SAA1 is upregulated in gastric cancer-associated fibroblasts possibly by its enhancer activation

Yoshimi Yasukawa


, Naoko Hattori , Naoko Iida , Hideyuki Takeshima ,

Masahiro Maeda , Tohru Kiyono , Shigeki Sekine , Yasuyuki Seto and Toshikazu Ushijima *
Division of Epigenomics, National Cancer Center Research Institute, Tokyo, Japan, Department of Gastrointestinal Surgery, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan, Division of Cell Culture Technology, National Cancer

Center Research Institute, Tokyo, Japan and


Department of Pathology and Clinical Laboratories, National Cancer Center

Hospital, Tokyo, Japan
*To whom correspondence should be addressed. Tel: +81 3 3542 2511, Fax: +81 3 5565 1753; Email: [email protected]

Cancer-associated fibroblasts (CAFs) tend to have tumor-promoting capacity, and can provide therapeutic targets. Even without cancer cells, CAF phenotypes are stably maintained, and DNA methylation and H3K27me3 changes have been shown to be involved. Here, we searched for a potential therapeutic target in primary CAFs from gastric cancer and a mechanism for its dysregulation. Expression microarray using eight CAFs and seven non-CAFs (NCAFs) revealed that serum amyloid A1 (SAA1 ), which encodes an acute phase secreted protein, was second most upregulated in CAFs, following IGF2. Conditioned medium (CM) derived from SAA1-overexpressing NCAFs was shown to increase migration of gastric cancer cells compared with that from control NCAFs, and its tumor-promoting effect was comparable to that of CM from CAFs. In addition, increased migration of cancer cells by CM from CAFs was mostly canceled with CM from CAFs with SAA1 knockdown. Chromatin immunoprecipitation (ChIP)-quantitative PCR showed that CAFs had higher levels of H3K27ac, an active enhancer mark, in the promoter and the two far upstream regions of SAA1 than NCAFs. Also, BET bromodomain inhibitors, JQ1 and mivebresib, decreased SAA1 expression and tumor-promoting effects in CAFs, suggesting SAA1 upregulation by enhancer activation in CAFs. Our present data showed that SAA1 is a candidate therapeutic target from gastric CAFs and indicated that increased enhancer acetylation is important for its overexpression.


Cancer-associated fibroblasts (CAFs), fibroblasts present in cancer tissues of many types of cancers, are one of the im- portant components of the tumor microenvironment. Previous studies reported that CAFs have duality in their functions. Some CAFs showed antitumorigenic roles in cancer initiation, progres- sion and metastasis (1). In contrast, many studies showed that CAFs can support growth, infiltration and migration of cancer cells (2,3). CAFs exert their tumor-supportive activity via secre- tion of multiple factors into the stroma, and these substances, including IGF2, WNT, IL6 and other chemokines or growth fac- tors, have recently attracted attention as therapeutic targets

of cancer (4–6). Nevertheless, such potential therapeutic tar- gets have not been fully explored yet in many types of cancers, including gastric cancer.
CAFs maintain their tumor-promoting capacity without the presence of cancer cells in their primary culture (7). As for the mechanism, the potential presence of mutations or copy number variations has been reported, but this is still waiting to be repro- duced (8–10). In contrast, recent studies reported that DNA methy- lation changes in CAFs, such as global hypomethylation (11,12), or aberrant methylation of specific genes (13,14), are important for the maintenance of CAF phenotypes. In addition, alterations of

Received: July 2, 2020; Revised: November 8, 2020; Accepted: December 4, 2020
© The Author(s) 2020. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected].


CAF cancer-associated fibroblast

Y.Yasukawa et al. | 181

Preparation of conditioned medium
A conditioned medium (CM) was prepared as described previously (5).

ChIP chromatin immunoprecipitation

Fibroblasts were seeded at a density of 5 × 10

cells per 10 cm dish in a

CM conditioned medium
NCAF non-cancer-associated fibroblast
PBS phosphate-buffered saline
qRT–PCR quantitative reverse transcription
TSS transcriptional start site
H3K27me3, which stably suppress gene transcription, were shown to be involved in CAF phenotypes, and a decrease of H3K27me3

MF medium and incubated for 24 h. After washing the cells with 1 × PBS (−), 4 ml serum-free phenol red-free RPMI1640 medium (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) was incubated with the cells for 24 h. For preparation of the negative CM without fibroblasts, serum- free RPMI1640 medium was incubated for 24 h in empty dishes. The cul- ture supernatant from three dishes was filtered and centrifuged at 4°C at 3500 rpm for 2 h using an Amicon Ultra-15 Centrifugal Filter (Merck Millipore, Burlington, MA). The condensed CM (the expected concentra- tion rate: 10-fold) was added with a serum-free RPMI1640 medium and used for cell migration and cell growth assays. For preparation of CM from

was responsible for increased secretion of cancer stem cell niche

cancer cells, N87 cells were seeded at a density of 1 × 10


cells per 10 cm

factors WNT5A , GREM1 and NOG , in gastric CAFs (5). Not only his- tone modifications near the promoter, but also enhancer alter- ations, resulting in transcriptional changes, have recently attracted great attention as mechanisms to maintain stable phenotypes in cancer and normal cells (15–17).
Here, we further investigated potential therapeutic targets secreted from CAFs, using primary CAFs and non-CAFs (NCAFs) from gastric cancer. In addition, we explored involvement of en- hancer activation as a mechanism of overexpression of a can- didate target.

Materials and methods
Primary fibroblasts
CAFs and their corresponding NCAFs were established from cancer and noncancerous tissues, respectively, of the same stomach surgi- cally resected from gastric cancer patients as described previously (5). Noncancerous tissue was distanced from the tumor margin by 5 cm or more. All gastric cancer patients had gastric mucosal atrophy, showing a history of Helicobacter pylori infection. The fibroblasts were maintained in a MF medium (Toyobo, Tokyo, Japan) and confirmed to be negative for myco- plasma infection with a MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland). All the primary fibroblasts were used within 10 pas- sages from the first culture. All the specimens were obtained with written informed consents, and the study was approved by the Institutional Review Boards at the National Cancer Center (2013-073).
Gastric cancer cell lines
MKN7, MKN45 and NUGC-3 cell lines were purchased from the Japanese Collection of Research Bioresources (Tokyo, Japan) on 18 September 2014. The N87 cell line was gifted by Dr K Yanagihara at the National Cancer Center on 11 August 2014. STR profiling analysis was performed with all cells by Takara Bio (Shiga, Japan) using the GenePrint 10 System (Promega, Madison, WI) on 1 July 2020. The authenticity of all cells was confirmed using CLASTR: the Cellosaurus STR similarity search tool (https://web. expasy.org/cellosaurus) (18). Cells were tested for mycoplasma infection with a MycoAlert Mycoplasma Detection Kit (Lonza).
Immortalization of NCAF and CAF
Lentiviral vector plasmids were constructed by recombination using the Gateway system (Invitrogen, Carlsbad, CA). Briefly, human TERT and a R24C mutant of CDK4 (CDK4 ) were cloned into entry vectors by BP reaction (Invitrogen) and then recombined with a lentiviral vector, CSII- CMV-RfA (a gift from Dr Hiroyuki Miyoshi via RIKEN BRC), by LR reaction (Invitrogen) to generate CSII-CMV-TERT and CSII-CMV-CDK4R24C (19). Recombinant lentiviruses with the vesicular stomatitis virus G glycopro- tein (VSV-G) were produced as described previously (20). As a control, CSII- CMV-EGFP was similarly prepared. NCAF23 or CAF19 was infected with CSII-CMV-CDK4R24C and CSII-CMV-TERT, or CSII-CMV-EGFP alone at a multiplicity of infection of 5 each using polybrene. The majority of the cells infected with CSII-CMV-EGFP showed EGFP positive. The cells which grew beyond a population doubling of 200 times were used as immortal- ized NCAFs and CAFs.

dish in RPMI1640 with 10% fetal bovine serum (FBS) and antibiotics, and culture supernatant was collected after 48-h incubation. In the treatment of NCAFs, the supernatant was diluted with equal amount of MF medium and added to NCAFs in subconfluent cultures.
Expression microarray analysis
Genome-wide analysis of mRNA expression was performed using an Agilent SurePrint G3 Human GE 8x60K v2 (Agilent Technologies, Santa Clara, CA). mRNA was labeled with a Low Input Quick Amp Labeling Kit (Agilent Technologies), purified by an RNeasy Mini Kit (QIAGEN, Venlo, The Netherlands) and hybridized to the microarray. The microarray was scanned with an Agilent G2565BA microarray scanner. The scanned data were processed using Feature Extraction Ver.10.7 software (Agilent Technologies), and analyzed using GeneSpring Ver.13 software (Agilent Technologies). The 75th percentile normalization was conducted for all sample data.
Quantitative reverse transcription PCR
Total RNA was extracted using Reliaprep (Promega). cDNA was synthesized using SuperScript IV Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA). qPCR was performed using the cDNA on the CFX con- nect Real-Time PCR Detection system (Bio-Rad Laboratories, Hercules, CA) and SYBR Green I (Takara). The copy number of a gene transcript was measured by comparing its amplification to that of a DNA standard, and normalized to the copy number of GAPDH. Each assay was performed at least in biological triplicate. The primer sets for quantitative reverse tran- scription PCR (RT–qPCR) are shown in Supplementary Table 1.
ELISA assay
All reagents and recombinant Human Serum Amyloid A1 standards of ELISA assay kit (Abcam, Cambridge, MA) were prepared as instructed. The CMs added with serum-free RPMI1640 medium were used as samples. All the standards and test samples were added into each well of a 96-well plate. Reactions with Biotinylated anti-Human SAA Antibody, HRP– Streptavidin solution, TMB One-Step Substrate Reagent and Stop solution (Abcam) were sequentially performed. The intensity of the color was im- mediately measured at 450 nm. All the samples were assayed in triplicate.
Paraffin block tissue samples were sectioned, deparaffinized and rehy- drated. After antigen retrieval in a buffer (Nichirei Biosciences Inc, Tokyo, Japan), sections were incubated in hydrogen peroxide for 13 min at room temperature, and then blocked with 10% goat serum for 30 min at room tem- perature. Sections were then incubated with anti-SAA1 antibody (ab201660, Abcam, Cambridge, MA) at 4°C overnight. After incubation with Histofine Simple Stain MAX-PO (MULTI) (Nichirei) as the secondary antibody, SAA1 expression was visualized by the addition of Histofine Simple Stain DAB so- lution (Nichirei), followed by the staining of nucleoli using Tissue-Tek Mayer Hematoxylin (Sakura Finetek Japan Co., Ltd., Tokyo, Japan).
Establishment of SAA1-overexpressing fibroblasts pCEP4 expression vector encoding SAA1 (pCEP4 Pig4) was purchased from Addgene (Watertown, MA) (21). pCEP4 encoding Chloramphenicol acetyltransferase (CAT ) as a control (pCEP4/CAT) was purchased from

182 | Carcinogenesis, 2021, Vol. 42, No. 2

Thermo Fisher Scientific. Plasmid DNA was isolated using a GenElute™ Plasmid Miniprep Kit (Sigma–Aldrich, St. Louis, MO). Immortalized NCAFs

respectively. CAFs were seeded per 10 cm dish, and subconfluent cells (80– 90%) were treated with JQ1 or mivebresib for 3 days at indicated concen-

were seeded at a density of 2 × 10


cells per well in a 24-well plate in a

trations. For CM preparation, the dishes were elaborately washed with PBS

MF medium, and on the next day transfected with pCEP4 Pig4 or pCEP4/ CAT using lipofectamine 2000 (Thermo Fisher Scientific). The medium was changed to Dulbecco’s modified Eagle’s medium (DMEM) (FUJIFILM Wako Pure Chemical Corporation) with 10% FBS without antibiotics, and again changed to a MF medium on the next day. After 6 days of incubation after transfection, 50 µg/ml of hygromycin (InvivoGen, San Diego, CA) was added to the medium to generate stable cell lines.
Establishment of SAA1-knockdown fibroblasts Synthesized oligonucleotides (Supplementary Table 2) for short hairpin RNA (shRNA) were annealed for 2 min at 95°C, and the DNA fragment was li- gated to a pGreenPuro™ shRNA cloning and expression lentivector (System Biosciences, Palo Alto, CA). 293TN cells were seeded at a density of 3 × 10 cells per 10 cm dish in a DMEM medium with 10% FBS and antibiotics, and on the next day transfected with the plasmid DNA vector. After 3 days of incubation, lentiviruses were collected. Immortalized CAFs were seeded at a density of 2 × 10 cells per well in a 24-well plate in a DMEM medium with 10% FBS without antibiotics, and on the next day transduced with the shRNA-expressing lentiviruses. The medium was changed to a MF medium on the next day. After 3 days of incubation after transduction, 0.8 µg/ml of puromycin (Sigma–Aldrich) was added to the medium, and after 5 days of incubation, the medium was changed to a MF medium without puromycin.
Cell migration assay
Cell invasion assay was performed using a Transwell Permeable Supports 6.5 mm 24 well (Corning, Corning, NY). The surface of the lower chamber was coated with 10 µg/ml collagen. A CM was added to the lower chamber, and N87 or NUGC-3 cells suspended in RPMI1640 with 0.5% FBS at 2.5 × 10 cells in 100 µl were applied to the upper chamber. After incubation at 37°C for 15 h, the nuclei of the cells that had migrated to the lower side of the upper chamber were stained with Hoechst 33342 (Thermo Fisher Scientific). Two random 10× fields per one upper chamber were captured using a Fluorescence Microscope BZ X710 (Keyence Corporation, Osaka, Japan).
Cell growth assay
N87 or NUGC-3 cells were seeded at 1 × 10 cells per well in a 96-well plate and cultured for 24 h. After washing with a serum-free RPMI1640 medium, the medium was changed to a CM. After 72-h incubation, cell viability was assessed using the Cell Counting Reagent SF (Nacalai Tesque, Kyoto, Japan).
Chromatin immunoprecipitation assay (ChIP assay) and ChIP-qPCR
Chromatin was cross-linked with 1% formaldehyde at room tempera- ture for 10 min and sonicated in SDS-lysis buffer (1% SDS, 50 mM Tris–HCl pH 8.0, 10 mM ethylenediaminetetraacetic acid pH 8.0, with protease inhibitor) using a Bioruptor (Cosmo Bio, Tokyo, Japan) for 25 cycles (NCAF) or 30 cycles (CAF) of 30-second sonication and 90-second interval. After centrifugation, the supernatant was collected as a crude chromatin solution. 30 µ g of sonicated chromatin was incubated at 4°C overnight with one of the following antibodies; 2 µ g of anti-H3K27ac (ab4729, Abcam), 2.4 µ g of anti-H3K4me1 (ab8895, Abcam), 2 µ g of anti- H3K27me3 (07-449, Merck Millipore), 2 µ g of anti-MED1 (17-10530, Merck Millipore), 4 µ g of anti-NF-κ B (17-10060, Merck Millipore) and 2 µ g of anti-BRD4 (A301-985A, BETHYL Laboratories, Montgomery, TX) anti- bodies. Chromatin was collected by Dynabeads Protein A (Thermo Fisher Scientific) ( 22 ), and cross-linking was reversed by digestion with proteinase K (Thermo Fisher Scientific) at 65°C overnight. DNA was purified by phenol and chloroform extraction and ethanol precipitation. Region-specific histone modifications or affinity with transcriptional mediators were analyzed by IP ratio in ChIP-quantitative PCR, which showed the ratio of immunoprecipitated DNA to input DNA. The primer sets for ChIP-qPCR are shown in Supplementary Table 1 .
Treatment with BET inhibitors
JQ1 and mivebresib were purchased from ChemScene (Monmouth Junction, South Brunswick, NJ) and Selleck Biotech (Tokyo, Japan),

(−) to remove the drug as much as possible, and the medium was changed to a MF medium on day 4, replaced with 4 ml serum-free phenol red-free RPMI1640 medium on day 5. After 24-h incubation, a CM was prepared as described previously.
Statistical analysis
Differences in the gene expression level, cell invasion, cell viability, IP ratio and the data of biochemical tests were analyzed using an unpaired Student’s t-test. The results were considered significant when a P-value <0.05 was obtained by two-sided tests. All calculations were performed using Microsoft Excel software (Microsoft Corp, Seattle, WA). Results SAA1 was identified as a gene upregulated in gastric CAFs To search for a therapeutic target from CAFs, we first performed expression microarray using eight CAFs and seven NCAFs. Unsupervised hierarchical clustering analysis of the 3000 most variable probes showed no clear difference between the CAFs and the NCAFs (Supplementary Figure 1). We then performed comparative transcriptome analysis of CAFs and NCAFs. The top two differentially expressed genes with the smallest P-values and the largest fold changes were insulin-like growth factor 2 (IGF2) and serum amyloid A1 (SAA1) (Figure 1A). Since IGF2 was previously shown to be upregulated in CAFs due to the de- crease of H3K27me3 (5), we here focused on SAA1. SAA1, serum amyloid A1, is a precursor of amyloid A protein, one of the four isotypes of serum amyloid A (23), and was reported as a serum marker for the presence of gastric cancer (24). To confirm the high expression of SAA1 in gastric CAFs, we conducted RT–qPCR using 13 CAF/NCAF pairs, and observed high expression of SAA1 in CAFs from both intestinal and dif- fuse types of gastric cancer, compared with those in NCAFs and gastric cancer cell lines (Figure 1B). An ELISA assay using CM derived from two CAF/NCAF pairs confirmed that SAA1 was secreted into the CM from CAFs (Figure 1C). Furthermore, immunohistochemistry using gastric cancer tissues showed SAA1 was expressed in fibroblasts adjacent to a tumor, but not in those separate from a tumor (Figure 1D). These data suggested that SAA1 could be involved in the characteristics of CAFs from gastric cancer. SAA1 expression in CAFs was involved in tumor invasion and migration To investigate the role of SAA1 upregulation in gastric CAF phenotypes, we established SAA1-overexpressing immortal- ized NCAF using a pCEP4 expression vector. RT–qPCR and an ELISA assay confirmed overexpression efficiency in SAA1- overexpressing NCAFs compared with CAT -overexpressing NCAFs (Supplementary Figure 2A and B). Influence on cell mi- gration was analyzed using gastric cancer cell lines, N87 and NUGC-3, and CMs derived from SAA1-overexpressing NCAFs and CAT-overexpressing NCAFs (control NCAFs). The CM de- rived from SAA1-overexpressing NCAFs showed a 2- to 4-fold stronger effect on cancer cell migration than the CM derived from control NCAFs (Figure 2A). We also performed cell growth assay using the two gastric cancer cell lines, and the same CMs, and found that the viable number of N87 cells increased when treated with the CM derived from SAA1-overexpressing NCAFs, although the CM derived from control NCAFs showed the same trend (Figure 2B). Y.Yasukawa et al. | 183 Figure 1. SAA1 was upregulated in CAFs from gastric cancer. (A) Volcano plot to compare expression in the seven CAFs and corresponding NCAFs. IGF2 and SAA1 were identified as the top two differentially expressed genes. (B) RT–qPCR of SAA1 expression in 13 CAF/NCAF pairs and two gastric cancer cell lines. SAA1 had higher expres- sion in CAFs than in corresponding NCAFs and gastric cancer cell lines. (C) Secretion of SAA1 into the CM from two CAF/NCAF pairs confirmed by ELISA assay. Larger amounts of SAA1 were secreted into CMs from CAFs than from NCAFs. (D) HE staining (left) and SAA1 immunohistochemistry (right) using cancer and noncancerous tissues from a gastric cancer patient. SAA1 was expressed in fibroblasts adjacent to a tumor (upper, arrowhead), but not in those separate from a tumor (lower) or tumor epithelial cells. Next, we established immortalized CAFs with SAA1 knockdown using an shRNA-expressing lentivirus. RT–qPCR and an ELISA assay confirmed knockdown efficiency in the CAFs (Supplementary Figure 2C and D). Increased migration of N87 and NUGC-3 cells by the CM from the CAFs with control knockdown was mostly canceled for gastric cancer cells with the CM derived from CAFs with SAA1 knockdown (Figure 2C). As for cell growth, the increase in the viable number of N87 cells 184 | Carcinogenesis, 2021, Vol. 42, No. 2 Figure 2. SAA1 from CAFs supported tumor invasion and migration. (A) Cancer cell migration assay using CM. A CM derived from SAA1-overexpressing NCAFs mark- edly promoted migration of both N87 and NUGC-3 gastric cancer cells compared with that from control NCAFs. (B) Cancer cell growth assay using CM. A CM derived from SAA1-overexpressing NCAFs increased the viable number of both N87 and NUGC-3 cells, although it showed no significant difference from the CM derived from control NCAFs. (C) Cancer cell migration assay using CM. Increased migration of both N87 and NUGC-3 cells by the CM from CAFs was mostly canceled with the CM from CAFs with SAA1 knockdown. (D) Cancer cell growth assay using CM. Increase in the viable number of N87 cells was moderately canceled when treated with the CM derived from SAA1-knockdown CAFs, although the difference in the growth of NUGC-3 cells was not significant. was moderately canceled when treated with the CM derived from CAFs with SAA1 knockdown (Figure 2D). These results con- firmed that SAA1 was one of the key regulators of CAF pheno- types, especially the tumor migration-promoting effect. SAA1 overexpression was likely to be due to its enhancer activation We next investigated the mechanism of how SAA1 was overexpressed in CAFs. DNA methylation levels obtained by an Y.Yasukawa et al. | 185 Figure 3. SAA1 was likely to be upregulated in CAFs by enhancer activation. (A) ChIP-qPCR of H3K27ac for two putative enhancer regions and a promoter region using four CAF/NCAF pairs. The three regions were more acetylated in CAFs than in NCAFs. (B) ChIP-qPCR of H3K4me1 for the two putative enhancer regions using four CAF/ NCAF pairs. H3K4me1 was increased more in CAFs than in NCAFs in almost all pairs. (C) ChIP-qPCR of H3K27me3 for the promoter region using four CAF/NCAF pairs. Some CAFs showed a decrease of H3K27me3 in the promoter compared with corresponding NCAFs, though not in all pairs. (D) ChIP-qPCR of MED1 for one putative en- hancer and promoter regions using two CAF/NCAF pairs. MED1 was increased both in the enhancer and promoter regions more in CAFs than in NCAFs. (E) ChIP-qPCR of NF-κB for one putative enhancer and promoter regions using two CAF/NCAF pairs. NF-κB was increased in CAFs more than in NCAFs, not only in the promoter, but also in the enhancer. Infinium HumanMethylation450 BeadArray in our previous study (5) (GEO accession number: GPL13534) were not different between CAFs (mean β value = 0.23) and NCAFs (0.31) in SAA1 TSS. Also, according to the UCSC database, no CpG island is located in the SAA1 promoter region. These showed that decreased DNA methy- lation in the SAA1 promoter region was unlikely to be involved in its 186 | Carcinogenesis, 2021, Vol. 42, No. 2 Figure 4. SAA1 can become a target of tumor stroma-directed epigenetic therapy. (A) BRD4 binding to enhancer region 2 assessed by ChIP-qPCR using two CAF/NCAF pairs. BRD4 binding level was increased in the enhancer in CAFs more than in NCAFs. (B) Preparation protocol of CM from CAF/NCAF with JQ1 treatment. JQ1 was added to CAF or NCAF in a 10-cm dish at a specified dose. The medium was changed to a MF medium after 72-h incubation and CM was prepared from the culture super- natant. (C) SAA1 expression in JQ1-treated CAF/NCAF19 assessed by RT–qPCR. SAA1 expression was decreased by JQ1 treatment of CAFs. (D) Effect of CM on cancer cell migration. Migration-promoting effect of the CM from CAFs was markedly suppressed by JQ1 treatment, whereas the same treatment did not show a clear effect for the CM from NCAFs. (E) Effect of CM on cancer cell viability. The viable number of N87 gastric cancer cells decreased more when they were exposed to CM derived from JQ1-treated CAFs than from mock-treated CAFs, although CM derived from JQ1-treated NCAFs showed the same effect. increased expression. According to the ChIP-Atlas data of human fibroblasts from fetal heart (GEO accession number: GSM2550214 and 2214081) ( 25), H3K27ac, a typical histone mark for active en- hancers, and H3K4me1, an established histone mark for poised and active enhancers, were present at 4.2 and 3.3 kbp upstream of the SAA1 promoter region, respectively, suggesting that an en- hancer activation might be involved in the SAA1 overexpression in CAFs (Supplementary Figure 3A and Supplementary Table 3). To analyze potential enhancer activation, ChIP-qPCR of four CAF/NCAF pairs was performed for the SAA1 upstream regions. It was shown that the two putative enhancer regions and the promoter region of SAA1 were more acetylated in CAFs than in NCAFs (Figure 3A). Also, we observed that H3K4me1 was in- creased in the two putative enhancer regions in CAFs more than in NCAFs (Figure 3B). H3K27me3, a histone mark for an inactive promoter, was decreased in CAFs in two of four CAF/NCAF pairs Y.Yasukawa et al. | 187 Figure 5. SAA1 was upregulated by epigenetic memory. (A) SAA1 and WNT5A expression after multiple passages in four pairs of CAF/NCAF57 assessed by RT–qPCR. SAA1 was still highly expressed even in CAFs at 17 passages. (B) ChIP-qPCR of H3K27ac using four pairs of CAF/NCAF57 the same as (A). Enhancer acetylation of SAA1 was still higher in CAFs than in NCAFs at 17 passages. (C) SAA1 expression in NCAF5 and NCAF19 with N87 cell-derived culture supernatant. SAA1 was moderately upregulated in NCAF19 compared with mock treatment, but not to a level as high as in corresponding CAFs. (Figure 3C). These results suggested that SAA1 was upregulated by increased enhancer acetylation. Going back to the ChIP-Atlas data, the SAA1 promoter re- gion had peaks with NF-κB and STAT3, compatible with the pre- vious reports showing their involvement in SAA1 upregulation (26–28) (Supplementary Figure 3B and Supplementary Table 3). In addition, the SAA1 upstream regions had two strong peaks with Mediator Complex Subunit 1 (MED1), known to be en- riched in promoters, enhancers and super-enhancers (29,30). To analyze the potential involvement of MED1 in the increased acetylation of the enhancer regions, ChIP-qPCR was performed in CAFs and NCAFs with antibodies against MED1 and NF-κB. MED1 and NF-κB were increased in both the putative enhancer and promoter regions more in CAFs than in NCAFs, indicating the involvement of these transcriptional mediators in the SAA1 upregulation in CAFs (Figure 3D and E). SAA1 can become a target of tumor stroma-directed epigenetic therapy Because BRD4 is known to be associated with enhancers (31), we performed ChIP-qPCR with an antibody against BRD4, and observed an increase of BRD4 in the SAA1 enhancer in CAFs (Figure 4A). Therefore, we tested the efficacy of JQ1 and mivebresib, which repress binding of BRD4 to its target (32, 33), to suppress the SAA1 enhancer activity (Figure 4B and Supplementary Figure 4A). JQ1 or mivebresib treatment re- duced SAA1 expression in CAF19 in a dose-dependent manner, whereas expression in the corresponding NCAF19 was not af- fected by the same treatment (Figure 4C and Supplementary Figure 4B), suggesting that the SAA1 enhancer was specifically activated in CAFs by BRD4 binding. Furthermore, treatment of CAFs with JQ1 or mivebresib canceled the increased migration and growth of a gastric cancer cell line, N87, by the CM from the CAFs (Figure 4D and E; Supplementary Figure 4C and D). These data indicated that SAA1 was upregulated by its enhancer ac- tivation, and can become a target of tumor stroma-directed therapy. SAA1 enhancer acetylation was maintained for epigenetic memory If the increased acetylation of SAA1 enhancer is involved in the stably maintained characteristics of CAFs, it should be retained even without cancer cells. To confirm this, we investigated whether SAA1 upregulation is diminished or not in a long-term culture. We cultured CAF/NCAF57 and passed them up to pas- sage 17 (P17). Then we performed RT–qPCR and ChIP-qPCR with an antibody against H3K27ac using P2, P7, P12 and P17 cells. Although the SAA1 expression level gradually declined from the P2 cells, it was still as high as that of WNT5A, one of the cancer stem cell niche factors, even in P17 cells (Figure 5A). Also, the SAA1 enhancer tended to have higher acetylation levels in CAFs than in NCAFs even in P12 cells (Figure 5B). On the other hand, we also analyzed the involvement of cancer cells in fibroblast education by treating NCAFs with gas- tric cancer (N87)-derived culture supernatant, and found that SAA1 was moderately upregulated in one of two NCAFs com- pared with mock treatment (Figure 5C). However, the SAA1 upregulation level was much lower than that in CAFs. Taken together, it was suggested that increased enhancer acetylation was the major mechanism of the epigenetic memory in CAFs while education by GC cells was also involved in the increased SAA1 expression in CAFs. 188 | Carcinogenesis, 2021, Vol. 42, No. 2 Discussion SAA1 was shown to be a potential therapeutic target in gastric CAFs. SAA1, along with SAA2, is an isotype of serum amyloid A, a lipoprotein known to be secreted from the liver during an acute-phase response ( 34 , 35 ). The level of SAA1 protein in serum is reported to increase in infection, inflammation ( 36 ), auto immune diseases and various malignant tumors, including lung ( 37 ), hepatocellular ( 38 ) and gastric can- cers ( 39 , 40 ). Although most studies assume that SAA1 is de- rived from the liver or normal epithelial cells, some studies showed that SAA1 was upregulated in cancer cells by IL6 or TNF-induced STAT3 and NF-κB signaling pathway ( 26 , 41 , 42 ), indicating that SAA1 in serum could be derived from tumor epithelial cells. Our present data showed that SAA1 was se- creted from CAFs into CM, suggesting a possibility that the high SAA1 serum level in gastric cancer patients was due to its secretion from CAFs. High expression of SAA1 in CAFs and its pro-tumorigenic activity were recently reported in pancre- atic ductal adenocarcinoma ( 43 ). Our study is the first report about SAA1 upregulation in gastric CAFs. As for a mechanism of how SAA1 increased invasion and migration of cancer cells, SAA1 upregulation is reported to activate integrin α5 and β 3, leading to tumor invasiveness and migratory capacity via Erk signaling ( 44 ). Further investigations are needed to confirm the downstream pathway of SAA1 secretion. As assessment of the contribution of SAA1 to CAF phenotypes, the tumor-promoting activity of CM from SAA1-overexpressing NCAFs (migrated cell area = 250 µm × 10 ; Figure 2A) was com- parable to that of CM from CAFs (migrated cell area = 200– 500 µm × 10 ; Figure 4D and unpresented data). This suggested that SAA1 is one of the major tumor-promoting factors from CAFs. At the same time, the CM from CAT-overexpressing NCAFs had a small but significant effect, and the tumor-promoting ef- fect of the CM was likely to lie on not only with SAA1 but also multiple secreted factors. NCAFs in gastric cancer patients have been exposed to chronic inflammation, which is known to in- duce epigenetic reprogramming potently (45), and may have already acquired a CAF-like nature. Overall, gastric CAFs can be- come a target of tumor stroma-directed therapy using an agent that targets enhancer activation. Enhancer acetylation was indicated as the mechanism of SAA1 overexpression in gastric CAFs, in comparison with NCAFs. As shown in Figure 3A and B, typical histone modifi- cations for active enhancers, H3K27ac and H3K4me1, were ob- served in the enhancers both in CAFs and NCAFs, showing that the SAA1 enhancers are functional also in NCAFs. However, the levels of both H3K27ac and H3K4me1 were more increased in CAFs, suggesting that SAA1 upregulation was due to the en- hancer activation. The enhancer may not be fully quiescent in NCAFs, possibly due to the fibroblasts’heterogeneity and prior exposure to chronic inflammation. JQ1 and mivebresib, BET bromodomain inhibitors, inhibited SAA1 upregulation and tumor-supportive capacity of CAFs, showing that the increased enhancer acetylation in CAFs can be targeted by the BET inhibitors. Especially, mivebresib is in clin- ical development, and, once the approach to CAFs is established, its clinical implementation will be facilitated. In other cancer types, the progression of pancreatic cancer was shown to be suppressed by JQ1 via inactivation of CAFs (46). At the same time, it was possible that residual amounts of BET inhibitors might have directly affected the migration capability or growth of GC cell lines. To address the concern, we used the CM from BET- inhibitor-treated NCAFs as a control, and observed a minimal effect on migration of GC cells (Figure 4D and Supplementary Figure 4C). The putative enhancer and the promoter regions were marked by MED1, a component of the mediator protein com- plex that directly interacts with regulatory proteins such as general transcription factors, and functions as a transcrip- tional coactivator of RNA polymerase II-dependent genes (47). In addition, NF-κB is known to be activated by prior exposure to chronic inflammation and is important in the formation of ac- tive enhancers in cancer cells (48,49). Both MED1 and NF-κB were shown to be increased in the putative enhancer and promoter regions in CAFs more than in NCAFs by ChIP-qPCR, indicating the involvement of MED1 and NF-κB in SAA1 upregulation in CAFs. Once established, multiple epigenetic mechanisms, DNA methylation changes, H3K27me3 changes and enhancer acetyl- ation are likely to be involved in CAF stable phenotypes. In conclusion, we found that SAA1 was upregulated in gastric CAFs potentially by enhancer activation, that the CAF pheno- type was stably maintained, and that it can be therapeutically targeted. Acknowledgements We thank Dr Hiroyuki Miyoshi (RIKEN, BioResource Center; pres- ently Keio University) for providing lentiviral constructs and packaging plasmids. Funding This research was supported by AMED under grant number JP20ck0106552, JSPS KAKENHI grant number JP18H02704. Conflict of Interest Statement: None declared. Data availability All data are incorporated into the article and its online Supplementary Material. References 1. Chandler, C. et al. (2019) The double edge sword of fibrosis in cancer. Transl. Res ., 209, 55–67. 2. Costa, A. et al. (2018) Fibroblast heterogeneity and immunosuppressive environment in human breast cancer. Cancer Cell, 33, 463–479.e10. 3. Ishii, G. et al. (2016) Phenotypic and functional heterogeneity of cancer- associated fibroblast within the tumor microenvironment. Adv. Drug Deliv. Rev., 99(Pt B), 186–196. 4. Unger, C. et al. (2017) Stromal-derived IGF2 promotes colon cancer progression via paracrine and autocrine mechanisms. Oncogene , 36, 5341–5355. 5. Maeda, M. et al. (2019) Cancer cell niche factors secreted from cancer- associated fibroblast by loss of H3K27me3. Gut, 69, 243–251. 6. Goulet, C.R. et al. (2019) Cancer-associated fibroblasts induce epithelial- mesenchymal transition of bladder cancer cells through paracrine IL-6 signalling. BMC Cancer, 19, 137. 7. Tao, L. et al. (2017) Cancer associated fibroblasts: an essential role in the tumor microenvironment. Oncol. Lett., 14, 2611–2620. 8. Pidsley, R. et al. (2018) Enduring epigenetic landmarks define the cancer microenvironment. Genome Res., 28, 625–638. 9. Du, H. et al. (2017) Genetic alterations and epigenetic alterations of cancer-associated fibroblasts. Oncol. Lett., 13, 3–12. 10. Hosein, A.N. et al. (2010) Breast carcinoma-associated fibroblasts rarely contain p53 mutations or chromosomal aberrations. Cancer Res., 70, 5770–5777. 11. Vizoso, M. et al. (2015) Aberrant DNA methylation in non-small cell lung cancer-associated fibroblasts. Carcinogenesis , 36, 1453–1463. 12. Jiang, L. et al. (2008) Global hypomethylation of genomic DNA in cancer- associated myofibroblasts. Cancer Res., 68, 9900–9908. 13. Lawrence, M.G. et al . (2020) Alterations in the methylome of the stromal tumour microenvironment signal the presence and severity of pros- tate cancer. Clin. Epigenetics, 12, 48. 14. Smith, E. et al. (2017) Fibroblasts derived from oesophageal adenocar- cinoma differ in DNA methylation profile from normal oesophageal fibroblasts. Sci. Rep., 7, 3368. 15. Hnisz, D. et al. (2013) Super-enhancers in the control of cell identity and disease. Cell, 155, 934–947. 16. Shukla, V. et al. (2008) Histone acetylation and chromatin signature in stem cell identity and cancer. Mutat. Res., 637, 1–15. 17. Tomazou, E.M. et al. (2015) Epigenome mapping reveals distinct modes of gene regulation and widespread enhancer reprogramming by the oncogenic fusion protein EWS-FLI1. Cell Rep., 10, 1082–1095. 18. Robin, T. et al. (2020) CLASTR: the Cellosaurus STR similarity search tool—a precious help for cell line authentication. Int. J. Cancer, 146, 1299–1306. 19. Sasaki, R. et al. (2009) Oncogenic transformation of human ovarian sur- face epithelial cells with defined cellular oncogenes. Carcinogenesis , 30, 423–431. 20. Miyoshi, H. (2004) Gene delivery to hematopoietic stem cells using lentiviral vectors. Methods Mol. Biol ., 246, 429–438. 21. Polyak, K. et al. (1997) A model for p53-induced apoptosis. Nature, 389, 300–305. 22. Takeshima, H. et al. (2009) The presence of RNA polymerase II, active or stalled, predicts epigenetic fate of promoter CpG islands. Genome Res ., 19, 1974–1982. 23. De Buck, M. et al. (2016) Structure and expression of different Serum Amyloid A (SAA) variants and their concentration-dependent func- tions during host insults. Curr. Med. Chem., 23, 1725–1755. 24. Chan, D.C. et al. (2007) Evaluation of serum amyloid A as a biomarker for gastric cancer. Ann. Surg. Oncol., 14, 84–93. 25. Oki, S. et al. (2018) ChIP-Atlas: a data-mining suite powered by full inte- gration of public ChIP-seq data. EMBO Rep., 19, e46255. 26. Yoon, S. et al. (2012) NF-κB and STAT3 cooperatively induce IL6 in starved cancer cells. Oncogene , 31, 3467–3481. 27. Kho, Y. et al. (2008) Induction of serum amyloid A genes is associated with growth and apoptosis of HC11 mammary epithelial cells. Biosci. Biotechnol. Biochem., 72, 70–81. 28. Ignacio, R.M.C. et al. (2019) Serum amyloid A predisposes inflammatory tumor microenvironment in triple negative breast cancer. Oncotarget, 10, 511–526. 29. Whyte, W.A. et al. (2013) Master transcription factors and mediator es- tablish super-enhancers at key cell identity genes. Cell , 153, 307–319. 30. Sabari, B.R. et al. (2018) Coactivator condensation at super-enhancers links phase separation and gene control. Science, 361:eaar3958. 31. Donati, B. et al. (2018) BRD4 and cancer: going beyond transcriptional regulation. Mol. Cancer, 17, 164. Y.Yasukawa et al. | 189 32. Zou, Z. et al. (2014) Brd4 maintains constitutively active NF-κB in cancer cells by binding to acetylated RelA. Oncogene , 33, 2395–2404. 33. Piha-Paul, S.A. et al. (2019) First-in-human study of mivebresib (ABBV- 075), an oral pan-inhibitor of bromodomain and extra terminal pro- teins, in patients with relapsed/refractory solid tumors. Clin. Cancer Res., 25, 6309–6319.
34. Urieli-Shoval, S. et al. (1998) Widespread expression of serum amyloid A in histologically normal human tissues. Predominant localization to the epithelium. J. Histochem. Cytochem ., 46, 1377–1384.
35. Uhlar, C.M. et al . (1994) Evolution of the serum amyloid A (SAA) protein superfamily. Genomics , 19, 228–235.
36. Lee, J.Y. et al. (2019) Serum amyloid A proteins induce pathogenic Th17 cells and promote inflammatory disease. Cell , 80, 79–91.e16.
37. Jovanović, D. et al. (2019) Soluble sPD-L1 and serum amyloid A1 as po- tential biomarkers for lung cancer. J. Med. Biochem., 38, 332–341.
38. Ni, X.C. et al. (2014) Serum amyloid A is a novel prognostic biomarker in hepatocellular carcinoma. Asian Pac. J. Cancer Prev., 15, 10713–10718.
39. Subbannayya, Y. et al. (2015) Identification of differentially expressed serum proteins in gastric adenocarcinoma. J. Proteomics, 127(Pt A), 80–88.
40. Liu, C. et al. (2012) Identification of serum amyloid A in the serum of gastric cancer patients by protein expression profiling. Oncol. Lett., 3, 1259–1262.
41. Choi, H. et al. (2017) Augmented serum amyloid A1/2 mediated by TNF- induced NF-κB in human serous ovarian epithelial tumors. Immune Netw., 17, 121–127.
42. Shinriki, S. et al. (2010) Aberrant expression of serum amyloid A in head and neck squamous cell carcinoma. J. Oral Pathol. Med., 39, 41–47.
43. Djurec, M. et al. (2018) Saa3 is a key mediator of the protumorigenic properties of cancer-associated fibroblasts in pancreatic tumors. Proc. Natl. Acad. Sci. USA , 115, E1147–E1156.
44. Lin, C.Y. et al. (2018) Serum amyloid A1 in combination with integrin αVβ3 increases glioblastoma cells mobility and progression. Mol. Oncol., 12, 756–771.
45. Chiariotti, L. et al. (2013) Epigenetic modifications induced by Helicobacter pylori infection through a direct microbe-gastric epithelial cells cross-talk. Med. Microbiol. Immunol., 202, 327–337.
46. Yamamoto, K. et al. (2016) Stromal remodeling by the BET bromodomain inhibitor JQ1 suppresses the progression of human pancreatic cancer. Oncotarget, 7, 61469–61484.
47. Borggrefe, T. et al. (2011) Interactions between subunits of the mediator complex with gene-specific transcription factors. Semin. Cell Dev. Biol ., 22, 759–768.
48. Hajmirza, A. et al . (2018) BET family protein BRD4: an emerging actor in
NFkappaB signaling in inflammation and cancer. Biomedicines, 6: 16. 49. Betancur, P.A. et al. (2017) A CD47-associated super-enhancer links pro-
inflammatory signalling to CD47 upregulation in breast cancer. Nat. Commun., 8, 14802.