Japanese ESCC cell lines (KYSE30 and KYSE140) were obtained from DSMZ, the German Resource Centre for Biological Material. Chinese ESCC cell lines (EC9706 and HKESC1) and normal esophageal epithelial cell lines (NE1) were kindly provided by Professor Tsao (Department of Anatomy, the University of Hong Kong, Hong Kong, China). All cell lines were authenticated by short-tandem repeat analysis. ESCC cells were cultured in DMEM (Life Technologies, Gaithersburg, MD, USA) supplemented with 10% FBS (Thermo Fisher Scientific, San Jose, CA, USA) and penicillin-streptomycin (Life Technologies, Gaithersburg, MD, USA) at 37ºC in 5% CO2. Normal epithelial cells were cultured in a 1:1 mixture of defined keratinocyte serum-free medium (Invitrogen, Mount Waverley, VIC, Australia) and EpiLife (Cascade Biologics, Portland, Oregon, USA) at 37ºC in 5% CO2.
BM cells for immunofluorescence (IF) staining or flow cytometry (FCM) analysis were collected by flushing femurs of euthanized nude mice with Hank's balanced salt solution (Thermo Fisher Scientific, San Jose, CA, USA). Leukocyte samples for FCM analysis were prepared by depleting erythrocytes. Briefly, blood was mixed with 20 volume of 1×FACS lysing solution (BD Biosciences, San Jose, CA, USA), incubated on ice for 10 minutes to lyse erythrocytes, and washed twice at 4°C with PBS.
Fibroblasts (normal fibroblasts [NFs]/CAFs from ESCC patients and induced-CAFs [iCAFs] from tumor-bearing mice with intravenous [i.v.] injections of FGFR2+ fibrocytes) were isolated from normal esophageal epithelium, ESCC tissue specimens or tumor xenografts respectively. In short, freshly collected tissues were cut into as small pieces as possible in sterile PBS, followed by collagenase digestion (Thermo Fisher Scientific, San Jose, CA, USA). The suspension was filtered through 20 µm stainless steel wire mesh to collect a single cell suspension. The filtrate was centrifuged and washed before being finally plated on 6 cm tissue culture dishes in 5 mL DMEM medium supplemented with 10% FBS. After culturing for 30 minutes at 37ºC, unadherent cells (mainly tumor cells) were removed to obtain pure fibroblasts. The adherent fibroblasts were subcultured for further study.
Fibrocytes (human and mouse) were purified from peripheral blood. Briefly, peripheral blood mononuclear cells (PBMCs) were isolated by centrifugation over Ficoll (GE Healthcare, Milwaukee, WI, USA) following the protocol of manufacturer and cultured on Fibronectin pre-coated tissue culture plates (BD Biosciences, San Jose, CA, USA) in DMEM supplemented with 10% FBS. After 2 days, the nonadherent cells (largely lymphocytes) were aspirated off, and the remaining adherent cells cultivated for 14 days. Over time, the contaminating monocytes died off, and fibrocytes appeared as spindle shaped cells (Supplementary Figure S1A). The crude fibrocytes were lifted by incubation in trypsin-EDTA 0.05% (Thermo Fisher Scientific, San Jose, CA, USA) and purified by immunomagnetic selection using anti-Col I microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). The resultant cultured, enriched fibrocyte populations were > 94% pure based on FGFR2 and CD34/CD45/Col I staining (Supplementary Figure S1B). Typically, between 2 and 8×104 fibrocytes were isolated per milliliter of peripheral blood.
Athymic BALB/C nude mice were housed under standard conditions and cared for according to the institutional guidelines for animal care. All animal procedures were performed under Hong Kong Animals (Control of Experiments) Ordinance license (Chap. 340) and with approval from the University of Hong Kong Animal Welfare Committee.
To determine the origin of FGFR2+ CAFs, ESCC tumor-bearing models were constructed using nude mice. 1×106 ESCC cells (KYSE30, EC9706, KYSE140, or HKESC1) were subcutaneous (s.c.) injected into the left hind legs of 4-week-old nude mice, respectively. Five mice per group were used for blood sample analysis. Besides, 24 mice per group were used for BM analysis. Day of inoculation of tumors in mice was considered to be day 0 (week 0). Mice were followed for 7 to 8 weeks after injection of tumor cells. Xenograft tumors were removed at week 4 after injection. The surgical procedure was mimicked by making a 1cm long skin incision in control mice without observable tumors (mice injected with PBS or KYSE140).
The mice used to perform in vivo chemotaxis assays were s.c. injected with 1×106 ESCC cells (KYSE30 or EC9706) one week before i.v. injection of FGFR2+ fibrocytes. To examine the effect of iCAFs on tumorigenicity, 1×106 ESCC cells (KYSE30 or EC9706) were s.c. injected into nude mice one week before i.v. injection of FGFR2+ fibrocytes. Tumor formation in nude mice was monitored over a 15-day period. Tumor size was measured every 3 days. The tumor volume was calculated by the formula V = 0.5×L×W2.
In FGF2 or CXCR4 inhibition assay, tumor-bearing mice were also established by s.c. injected with 1×106 ESCC cells (KYSE30 or EC9706). For FGF2 inhibition, mice were i.v. injected with 4 µg/kg FGF2 neutralizing antibody every day. Tumor size was monitored every 4 days over a 20-day period. Each tumor xenograft was dissected, weighed, fixed in 4% paraformaldehyde, embedded in paraffin, and processed for hematoxylin/eosin staining and immunohistochemistry (IHC) analysis. For CXCR4 inhibition, mice were treated with 1, 5, or 10 µg CXCR4 neutralizing antibody every day over a 7-day period. Each tumor xenograft was collected and processed for IHC staining. The mice treated with PBS or isotype-matched immunoglobulins were served as controls. Neutralizing antibodies used in these studies were listed in Supplementary Table S1.
Human blood samples and tissue specimens
Blood samples of 18 healthy controls and 144 ESCC cases were kindly provided by Sun Yat-Sen University Cancer Center (Guangzhou, Guangdong, China). All blood samples from ESCC patients were collected before surgical resection. 73 primary ESCC tumors and 27 adjacent nontumorous tissues from the proximal resection margins were collected immediately after surgical resection at Sun Yat-Sen University Cancer Center. 40 pairs of ESCC RNA samples and their corresponding RNA samples from nontumorous tissue were kindly provided by the First Affiliated Hospital, ZhengZhou University (Zhengzhou, Henan, China). No patients recruited in this study have received any preoperative treatment. Informed consent was obtained from all patients before the collection of blood samples and tissue specimens, and the study was approved by the Committees for Ethical Review of Research involving Human Subjects at Sun Yat-Sen University and Zhengzhou University.
Flow cytometry analysis was conducted using antibodies with fluorescein conjugates. Cells were incubated in PBS containing 2% FBS with either FITC-conjugated primary antibody and/or PE-conjugated primary antibody. Isotype-matched mouse and/or rabbit immunoglobulin were served as controls to gate positive cells. Samples were analyzed on BD LSR Fortessa Analyzer (BD Biosciences, San Jose, CA, USA), and data were analyzed using FlowJo software (Tree Star, San Carlos, CA, USA). Antibodies used in this study were listed in Supplementary Table S2.
Before processing, fibrocytes and fibroblasts (NFs, CAFs and iCAFs) were seeded on cover slips at 25%-50% confluency, and BM cells were centrifuged onto glass slides using a Cytospin centrifuge (800 g for 5 minutes). The slides were fixed in 4% paraformaldehyde for 10 minutes and permeabilized in 0.05% Triton X-100 for 1–5 minutes. After washing in PBS, primary antibodies with or without fluorescein conjugates were added and the slides were incubated at 4°C overnight in a dark humidified chamber. On the second day, primary antibodies were washed away. The slides treated with primary antibodies without fluorescein conjugates were incubated with FITC- or PE-conjugated secondary antibodies for 30 minutes. The nucleus was stained by DAPI. All slides were examined at a magnification of 40× in a fluorescence microscope (Carl Zeiss LSM 710). Primary antibodies used in this study were listed in Supplementary Table S2.
Standard streptavidin-biotin-peroidase complex method was used for IHC staining. Briefly, xenograft and organ sections were deparaffinized in xylene and rehydrated in graded alcohols and distilled water. Slides were heated for antigen retrieval in 1× retrieval buffer (Dako, Palo Alto, CA, USA). The EnVision Plus System (Dako, Palo Alto, CA, USA) was used for IHC according to the manufacturer’s instructions. Sections were incubated with anti-FGFR2 or anti-Luciferase antibody overnight at 4°C. Stained slides were imaged on an Aperio Scanscope CS imager (Leica Biosystems, Newcastle, UK). The immunostaining area of FGFR2 or Luciferase was quantified by Image-Pro Plus 6.0 software (Media Cybernetics, Silver Spring, MD, USA). Primary antibodies used in this study were listed in Supplementary Table S2.
Generation of target gene-overexpression/knockdown cells
GFP-expression plasmid (pLenti6-CMV-MCS-GFP-SV-puro, Invitroge), Luciferase-expression plasmid (pLV-EF1a-Firefly Luciferase-IRES-Bsd, GeneCopoeia) or pLV-EF1a empty vector with blasticidin selectable marker (GeneCopoeia, Rockville, MD, USA) were packaged using Lentiviral Packaging mix (GeneCopoeia, Rockville, MD, USA) and used to infect ESCC cell lines or FGFR2+ fibrocytes to establish ESCC cells constitutively expressing GFP and fibrocytes constitutively expressing Luciferase or resistant to blasticidin. Stable clones were selected using puromycin or blasticidin and subjected to in vitro experiments.
To generate CXCR4, YAP, or TEADs knockdown cells, oligonucleotides directly against these genes were cloned into pLKO.1, respectively (short hairpin RNAs [shRNAs] of TEADs were designed in a region identical in TEAD1, 2, 3 and 4). The sequences of the oligonucleotides were summarized in Supplementary Table S3. Plasmids were propagated in and purified from Stbl3 competent cells (Invitrogen, Mount Waverley, VIC, Australia). shRNA expression plasmid and lentiviral packaging plasmids were co-transfected into HEK293-T cells for virus production using Lipofectamine 2000 (Invitrogen, Mount Waverley, VIC, Australia). Fibrocytes or CAFs were infected with lentiviral media in the presence of 10 µg/ml polybrene (Sigma-Aldrich, St Louis, MO, USA) overnight at 37ºC in 5% CO2. The cells with stable target gene silencing were selected using puromycin and then collected for RNA extraction and/or subjected to in vitro experiments. The cells infected with non-targeting shRNA were served as controls.
In vivo chemotaxis assay
To evaluate the chemotaxis capability of FGFR2+-circulating fibrocytes, Luciferase-expressing fibrocytes (2 ×105 per mice) were i.v. injected into ESCC tumor-bearing mice (tumors were formed by GFP-expressing KYSE30 or EC9706). After 24 hours, fibrocyte infiltration (luminescent signals) and tumor burden (fluorescence signals) were assessed via in vivo fluorescence and bioluminescence measurement using IVIS Imaging System (Perkin Elmer, Heidelberg, Germany). Before acquisition of bioluminescence images, D-luciferin (50 mg/kg) was intraperitoneally injected to tumor-bearing mice. All of the bioluminescence images were acquired using the same parameters: Excitation filter (Block), Emission filter (Open), Exposure time = 5 minutes, Binning = Medium, F/Stop = 1. Sequentially, fluorescence images were acquired using the same parameters besides the Excitation filter (465 nm), Emission filter (520 nm), Exposure time = 1 s. Image analysis was performed with Living Image 4.4. A region of interest (ROI) was manually selected over signal intensity. The area of the ROI was kept constant. The in vivo chemotactic index was calculated as relative radiance (Radiance of Luminescent in ROI/Radiance of Fluorescence in ROI). The nontumor-bearing mice i.v. injected with fibrocytes were used as control. The tumor-bearing mice i.v. injected with PBS were used to discard non-specific background signals in bioluminescence detection. Three nude mice will be evaluated for each condition.
In vitro chemotaxis assay
Chemotaxis of fibrocytes was assayed in a transwell system (Corning, Cambridge, MA, USA) using 8 µm polycarbonate membranes precoated with Matrigel (BD Biosciences, San Jose, CA, USA). 800 µL serum-free DMEM with ESCC cells (KYSE30 or EC9706, 1×105 per well) or CXCL12 recombinant proteins (0.01, 0.1 or 1 ng/mL) was added to the lower chamber as chemotactic stimulus. Fibrocytes (5 ×103 per well) suspended in 500 µL serum-free DMEM alone or with CXCR4 neutralizing antibody (0.2, 1, or 5 µg/mL) were loaded into the upper chamber and incubated for 24 hours at 37°C, 5% CO2. Fibrocytes migration toward DMEM alone or DMEM with isotype-matched immunoglobulin was used as the negative control. Invasive fibrocytes were fixed, stained and counted under a microscope. The chemotactic index was calculated as the number of cells invaded per 40× field. Three independent experiments were done with triplicates each time. Recombinant proteins and neutralizing antibodies used in this study were listed in Supplementary Table S1.
In vivo differentiation assay
To induce FGFR2+ fibrocytes differentiation in vivo, 2×105 FGFR2+-blasticidin resistant-fibrocytes were i.v. injected into ESCC tumor-bearing mice. One week later, exogenous cells were isolated from freshly collected xenografts. The isolation process was similar to that of fibroblast isolation except that the isolated cells were treated with blasticidin (Sigma-Aldrich, St Louis, MO, USA) for selection of exogenous cells. These cells were collected for RNA extraction after 4-day selection, characterization was performed by semi-quantitative PCR (qPCR), which detects the expression of fibrocyte, fibroblast and CAF marker genes. FGFR2+ fibrocytes cultured in normal culture medium and FGFR2+-blasticidin resistant-fibrocytes cultured in medium with blasticidin were served as controls. Three independent experiments were done.
In vitro differentiation assay
To induce FGFR2+ fibrocytes differentiation in vitro, we used three distinct co-culture systems. For system I, FGFR2+ fibrocytes were seeded into 6-well plate at a density of 2×104 per well and cultured in conditioned medium from KYSE30. After 48 hours, the cells were collected for RNA extraction. Fibrocytes treated with normal culture medium or conditioned medium from NE1 were used as control. In system II, FGFR2+ fibrocytes and KYSE30 cells were co-cultured for 48 hours in the same well, but physical interaction was prevented by growing the KYSE30 cells in Millicell hanging inserts (Millipore, Bedford, MA, USA), which are inserted into the wells of 6-well plates and sit 1-2mm above the layer of fibrocytes. The bottom of the insert is formed by a cellulose membrane containing 0.45 µm pores, which allow for diffusion of soluble effectors. Fibrocytes cultured in normal culture medium or co-culture with NE1 were served as control. For system III, directly co-cultured of FGFR2+-blasticidin resistant-fibrocytes and KYSE30 cells were established in the same well at ratio of 1:5 (fibrocyte/KYSE30). After 48h, the culture medium was complete removed, whereas, the cells were continually cultured in medium with blasticidin for 48 hours before RNA extraction. Over time, all KYSE30 cells died off, and fibrocytes were left. Fibrocytes cultured in normal culture medium, blasticidin resistant-fibrocytes cultured in medium with blasticidin or co-culture with NE1 were used as control. For preparation of conditioned medium or feeder cells in co-culture systems, the KYSE30 and NE1 cells were mitotically inactivated in a medium containing 10 µg/mL mitomycin C for 24 hours. The cell identities in differentiation assays were accessed as described previously. All Experiments were done in triplicate.
Total RNA was extracted from cells and frozen tissues by the TRIzol reagent (Invitrogen, Mount Waverley, VIC, Australia). Reverse transcription of total RNA was done using a PrimeScript RT reagent Kit (Clontech, Palo Alto, CA, USA), and complementary DNA (cDNA) was subjected to qPCR using SYBR Green master mix (Roche, Lilleroe, Denmark) with housekeeping gene β-actin as an internal control. qPCR was performed on HT7900 system (Applied Biosystems, Foster City, CA, USA). Primers used were listed in Supplementary Table S4.
XTT cell proliferation assay
To examine the effect of conditioned medium from iCAFs on growth of ESCC cells, KYSE30 or EC9706 cells were seeded into 96-well plate at a density of 1.0×103 per well and cultured in conditioned medium from iCAFs. Conditioned medium from fibrocytes or normal culture medium was used as control. The cell growth rate was determined by XTT Cell Proliferation Kit II (Roche, Lilleroe, Denmark) according to the manufacturer’s instruction. Three independent experiments were done with triplicates each time.
Transwell invasion assay
To examine the effect of conditioned medium from iCAFs on invasion of ESCC cells, KYSE30 or EC9706 cells (5×104) were suspended in 0.5 ml serum-free conditioned medium from iCAFs and loaded on the upper invasion chamber coated with Matrigel (Corning, Cambridge, MA, USA). The lower chamber was filled with normal culture medium supplemented with 10% FBS.
After 48 hours, invasive cells were fixed, stained and counted under a microscope. Results are expressed as the number of cells invaded per 40× field. Three independent experiments were done with triplicates each time.
Enzyme-linked immunosorbent assay (ELISA)
FGF2 concentrations in serum and CXCL12 concentrations in plasma or tissue homogenates were measured by FGF2/CXCL12 ELISA Kit (Cat# KHG0021, Life Technologies, Gaithersburg, MD, USA; Cat# DSA00, R&D Systems, Minneapolis, MN, USA) according to the manufacturer’s instruction, respectively. All Experiments were done in triplicate.
Rna Extraction And Illumina Messenger Rna Library Preparation
Total RNA of fibrocytes and CAFs was isolated using the mirVana miRNA Isolation Kit (Applied Biosystems, Foster City, CA, USA). Extracted total RNA were treated with the DNA-free Kit (Ambion, Austin, TX, USA). Large 18S and 28S ribosomal RNAs (rRNAs) were removed from total RNA with the RiboMinus Transcriptome Isolation Kit (Invitrogen, Mount Waverley, VIC, Australia). The rRNA-depleted RNA was precipitated with Pellet Paint (Novagen Inc, Madison, WI, USA) and checked on the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). The rRNA depleted RNA was fragmented by incubation for 5 minutes at 94°C in 5× Array Fragmentation Buffer (Ambion, Austin, TX, USA). The reaction was stopped by chilling the tube on ice and precipitated with Pellet Paint. cDNA was synthesized with the SuperScript Double Stranded cDNA Synthesis Kit (Invitrogen, Mount Waverley, VIC, Australia) using random hexamers. The reaction was purified on a QiaQuick PCR column (Qiagen, Hilden, Germany). Double-strand cDNA fragments were repaired with DNA Terminator End Repair Kit (Lucigen, Middleton, WI, USA) incubated for 30 minutes at 30°C and purified on a QiaQuick PCR column. The Klenow 3’ to 5’ exo-polymerase (NEB, Ipswich, MA, USA) was used to add a single “A” base to the 3’ end of blunt phosphorylated DNA fragments by incubation for 30 minutes at 30°C. Following purification, an Illumina PE Adapter (Illumina, San Diego, CA, USA) was ligated to the end of DNA fragments with the Quick Ligation Kit by incubation for 15 minutes at room temperature. The reaction was purified on a QiaQuick PCR column. 300 to 320-base pair fragments were excised from a 2% low-melting agarose gel. Fragments were enriched by 10 cycles using AccuPrime Pfx DNA Polymerase (Invitrogen, Mount Waverley, VIC, Australia). PCR product was run on Novex 8% TBE polyacrylamide gel (Invitrogen, Mount Waverley, VIC, Australia) and stained with SYBR Gold (Invitrogen, Mount Waverley, VIC, Australia). The gel slice containing the 300 to 320-base pair fragment was excised and the DNA purified using the QiaQuick Gel Extraction Kit (Qiagen, Hilden, Germany). The concentration of gel-purified DNA fragments was measured using an ND-1000 UV/Vis spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA).
Cluster generation and sequencing were performed using the Standard Cluster Generation Kit v4 and TruSeq SBS Kit v3 on the Illumina Cluster Station and GAIIx following the manufacturer’s instructions (Illumina, San Diego, CA, USA). cDNA libraries from 2 paired fibrocytes and their corresponding CAFs (F1/C1 and F2/C2) were sequenced with 38-base single reads. Raw data from the GAIIx were analyzed with the Illumina Real Time Analysis (RTA) v1.6. A PhiX 174 control lane was included in each Solexa run for matrix, phasing, and error rate estimations as recommended by the manufacturer. The error rate of the PhiX control was < 0.1% for all the sequencing runs. Sequencing reads with polymer tracks longer than 30 base pairs or with primer sequences longer than 36 base pairs were filtered. Ribosomal RNA sequences were also removed by aligning to 28S (NCBI RefSeq accession number NR_003287.2), 18S (NCBI RefSeq accession number NR_003286.2), human ribosomal DNA complete repeating unit (HSU13369), and mitochondrial ribosomes (Ensembl transcript ID ENST00000387347 and ENST00000389680) with at least 95% similarity. The clean and high-quality reads were then mapped against human genome assembly (NCBI Build 37.1) using CLC Genomics Workbench (CLC Bio, Aarhus, Denmark). The 9 alternate assembly loci were excluded from the reference. Only sequencing reads with at most 2 mismatches per loci and a maximum of 10 mappable locations were retained for calculation of gene expression. All genes mapped by at least 10 reads in any of the 4 samples were considered. The expression level of each gene was measured by the number of reads per gene per kilobase exon per million mapped reads (RPKM) by normalizing the number of exon reads to the length of exons within that gene and per million mapped reads. The differential expression analysis was performed using Baggerley’s test. The genes significantly differentially expressed were extracted with a Benjamini-Hochberg corrected P-value < 0.005 for downstream GO and KEGG pathway analysis.
Gene set enrichment analysis (GSEA)
RNA-seq data was processed and analyzed using the GSEA software, developed by the Broad Institute of MIT and Harvard (USA) and available at www.broadinstitute.org, following the program guidelines. The specific settings applied in all analyses are: Number of Permutations (1000), Permutation Type (Gene set), Enrichment statistic (Weighted), Metric for ranking genes (t Test). The rest of the fields were left as defaulted. Values in the tables represent the Normalized Enrichment Score and Nominal P-value of each gene set. The list of the specific gene sets analyzed and their sources are available in Supplementary Table S5.
Motif enrichment analysis
Putative promoter and enhancer regions of CAF-specific extracellular factor encoded genes were identified by FANTOM5 CAGE database, VISTA Enhancer Browser and ENCODE Encyclopedia. Motif enrichment in these regulatory regions was analyzed using HOMER version 4.9, a suite of tools for regulatory element analysis in genomics applications. Sequences of the two region sets (promoters and enhancers of CAF-specific extracellular factor encoded genes) were compared to background sequences, matched for GC content and auto-normalized to remove bias from lower-order oligo sequences. After masking repeats, motif enrichment was calculated using the cumulative binomial distribution by considering the total number of target and background sequence regions containing at least one instance of the motif. The top scoring motifs resulting from each search were combined, remapped and ranked according to enrichment in our promoter and enhancer region set after filtering redundant motifs. Their best-matching transcription factors were found by alignment. The predicted transcriptional regulatory networks were than constructed by Cytoscape (Supplementary Table S6).
Denatured protein samples were loaded into the wells of 10% SDS-PAGE gel along with molecular weight marker. After electrophoresis, proteins were transferred to PVDF membranes with a constant voltage of 100V for 2 hours. The PVDF membranes were blocked with 5% non-fat milk for 1 hour and incubated with primary antibody at 4℃ overnight. After washing in TBS-T, the PVDF membranes were incubated for 1 hour with corresponding secondary antibody. Immune detection was performed using ECL Western Blotting Detection Kit and Hyperfilm. Primary antibodies used in this study were listed in Supplementary Table S2.
All statistical analyses were performed using SPSS statistical package for Windows, version 13.0. Data are represented as the mean ± SD from at least three independent experiments or 3 mice. The significance of the difference between groups was evaluated with the Student's t-test or χ2 test. P-value < 0.05 was considered significant. In experiments where there were multiple groups or variables, these were compared to the control for each sample using a 2-sample t-test. Due to the amount of data presented in this manuscript, we have highlighted the statistical significance for the most relevant comparisons.