. Geng Y, Dong Y, Yu M, Zhang L, Yan X, Sun J, et al. Follistatin-like 1 (Fstl1) is a bone morphogenetic protein (BMP)4 signaling antagonist in controlling mouse lung development. Proc Natl Acad Sci USA 2011; 108: 7058-63.

12. Sylva M, Li VS, Buffing AA, van Es JH, van den Born M, van der Velden S, et al. The BMP antagonist follistatin-like 1 is required for skeletal and lung organogenesis. PLoS One 2011; 6: e22616. 13. Xu J, Qi X, Gong J, Yu M, Zhang F, Sha H, et al. Fstl1 antagonizes BMP signaling and regulates ureter development. PLoS One 2012; 7: e32554. 14. Yang Y, Liu J, Mao H, Hu YA, Yan Y, Zhao C. The expression pattern of follistatin-like 1 in mouse central nervous system development. Gene Expr Patterns 2009; 9: 532-40. 15. Shimano M, Ouchi N, Nakamura K, van Wijk B, Ohashi K, Asaumi Y, et al. Cardiac myocyte follistatin-like 1 functions to attenuate hypertrophy following pressure overload. Proc Natl Acad Sci USA 2011; 108: 899-906. 16. Ogura Y, Ouchi N, Ohashi K, Shibata R, Kataoka Y, Kambara T, et al. Therapeutic impact of follistatin-like 1 on myocardial ischemic injury in preclinical models. Circulation 2012; 126: 1728-38. 17. Oshima Y, Ouchi N, Sato K, Izumiya Y, Pimentel DR, Walsh K. Follistatin-like 1 is an Akt- regulated cardioprotective factor that is secreted by the heart. Circulation 2008; 117: 3099- 108. 18. Wei K, Serpooshan V, Hurtado C, Diez-Cunado M, Zhao M, Maruyama S, et al. Epicardial FSTL1 reconstitution regenerates the adult mammalian heart. Nature 2015; 525: 479-85. 19. Tanaka M, Ozaki S, Osakada F, Mori K, Okubo M, Nakao K. Cloning of follistatin-related protein as a novel autoantigen in systemic rheumatic diseases. Int Immunol 1998; 10: 1305- 14. 20. Wu J, Dong Y, Teng X, Cheng M, Shen Z, Chen W. Follistatin-like 1 attenuates differentiation and survival of erythroid cells through Smad2/3 signaling. Biochem Biophys Res Commun 2015; 466: 711-6. 21. Ni S, Miao K, Zhou X, Xu N, Li C, Zhu R, et al. The involvement of follistatin-like protein 1 in osteoarthritis by elevating NF-κB-mediated inflammatory cytokines and enhancing fibroblast like synoviocyte proliferation. 2015; 17: 91. 22. Fan N, Shu H, Wang Y, Wang Y, Zhang L, Xia Z, et al. Follistatin-like 1: a potential mediator of inflammation in obesity. Mediators Inflamm 2013; 2013: 752519. 23. Chan QK, Ngan HY, Ip PP, Liu VW, Xue WC, Cheung AN. Tumor suppressor effect of follistatin- like 1 in ovarian and endometrial carcinogenesis: a differential expression and functional analysis. Carcinogenesis 2009; 30: 114-21. 25

24. Zhao W, Han HB, Zhang ZQ. Suppression of lung cancer cell invasion and metastasis by connexin43 involves the secretion of follistatin-like 1 mediated via histone acetylation. Int J Biochem Cell Biol 2011; 43: 1459-68. 25. Zhou X, Xiao X, Huang T, Du C, Wang S, Mo Y, et al. Epigenetic inactivation of follistatin-like 1 mediates tumor immune evasion in nasopharyngeal carcinoma. Oncotarget 2016; 7: 16433- 44. 26. Tan X, Zhai Y, Chang W, Hou J, He S, Lin L, et al. Global analysis of metastasis-associated gene expression in primary cultures from clinical specimens of clear-cell renal-cell carcinoma. Int J Cancer 2008; 123: 1080-8. 27. Reddy SP, Britto R, Vinnakota K, Aparna H, Sreepathi HK, Thota B, et al. Novel glioblastoma markers with diagnostic and prognostic value identified through transcriptome analysis. Clin Cancer Res 2008; 14: 2978-87. 28. Kudo-Saito C, Fuwa T, Murakami K, Kawakami Y. Targeting FSTL1 prevents tumor bone metastasis and consequent immune dysfunction. Cancer Res 103; 73: 6185-93. 29. Su S, Parris AB, Grossman G, Mohler JL, Wang Z, Wilson EM. Up-regulation of follistatin-like 1 by the androgen receptor and melanoma antigen-A11 in prostate cancer. Prostate 2017; 77: 505-16. 30. Bae K, Park KE, Han J, Kim J, Kim K, Yoon KA. Mitotic cell death caused by follistatin-like 1 inhibition is associated with up-regulated Bim by inactivated Erk1/2 in human lung cancer cells. Oncotarget 2016; 7: 18076-84. 31. Guo J, Liang W, Li J, Long J. Knockdown of FSTL1 inhibits oxLDL-induced inflammation responses through the TLR4/MyD88/NF-κB and MAPK pathway. Biochem Biophys Res Commun 2016; 478: 1528-33. 32. Murakami K, Tanaka M, Usui T, Kawabata D, Shiomi A, Iguchi-Hashimoto M, et al. Follistatin- related protein/follistatin-like 1 evokes an innate immune response via CD14 and toll-like receptor 4. FEBS Lett 2012; 586: 319-24. 33. Zhang ZM, Zhang AR, Xu M, Lou J, Qiu WQ. TLR-4/miRNA-32-5p/FSTL1 signaling regulates mycobacterial survival and inflammatory responses in Mycobacterium tuberculosis-infected macrophages. Exp Cell Res 2017; 352: 313-21. 34. Kawai T, Akira S. Signaling to NFκB by toll-like receptors. Trends Mol Med 2007; 13: 460-9. 35. Tanaka M, Murakami K, Ozaki S, Imura Y, Tong XP, Watanabe T, et al. DIP2 disco-interacting protein 2 homolog A (Drosophila) is a candidate receptor for follistatin-related 26

protein/follistatin-like 1 – analysis of their binding with TGF-β superfamily proteins. FEBS J 2010; 277: 4278-89. 36. Shimada Y, Imamura M, Wagata T, Yamaguchi N, Tobe T. Characterization of 21 newly established esophageal cancer cell lines. Cancer 1992; 69: 277-84. 37. Guan XY, Sham JST, Tang TC, Fang Y, Huo KK, Yang JM. Isolation of a novel candidate oncogene within a frequently amplified region at 3q26 in ovarian cancer. Cancer Res 2001; 61: 3806-9. 38. Schwab M. Amplification of oncogenes in human cancer cells. Bioessays 1998; 20: 473-9. 39. Flemming A. Cancer stem cells: Targeting the root of cancer relapse. Nat Rev Drug Discov 2015; 14: 165. 40. Ming XY, Fu L, Zhang LY, Qin YR, Cao TT, Chan KW, et al. Integrin α7 is a functional cancer stem cell marker in oesophageal squamous cell carcinoma. Nat Commun 2016; 7: 13568. 41. Dong Y, Geng Y, Li L, Li X, Yan X, Fang Y, et al. Blocking follistatin-like 1 attenuates bleomycin-induced pulmonary fibrosis in mice. J Exp Med 2015; 212: 235-52. 42. Huang RL, Yuan Y, Zou GM, Liu G, Tu J, Li Q. LPS-stimulated inflammatory environment inhibits BMP-2-induced osteoblastic differentiation through crosstalk between TLR4/MyD88/NF-κB and BMP/Smad signaling. Stem Cells Dev 2014; 23: 277-89. 43. Shibanuma M, Mashimo J, Mita A, Kuroki T, Nose K. Cloning from a mouse osteoblastic cell line of a set of transforming-growth-factor-beta 1-regulated genes, one of which seems to encode a follistatin-related polypeptide. Eur J Biochem 1993; 217: 13-9. 44. Shi DL, Shi GR, Xie J, Du XZ, Yang H. MicroRNA-27a inhibits cell migration and invasion of fibroblast-like synoviocytes by targeting follistatin-like protein 1 in rheumatoid arthritis. Mol Cells 2016; 39: 611-8. 45. Tsou PS, Wren JD, Amin MA, Schiopu E, Fox DA, Khanna D, et al. Histone deacetylase 5 is overexpressed in scleroderma endothelial cells and impairs angiogenesis via repression of proangiogenic factors. Arthiritis Rheumatol 2016; 68: 2975-85. 46. Wilson DC, Marinov AD, Blair HC, Bushnell DS, Thompson SD, Chaly Y, et al. Follistatin-like protein 1 is a mesenchyme-derived inflammatory protein and may represent a biomarker for systemic-onset juvenile rheumatoid arthritis. Arthritis Rheum 2010; 62: 2510-6. 47. Torres S, Bartolome RA, Mendes M, Barderas R, Fernandez-Acenero MJ, Pelaez-Garcia A, et al. Proteome profiling of cancer-associated fibroblasts identifies novel proinflammatory signatures and prognostic markers for colorectal cancer. Clin Cancer Res 2013; 19: 6006-19. 27

48. Chen SX, Xu XE, Wang XQ, Cui SJ, Xu LL, Jiang YH, et al. Identification of colonic fibroblast secretomes reveals secretory factors regulating colon cancer cell proliferation. J Proteomics 2014; 110: 155-71.

Figure Legends

Figure 1. FSTL1 is frequently overexpressed in ESCC. (A, left) RPKM value of FSTL1 in 3 pairs of non-tumor (N) and ESCC (T) clinical samples examined by RNA-Seq (patients 16, 18 and 19). (A, right) Validation of FSTL1 expression by RT-qPCR in the same 3 ESCC and their adjacent non- tumor samples. (B, left) FSTL1 expression in 73 matched non-tumor (NT) and primary ESCC as detected by RT-qPCR. p<0.001. (B, right) Waterfall plot showing fold change expression of FSTL1 in the 73 ESCC clinical samples as compared to their non-tumor counterpart. (C) Representative IHC staining of FSTL1 in matched non-neoplastic squamous epithelium (non-tumor) and poorly differentiated ESCC tissue from 3 patient samples. Scale bar 100 µm. Bar chart summary of the distribution of FSTL1 expression levels in non-tumor versus ESCC examined by IHC on tissue microarrays (n=619 total cases; p<0.001). (D) Kaplan-Meier survival analysis comparing the cumulative survival rate of all patients with different FSTL1 expression levels (p=0.0021). (E) Box plot showing expression of FSTL1 in the sera collected from 30 healthy normal individuals and 104 ESCC patients (p<0.001).

Figure 2. Overexpression of FSTL1 in ESCC is in part a result of gene amplification. (A) Measurement of proteomic (endogenous and secretory) FSTL1 expression levels in a panel of ESCC cell lines by Western blot. (B) The ideogram of chromosome 3 shows localization of FSTL1 and the BAC clones used. (C) Detection of DNA copy number change of FSTL1 by dual-color FISH in absent/low FSTL1 expressing ESCC cell lines, high FSTL1 expressing ESCC cell lines and (D) formalin-fixed paraffin-embedded non-tumor and ESCC clinical samples with known FSTL1 overexpression in the ESCC counterpart. BAC probe to FSTL1 and the control reference probe to the centromere of chromosome 3 are represented by green and red signals, respectively. Nuclei in blue. Scale bar 5 µm. PBMC from healthy individual as control. (E) Bar chart summary of the distribution of normal and abnormal copy number in non-tumor versus ESCC examined by FISH (n=96 total cases; p<0.001). (F) Bar chart summary of the distribution of FSTL1 expression levels in normal versus abnormal copy number samples examined by FISH (n=96 total cases; p<0.0001).

Figure 3. FSTL1 contributes to aggressive cancer phenotype in ESCC cells in vitro. (A) Validation of FSTL1 knockdown (EC109) and overexpression (KYSE150-Luc) at proteomic (endogenous and secretory) levels by Western blot. NTC for non-target control and EV for empty vector. β-actin and 29

silver staining as loading control. (B) Growth curve of indicated stable cell lines by XTT cell proliferation assay. (C) Quantification of colonies induced by the indicated stable cell lines. Quantification of number of cells that (D) migrated through a membrane or (E) invaded through a Matrigel-coated membrane. (F) Quantification of spheroids induced by the indicated stable cell lines. (G) Percentage of Annexin V positive cells in the indicated stable cell lines following cisplatin treatment. (H) Representative IHC images and bar chart summary of FSTL1(+) cells in KYSE520 xenograft tumors treated with varying dosage of cisplatin. Scale bars in panels D to H at 100 µm. Scale bar in foci formation image 5 mm.

Figure 4. FSTL1 contributes to augmented tumorigenic and metastatic potential in ESCC cells in vivo. (A) Representative xenograft tumors and H&E images of resected tumors derived from mice subcutaneously injected with KYSE150-Luc cells with or without FSTL1 stably overexpressed. Graph showing the volume of tumors generated in each group (n=9). (B, top) Representative ex vivo imaging of lungs harvested from mice that received tail vein injections of KYSE150-Luc cells with or without FSTL1 stably overexpressed (n=10). (B, bottom) Luciferase signals shown as dot plot. (C) Graph showing the number of metastatic foci in KYSE150-Luc EV control group as compared to the same cells with FSTL1 stably overexpressed (n=10). (D) Representative H&E images of resected lung sections derived from mice injected via tail vein with KYSE150-Luc cells with or without FSTL1 stably overexpressed. All scale bars at 100 µm.

Figure 5. Secretory FSTL1 contributes to aggressive features in ESCC cells. Quantification of cells that migrated through a membrane or invaded through a Matrigel-coated membrane following treatment with (A) conditioned medium collected from KYSE150-Luc cells with FSTL1 stably overexpressed or (B) PBS control or human recombinant FSTL1 (rFSTL1) or (C) normal goat IgG control or anti-FSTL1 neutralizing antibody. (D) Relative growth curve of xenograft tumors formed following subcutaneous injection of KYSE150-Luc cells treated with conditioned medium collected from KYSE150-Luc cells with EV (EV CM) or FSTL1 stably overexpressed (FSTL1 CM). Graphs showing tumor weight and volume of tumors generated in each group (n=5). Representative H&E images of resected tumors derived from mice in each group. (E) Graph of average tumor volumes of mice along treatment course. Treatment administered at days 0, 2, 4, 6, 8 and 10. Bar graph showing tumor weight generated in each group (n=5). All scale bars at 100 µm. 30

Figure 6. FSTL1 drives ESCC through deregulated NFκB and SMAD pathways crosstalk. (A) Hierarchical cluster heatmap analysis of gene expression profiles in EV vs. FSTL1-overexpressing KYSE150-Luc ESCC cells. Each cell in the matrix represents a particular expression level, where red and green indicates high and low gene expression, respectively. (B) GSEA comparison of differentially expressed genes in EV vs. FSTL1-overexpressing KYSE150-Luc ESCC cells identified enrichment of TNFα signaling via NFκB, epithelial mesenchymal transition (EMT) and TGFβ signaling gene sets. NES, normalized enrichment score; FDR q, false discovery rate q value. (C-D) IPA prediction of 33 differentially expressed genes identified in our microarray that are known to be regulated by NFκB in FSTL1-overexpressing ESCC cells. Of these 33, 10 were also consistently found to be enriched in the GSEA TNFα signaling via NFκB gene set and thus were chosen for validation by RT-qPCR. (E) Western blot analysis for expression of p-IκBα, total IκBα, p-SMAD1 and total SMAD1 in total lysate; as well as p-p65 and total p65 in the nuclear sub-fractionated protein lysate, in the indicated cells. β-actin and histone H3 as loading controls.

Figure 7. (A) Western blot analysis for FSTL1, p-IκBα, total IκBα, p-SMAD1 and total SMAD1 in total lysate; as well as p-p65 and total p65 in the nuclear sub-fractionated protein lysate, in KYSE150-Luc cells with or without FSTL1 stably overexpressed, in the presence or absence of IMD-0354. DMSO was used as a control for IMD-0354. β-actin and histone H3 as loading controls. (B) Quantification of number of cells that migrated through a membrane, invaded through a Matrigel-coated membrane, formed spheroids or formed colonies in KYSE150-Luc cells with or without FSTL1 stably overexpressed, in the presence or absence of IMD-0354. (C) Western blot analysis for FSTL1, p-IκBα, total IκBα, p-SMAD1 and total SMAD1 in total lysate; as well as p-p65 and total p65 in the nuclear sub-fractionated protein lysate, in KYSE150-Luc cells with or without FSTL1 stably overexpressed, in the presence or absence of C34. β-actin and histone H3 as loading controls. (D) Quantification of number of cells that migrated through a membrane, invaded through a Matrigel-coated membrane, formed spheroids or formed colonies in KYSE150-Luc cells with or without FSTL1 stably overexpressed, in the presence or absence of C34. Scale bars in migration, invasion and spheroid images at 25 µm. Scale bars in foci formation image at 5 mm. (E) Cartoon diagram illustrating the proposed FSTL1-regulated oncogenic and metastatic mechanism in ESCC tumorigenesis. RE stands for responsive element.

FSTL1 promotes metastasis and chemoresistance in esophageal squamous cell carcinoma through NF κB-BMP signaling crosstalk

Marco Chi-Chung Lau, Kai-Yu Ng, Tin Lok Wong, et al.

Cancer Res Published OnlineFirst September 7, 2017.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-17-1411

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2017/09/07/0008-5472.CAN-17-1411.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet been Manuscript edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2017/09/07/0008-5472.CAN-17-1411. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.