Introduction

Colorectal cancer (CRC) is the third most common cancer and fourth leading cause of cancer-related deaths worldwide [1]. Despite improvements in surgical techniques, fatal disease recurrence occurs in 20–25 % of curatively treated patients [2]. Patients’ prognosis depends on the stage or anatomic extent of disease, based on the Union for International Cancer Control (UICC) tumor-node-metastasis (TNM) classification. The function of TNM staging has expanded from predicting prognosis to informing treatment choices [3, 4]. Several large randomized trials have suggested that all node-positive patients should receive adjuvant chemotherapy, while the value of adjuvant therapy for node-negative cases is controversial [3, 5]. The predictive value of TNM staging for identifying node-negative patients at risk of early recurrence in CRC is therefore limited. Accordingly, extensive research has been devoted to studying clinicopathological features and/or predictive molecular factors that may supplement TNM classification for predicting prognosis and recurrence in patients with CRC undergoing curative surgery.

The systemic inflammatory response (SIR) status is thought to be secondary to hypoxia or tumor necrosis and is associated with anti-apoptotic characteristics of cancer cells [6]. Preoperative serum SIR has been reported as a predictive biomarker of early recurrence in CRC patients treated with curative surgery [7], while neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), and C-reactive protein (CRP) are considered to be indicators of SIR status.

The molecular status of CRC, including microsatellite instability (MSI) and tumor-infiltrating lymphocytes (TILs), is also associated with high risk of recurrence in CRC patients treated with curative surgery. TILs can act as an indicator of the host’s tumoral immune response and have been associated with recurrence and improved clinical outcome in CRC, as well as providing an attractive target for immunotherapy [8–12]. MSI represents an alternative pathway of colorectal carcinogenesis, in which tumors arise as a result of mutation or hypermethylation in the DNA mismatch repair system. MSI status has been reported to be an independent prognostic predictor of improved survival and time to recurrence [13].Therefore, both SIR-related host and local tumoral factors should be investigated to determine the factors predicting early recurrence in CRC patients treated with curative surgery.

This study therefore aimed to elucidate the clinical significance of both preoperative serum SIR and local tumor TILs and MSI status to identify biomarkers of recurrence risk in CRC patients treated with curative surgery.

Methods

Patients and Specimens

A total of 157 patients with stage I, II, or III primary CRC who underwent surgical resection at the Department of Gastrointestinal and Pediatric Surgery, Mie University, Mie, Japan, from 2007 to 2011 were analyzed. Patients with the following criteria were excluded: preoperative radiotherapy or chemotherapy, diagnosis of multiple CRCs, history of cancer in another organ, familial CRC, and inflammatory bowel disease. All patients were classified postoperatively based on the UICC TNM classification [14], using 10 % formalin-fixed, paraffin-embedded (FFPE) specimens. The histomorphology of the primary tumors and lymph nodes was confirmed by the Department of Pathology, Mie University, Mie, Japan. Written informed consent was obtained from all patients according to the guidelines approved by our institutional research board.

Laboratory Measurements of Neutrophils, Lymphocytes, CRP, and Carcinoembryonic Antigen

Neutrophils, lymphocytes, CRP, and carcinoembryonic antigen (CEA) were measured in routine blood samples obtained within 1 week before operation. Patients were divided into two groups using an NLR cutoff value of 3, based on a previous report [15]. Patients were also categorized according to PLR ≤150 or >150, as reported previously [16]. The cutoff value for CRP was 0.5 mg/dl (≤0.5 and >0.5 mg/dl) and for CEA was 5 ng/µl (≤5 and >5 ng/µl), according to the normal range at our institute.

Immunohistochemistry

FFPE specimens were sliced into 5-µm sections and subjected to immunohistochemical analysis to detect Foxp3 and CD8 expression. After deparaffinization and dehydration, the sections were placed in 10 mM sodium citrate buffer (pH 6.0) and autoclaved at 121 °C for 10 min for antigen retrieval. The sections were incubated in 3 % hydrogen peroxide for 10 min, blocked, and incubated in normal goat serum (Vector Laboratories Inc, Burlingame, CA, USA) overnight at 4 °C for Foxp3 and for 1 h at room temperature for CD8. The primary antibodies used were monoclonal mouse anti-human Foxp3 antibody (clone: 236A/E, Abcam, Cambridge, UK; dilution 1:100) for regulatory T cells and monoclonal rabbit anti-human CD8 (clone: EP1150, GeneTex, San Antonio, TX, USA; dilution 1:1000) for cytotoxic T cells. Antibody binding was visualized using Envision reagents (Dako REAL EnVision Detection System; peroxidase/DAB+, Dako Cytomation, Glostrup, Denmark). All the sections were counterstained with hematoxylin–eosin prior to dehydration and mounting.

Scoring Foxp3- and CD8-Positive T Cells

Foxp3- and CD8-positive T cells were counted using a scanner system under a Biorevo BZ-9000 microscope (Keyence, Osaka, Japan). Each slide was scanned microscopically, and intratumoral Foxp3- and CD8-positive T cells were photographed at a magnification of 400× in three representative high-power fields. Foxp3- and CD8-positive T cells with any detectable staining above background levels were considered positive.

KRAS/BRAF Mutation and MSI Analysis

FFPE sections (10 µm thick) from 157 surgical CRC patients were evaluated for mutations. Hematoxylin–eosin-stained FFPE sections were microdissected to extract DNA from the tumor cells. Genomic DNA was extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen, Tokyo, Japan) according to the manufacturer’s protocol. DNA quantity and quality were assessed using a NanoDrop 1000 spectrophotometer (NanoDrop Technologies, Houston, TX, USA). KRAS (exons 2 and 3) and BRAF (V600E) mutations were analyzed by pyrosequencing using the primers listed in Table 1. Reactions were run on a PyroMark Q96 ID system (Qiagen). MSI status was determined by polymerase chain reaction analyses of five mononucleotide repeat microsatellite markers (BAT-25, BAT-26, NR-21, NR-24, and NR-27), as recommended previously [17]. The primer sequences have been described previously [17]. Tumors with instability at more than three of these markers were classified as showing MSI, and those with instability at fewer than two markers as showing microsatellite stability (MSS).

Table 1 Primer sequences for KRAS and BRAF mutation analysis
Full size table

Statistical Analysis

The associations between SIR status, TILs, MSI status, and clinicopathological features were analyzed using Chi-square tests. Disease-free survival (DFS) curves were analyzed using the Kaplan–Meier method, and differences were examined using log-rank tests. Cox’s proportional hazards regression tests were used to estimate univariate and multivariate hazard ratios for recurrence. Multivariate survival analyses were performed using factors identified as significant in univariate analyses. All P values were two-sided, and P < 0.05 was considered statistically significant. All statistical analyses were carried out using JMP version 10 (SAS Institute, Cary, NC, USA).

Results

Patient Characteristics

The study included a total of 90 men and 67 women with an average age of 66.9 years (range 35–89 years). The median follow-up time was 20.5 months (range 0.2–62.4 months). All patients were classified postoperatively based on the histopathological analysis using the UICC TNM classification criteria: stage I (n = 48), stage II (n = 55), and stage III (n = 54). Twenty-nine patients (18.4 %) developed recurrence.

Immunohistochemical Analysis of Tumor-Infiltrating Foxp3- and CD8-Positive T Cells

Intratumoral Foxp3- and CD8-positive T cells were detected (Fig. 1), with median numbers of 11 (range 0–99) and 15 (range 1–144), respectively. Patients were classified into high- and low-expressing groups for each antigen using the median values as cutoff points.

Fig. 1
figure 1

Immunohistochemical staining of Foxp3- and CD8-positive TILs. a Foxp3-positive TILs/low density (original magnification ×400). b CD8-positive TILs/low density (original magnification ×400). c Foxp3-positive TILs/high density (original magnification ×400). d CD8-positive TILs/high density (original magnification ×400). e, f Intratumoral densities of positively stained cells were measured using an automatic image analysis system

Full size image

KRAS/BRAF Mutation and Tumor MSI Analysis

KRAS and BRAF mutations were analyzed in CRCs from 152 patients. Fifty patients (32.9 %) had KRAS codon mutations, and five (0.03 %) harbored BRAF mutations. A total of 98 patients (64.5 %) had no mutations in either the BRAF or KRAS gene (wild type). MSI was analyzed in 151 patients, of whom 142 showed MSS and nine showed MSI.

Associations Between SIR Status as Host Factor and Clinicopathological Features

We analyzed the association between preoperative SIR status and clinicopathological features (Table 2). High NLR was significantly associated with lymphatic invasion (P = 0.0371) and recurrence (P = 0.0062), while high PLR was also associated with recurrence (P = 0.006). In addition, high CRP level was significantly associated with T classification (P = 0.0004), recurrence (P = 0.0016), and TNM stage (P = 0.0091).

Table 2 Correlations between SIR status and clinicopathological features in CRC patients treated with curative surgery
Full size table

Associations Between TILs, MSI, and CEA as Tumor Factors and Clinicopathological Features

We analyzed the correlations between TILs, MSI, and CEA as tumor factors and clinicopathological features (Table 3). A low CD8-positive T cell count was significantly associated with venous invasion (P = 0.0185) and recurrence (P = 0.0084). However, Foxp3-positive T cell was not associated with clinicopathological features. MSI was significantly associated with undifferentiated histology (P = 0.022). High CEA levels were significantly associated with age (P = 0.0488), T classification (P < 0.0001), lymphatic invasion (P = 0.0133), venous invasion (P = 0.0234), recurrence (P = 0.0015), and TNM stage (P = 0.0006).

Table 3 Correlations between tumor factors including TILs, MSI, and CEA and clinicopathological features in CRC patients treated with curative surgery
Full size table

Associations Between MSI and TILs

We also investigated the associations between MSI and TILs. There was a trend toward more infiltrating Foxp3-positive T cells in patients with MSI, though the association was not significant (103 cells in MSI vs. 74 cells in MSS; P = 0.0585) (Fig. 2a), and significantly more infiltrating CD8-positive T cells were detected in patients with MSI (130 cells in MSI vs. 73 cells in MSS; P = 0.0001) (Fig. 2b).

Fig. 2
figure 2

Numbers of TILs in MSI and MSS tumors. a Numbers of Foxp3-positive T cells in MSI and MSS tumors. b Numbers of CD8-positive T cells in MSI and MSS tumors

Full size image

Few Infiltrating CD8-Positive T Cells and High CRP Levels Are Independent Predictors of Recurrence in CRC Patients Treated with Curative Surgery

We analyzed DFS according to SIR status. Kaplan–Meier analysis showed that patients with high NLR, PLR, and CRP as host factors had significantly poorer DFS than patients with lower levels (P = 0.0232, P = 0.0233, P = 0.0017, respectively) (Fig. 3a–c). We also investigated DFS according to tumor status, including TILs, MSI, and preoperative CEA. Patients with fewer infiltrating Foxp3-positive T cells had poorer DFS than those with more infiltration, though the difference was not significant (P = 0.2568) (Fig. 3d), while patients with fewer infiltrating CD8-positive T cells had significantly poorer DFS than those with more infiltration (P = 0.0028) (Fig. 3e). In contrast, there was no association between MSI and DFS (Fig. 3f). High preoperative CEA levels were significantly associated with early recurrence (P = 0.0010) (Fig. 3g).

Fig. 3
figure 3

Kaplan–Meier curves for DFS classified according to SIR status, TILs, MSI status, and CEA levels in CRC patients treated with curative surgery. a DFS according to NLR. b DFS according to PLR. c DFS according to CRP levels. d DFS according to density of Foxp3-positive T cells. e DFS according to density of CD8-positive T cells. f DFS according to MSI status. g DFS according to CEA levels. h DFS according to combined score of few infiltrating CD8-positive T cells and high CRP levels

Full size image

We also performed univariate analysis to detect important factors associated with DFS. T3/4 classification (P < 0.0001), lymphatic node metastasis (P = 0.0028), poor differentiation/mucinous (P = 0.0120), lymphatic invasion (P = 0.0307), venous invasion (P = 0.0002), high CEA levels (P = 0.0010), high CRP levels (P = 0.0044), high NLR (P = 0.0266), high PLR (P = 0.0182), and few infiltrating CD8-positive T cells (P = 0.0026) were all significant predictive factors for poor DFS (Table 4). Multivariate analysis of these factors identified few infiltrating CD8-positive T cells (hazard ratio (HR) = 4.3, P = 0.0059) as a tumor factor and high CRP (HR = 3.07, P = 0.0145) as a host factor, as independent predictive markers for recurrence in CRC patients treated with curative surgery (Table 4).

Table 4 Univariate and multivariate analyses of DFS in CRC patients treated with curative surgery
Full size table

Combination of Few Infiltrating CD8-Positive T Cells and High CRP Levels Increase the Predictive Accuracy for Early Recurrence in CRC Patients Treated with Curative Surgery

We assessed the effect of combining independent predictive factors (preoperative CRP and intratumoral CD8 T cells) on the predictive accuracy for early recurrence in CRC patients treated with curative surgery. We divided the patients into three groups based on the following scores: high CRP, +1; low CRP, 0; few infiltrating CD8-positive T cells, +1; more infiltrating CD8-positive T cells, 0. Sixteen patients had a score of 2, 77 had a score of 1, and 59 had a score of 0. Kaplan–Meier analysis showed that patients with a score of 2 had significantly poorer DFS than the other two groups (P < 0.0001) (Fig. 3h) and that the risk of recurrence was increased when both independent factors were combined (HR = 5.26).

Discussion

The UICC staging system provides the most reliable indication of prognosis and is useful for discriminating between patients with early-stage disease and those with advanced disease. However, its ability to predict prognosis in patients with intermediate levels of tumor invasion (stage II or stage III) is less accurate. Sensitive biomarkers are therefore needed to allow the targeting of postoperative adjuvant chemotherapy to those patients at highest risk of early relapse, with a resultant improvement in survival. This study provides the first comprehensive analysis for identifying predictive biomarkers based on both the host immune response and the local tumor molecular status. We showed that high SIR status, including NLR, PLR, and CRP levels as host factors, and few intratumoral CD8-positive T cells and high CEA levels as tumor factors, was significantly associated with early recurrence in patients with surgically resected CRC. In addition, multivariate analysis revealed that high CRP levels and few intratumoral CD8-positive T cells were independent predictors of recurrence. Furthermore, the combination of both host and tumor factors increased the predictive accuracy for determining recurrence risk in CRC patients treated with curative surgery. This new predictive test, which is based on the widely available results of CRP assays, combined with CD8 immunohistochemistry that is feasible in most laboratories, represents a step forward for accurately identifying curatively treated CRC patients at risk of recurrence in the clinical setting.

Cancer progression depends on complex interactions between the tumor, its microenvironment, and the host immune response. Inflammation has been implicated in the pathogenesis of many adult malignancies and is now recognized as a hallmark of tumorigenesis [18]. Park et al. [19] showed that the host inflammatory response to CRC influenced disease recurrence and survival via the SIR. However, the tumor may also encourage the inflow of inflammatory lymphocytes, resulting in cell destruction within the surrounding tissue, generating a more widespread, nonspecific inflammatory response [20]. Several studies have shown an association between the local inflammatory response, indicated by increased T cell tumor infiltration, and improved prognosis and recurrence in CRC [8, 21, 22]. It is therefore important to carry out a comprehensive analysis of biomarkers of early recurrence based on both systemic and local inflammatory factors in patients with CRC treated with curative intent.

SIR status, reflected by CRP, NLR, and PLR, is thought to be a surrogate indicator of the host immune response to tumors and has been shown to act as a biomarker of outcome in a variety of malignancies [23–25].

CRP is an essential acute-phase reactant that acts as a surveillance molecule for activation of the adaptive immune system. It is synthesized in hepatocytes and is upregulated by cytokines such as interleukin-6 and tumor necrosis factor-α [26]. Several studies have demonstrated associations between high CRP levels and increased risk of early recurrence and poor outcome in CRC patients treated with curative surgery [27–29]. However, although CRP represents an excellent biomarker for oncological outcome, we should be aware of its limitations; CRP levels cannot discriminate between patients with several cancers, including CRC, and those with inflammatory conditions such as inflammatory bowel disease, collagen disease, rheumatic disease, and cardiovascular disease [30].

Patients with high NLRs also have the balance tipped in favor of pro-tumor inflammatory status, meaning that a high NLR is a predictor of recurrence in CRC patients treated with curative surgery [31, 32]. Both thrombocytosis and lymphocytopenia have also been shown to correlate with the degree of host systemic inflammation, and the ratio of these factors (PLR) might reflect a novel inflammatory factor incorporating these two individual hematologic factors [33]. Szkandera et al. [34] found that PLR was a predictive marker of recurrence in CRC patients treated with curative surgery. Collectively, SIR status thus comprises potential biomarkers for predicting early recurrence in CRC, though no previous studies have compared the predictive values of CRP level, NLR, and PLR in terms of CRC recurrence. However, the current multivariate analysis demonstrated that high CRP level was the best component of SIR for predicting recurrence in CRC patients treated with curative surgery.

CRCs are generally immunogenic and often infiltrated by T cells [35], suggesting that clinical outcome and recurrence may depend on the status of the local adaptive immune response. Upon antigenic stimulation, CD8-positive T cells differentiate into effector cells that kill tumor cells by releasing toxic granules such as granzyme B and perforins [36, 37]. A high density of TILs in CRC tissue is associated with increased tumor cell apoptosis [38]. However, regulatory T cells expressing the Foxp3 transcription factor may block the adaptive immune response [39]. Foxp3-positive T cells can suppress host-mediated antitumor immunity and tumor-specific cytotoxicity, suggesting regulatory T cell deletion as a potential therapeutic strategy [40]. Recent evidence has also indicated that signaling by the T cell chemoattractant CCL5 can recruit Foxp3-positive T cells to tumors and enhance their ability to kill CD8-positive T cells, thereby providing a mechanism of immune escape [41]. A high density of infiltrating Foxp3-positive T cells has thus been shown to be associated with an adverse prognosis in some tumor types [39, 42]. However, several reports have indicated favorable impact of infiltrating Foxp-3 T cells in other tumors, including CRC [43–47], and the role of Foxp3-positive T cells thus remains controversial. The results of the present study revealed that, as components of the local inflammatory response, infiltrating CD8-negative T lymphocytes, but not Foxp3-positive T lymphocytes, were significantly associated with poor DFS and were an independent predictive factor for early recurrence in CRC patients treated with curative surgery.

The majority of hereditary nonpolyposis CRC tumors are characterized by MSI, while only 10–15 % of sporadic CRC cases display MSI, predominantly caused by epigenetic hypermethylation of the MLH1 mismatch repair gene. A previous systematic review of the prognostic and predictive values of MSI status found that MSI was associated with better overall survival and DFS [48]. Closer analysis of the clinical data suggested no benefit from 5-fluorouracil treatment in patients with MSI CRC, supported by the fact that patients with MSI tumors have a better prognosis than those with MSS [49]. The better outcome in patients with MSI may be partly attributable to the generation of neoantigens as a result of mutational frameshifts within the coding regions of specific genes, as a consequence of inactivation of DNA mismatch repair in CRC epithelial cells. This attracts specific immune cells that help to contain the tumor and limit metastasis [50, 51]. Several previous studies in CRC patients have demonstrated an association between MSI and increased intraepithelial CD8-positive T cells compared with patients with MSS [51–53]. In addition, the density of intratumoral Foxp3-positive T cells is significantly higher in MSI compared with MSS CRC, paralleling the enhanced number of CD8-positive cells [54]. Our current study found no significant relationship between MSI status and recurrence, possibly because of the small sample size; however, we did demonstrate a significant association between TILs and MSI status in CRC.

In conclusion, this study demonstrates that recurrence in CRC patients treated with curative surgery is affected by both systemic and local inflammatory statuses. Host inflammatory status, represented by preoperative serum CRP levels, and local tumor inflammatory status, represented by CD8-positive infiltrating T cells, were independent predictive factors for recurrence in curatively resected CRC patients. The combination of CRP level and intratumoral CD8-positive T cells could thus be used to identify CRC patients who require adjuvant chemotherapy after curative surgery.