Established brous peritoneal metastasis in an immunocompetent mouse model accurately reects the real clinical immune microenvironment of gastric cancer

Peritoneal metastasis (PM) in gastric cancer (GC) is characterized by diffusely inltrating and proliferating cancer cells accompanied by extensive stromal brosis in the peritoneal space. The prognosis of GC with PM is still poor regardless of the various current treatments. In order to elucidate the cause of diculties in PM treatment, we compared the tumor immune microenvironment (TME) in primary and PM lesions in GC. In addition, a PM model with brous stroma was constructed using immunocompetent mice to determine whether its TME was similar to that in patients. independent of tumor PD-L1 status in this trial. These results suggest that other components of TME, such as cancer-associated broblasts (CAFs) and tumor-associated macrophages (TAMs), might affect the immune system in PM. To obtain a favorable outcome using ICIs, tailoring combined therapy with modication of the TME will need to be considered. In the present study, we compared the TME in primary and peritoneal lesions to elucidate the cause of diculties in treatment of metastatic GC. Additionally, a PM model with brous stroma features similar to those of human clinical samples was established using immunocompetent mice. The model was used to conrm whether the TME was similar to clinical PM for the development of new treatment strategies.


Background
Gastric cancer (GC) is one of the most common malignancies worldwide, with an estimated 1,033,701 new cases and 782,685 associated deaths in 2018 according to the International Agency for Research on Cancer [1]. Peritoneal metastasis (PM) is the most common mode of metastasis in gastric cancer and a critical indicator of poor prognosis [2][3][4]. Although various approaches to the treatment of PM have been assessed, including systemic and/or intraperitoneal chemotherapy [5][6][7][8], hyperthermic intraperitoneal chemotherapy (HIPEC) [9,10], and aggressive surgery (peritonectomy) [11,12], satisfactory outcomes have not been achieved. PM is characterized by rapid in ltration and proliferation of cancer cells accompanied by extensive stromal brosis, causing potentially fatal disorders such as bowel obstruction, hydronephrosis, and jaundice [13]. Fibrosis interferes with drug delivery due to high intratumoral pressure [14]. These brous stroma are conducted by crosstalk between cancer cells and cancer-associated broblasts (CAFs). CAFs also induce endothelial mesenchymal transduction (EMT) in cancer cells, result to produce the invasion ability and chemo-resistance in cancer cells. Besides, cytokines and chemokines derived from CAFs interact tumor immune microenvironment (TME), which affect not only chemosensitivities but also e cacy of immune checkpoint inhibitors (ICIs).
Recently, ICIs have become a major tools of cancer therapy tool. ICI therapies have been shown to be extremely effective against numerous malignancies, such as melanoma, renal cell carcinoma, non-small cell lung carcinoma (NSCLC), and Hodgkin disease [15][16][17][18]. The anti-programmed death-1 (PD-1) antibody, nivolumab, showed an 11% of response rate and 40% of disease control rate for late line of metastatic GC patients in the ATTRACTION2 trial [19]. Generally, the e cacy of anti-PD-1 antibody is believed to depend on the tumor microenvironment with PD-ligand 1 (PD-L1) expression in cancer cells and the presence of tumor-in ltrating lymphocytes (TILs), mediating adaptive immune resistance [20]. This tumor immune microenvironment (TME) phenotype was found in 10-11% of Japanese GC patients [21,22], which is consistent with data from the ATTRACTION2 trial [19]. However, the antitumor response to nivolumab was independent of tumor PD-L1 status in this trial. These results suggest that other components of TME, such as cancer-associated broblasts (CAFs) and tumor-associated macrophages (TAMs), might affect the immune system in PM. To obtain a favorable outcome using ICIs, tailoring combined therapy with modi cation of the TME will need to be considered. In the present study, we compared the TME in primary and peritoneal lesions to elucidate the cause of di culties in treatment of metastatic GC. Additionally, a PM model with brous stroma features similar to those of human clinical samples was established using immunocompetent mice. The model was used to con rm whether the TME was similar to clinical PM for the development of new treatment strategies.

Patients and resource of samples
We performed immunostaining of paired primary lesions and peritoneal lesions from 28 GC patients with PM at our hospital from 2009 to 2016, and compared their immune environments. Tissue samples from primary lesions were obtained from biopsy specimens during upper gastrointestinal endoscopy, and peritoneal lesion were obtained from sampling during staging laparoscopy. All specimens were obtained before chemotherapy. Prior to this research, written informed consent was obtained from each patient. This study was approved by the Institutional Review Board of Kanazawa University Graduate School of Medical Science (study permission number 2789).
Cell lines and cell culture YTN16 is a gastric cancer cell line transplantable into C57BL/6 mice. YTN16 cells were established from subcutaneous tumors by injection of primary cultured cells derived from a mouse gastric adenocarcinoma. Mouse gastric tumors were established in p53 heterozygous knockout C56BL/6 mice by addition of N-Methyl-N-nitrosourea (MNU) to the animals' drinking water [23]. The resulting tumor cells were cultured in high-glucose Dulbecco's modi ed Eagle medium (DMEM, Sigma-Aldrich Japan, Tokyo, Japan) containing 1.0mL/L MITO (Coning Japan, Tokyo), 10mL/L L-Glutamine, 10mL/L Penicillin/Streptomycin and 10% fetal bovine serum (FBS), on plastic dishes coated with Type I collagen solution (0.5% Atelocollagen Acidic Solution IPC-50; Koken Co., Ltd., Japan) at 37°C in 5% CO 2 atmosphere.
Mouse intestinal myo broblast cell lines (LmcMFs) derived from mouse colonic mucosa were established by Takashi Ohama and Koichi Sato, Laboratory of Veterinary Pharmacology, Joint Faculty of Veterinary Medicine, Yamaguchi University [24]. Cells were cultured in DMEM containing 10% FBS at 37°C in 5% CO 2 atmosphere.

Co-culture
Indirect co-cultures were established as follows. YTN16 cells were seeded on a 6-well plate, while LmcMF cells were seeded in 1-μm pore-size Boyden chambers (BD Falcon, Franklin Lakes, NJ), both at a density of 1×10 5 cells per well or chamber in DMEM or high-glucose DMEM containing 10% FBS, respectively. After 24 h, the cells were washed twice with PBS, the chambers were placed into the wells of the plates, and the plates were incubated for 5 days in 2 mL of DMEM or high-glucose DMEM.

Mouse allograft model
The animal use proposal and experimental protocol (AP-183944) was reviewed and approved by the Animal Care and Use Committee of Kanazawa University. All animal experiments were performed in accordance with the standard guidelines of Kanazawa University. Female C57BL/6J mice (20 g, 6-8 weeks old) were purchased from Charles River Laboratories, Inc., Yokohama, Japan. The mice were housed with a 12h day-night cycle in a temperature-(21°C) and humidity-(50%) controlled room of the animal experimental institute. All mice were kept in individually ventilated cages, fed with sterile standard food and water ad libitum. Mice were randomly distributed into YTN16 inoculated group (n = 6) and YTN16 with LmcMF co-inoculated group (n = 6). To establishing peritoneal metastatic models, 1×10 7 of YTN16 cells alone in 1 mL of high-glucose DMEM were inoculated intraperitoneally under iso urane anesthesia on day 0 as YTN16 inoculated group. YTN16 cells were co-cultured with an equivalent number of LmcMFs for 5 days, and a total of 1×10 7 cells in 1 mL of the same medium were then inoculated same manner as YTN16 inoculated group. In this study, total inoculated cell counts were aligned for comparing tumor weights because tumors consisted of both cancer cells and stroma cells including LmcMFs. After inoculation, on day 14, the mice were euthanized with iso urane and cervical dislocation, and tumors were removed for weight calculation and immunohistochemical examination.

Immunohistochemistry
Tumor specimens were xed in 10% neutral buffered formalin and embedded in para n. Sections were stained with hematoxylin and eosin (H&E) and Azan stain for assessment of brosis, while the expression of various antigens was assessed immunohistochemically. Depara nizing sections were pretreated by autoclaving in 10% citric acid buffer at 120 °C for 15 min. Following treatment with protein block serum (Dako Co., Kyoto, Japan) for 10 min and incubation with 2% skim milk for 30 min to block nonspeci c reactions, sections were incubated with primary antibody at 4 °C overnight. Information on the antibodies used is listed in Table 1. After the sections were washed in PBS, immunoreactivity was visualized using EnVision reagent (Dako Co.), and the slides were developed with diaminobenzidine and counterstained with hematoxylin. All sections were examined using a uorescence microscope (Olympus, Tokyo, Japan). The degree of brosis was calculated as a percentage of brosis within the whole section in all samples using a BZ-9000 BZII microscope (Keyence, Osaka, Japan).

Quanti cation of immunostaining parameters
Data were obtained by manually counting positively stained cells in ve non-overlapping intratumoral elds. Due to the small size of specimens, clinical specimens were analyzed under x400 magni cation to avoid counting the non-tumoral elds. Stained cells in mouse model tumors were accessed under ×100 magni cation for CD4 + and CD8 + cells, and under x200 magni cation for CD163 + cells. All immunostaining was interpreted by two independently (JK and TY).

Statistical analysis
Statistical analyses were conducted using SPSS statistical software, version 23 (IBM Corp., Armonk, NY, USA). Comparison of peritoneal tumor weight in each group was made using Student's t-test. Differences among the cell count and brotic area data sets in each group were evaluated using the Mann-Whitney Utest. P values less than 0.05 indicated statistical signi cance.

Patient characteristics
Patient characteristics are shown in Table 2. The median age of the study population was 66 years (range, 27-83 years). Overall, 17 men and 11 women were enrolled in the study. Twenty-ve of the twentyeight patients (89%) had diffuse type GC (Bormann 3 or 4), while the other three patients showed intestinal type GC based on the Lauren classi cation. According to the Japanese Classi cation of Gastric Carcinoma 15th edition, the P statuses were P1a in 11 cases, P1b in 4 cases, and P1c in 13 cases.

Immunostaining of GC primary and peritoneal lesions
We performed immunostaining of biopsy specimens of primary and peritoneal lesions to analyze the microenvironment. We used antibodies against CD4 and CD8 as markers for lymphocytes and the CD163 antibody as a marker for M2 macrophages. Representative microscopic views of the peritoneal lesions are shown in Figure 1a. There was no signi cant difference in the number of CD4 + cells between primary and peritoneal lesions (8.9 ± 5.5 vs. 7.4 ± 7.7). The number of CD8 + cells observed in the peritoneal lesions was signi cantly lower than in the primary lesions (28.3 ± 12.7 vs. 8.7 ± 7.2, P=0.01). CD163 positive cells in the peritoneal lesion were signi cantly higher than in the primary lesion (18.4 ± 2.5 vs. 32.1 ± 5.0, P=0.016) (Figure 1b).

Establishment of brous allograft model
All mice were successfully established peritoneal tumors in each group and were analyzed using their datasets of day 14. YTN16 cells were co-cultured with LmcMF and co-inoculated intraperitoneally into mice. As shown in Figure 2a and 2b, the tumor weight of peritoneal tumors from mice co-inoculated with YTN16 and LmcMF was signi cantly higher than that of tumors from mice inoculated with YTN16 alone (0.57 ± 0.33 g vs. 1.25 ± 0.25 g, P=0.045). Microscopic ndings of peritoneal tumors in co-inoculated mice showed invasive progression, but not in mice inoculated with YTN16 alone (Figure 2c).
The degree of brosis in the tissues was compared by Azan staining ( gure 3a). Areas that stained blue were regarded as brosis, and the percentage of stained tissue was compared. In the YTN16 control group, some brosis was observed, but in the co-inoculated group (YTN16 and LmcMF), the median brotic area was signi cantly increased in peritoneal tumors (1.4%; range 0.13-3.38 vs. 19%; range 12.8-25.3, P<0.05) (Figure 3b). Co-inoculated LmcMF cells that were pretreated with PKH staining were con rmed as a component of brous tumors (data not shown).

Discussion
In this clinical analysis, the TME in peritoneal lesions was quite different from the TME in primary lesions.
There were signi cantly fewer CD8 + cells corresponding to CTL in peritoneal lesions than in primary lesions. The presence of CD8 + cells in the tumor in ltrate prior to the onset of chemotherapy could predict pathological complete response [25,26]. These results indicate that it is di cult to treat with not only ICIs but also chemotherapy in TME of PM. Stromal broblasts so called CAFs suppress CD8 + cells function via PDL2 and FAS-ligand, and enhance secretion of interleukin (IL)-6 via crosstalk with cancer cells resulting decreasing accumulation of CD8 + cells [27,28]. Accordingly, it is important to suppress the function of CAFs for recruitment of CD8 + cells.
There was no signi cant difference in the number of CD4 + cells between primary and peritoneal lesions.
CD4 + cells include helper T cells, monocytes, and macrophages, which have various functions in tumor progression. CD4 in ltration status is not statistically associated with prognosis [29], but may affect the e cacy of ICIs. CD163 + cells corresponding to M2 macrophages were found to be signi cantly higher in PM than in primary lesions. Although macrophages are thought to function as both antitumor agents (M1 macrophages) and protumor agents (M2 macrophages) [30], protumoral tumor-associated macrophages (TAMs) are believed to exhibit characteristics similar to M2 macrophages [31]. Monocytes migrate into peritoneal lesions using the C-C chemokine ligand 2 (CCL2) derived from CAFs and differentiate into the M2 phenotype by macrophage colony-stimulating factor (M-CSF) and IL-6 secreted from CAFs and cancer cells [32]. These M2 macrophages contribute to tumor angiogenesis, immune suppression, and metastasis [33,34]. The high density of M2 macrophages in PM may contribute to the low e cacies of ICIs treatments for metastatic GC, including PM.
Low-dose paclitaxel changed macrophage phenotype from M2 to M1 through the toll-like receptor (TLR)-4, resulting in tumor growth inhibition [35,36]. This theory is consistent with taxan-containing therapies showing superior results than other regimens for treating gastric cancer [37,38]. Half of the patients with unresectable GC have PM at the time of second-line chemotherapy induction [2,38].
Because ICI was used after treatment failure of taxan-containing therapies, the phenotype of macrophages may have changed from M1 to M2 in the PM TME. Therefore, another strategy for reprogramming macrophages from the M2 to M1 phenotype is needed.
We established a brous peritoneal tumor model by co-inoculating the mouse gastric cancer cell line YTN16 and the mouse myo broblast cell line LmcMF into immune-competent C57BL6/J mice. LmcMF played as CAFs by crosstalk with YTN16 resulted in creating much more brous tumor than tumors inoculated with YTN16 alone. Tissue brosis may interfere with drug delivery and immune cells in ltration due to high intratumor pressure [14]. In the YTN16 and LmcMF co-inoculated tumor model, we observed less in ltration of CD8 + cells, but more in ltration of M2 macrophages. In addition, the tumors were hyperprogressive and more invasive compared to YTN16 inoculated tumors. These results indicate that CAFs interfere accumulation of CD8 + cells and induce recruitment and M2 polarization of macrophages. Thus, both CAFs and TAMs contribute to tumor progression and immunosuppression.
This study has some limitations. First, the number of clinical samples was small, due to di culty of obtaining non-treated paired specimens from GC patients with PM. Second, TME differences between solid peritoneal tumors were investigated for only one cell line. However, to our knowledge, this is the rst report of the establishment of a brous peritoneal model in immunocompetent mice.

Conclusion
In this study, we have clari ed the TME of PM in GC and successfully established a mouse model that closely resembles human clinical ndings. By using this model, it could be possible to develop new treatment strategies for PM in GC through anti-CAFs therapy. Informed consent or substitutes for it was obtained from all patients for their inclusion in this study.

Abbreviations
All institutional and national guidelines for the care and use of laboratory animals were followed. Animals were treated in accordance with the Fundamental Guidelines for the Proper Conduct of Animal Experiments and Related Activities in Academic Research Institutions, under the jurisdiction of the Ministry of Education, Culture, Sports, Science, and Technology of Japan. All animal experiments were approved by the Committee on Animal Experimentation of Kanazawa University.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interest
The authors declare that they have no con ict of interest.