Interleukin 10 level in the peritoneal cavity is a prognostic marker for peritoneal recurrence of T4 colorectal cancer

Peritoneal recurrence (PR) is a major relapse pattern of colorectal cancer (CRC). We investigated whether peritoneal immune cytokines can predict PR. Cytokine concentrations of peritoneal fluid from CRC patients were measured. Patients were grouped according to peritoneal cancer burden (PCB): no tumor cells (≤ pT3), microscopic tumor cells (pT4), or gross tumors (M1c). Cytokine concentrations were compared among the three groups and the associations of those in pT4 patients with and without postoperative PR were assessed. Of the ten cytokines assayed, IL6, IL10, and TGFB1 increased with progression of PCB. Among these, IL10 was a marker of PR in pT4 (N = 61) patients based on ROC curve (p = 0.004). The IL10 cut-off value (14 pg/mL) divided patients into groups with a low (7%, 2 of 29 patients) or high (45%, 16 of 32 patients) 5-year PR (p < 0.001). Multivariable analysis identified high IL10 levels as the independent risk factor for PR. Separation of patients into training and test sets to evaluate the performance of IL10 cut-off model validated this cytokine as a risk factor for PR. Peritoneal IL10 is a prognostic marker of PR in pT4 CRC. Further research is necessary to identify immune response of intraperitoneal CRC growth.

IL6, IL10, and TGFB1 in ascites increase with increasing peritoneal cancer burden. All pT4 (39 patients) and M1c (19 surgical and 7 nonsurgical patients) patients and only the initial 78 (double the number of patients in the pT4 group) ≤ pT3 patients from cohort I were included in the analysis of ten cytokines. Therefore, samples from a total of 143 patients (102 male and 41 female) were eligible for analysis of immune cytokines and peritoneal tumor burden. The mean patient age was 66.5 ± 11.9 years ( Table 1).
Of the ten cytokines examined, the levels of IL (interleukin)4, IL6, IL10, IL12 (IL12p70) and TGFB1 (transforming growth factor beta-1) increased with increasing peritoneal cancer burden (p < 0.005 by Kruskal-Wallis test) although the subgroup analyses mostly were not significant, whereas the levels of IL2, IL5, IL17A (IL17A homodimer), IFNG (interferon-gamma) and TNF (tumor necrosis factor) did not. In most cases, the levels of IL4 and IL12 were lower than the range covered by the reference standards (7.8 ~ 500 pg/mL), rendering the measurements meaningless; therefore, of the ten cytokines tested, IL6, IL10, and TGFB1 tended to increase with peritoneal cancer burden, and IL6 and IL10 were significantly different between the pT4 and M1c groups ( Table 1; Fig. 2).
Although the levels of ascitic IL6 and IL10 between TNM stage III and IV patients were significantly different, they were not different when M1c group patients were excluded (Table 2; Fig. 3). These data indicate that concentrations of IL6 and IL10 in ascitic fluid are affected by peritoneal tumor burden regardless of nodal metastasis or hematogenous distant metastasis.
Patients with pT4 disease are at high risk of PR. PR by June 2019 in patients who underwent surgery with curative intent during the study period (August 1, 2009, to June 30, 2017) was surveyed based on data from medical records, regardless of ascites harvest (Table 3). There were 45 (5.8%) PRs and 112 (14.5%) systemic recurrences (SRs) among 775 patients undergoing curative surgery who had neither preoperative chemotherapy nor radiotherapy and who had neither retroperitoneal (Rb) rectal cancer nor M1c stage disease.
As expected, PR was most common (27 of 97, 27.8%) in pT4 group patients and was higher in patients with SR (14 of 25, 56%) than in those without SR (13 of 72, 18.1%) (p < 0.001, two-tailed χ 2 test). PR also occurred in patients with T1 and T3 tumors; however, the frequency was very low, and most were accompanied by SR (100% in T1 (2 of 2) and 81.3% in T3 (13 of 16)). There was no difference in the PR rates between patients with pT4 with (29.0%, 18 of 62) and without (25.7%, 9 of 35) harvestable ascites (p = 0.726, two-tailed χ 2 test). Therefore, patients with pT4 tumors were the most appropriate group in whom to investigate PR with respect to the mechanism and frequency.
Ascitic IL10 is a prognostic marker of PR in pT4 group patients. Of the 79 patients in the pT4 group with available ascites (cohorts I and II; Fig. 1), one did not have the data on three cytokines (TGFB1, IL6 and IL10), 14 underwent palliative resection (without resection of distant metastasis), one was an operative mortality and two were lost to follow-up (less than 1 month). Of the remaining 61 patients who underwent curative resection, eighteen experienced PR at a median of 9 (range 1-48) postoperative months, with a median follow-up period of 39 (range 1-87) months. The median IL10 concentration in patients with pT4 tumors who experienced PR (27.8 pg/mL) was significantly higher than that in patients who did not (12.1 pg/mL; p = 0.004, Mann-Whitney U test). However, although the median IL6 concentration in pT4 group patients who experienced PR (294 pg/mL) was higher than that in pT4 group patients who did not (154 pg/mL), the difference was not significant (p = 0.066). Similarly, TGFB1 concentrations did not differ between patients with (292.7 pg/mL) and without (211.0 pg/mL) PR (p = 0.267, in cohort I only; determination of TGFB1 levels in cohort II was not possible due to technical problems). ROC curve analysis generated AUC (area under the curve) values of 0.733 for IL10 (p = 0.004), 0.651 for IL6 (p = 0.066), and 0.628 for TGFB1 (p = 0.267), and the calculated cut-off of IL10 concentration for predicting PR was 14 pg/mL (Fig. 4). The cumulative PR rates were 6.9% (2/29) for patients with low IL10 levels (≤ 14 pg/mL) and 45% (16/32) for those with high IL10 levels (> 14 pg/mL; p < 0.001, logrank test). Furthermore, curves for PR according to time showed a significant difference between patients with high and low IL10 levels.
We next performed Cox regression analysis (backward Wald test), including other clinicopathological factors that are considered to be risk factors for recurrence, to examine whether increased IL10 was an independent risk factor for PR in pT4 patients (Table 4). Because IL6 and TGFB1 were highly correlated with IL10 (Pearson's r = 0.640, p < 0.001 for IL6; Pearson's r = 0.646, p < 0.001 for TGFB1), they were not included in the analysis. To investigate the relationship between peritoneal tumor burden and peritoneal immune characteristics, ascites was collected from patients undergoing surgery for colorectal adenocarcinoma since August 1, 2009. Patients with the following peritoneal conditions that could have influenced the results were excluded: patients with extraperitoneal rectal cancer (Rb rectal cancer) determined by pelvic MRI and operative findings (the lower margin located below the anterior peritoneal reflection); those who had undergone preoperative chemotherapy or radiotherapy; and those with intestinal perforation, abscess, leukocytosis, or fever (over 37.3 °C, two or more consecutive times at 4-h intervals within 24 h). Patients who did not agree to take part in the study were also excluded (primary exclusion). By January 31, 2014 (cohort I), we collected ascites from a sufficient number of patients to identify trends in cytokine distribution in three groups of patients: ≤ pT3, pT4, and M1c. Therefore, we did not attempt ascites sampling from patients with probable T1 or T2 (clinical stage T1 or T2) stage tumors from February 1, 2014 (cohort II), as it was unnecessary to have ascites samples from so many patients with ≤ pT3, while patients with clinical T3 tumors were not excluded because it was possible that those tumors would be classified as T4 on pathological examination. Ascites harvest was attempted in all other patients who were not subject to primary exclusion; however, there were some failures of ascites sampling due to adhesions, insufficient ascites, or blood contamination of ascites, and these patients were also excluded (secondary exclusion). Furthermore, patients undergoing palliative resection (R1 or R2 resection), those with operative mortality (who died within 30 postoperative days), and those for whom no postoperative surveillance imaging was conducted (follow-up loss) were not surveyed for peritoneal recurrence (tertiary exclusion in the pT4 group). Although this was a relatively small study, we divided the data set of 61 patients into training (cohort I, n = 32) and test (cohort II, n = 29) groups to validate the classification performance of the IL10 cut-off model. The optimal cut-off value derived from cohort I was 13.5 pg/mL (AUC = 0.736, p = 0.022), which could distinguish between the low and high recurrence groups (p = 0.004) and divide cohort II into low and high recurrence groups (p = 0.031). Moreover, IL10 was also a significant factor for predicting PR, with a higher AUC value, when used for validation in patients with PR and without SR (Fig. 5).

Discussion
The incidence of synchronous or metachronous PC is not well known; however, Santvoort et al. 13 reported that 23% of T4 CRC patients had synchronous PC and 21% had metachronous PC, and Segelman et al. 14 reported that 27.7% of T4 CRC patients had metachronous PC. In our series, there were 28 patients with synchronous PC of the 153 T4 patients (18.3%) who underwent resection of the primary lesion. Moreover, 27 of the 97 T4 patients (27.8%) who underwent curative resection were diagnosed with PR ( Table 3). The overall incidences of synchronous PC and later PR were not very different from those previously reported. However, the risk factors of PR in patients of stage pT4 is poorly understood. Nagata et al. reported that poor differentiation, lymph node metastasis and preoperative CEA were independent risk factors for peritoneal recurrence in a larger cohort. We did not find such results, probably due to the small size of the patient cohort 15 .
If carcinoma cells escape immune surveillance by immunoediting, they can form a tumor; if not, they are destroyed by the immune system 12 . The mechanisms by which cancer cells escape immune surveillance include loss of antigenicity, loss of immunogenicity and suppression of antitumor immune responses 16 . Factors that suppress the immune response include IL10, TGFB1, indoleamine dioxygenase (IDO), soluble Fas ligand, and cellular components such as regulatory T-cells and myeloid-derived suppressor cells 6 . Activation of immune www.nature.com/scientificreports/ checkpoints can also suppress the immune response 17 . In this study, increased levels of IL10, IL6 and TGFB1 correlated with an increased tumor burden. IL10 is produced not only by immune cells but also by cancer cells themselves 18 . Many studies have examined the ability of IL10 to suppress antitumor immunity. For example, IL10 secreted by peritoneal monocytes downregulates cytokine production and T-cell proliferation in ovarian cancers 19 . Patients with more advanced CRC have higher serum IL10 levels 20 , and serum IL10 has been shown to affect the prognosis of colon cancer patients 21 . In addition, Giacomelli et al. 22 reported higher recurrence rates in patients with persistently high serum IL10 levels. However, those studies were based on measurements of IL10 in the serum, whereas our study is the first to measure IL10 levels in ascites, where peritoneal carcinoma cells grow, and to observe the prognosis of patients with PR. The IL10 levels presented herein are supported by other studies showing similar IL10 levels in ascites 23 . As an IL10 ELISA is far simpler and more convenient than detecting and quantitatively measuring free peritoneal cancer cells, so this may be a preferable method for assessing the risk of PR.
IL6, a multipotent proinflammatory cytokine, is known to be expressed in colon cancer tissues 24,25 and plays a role in proliferation, metastasis and angiogenesis 26,27 . Because the immune response is a complex network of immune cells and molecules, IL10 and IL6 are only limited aspects of the immunosuppressive peritoneal   28 , should be examined through further investigations of this model. The amount of peritoneal fluid varies from patient to patient but generally increases with increasing peritoneal tumor burden. Hydration before surgery made ascites sampling possible in most pT4 patients and in approximately half of ≤ pT3 patients unless pelvic adhesion or bleeding prohibited sampling. The minimum volume of ascites needed to measure the ten cytokines was 1.5-3 mL, and patients with ascites volumes less than that were excluded by secondary exclusion. Nevertheless, it is not evident whether IL10 is related to peritoneal tumor growth in patients without ascites.
There was doubt that the low IL10 levels in ≤ pT3 patients was due to a dilution effect in patients with less ascites. However, this is unlikely because hydration was performed in all surgical patients and some other cytokines did not show similar tendencies. Moreover, we could acquire sequential samples from three M1c patients who needed repeated aspirations to relieve abdominal distension. The IL10 concentrations were always higher in the later samples (more progressed PC) for all 3 patients. This finding supports that progression of peritoneal tumor burden is accompanied by increases in ascitic IL10 levels.
We evaluated the ascitic IL10 level at the time of laparotomy opening, not laparotomy closure, for postoperative peritoneal recurrence. Because ascites at the time of closing laparotomy is not only ascites but instead a mixture of ascites, blood and irrigation saline, it has little significance for the immune status of the peritoneal cavity. Therefore, the elevated peritoneal IL10 levels found in this study are thought to imply latent and microscopic peritoneal tumor implants containing tumor cells as well as immune cells, which could be accidentally eradicated within the removed surgical specimens or with postoperative chemotherapy and which would otherwise become peritoneal recurrence.
This study has some limitations. We examined only a small number of patients, and we did not include the assessment of peritoneal cancer cells themselves. In addition, we cannot explain the high IL10 levels in some T3 or lower patients. Finally, a practical cut-off value for IL10 and a standardized and effective way of acquiring ascites are needed. Despite these limitations, the present study is the first to measure the concentration of immune cytokines in ascites, the fluid that forms the microenvironment for progressing tumors in the peritoneal cavity.
This is a good human model for studying the immune response to colorectal tumor growth, in which ascites is ready for protein assay and cellular analysis. The assay of additional immune proteins (such as cytokines, chemokines, and growth factors), identification of the original cells of significant proteins, investigation of differences in detailed immune characteristics between pT4 patients with and without peritoneal recurrence and characterization of the spatial arrangement of each immune cell and cancer cell within peritoneal seeding nodules will provide much information on the immune response to cancer growth as well as immune suppression. Most of the patients in this study were microsatellite stable because we did not sort the patients according to MSI status. Therefore, we anticipate that further studies of this model will supply evidence of immunotherapy for microsatellite-stable colorectal cancers, which is not indicated with the current immunotherapeutic, anti-PD-1.

Conclusion
Peritoneal IL10 concentration correlates with peritoneal tumor burden in patients with CRC. Ascitic IL10 is a prognostic marker of PR in patients with stage T4 CRC following curative-intent resection. More immune factors, including immune cell functions, should be explored in this model with a larger cohort to better understand the immunological characteristics that affect intraperitoneal CRC growth.

Materials and methods
Ascites samples were collected prospectively from patients with CRC (adenocarcinoma) who underwent surgery at the Seoul National University Boramae Medical Center since August 2009. Patients undergoing surgery from August 1, 2009, to June 30, 2017, were enrolled and surveyed for recurrence until June 30, 2019 (Fig. 1). Ascites collection. To facilitate ascites sampling, the patients were supplemented with intravenous fluid the day before surgery to avoid dehydration during the fasting or bowel preparation stages. After general anesthesia, the operating table was tilted into the reverse Trendelenburg position to allow the ascites to run into the Douglas pouch. Care was taken to ensure that blood or tissue fluid from the incision site did not flow into the peritoneal cavity during laparotomy incision or laparoscopic port insertion. As soon as the peritoneal cavity was opened, ascites samples were aspirated from the Douglas pouch and transferred to polypropylene tubes. Fibrin materials Table 3. Peritoneal and systemic recurrence rates * during study period according to T stages in curatively resected patients. *Recurrence of patients excluding palliative resection, preoperative chemo-or radiotherapy, retroperitoneal rectal cancer, M1c stage, operative mortality and follow up less than 1 month regardless of ascites harvest. † PR: peritoneal recurrence; SR: systemic recurrence. ‡ One patient with ascites but without IL10 data was included.   Laparoscopic surgery 9 24  www.nature.com/scientificreports/ and cellular debris were removed by centrifugation, and ascites was transferred to Eppendorf tubes, which were frozen at − 80 °C. Only ascites (not peritoneal irrigation fluid) was used. Patients whose tumors were located below the peritoneal reflection (Rb rectal cancer), those who had undergone preoperative chemotherapy or radiotherapy, and those in whom the ascitic cytokines could have been affected by inflammation other than that caused by the cancer itself (such as intestinal perforations, peritumoral abscesses, fever, or leukocytosis) were excluded (primary exclusion). All the other patients were candidates for ascites sampling. However, some patients had pelvic adhesions prohibiting ascites collection, others had insufficient amounts of ascites fluid, and others presented bleeding during ascites collection, which can affect the concentrations of ascitic cytokines. These patients were also excluded from ascites collection (secondary exclusion) (Fig. 1).

Tumor location
Additionally, we collected ascites from M1c patients who were not surgical candidates but required aspiration of malignant ascites to reduce abdominal distension to include a sufficient number of patients with macroscopic peritoneal tumors.
Patient grouping for the assessment of changes in cytokines. The patients were classified into three groups according to the extent of tumor exposure and growth in the peritoneal cavity (based on pathological results) as follows: no tumor cells (pT3 or lower T stages), microscopic tumor cells (pT4), and gross tumors (M1c). In the ≤ pT3 group patients, the primary carcinoma had not penetrated the serosa and there was no peritoneal seeding. In the pT4 group patients, carcinomas were exposed through the serosa of the colon without peritoneal seeding. In the M1c group, there were patients with a few localized peritoneal seeding nodules around the primary lesion or with multiple peritoneal seeding nodules throughout the peritoneum. Peritoneal metastatic carcinoma lesions in the M1c group were confirmed by pathological examination during the operation. Pathological stages were classified according to the 8th edition of the AJCC cancer staging manual. We reviewed pathologic slides of some patients from an earlier period of the study to clarify N1c and T4ab.
Patient follow-up and recurrence. The patients were treated and followed up regularly after surgery.
Postoperative chemotherapy was recommended and performed when indicated according to the NCCN (National Comprehensive Cancer Network) guidelines. However, some patients rejected chemotherapy. If the patient had even one cycle of scheduled chemotherapy, he or she was considered to have received chemotherapy. Serum carcinoembryonic antigen (CEA) was checked, and an abdominal computed tomography (CT) scan was conducted three or four times per year for patients with ≥ TNM stage II for the first 2 years; this was repeated twice a year for the next 3 years if there was no evidence of recurrence. PR was determined as follows: by surgical biopsy; when at least two serial images (CT or positron emission tomography scan) indicated the growth Table 4. Factors contributing to peritoneal recurrence in pT4 patients after curative resection (n = 61). IL interleukin, A ascending, T transverse, D descending, R rectum, (Q1, Q3) (first quartile, third quartile), WD well differentiated, MD moderately differentiated, PD poorly differentiated, UD undifferentiated, Muc mucinous. *Age and BMI had normal distribution (p > 0.05, Shapiro-Wilk test) but tumor size did not. (p < 0.05 in PR(-)). † The cut-off value obtained using the maximum value of Youden's index (sensitivity = 0.889, specificity = 0.628). ‡ p < 0.1, Univariable Cox regression. § p < 0.05, Multivariable Cox regression. ¶ Conversion cases from laparoscopic to open surgeries were included. **Cox regression with Firth's correction. † † Data from two patients were missed in BMI and four data were missed in CEA. www.nature.com/scientificreports/ of a mass suggestive of PR; when a peritoneal mass appeared in a patient with elevated serum CEA levels but without accompanying distant metastasis; or when the size and number of recurrent masses were reduced by chemotherapy. Time to PR was defined as the time of the first recognition of a mass in imaging studies, which was determined to be PR. SR was determined similarly using CEA, imaging modalities, and surgical biopsy.

Statistical analysis.
To examine the normality assumption for continuous variables (cytokines), the Shapiro-Wilk test was performed. The cytokine levels among the groups were compared using the Kruskal-Wallis test, and the Mann-Whitney U test was used for post-hoc analysis. The risks of PR with T stages were compared using the χ 2 test. To examine the ability of IL10 to predict PR, receiver operating characteristic (ROC) curve analysis was performed, and the cut-off value for IL10 was determined based on the maximum value of the Youden Index (J = sensitivity + specificity -1). Peritoneal disease-free survival was calculated using the Kaplan-Meier method, and the groups were compared using the log-rank test. To assess which factors were associated with PR, univariable and multivariable Cox regression models were applied, and we used Firth's bias-correcting penalized maximum likelihood method 29 for TNM stage due to the small sample size. Factors considered in the multivariable Cox regression model were selected from the univariable model (p < 0.1). In addition, to assess the proportional hazards assumption, Grambsch and Therneau's test based on Schoenfeld residuals was used 30 . All statistical analyses were performed using SPSS version 20 (IBM Inc., Somers, NY, USA) and SAS software, version 9.4 (SAS Institute, Cary, NC, USA), with p < 0.05 considered significant. For multiple comparisons, p values were adjusted using Bonferroni correction using significance values derived by dividing the p value by the number of tests.

Data availability
All data generated or analyzed during this study are included in this published article as Supplementary Information files.