Tumor-associated Macrophages as Prognostic and Predictive Biomarkers for Postoperative Adjuvant Chemotherapy in Patients with Stage II Colon Cancer

Colon cancer is common around the world. Approximately one-quarter of patients with colon cancer present with stage II disease (1). Approximately 15%–25% of these patients suffer from systemic recurrence (including local recurrence and distant metastasis) and die even after surgery with curative intent (2, 3). Fluorouracil (FU)-based adjuvant chemotherapy after surgery has been widely used to reduce colon cancer recurrence and improve survival (4, 5). However, its efficacy for stage II disease remains controversial (5–7), with an improved survival rate of no more than 5% on average (8–10). Considering the adverse events, cost, and inconvenience, it is important to identify patients with a high risk of recurrence who can benefit from adjuvant chemotherapy. Thus, accurate prognostic factors, such as traditional clinicopathologic high-risk factors (T4 stage, poorly differentiated histology, lymphatic/vascular invasion, perineural invasion, bowel obstruction, localized perforation, positive margins, and <12 lymph nodes examined) and microsatellite instability (MSI) status, were always expected to be predictive for the efficacy of adjuvant chemotherapy. According to these widely recognized prognostic factors, patients with stage II colon cancer after surgery are divided into three groups according to recurrence risk by the National Comprehensive Cancer Network (NCCN) guidelines (8): low-risk group as T3 (MSI-H) with no high-risk factors; mid-risk group as T3 (MSS/MSI-L) with no high-risk factors; and high-risk group as T3 with high-risk factors or T4. It has been shown that patients in the low-risk group have a good prognosis and do not benefit from adjuvant chemotherapy (11, 12). However, for the remaining 85%–90% of patients in the mid-risk and high-risk groups, clinicopathologic high-risk factors combined with MSI status cannot identify patients who benefit from adjuvant chemotherapy (5–7). There is still a need for better prognostic and predictive biomarkers.

It is well known that the immune microenvironment significantly affects tumor development. Among the various immune cells recruited to the tumor site, tumor-associated macrophages (TAM) are particularly abundant at all stages of tumor progression and play an important role. TAMs are macrophages that infiltrate in tumor tissues. Generally, macrophages are divided into two phenotypes (13–15): the classically activated (M1) type, with iNOS, CD86, CD169, etc. as markers, and the alternatively activated (M2) type, with CD163, CD206, CD204, etc. as markers. M1 TAMs in tumor tissues stimulate tumor immunity and suppress tumor progression. In contrast, M2 TAMs enhance tumor cell invasion, motility, and intravasation, stimulate angiogenesis, suppress the immune response, and prevent tumor cell attack by natural killer and T cells (14–16). There are also studies of TAMs as prognostic factors for colorectal cancer (17–19). However, no clinical evidence for TAMs predicting the efficacy of postoperative adjuvant chemotherapy has been reported.

In this study, we detected TAMs and M2 subtype that infiltrated in the tumor tissues of patients with stage II colon cancer. We assessed the prognostic and predictive accuracy of TAMs as biomarkers for postoperative adjuvant chemotherapy. We also compared their prognostic and predictive efficacy with traditional clinicopathologic high-risk factors.

This study retrospectively enrolled two independent cohorts of consecutive patients at different time periods from the General Surgery Department, Zhongshan Hospital, Fudan University (Shanghai, China). The primary cohort was comprised of patients admitted from July 2009 to June 2012 to define the cut-off value of TAMs and to determine their prognostic and predictive efficacy. The validation cohort was comprised of patients admitted from July 2012 to December 2013 to verify the cut-off value and to confirm the prognostic and predictive efficacy of TAMs. All hypotheses were developed on the primary cohort and then tested on the validation cohort.

In both cohorts, the inclusion criteria were the same, as follows: ages between 18 and 80 years; pathologically confirmed colon adenocarcinoma, mucinous adenocarcinoma, or signet-ring cell carcinoma with stage II disease (T3-4, N0, M0) according to the AJCC/UICC TNM staging system 8th edition; and radical (R0) resection of the primary tumor. The exclusion criteria were also the same in both cohorts, as follows: emergency surgery because of an acute intestinal obstruction, bleeding or perforation; evidence of distant metastases; preoperative cancer therapy; hereditary colorectal cancer, such as familial adenomatous polyposis, Lynch syndrome, and MYH-associated polyposis; multiple primary tumors; a history of other malignancies (except for adequately treated basocellular carcinoma of the skin or in situ carcinoma of the cervix uteri); tissue specimen unavailable; and follow-up data unavailable. Colon cancer was defined as a tumor localized ≥15 cm away from the anal verge. The application of postoperative adjuvant chemotherapy was performed according to high-risk factors for systemic recurrence (8), physical status, treatment tolerance, and patient preference. No targeted agents were used in postoperative adjuvant chemotherapy. All clinicopathologic and follow-up data were obtained from a prospective database. This study was approved by the institutional review board of Zhongshan Hospital, Fudan University (Shanghai, China), and was carried out in accordance with the Declaration of Helsinki. All patients provided written and oral informed consent.

Follow-up principles were based on the Chinese guidelines for colorectal cancer. Patients were asked to come to the hospital for documented visits and the following examinations at the outpatient department: medical history; physical examination; serum carcino-embryonic antigen (CEA) level; abdominal ultrasound every 3 months for 2 years, then every 6 months for 5 years, then every year after 5 years; chest/abdominal/pelvic CT scan every 6 months for 2 years, then every year after 2 years; and colonoscopy at 6 months after primary tumor resection, then every year for 5 years, then every 2 years after 5 years. For patients who did not report on time, phone calls and letters were used.

In this study, a retrospective chart review was performed. Recurrence included both local recurrence and distant metastases. Once recurrence was suspected, detailed examinations, including contrast CT/MRI, PET-CT, colonoscopy, and biopsy (if necessary), were conducted for confirmation. The primary endpoint, disease-free survival (DFS), was defined as the time between surgery and the first recurrence, a new occurrence of colorectal cancer, or death from any cause. The overall survival (OS) was defined as the time between surgery and death. The estimated sample size for the primary cohort was 495 patients. Details are shown in Supplementary Information S1. According to the actual sample size and study results, post hoc power analysis was also conducted as described in Supplementary Information S2. The actual power was in line with expectations.

In the primary cohort, a tissue microarray (TMA) of tumor and normal mucosa tissue was constructed using formalin-fixed paraffin-embedded (FFPE) surgical specimens as described previously (20, 21). For each patient, two cores of tumor tissues were taken from two paraffin blocks of different areas of one tumor. In the validation cohort, a TMA of only tumor tissue was constructed in the same method as the primary cohort for each patient. No normal mucosa tissue was used in the validation cohort. For further clinical application, in the validation cohort, two normal FFPE sections of tumor tissue were taken from the two paraffin blocks used for tumor cores. Details are shown in Supplementary Information S3.

IHC was subsequently conducted on the TMA for both the primary and validation cohorts and on normal FFPE sections for the validation cohort. Primary antibodies against human CD68 (Clone KP1, 1:200; Abcam) and CD206 (Clone 5C11, 1:200; Abcam) were used to detect macrophages and M2 subtypes, respectively. The IHC protocols were performed as described previously (20, 21). Details are shown in Supplementary Information S3.

In most solid tumors as well as in colorectal cancer, tumor area is composed of tumor nest and stroma (22, 23). TAMs mainly infiltrate in the tumor stroma (24). In this study, only TAMs that infiltrated in the tumor stroma were counted and corrected according to the stroma area. TAMs that infiltrated in the tumor nest were not analyzed because of very low density. Details are shown in Supplementary Information S3.

The infiltration density of macrophages per field was evaluated by two independent pathologists who were blinded to the patients’ clinical data using Image-Pro Plus 6.0 (Media Cybernetics Inc.) for assistance. For each tissue core or normal section, three randomized fields of positive-stained cells were counted under a high-power field (HPF) of 200×. The density of macrophages was calculated as the mean number of all fields from cores or normal sections. Details are shown in Supplementary Information S3. The stability and repeatability of the detection method for TAMs were evaluated, and showed good performance (details are shown in Supplementary Information S4). The infiltration densities of CD68⁺ TAMs, CD206⁺ TAMs, and the ratio of CD206⁺ TAMs/CD68⁺ TAMs (CD206/CD68 ratio) were calculated as prognostic and predictive biomarkers.

A previous study (25) demonstrated high concordance between the IHC-based mismatch repair (MMR) test and the PCR-based MSI test. In this study, MSI status was based on the IHC testing of MMR with four markers (25, 26): MLH1 (Clone G168-15, 1:50; BD Pharmingen), MSH2 (Clone FE11, 1:200; Invitrogen), MSH6 (Clone 44, 1:100; BD Pharmingen), and PMS2 (Clone A16-4, 1:100; BD Pharmingen). Patients with tumor tissue that exhibited positive staining for all these markers were considered MSS or MSI-L. Patients with negative staining for at least one marker were considered MSI-H.

Patient baseline characteristics and disease factors were summarized using descriptive statistics. Categorical variables were compared using the two-sided Pearson χ² test or Fisher exact test as appropriate. Continuous variables were compared using a t test (normal distribution) or the Wilcoxon rank test (abnormal distribution) as appropriate. The correlation analysis was conducted using Spearman rank correlation test. Summary statistics on time-to-event variables, such as DFS and OS, were calculated according to the Kaplan–Meier method and compared by the log-rank test. Cox regression was used for univariate and multivariate analyses with HRs and 95% confidence intervals (CI). Only factors with P < 0.1 in the univariate analysis were included in the multivariate analysis. Interaction analysis was also conducted using Cox regression. To compare the prognostic efficacy of different biomarkers, Harrell concordance index (C-index, calculated using the Hmisc package, Soft R version 2.11.1) was used (27). The higher the C-index, the more effective the biomarker is. Comparisons between different prognostic factors were performed with the rcorrp.cens function in the Hmisc package. To evaluate the stability and repeatability of the detection method for TAMs, inter-rater agreement (kappa) was used for categorical parameters, and intraclass correlation coefficient (ICC) was used for continuous variables. All P values were two-sided and considered significant when <0.05.

The cut-off values of TAMs as prognostic biomarkers were defined according to the DFS data of the primary cohort and then applied in the validation cohort. To obtain the best prognostic efficacy, X-Tile Software (Yale University, version 3.6.1) was used as described previously (28).

In this study, the primary cohort consecutively enrolled 521 eligible patients, and the validation cohort consecutively enrolled 314 eligible patients. The flow diagram of cohort selection is shown in Fig. 1. By June 2016, the median follow-up time in the primary cohort was 69.0 months (range, 13.0–84.0; interquartile range, IQR, 54.0–76.0). By June 2017, the median follow-up time in the validation cohort was 55.0 months (range, 15.0–60.0; IQR, 50.0–58.0). In the primary cohort, 103 (19.8%) patients suffered from recurrence (local recurrence or distant metastases) and 70 (13.4%) patients died, including 6 (1.2%) deaths without evidence of recurrence. In the validation cohort, 59 (18.8%) patients suffered from recurrence and 38 (12.1%) patients died, including 2 (0.6%) without recurrence. The basic demographics and clinicopathologic characteristics of both cohorts are shown in Table 1. No significant difference was observed. In both cohorts, postoperative adjuvant chemotherapy was not balanced. Significantly more patients with clinicopathologic high-risk factors received adjuvant chemotherapy. Details are shown in Supplementary Table S1.

Positive IHC staining of CD68⁺ and CD206⁺ macrophages that infiltrated in primary tumor tissues and normal tissues is shown in Supplementary Fig. S1. In the primary cohort, CD68⁺ macrophage density, CD206⁺ macrophage density, and the ratio of CD206⁺/CD68⁺ macrophages were significantly higher in tumor tissues than in normal tissues. Details are shown in Supplementary Table S2.

For CD68⁺ TAM density, ≥118/HPF was defined as high and <118/HPF was defined as low. For CD206⁺ TAM density, ≥74/HPF was defined as high and <74/HPF was defined as low. For the CD206/CD68 ratio, ≥0.77 was defined as high and <0.77 was defined as low. Details are shown in Supplementary Information S5, S6, and S7. The C-index analysis of different cut-off values for TAMs is shown in Supplementary Fig. S2.

According to the cut-off value, 148 (28.4%) patients in the primary cohort and 75 (23.9%) patients in the validation cohort were defined as having high CD68⁺ TAM density; 102 (19.6%) patients in the primary cohort and 73 (23.2%) patients in the validation cohort were defined as having high CD206⁺ TAM density; and 86 (16.5%) patients in the primary cohort and 65 (20.7%) patients in the validation cohort were defined as having a high CD206/CD68 ratio.

In both the primary and validation cohorts, low CD68⁺ TAM density was significantly associated with perineural invasion (primary cohort P = 0.040, validation cohort P = 0.042). High CD206⁺ TAM density was significantly associated with poor differentiation (P = 0.035, P = 0.032). A high CD206/CD68 ratio was significantly associated with poor differentiation (P = 0.001, P = 0.003), pathologic T4 stage (P = 0.002, P = 0.004), lymphatic/vascular invasion (P = 0.002, P = 0.003), and perineural invasion (P = 0.002, P = 0.003). Significantly more patients in the high-risk group had a higher CD206/CD68 ratio than those in the mid-risk and low-risk groups (P = 0.002, P = 0.007). In addition, more patients with a high CD206/CD68 ratio received postoperative adjuvant chemotherapy in the primary cohort (P = 0.014), but not in the validation cohort (P = 0.451). Details of the correlation analysis, chemotherapy regimen, and DFS events are shown in Supplementary Tables S3, S4, and S5.

In the primary cohort, CD68⁺ TAM density was not a significant prognostic biomarker for DFS (P = 0.135) or OS (P = 0.739). However, patients with high CD206⁺ TAM density had significantly worse DFS (P < 0.001) and OS (P < 0.001) than those with low density. Patients with a high CD206/CD68 ratio also had significantly worse DFS (P < 0.001) and OS (P < 0.001) than those with a low CD206/CD68 ratio. Details are shown in Fig. 2. In the validation cohort, the results were similar. CD68⁺ TAM density was still not a significant prognostic biomarker for DFS (P = 0.624) or OS (P = 0.355). Patients with high CD206⁺ TAM density had significantly worse DFS (P = 0.005) and OS (P = 0.009). Patients with a high CD206/CD68 ratio also had significantly worse DFS (P < 0.001) and OS (P < 0.001). Details are shown in Supplementary Fig. S3.

The efficacy of prognostic biomarkers (CD68⁺ TAM density, CD206⁺ TAM density, and CD206/CD68 ratio) was compared. In the primary cohort, the efficacy of the CD206/CD68 ratio as a prognostic biomarker for DFS (C-index = 0.632) was significantly better than CD68⁺ TAM density (C-index = 0.524; P < 0.001) and CD206⁺ TAM density (C-index = 0.605; P = 0.001). For OS, the CD206/CD68 ratio (C-index = 0.654) was also significantly better than CD68⁺ TAM density (C-index = 0.512; P < 0.001) and CD206⁺ TAM density (C-index = 0.633; P = 0.009). Details are shown in Supplementary Table S6. In the validation cohort, the CD206/CD68 ratio was also significantly better than CD68⁺ TAM density and CD206⁺ TAM density for both DFS and OS. Details are shown in Supplementary Table S7.

The CD206/CD68 ratio was also compared with clinicopathologic high-risk prognostic factors. In the primary cohort, for both DFS and OS, the efficacy of the CD206/CD68 ratio was significantly better than primary tumor differentiation combined with MSI status, pathologic T stage, lymph nodes examined, lymphatic/vascular invasion, perineural invasion, and MSI status. Details are shown in Supplementary Table S6. In the validation cohort, the CD206/CD68 ratio was also significantly better than these clinicopathologic high-risk factors for both DFS and OS. Details are shown in Supplementary Table S7.

In both the primary and validation cohorts, univariate analysis showed that DFS was associated with tumor differentiation combined with MSI status (primary cohort HR = 1.744, P = 0.006; validation cohort HR = 2.131, P = 0.004), pathologic T stage (HR = 3.375, P < 0.001; HR = 2.995, P < 0.001), number of lymph nodes examined (HR = 1.903, P = 0.011; HR = 1.841, P = 0.091), lymphatic/vascular invasion (HR = 2.658, P < 0.001; HR = 2.465, P = 0.002), perineural invasion (HR = 2.679, P < 0.001; HR = 2.459, P = 0.005), MSI status (HR = 0.427, P = 0.021; HR = 0.359, P = 0.047), and CD206/CD68 ratio (HR = 4.158, P < 0.001; HR = 3.730, P < 0.001). For these factors included in the multivariate analysis, only pathologic T stage (HR = 2.613, P < 0.001; HR = 2.093, P = 0.012) and CD206/CD68 ratio (HR = 3.346, P < 0.001; HR = 2.867, P < 0.001) were identified as independent prognostic factors in both cohorts. Lymphatic/vascular invasion was considered an independent factor in the primary cohort (HR = 1.621; P = 0.044) but not the validation cohort (HR = 1.356; P = 0.355). Details are shown in Supplementary Tables S8 and S9.

In the primary cohort, postoperative adjuvant chemotherapy had no significant benefit on DFS (P = 0.512) or OS (P = 0.806) for all patients. For patients with a low CD206/CD68 ratio, adjuvant chemotherapy had no benefit on DFS (P = 0.986) or OS (P = 0.706). However, for patients with a high CD206/CD68 ratio, adjuvant chemotherapy significantly improved the DFS rate from 38.9% to 68.0% at 3 years and from 33.1% to 66.0% at 5 years (P = 0.003) and the OS rate from 75.0% to 82.0% at 3 years and from 47.8% to 75.8% at 5 years (P = 0.029). The interaction analysis between the CD206/CD68 ratio and adjuvant chemotherapy was significant for DFS (P = 0.023 for interaction) and potentially significant for OS (P = 0.063 for interaction), revealing that the benefit of adjuvant chemotherapy in patients with a high CD206/CD68 ratio was superior to that in patients with a low CD206/CD68 ratio. Details are shown in Fig. 3.

In the validation cohort, the results were similar. For all patients, postoperative adjuvant chemotherapy had no significant benefit on DFS (P = 0.216) or OS (P = 0.293). Patients with a low CD206/CD68 ratio still did not benefit from postoperative adjuvant chemotherapy regarding DFS (P = 0.868) or OS (P = 0.556). For patients with a high CD206/CD68 ratio, the benefit of adjuvant chemotherapy was significant on DFS (P = 0.007) and OS (P = 0.011). The interaction analysis was significant for both DFS (P = 0.030 for interaction) and OS (P = 0.024 for interaction). Details are shown in Supplementary Fig. S4.

However, in both the primary and validation cohorts, neither CD68⁺ TAM density nor CD206⁺ TAM density could significantly identify patients benefiting from adjuvant chemotherapy. No significant interaction was observed between CD68⁺ TAM density, CD206⁺ TAM density, and adjuvant chemotherapy. Details for the primary cohort are shown in Supplementary Figs. S5 and S6. Details for the validation cohort are shown in Supplementary Figs. S7 and S8.

Interaction analyses were conducted to evaluate clinicopathologic high-risk factors in predicting the efficacy of postoperative adjuvant chemotherapy (as shown in Fig. 4 and Supplementary Table S10). In both the primary and validation cohorts, no significant interaction was observed between adjuvant chemotherapy and traditional clinicopathologic high-risk factors. The traditional mid-/high- risk groups could not identify patients who would benefit from adjuvant chemotherapy (more details are shown in Supplementary Figs. S9 and S10). Only the CD206/CD68 ratio was identified as a significant predictive factor.

In subgroup analysis, patients were divided into three subgroups according to the NCCN guidelines (8): low-risk group as T3 (MSI-H) with no high-risk factors, mid-risk group as T3 (MSS/MSI-L) with no high-risk factors, and high-risk group as T3 with high-risk factors or T4. Interaction analysis was conducted in each subgroup.

In the primary cohort, for patients in the low-risk group (n = 35, 6.7%), no recurrence or death occurred. The interaction analysis was unavailable. In the mid-risk group (n = 241, 46.3%), no significant benefit of adjuvant chemotherapy for DFS or OS was observed in patients with low or high CD206/CD68 ratios. There was no significant interaction for DFS (P = 0.412 for interaction) or OS (P = 0.547 for interaction). In the high-risk group, a significant benefit of adjuvant chemotherapy was observed for both DFS (P = 0.001) and OS (P = 0.039) in patients with a high CD206/CD68 ratio but not in patients with a low CD206/CD68 ratio. The interaction analysis was significant for DFS (P = 0.024 for interaction) but not for OS (P = 0.136 for interaction). Details are shown in Fig. 5.

In the validation cohort, the results were similar. In the low-risk group (n = 26, 8.3%), no recurrence or death occurred. In the mid-risk group (n = 152, 48.4%), no significant benefit of adjuvant chemotherapy for DFS or OS was observed. In the high-risk group (n = 136, 43.3%), the benefit of adjuvant chemotherapy was significant for DFS (P = 0.010) and potentially significant for OS (P = 0.072) in patients with a high CD206/CD68 ratio but not in patients with a low CD206/CD68 ratio. The interaction analysis was significant for DFS (P = 0.046 for interaction) but not for OS (P = 0.091 for interaction). Details are shown in Supplementary Fig. S11.

In validation cohort, normal sections were also used for IHC staining, and were compared with TMA. Results of high ICC and kappa showed that normal sections were in good consistency with TMA (as shown in Supplementary Information S8). The CD206/CD68 ratio detected using normal sections was also confirmed effective prognostic and predictive biomarker. Details are shown in Supplementary Figs. S12, S13, and S14.

For stage II colon cancer, most patients can be cured by radical resection alone (29). Only approximately 15%–25% of patients suffering from recurrence may benefit from postoperative adjuvant chemotherapy. Thus, prognostic factors are expected to identify patients with a high risk of recurrence, which can be predictive of the efficacy of postoperative adjuvant chemotherapy. To identify these patients, traditional clinicopathologic high-risk factors are too broad, classifying more than half of patients as having a high risk of recurrence needing adjuvant chemotherapy (6, 7). Thus, the benefit of adjuvant chemotherapy is easily concealed and still not significant (5, 7). There has been some progress in molecular biomarkers, such as CDX-2 (30) and the multi-miRNA model (1), but more reliable evidence is still needed. MSI status has proven accurate and effective enough in identifying approximately 10%–15% of patients who do not benefit from FU-based adjuvant chemotherapy (11, 12). However, for the remainder of patients, there is still a need for better predictive biomarkers for adjuvant chemotherapy.

For stage II colon cancer, few patients suffer from local recurrence, which means that surgery is always effective. However, for patients suffering from distant metastases, tiny metastatic lesions could form at early time before primary tumor resection. And M2 TAMs play an important role in releasing circulating tumor cells (CTC). Studies have proved that M2 TAMs promote tumor cell vessel directional migration and invasion by the paracrine loop of tumor-derived CSF-1 and TAM-derived EGF/EGF-like ligands (31, 32) and secrete osteonectin (33), cathepsin (34), TGFβ (35), etc. M2 TAMs also promote tumor angiogenesis by releasing VEGF (36). Through these pathways, CTCs are significantly increased, leading to distant metastases. Thus, tiny metastatic lesions could be effectively predicted by M2 TAMs as biomarkers and eliminated by postoperative adjuvant chemotherapy.

CD206⁺ M2 TAMs identified by the IHC method have been confirmed as significant prognostic biomarkers for pancreatic adenocarcinoma (37, 38), renal cell carcinoma (39), gastric cancer (40), hepatocellular carcinoma (41, 42), and some other malignant tumors. In this study, we also confirmed CD206⁺ TAM density as a significant prognostic biomarker for stage II colon cancer. However, CD206⁺ TAM density could still not significantly predict adjuvant chemotherapy for stage II colon cancer. It should be noted that TAMs have two aspects (14–16): immune stimulation and immune inhibition. CD206⁺ TAM density could only reflect the inhibitory factors as the M2 subtype, but not the stimulatory factors as the M1 subtype. Thus, the prognostic efficacy was not accurate enough. As an improvement, we detected the CD206/CD68 ratio as the proportion of M2 TAMs in total TAMs, reflecting both immune stimulatory and inhibitory factors. The results showed that the CD206/CD68 ratio was a better prognostic factor than CD206⁺ TAM density and traditional clinicopathologic high-risk factors. More importantly, the CD206/CD68 ratio could successfully identify patients who really benefited from postoperative adjuvant chemotherapy. These results were consistent in two independent cohorts of the same medical center.

However, there were still limitations to this study. First, to conduct a real-world study, we enrolled consecutive patients to comprise the study cohort. Thus, there were imbalances at baseline, especially in the application of adjuvant chemotherapy and in the variation in regimens. Significantly more patients with high-risk factors received adjuvant chemotherapy. This imbalance could have interfered with the results. However, predictive biomarkers must face it in clinical applications. Second, the two cohorts of patients in this study came from the same medical center, lacking external validation. Thus, the extrapolation of results was limited. Third, our detection method of TAMs was not automated as it still needed the involvement of pathologists. Bias in manual detection could affect the results. Our detection method has been evaluated, and showed good stability and repeatability. However, the best way to solve this problem is to develop an automated detection method. And this would also significantly enhance the clinical applicability. Fourth, the predictive efficacy was significant for DFS in the interaction analysis but not significant for OS. For colon cancer, liver metastases play a major part in DFS events (67.0% in the primary cohort and 72.9% in the validation cohort in this study). However, some liver metastases can be cured by radical surgery, interfering with long-term survival.

In summary, our study showed that the CD206/CD68 ratio of TAMs could effectively classify patients with stage II colon cancer into groups with a low and high risk of tumor recurrence. Moreover, the CD206/CD68 ratio could identify patients who benefited from postoperative adjuvant chemotherapy. Therefore, patients will be treated with adjuvant chemotherapy more accurately, thus improving efficacy, reducing adverse events, and saving financial costs. At the same time, TAM detection using IHC methods is cheap, not time consuming, and can be easily conducted in the laboratories of most hospitals. Most patients can afford TAM detection for individual medical management. With further large-scale clinical verification, the CD206/CD68 ratio will aid in precision treatment for patients with stage II colon cancer.

No potential conflicts of interest were disclosed.

Conception and design: Q. Feng, W. Chang, G. He, P. Zheng, Y. Wei, Y. Tu, X. Qin, J. Xu

Development of methodology: Q. Feng, Y. Mao, G. He, P. Zheng, Y. Wei, L. Ren, D. Zhu, Y. Tu, J. Xu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Feng, Y. Mao, G. He, P. Zheng, W. Tang, Y. Wei, L. Ren, D. Zhu, Y. Tu, J. Xu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Feng, Y. Mao, G. He, P. Zheng, Y. Wei, D. Zhu, Y. Tu, J. Xu

Writing, review, and/or revision of the manuscript: Q. Feng, Y. Mao, G. He, Y. Wei, L. Ren, Y. Tu, J. Xu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Feng, W. Chang, Y. Mao, Y. Wei, D. Zhu, M. Ji, J. Xu

Study supervision: Q. Feng, G. He, Y. Wei, Y. Tu, J. Xu

This work was supported by National Natural Science Foundation of China (grant No. 81602040, to Q. Feng), National Natural Science Foundation of China (grant No. 81602035, to W. Chang), National Natural Science Foundation of China (grant No. 81472228, to J. Xu), and Science and Technology Project of Xiamen (grant No. 3502Z20154040, to Y. Tu).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.