CD86⁺/CD206⁺ tumor-associated macrophages predict prognosis of patients with intrahepatic cholangiocarcinoma

Clinical patients

A total of 322 ICC patients who underwent surgical resection and were pathologically diagnosed with ICC at Zhongshan Hospital, Fudan University (Shanghai, China) between May 2005 and April 2006 were enrolled in the study (Liu et al., 2016). Post-operative follow-up was consistent with harmonized standard. The clinicopathological and baseline demographic characteristics of the patients were retrospectively collected. OS was defined as the time frame from the date of operation to death, and recurrence-free survival time (RFS) was defined as the time interval from the date of surgery to recurrence (Dong et al., 2016). All patients were followed until recurrence of disease, death or lost to follow-up. This study was approved by the Clinical Research Ethics Committee of Zhongshan Hospital (No. y2017-179) and all patients provided written informed consent prior to the study. The consent inform is mainly concerned with informing patients of collecting their clinicopathological data without patient privacy exposure and obtaining biological samples for protein or genetic testing.

Immunohistochemistry

Protocols and details of tissue microarray (TMA) sections and immunohistochemistry (IHC) were performed as previously described (Liu et al., 2016). TMAs are produced using formalin-fixed paraffin-embedded (FFPE) ICC tissue samples. Representative regions were premarked in the paraffin-embedded tumor tissue. Sample cores were selected and taken from each representative ICC tissue. Serial sections (4 μm thick) were placed on slides coated with 3-aminopropyltriethoxysilane. IHC was performed using a two-step protocol. Paraffin sections were first deparaffinized and hydrated. After microwave antigen retrieval and neutralization of endogenous peroxidase, slides were preincubated with blocking serum and then incubated overnight with primary antibodies (anti-CD68 antibody, anti-CD86 antibody, Abcam; anti-CD206 antibody, Abcam; anti-CD31 antibody, Abcam; anti- LYVE1 antibody, diluted at 1:100, Abcam). Subsequently, the sections were serially rinsed, incubated with second antibodies. Positive staining was visualized with DAB (3, 3-diaminobenzidine), then counterstained with hematoxylin. The IHC staining of the TMA sections was evaluated manually under the microscope (DP73; Olympus, Tokyo, Japan). The positive staining cells were counted by Image Pro Plus 6.0 analysis software (Media Cybernetics, Bethesda, MD, USA) and the detailed procedure is described elsewhere (Brey et al., 2003).The positive staining TAMs cells were identified as brown colored, irrespective of color depth, and were counted in five randomly selected visual fields under high magnification (×200). Representative low or high immunostaining density of CD68⁺, CD86⁺ and CD206⁺ TAMs was recorded. The immunohistochemical staining was evaluated independently by two experienced researchers blinded to the patients’ clinical characteristics and outcomes. The intra-observer reproducibility was tested by obtaining statistical κ-scores (Landis & Koch, 1977). Based on the median values of CD68, CD86 and CD206 positive TAMs, the ICC patients were classified into high and low subgroups (Kuang et al., 2011).

Multiple immunofluorescence labeling of formalin-fixed paraffin-embedded tissue

Multiple immunofluorescence labeling of FFPE tissue was performed as mentioned previously (Robertson & Isacke, 2011). In brief, slides were dewaxed and rehydrated. Then transfer the slide rack into prewarmed target retrieval solution. After rinsed with PBS and immunofluorescence buffer, the samples were incubated with prediluted primary antibody, then washed with 1× PBS and incubated with corresponding secondary horseradish peroxidase-conjugated polymer for antibody conjugation. Fluorescein tyramide signal amplification plus (PerkinElmer, Waltham, MA, USA) was used for amplifying signals. After serials rinsing, slides were placed in specific retrieval buffer in microwave to remove redundant antibodies before the sequenced staining. Nuclei were counterstained with 4′-6-diamidino-2-phenylindole (DAPI) for 5 min. All slides were examined under a laser confocal microscope (Leica TCS SP5; Leica, Wetzler, Germany).

Statistical analysis

Statistical analyses were performed using SPSS software (20.0; SPSS, Inc., Chicago, IL, USA) and R software version 3.3.2. Univariate predictors of the status of immunohistochemical biomarkers (CD68, CD86 and CD206) were evaluated using the Pearson χ² test. OS and RFS curves were depicted using Kaplan–Meier method and compared with log-rank test. A Cox proportional hazards model was used to identify potential predictors and to assess the relationship between immunohistochemical biomarkers and survival. The statistically significant prognostic indicators for OS and RFS were selected to build nomogram models, using the “rms” package (R Foundation for Statistical Computing, Vienna, Austria). Furthermore, calibration plots and area under the receiver operating characteristic (ROC) curves (AUC) were used to evaluate the performance of the nomograms with “pROC” package (R Foundation for Statistical Computing, Vienna, Austria). A two-sided value of p < 0.05 was considered statistically significant.

Immunohistochemical and immunofluorescence characterizations of tumor-associated macrophages in ICC Patients

As shown in Fig. 1, the positive staining of CD68, CD86 and CD206 were observed mostly in the cytoplasm of TAMs. Figure S1 showed the immunofluorescence staining and the colocalization of CD68⁺, CD86⁺ and CD206⁺ macrophages in the same image. The average levels of CD68 positive staining cells (median, 96 cells/field) was higher than that of CD86 positive staining cells (median, 57 cells/field) and CD206 positive staining cells (median, 61 cells/field, Fig. 2).

Correlations between intratumoral expression of TAMs and clinicopathologic characteristics in ICC patients

Based on the immunohistochemical findings of TAMs in 322 ICC patients, the potential correlation between TAMs and patients’ clinical characteristics was evaluated. The associations were shown in Table 1. CD68⁺ TAMs had no association with patients’ clinicopathologic features. Low intratumoral infiltration of CD86⁺ TAMs positively correlated with higher preoperative CA-199 (p = 0.014), appearance of lymph node metastasis (p = 0.012), presence of liver cirrhosis (p = 0.008) and advanced TNM staging (p = 0.046); while high CD206⁺ TAMs infiltration was significantly associated with presence of lymph node metastasis (p = 0.005), vascular invasion (p = 0.038) and high-grade TNM staging (p = 0.001). To investigate the relationship among tumorinfiltrating M2 macrophages and angiogenesis and lymph node metastasis, IHC experiments were performed for the evaluation of microvessel density (MVD) and lymphatic microvessel density (LMVD) via staining of CD31 and LYVE-1 in 15 samples of both groups. As shown in Fig. S2A, Case 4 represented a sample with high density of M2-TAMs, MVD and LMVD while case 11 showed a sample with low density of M2-TAMs, MVD and LMVD. It revealed that the MVD was significantly higher in the CD206 high tissues compared with the CD206 low tissues (10.76 ± 0.66 vs. 7.47 ± 0.40, p = 0.0002), which indicated that M2 macrophages may stimulate angiogenesis (Fig. S2B). But there is no significant difference between the two groups in LMVD (p > 0.05) (Fig. S2C).

Survival analysis with a single macrophages immunomarker (CD68, CD86 and CD206) in ICC patients

The prognostic value of single macrophages immune marker (CD68, CD86 and CD206) in this cohort of 322 ICC patients was further investigated. The result demonstrated that, as a single variable, CD68⁺ TAMs density had no prognostic value (Figs. 3A and 3B). Patients with low CD86⁺ TAMs infiltration had a markedly worse median overall OS and shorter RFS (OS, 42.6 months; RFS, 37.0 months) when compared to those with high CD86⁺ TAMs expression (OS, 58.2 months, p = 0.010; RFS, 56.1 months, p = 0.008) (Figs. 3C and 3D). In contrast, patients with low CD206⁺ TAMs presence had a longer median OS and RFS (OS, 55.1 months; RFS, 50.6 months) than those with high CD206⁺ TAMs density (OS, 40.1 months, p = 0.003; RFS, 39.1 months, p = 0.021) (Figs. 3E and 3F). Collectively, these data indicated that CD86 and CD206 had a prognostic value in predicting survival and tumor recurrence of ICC patients.

Combined analysis of CD86 and CD206 improves predictive value of ICC patient outcome

Integrated analysis of CD86 and CD206 provided a more powerful prediction for ICC patient outcomes. Patients were classified into four subgroups based on intratumoral CD86⁺ and CD206⁺ TAMs number (I: CD86^high/CD206^low, II: CD86^low/CD206^low, III: CD86^high/CD206^high, IV: CD86^low/CD206^high). Subgroup comparisons showed that patients in the CD86^low/CD206^high group had the worst prognosis and the greatest risk of tumor recurrence. Conversely, ICC patients in the CD86^high/CD206^low group had the best prognosis (Figs. 4A and 4B). The median OS of Group I, II, III and IV were 62.8, 47.3, 48.7 and 31.3 months (p < 0.001), respectively. While the median RFS of Group I, II, III and IV were 58.9, 41.0, 48.5 and 26.7 months (p = 0.002) respectively. Taken together, these results evidently indicated that integrated analysis of immunomarkers CD86 and CD206 could serve as a more powerful predictor of prognosis in ICC patients.

As summarized in Tables 2 and 3, multivariable analyses revealed that apart from CA-199, lymph node metastasis and tumor multiplicity, CD86⁺/CD206⁺ TAMs model remained to be the independent prognostic indicator for both OS (p = 0.003) and RFS (p = 0.005). Collectively, these data suggested that integrated analysis of CD86⁺/CD206⁺ TAMs model is a valuable predictor for poor prognosis in ICC patients.

CD86⁺/CD206⁺ TAMs model predicts prognosis of carbohydrate antigen 199 negative ICC Patients

Notably, several ICC prognostic studies indicated that preoperative serum CA-199 level was a significant prognostic indicator for ICC patients after hepatectomy (Bergquist et al., 2016; He et al., 2018; Juntermanns et al., 2010; Kondo et al., 2014; Zheng et al., 2017). In general, high serum tumor marker CA-199 level was markedly associated with advanced tumor staging, increased risk for tumor recurrence, and poor OS (Bergquist et al., 2016; Juntermanns et al., 2010; Yoo et al., 2015). Resected patients with CA19-9 elevation had similar peri-operative outcomes but decreased long-term survival (Bergquist et al., 2016). However, some ICC patients with low preoperative serum level of CA-199 was associated with rapid disease progression (He et al., 2018). Worse yet, a reliable tumor marker to distinguish the prognosis of CA-199 negative ICC patients remains limited. To investigate the prognostic effect of CD86⁺/CD206⁺ TAMs model in CA-199 negative patients (cut-off value 36 U/ml) (Chen et al., 2015), 163 patients from the 322 ICC patient cohort were enrolled. In this CA-199 negative cohort, patients in CD86^high/CD206^low group had the best prognosis, while CD86^low/CD206^high group patients had the worst prognosis. The median OS of Group I, II, III and IV were 72.2, 57.4, 56.9 and 30.1 months (p = 0.002), respectively (Fig. 5A) and the median RFS of Group I, II, III and IV were 65.2, 46.0, 56.7 and 25.9 months (p = 0.005), respectively (Fig. 5B).

Construction and validation of the nomogram

Independent prognostic indexes were identified through univariate analysis and forward stepwise (Bendel & Afifi, 1977 ) multivariable regression analysis. CA-199 (p = 0.014), lymph node metastasis (p = 0.029), tumor multiplicity (p = 0.007) and CD86⁺/CD206⁺ TAMs model (p = 0.003) were utilized to construct the OS nomogram (Fig. 6A). On the other hand, lymph node metastasis (p < 0.001), tumor multiplicity (p = 0.002) and CD86⁺/CD206⁺ TAMs model (p = 0.005) were selected to build the RFS nomogram (Fig. 6B). The summed-up points of each nomogram variable score indicated a more accurate hierarchical prediction of patient outcomes. Higher total points are corresponding to worse OS and RFS probability.

Internal validation of the nomograms was carried out by calibration plots with bootstrap sampling (m = 100, n = 1,000) (Harrell, Lee & Mark, 1996). The C-indexes for the OS and RFS nomograms were 0.649 and 0.635, respectively. Both calibration plots were closely related to the 45-degree line (Figs. 6C–6F). Additionally, the ROC was performed to further evaluate the nomograms (Linden, 2006; Robin et al., 2011). The AUC was 0.6922 (95% CI [0.6312–0.7531]) for the OS nomogram (Fig. 6G) and 0.6351 (95% CI [0.5736–0.6966]) for the RFS nomogram (Fig. 6H). These results illustrated that the predicted and observed survival probabilities were in good concordance, and the goodness-of-fit was favorable.