Introduction

Colorectal cancer is the third most malignant tumor worldwide, defined by its high morbidity and mortality. Neoadjuvant chemoradiotherapy (NCRT) is a standard strategy recommended by the National Comprehensive Cancer Network guidelines for treatment of stage II/III rectal cancer. This approach has the potential to increase rates of pathologic complete response (pCR) and sphincter preservation. It has previously been shown that NCRT not only can reduce local recurrence rates but also can improve overall survival in patients with locally advanced rectal cancer. Nonetheless, only 5–40 % of the patients achieve a pCR to NCRT [1]. Therefore, it is important to understand this dichotomy in patient response to NCRT and differentiate between patients who will or will not benefit from this treatment. Recently, there have been various studies investigating markers associated with NCRT sensitivity in rectal cancer patients. However, the markers identified have limitations in clinical application. Therefore, further identification of biomarkers predicting the response to NCRT is required to establish which patients will most likely benefit from this treatment.

Angiogenesis is an essential process in malignant transformation, as the formation of new blood vessels ensures that the growing tumor receives an adequate oxygen supply. The genes associated with angiogenic pathways have been studied in many tumor types and provide a guideline for targeted therapy. Chemokines, a family of small cytokines secreted by cells, participate in pleiotropic functions including angiogenesis. In particular, the chemokine CXCL10 has been reported to play a role in angiostasis and have anti-tumor effects. In addition, CXCL10, as an anti-angiogenesis factor, has been studied in numerous diseases including colorectal cancer [2, 3]. In our previous study, we also demonstrated that high expression of CXCL10 was related to improved survival in colorectal cancer patients [4].

It has recently been shown that radiation has a considerable effect on chemokine expression, although changes in chemokine expression before radiation have not been well investigated. It seems probable that CXCL10 expression affects the chemoradiosensitivity of rectal cancer to NCRT. In this study, we measured the messenger RNA (mRNA) levels of CXCL10 in order to investigate the association between its expression and tumor response to NCRT. We hypothesized that CXCL10 would be a predictive marker for patient outcome after NCRT.

Materials and methods

Patients

This study included 95 patients, selected between 2007 and 2013 in Fudan University Shanghai Cancer Center. CXCL10 mRNA level analysis was performed in 63 patients; meanwhile, CXCL10 protein expression was analyzed in 53 patients. Among these, 21 patients underwent the two analyses. All patients had histologically confirmed rectal adenocarcinoma within 15 cm from the anal verge. The clinical T and N stages were identified by magnetic resonance imaging. Measurements of carcinoembryonic antigen (CEA) level, complete blood count, serum chemistry tests, colonoscopy, and abdominal and chest computerized tomography were performed before treatment. This study was approved by the institutional ethical committee, and all patients gave written informed consent allowing their tissues to be used in this study.

Treatment

All patients received NCRT. Radiation was given with a total dose of 44, 50, or 55 Gy in 25 fractions. During radiotherapy, patients also received a chemotherapy regimen consisting of oxaliplatin (50 mg/m2, qw) and capecitabine (625 mg/m2, bid). Surgery was performed 6–8 weeks following completion of NCRT.

Biopsies and pathological assessment

Pretreatment biopsies of rectal carcinoma were stored in RNAlater (Ambion/Applied Biosystems, Oslo, Norway). Pathological characteristics of post-treatment tumor include T and N stages, vascular and lymphatic invasion, and perineural invasion. Pathologic complete response was defined as the absence of viable tumor cells in the rectal wall and in any of the resected lymph nodes. Tumor downstaging was determined by comparing pretreatment T and N stages with the pathological stage of the surgical specimen [5]. To compare CXCL10 mRNA expression and protein expression with patient outcome to NCRT, patients were divided into two groups, pCR group and non-pCR (npCR) group, according to tumor response.

RNA extraction and real-time quantitative PCR

Total RNA was extracted from the pretreatment biopsies using an All-Prep RNA/DNA/Protein Mini Kit (Qiagen, GmbH, Hilden, Germany) according to the manufacturer’s instructions. Reverse transcription was performed using a PrimeScript RT reagent kit (TaKaRa, Otsu, Shiga, Japan) with 15-μl volume in each reaction, including 300 ng total RNA. Then, a 1-μl cDNA sample was used to perform a real-time quantitative polymerase chain reaction (RT-qPCR). To determine the amount of RNA in each sample, a standard curve was constructed. Quantification standards were prepared by the cloning of genes of interest. Specific primers and probes for CXCL10 and an internal control (ß-actin [ACTB]) were designed according to GenBank sequences (NM_001565.3 and NM_001101.3, respectively) using the Universal Probe Library Assay Design Centre via ProbeFinder software (Roche Diagnostics, Indianapolis, IN, USA). The forward (F) and reverse (R) primers were as follows: CXCL10: F, 5′-CAAATCTGCTTTTTAAAGAATGCTC-3′; R, 5′-AAGAATTTGGGCCCCTTG-3′; ACTB: F, 5′-ATTGGCAATGAGCGGTTC-3′; R, 5′-TGAAGGTAGTTTCGTGGATGC-3′. Raw expression data for CXCL10 mRNA levels were adjusted based on ACTB levels. RT-qPCR assays were carried out on the sequence detection system (ABI Prism 7900 HT; Applied Biosystems, Foster City, CA, USA) and conducted in triplicate for each sample to ensure experimental accuracy. The mean value was used for calculation. The cycling conditions were as follows: 10 min at 95 °C for 1 cycle; and 15 s at 95 °C, 30 s at 57 °C, and 30 s at 72 °C for 40 cycles.

Immunohistochemistry

Formalin-fixed paraffin-embedded tissue sections from preoperative biopsies were deparaffinized in xylene (10 min for three times) and rehydrated in ethanol series (75, 95, and 100 %, each for 5 min). Slides were treated with 3 % methanol-peroxide for 10 min to block endogenous proxidases and then incubated in a boiling 10-mM sodium citrate buffer for 10 min to retrieve antigen. Subsequently, we applied polyclonal rabbit anti-human CXCL10 antibody (1:300, Abcam, Cambridge, MA, USA) as a primary antibody overnight at 4 °C. Detection was done with a kit (GeneTech, Shanghai, China) for 30 min at room temperature. Finally, antibody binding was visualized with DAB and counterstained with hematoxylin.

Scoring of slides

The individual tissue cores for each slide (×200 magnification) were viewed using Aperio ImageScope (Aperio Technologies, CA, USA, version 11.2.0.780) and scored by applying the Positive Pixel Count Algorithm (version 9.1). The CXCL10 staining of the tumor tissue was marked with a pen tool for our later analysis. Data was expressed as positivity (number of positive pixel count / total number of positive+negative pixel count).

Statistics

CXCL10 mRNA expression and protein expression were respectively compared between the pCR and npCR groups using the Student’s t test. Then, the data were ranked based on percentile groups: low gene expression or low positivity for cases below the 50th percentile and high gene expression or high positivity for cases above the 50th percentile. Next, we used the chi-square test to compare differences between the pCR and npCR groups. The diagnostic performance of CXCL10 mRNA expression was assessed by means of receiver operating characteristic (ROC) curve analysis. All tests were two-tailed, and the significance level was set to 0.05. The analyses were performed using SPSS 13.0 software (SPSS Inc., Chicago, IL, USA).

Results

Clinical characteristics compared with tumor response

Clinical parameters, including age, gender, tumor distance from anal verge, pretreatment T and N stages, vascular invasion, and perineural invasion, were compared between tumor response groups. No significant differences were found between the pCR and npCR groups (Table 1).

Table 1 Clinical characteristics compared with tumor response
Full size table

Tumor response and CXCL10 mRNA expression

First, we used the normalized data as previously described to directly analyze the difference in CXCL10 mRNA expression between the pCR (19 patients) and npCR (44 patients) groups using the Student’s t test (Fig. 1). The mean mRNA expression level of CXCL10 in the pCR group was significantly higher than that in the npCR group (p = 0.010). We then performed two-category analysis to compare the distribution of high and low CXCL10 mRNA expression in the tumor response groups using the chi-square test (Table 2). The results showed that 73.7 % of the patients in the pCR group had high CXCL10 mRNA expression, and 61.4 % of the patients in the npCR group had low CXCL10 mRNA expression, and these differences were statistically significant (p = 0.011). This result indicates that patients with high CXCL10 mRNA expression are more likely to respond to NCRT (odds ratio 4.447; 95 % confidence interval [CI], 1.356–14.586). The diagnostic performance of CXCL10 mRNA expression, as assessed by the ROC curve (Fig. 2), showed an area under the curve of 0.720 (95 % CI, 0.573–0.867).

Fig. 1
figure 1

CXCL10 mRNA expression in the pCR and npCR groups. The comparison was conducted using the Student’s t test (p = 0.010). pCR pathological complete response, npCR non-pathological complete response, ACTB ß-actin (internal control)

Full size image
Table 2 Chi-square test for tumor response and CXCL10 mRNA expression
Full size table
Fig. 2
figure 2

Diagnostic performance of CXCL10 mRNA expression. The receiver operating characteristic (ROC) curve shows the combined sensitivity and specificity across different values of the predictive index, with an area under the ROC curve (AUC) of 0.720 (95 % confidence interval, 0.573–0.867)

Full size image

Clinical characteristics and CXCL10 mRNA expression

The t test used to analyze clinical parameters and CXCL10 mRNA expression (Table 3) showed an association between tumor distance from the anal verge and CXCL10 expression level (p = 0.028). In addition, there was a statistically significant association between post-operative T stage and CXCL10 mRNA expression (p = 0.024).We then analyzed the difference in clinical parameters between the two CXCL10 mRNA expression groups using the chi-square test. As shown in Table 4, the CXCL10 mRNA expression level was not statistically related to any of the clinical parameters (p > 0.05), with the exception of post-operative T stage (p = 0.016).

Table 3 T test for significant associations between clinical characteristics and CXCL10 mRNA expression
Full size table
Table 4 Chi-square test for significant associations between clinical characteristics and CXCL10 mRNA expression
Full size table

Tumor response and CXCL10 protein expression

Immunohistochemistry results of 53 pretreatment biopsies of rectal cancer patients, with 13 in the pCR group and 40 in the npCR group, were chosen to be analyzed. CXCL10 protein expression was mostly detected in the cytoplasm of rectal cancer cells, and partly in the tumor stroma (Fig. 3). Positivity of CXCL10 protein expression ranged from 0.01 to 62.88 %. Student’s t test showed no significant difference between the pCR and npCR groups in CXCL10 protein expression (Fig. 4). Still, we performed two-category analysis to compare the distribution of high and low positivity of CXCL10 protein expression in the tumor response groups using the chi-square test. As shown in Table 5, 53.8 % of the patients had high positivity of CXCL10 protein expression in the pCR group, and 50.0 % had low positivity in the npCR group, although there were no statistically significant differences (p > 0.05) (Table 5).

Fig. 3
figure 3

Immunohistochemistry analysis of CXCL10. a (×200 magnification) and b (×400 magnification): low positivity of CXCL10 protein expression in pretreatment biopsies. c (×200 magnification) and d (×400 magnification): high positivity of CXCL10 protein expression in pretreatment biopsies. CXCL10 expression was seen both in the cytoplasm of tumor cells and tumor stroma

Full size image
Fig. 4
figure 4

Positivity of CXCL10 protein expression in the pCR and npCR groups. The comparison was conducted using the Student’s t test (p > 0.05). pCR pathological complete response, npCR non-pathological complete response

Full size image
Table 5 Chi-square test for tumor response and positivity of CXCL10 protein expression
Full size table

Correlation between CXCL10 mRNA levels and protein expression

The CXCL10 mRNA levels and protein expression of 21 patients were examined in our research. There was no significantly statistical correlation found between the mRNA levels and protein expression (r = −0.070, p > 0.05) (Fig. 5).

Fig. 5
figure 5

Correlation between CXCL10 mRNA levels and protein expression. Pearson correlation value (r = −0.07) was calculated when performing the correlation analysis (p > 0.05)

Full size image

Discussion

In recent years, NCRT has improved the treatment outcomes for rectal cancer patients. However, some patients still fail to respond to this therapy. Numerous studies have been carried out to discover the association between tumor response to NCRT and pretreatment clinical characteristics, such as age, gender, CEA level, tumor position, tumor differentiation, and clinical staging. However, few factors have been verified as significant predictors. Likewise, in this study, no clinical parameters were found to be significantly different between the pCR and npCR groups. Therefore, it is important to identify molecular biomarkers, rather than disease characteristics, that may predict treatment response. Several genes have been proposed as predictive biomarkers associated with chemoradiosensitivity in rectal cancer, including EGFR, CD133, and thymidylate synthase, in addition to circulating free DNA and RNA [6–10]. A number of microarray studies have also identified genes with a predictive value [11–13]. However, most of these results remain controversial.

To our knowledge, the mRNA expression of CXCL10 has never been reported as a predictive biomarker for tumor response to NCRT in rectal cancer patients. As an angiostatic chemokine, CXCL10 plays a key role in the course of tumor growth, especially in the process of new vessel formation. Yates-Binder et al. demonstrated that a CXCL10-derived peptide can inhibit vascular endothelial growth factor (VEGF)-induced endothelial motility and tube formation. They also showed that this peptide could both prevent vessel formation and induce involution of newly formed vessels [14]. Further, Sato et al. demonstrated that CXCL10 levels showed a significant inverse correlation with VEGF levels in uterine cervical cancers. They concluded that CXCL10 might act through suppression of angiogenesis associated with VEGF [15]. The angiostatic effects of CXCL10 may be mediated by the p38 mitogen-activated protein kinase signaling pathway [16]. In the present study, we measured the mRNA levels of CXCL10 in rectal cancer patients by RT-qPCR and found high CXCL10 expression in patients who achieved pCR. A number of studies have reported that treatment with anti-angiogenic agents may increase the benefit of chemotherapy and radiotherapy [17, 18]. We hypothesized that high expression of CXCL10 in rectal cancer patients before treatment was indicative of angiostasis. Therefore, fewer abnormal vessels would form in the developing tumor, resulting in reduced vessel density. Similarly, a study by Yang et al. concluded that CXCL10 overexpression could lead to fewer abnormal vessels and reduce tumor growth in an ovarian cancer xenograft model [19]. One explanation for this result is that reduced vessel formation may lead to decreased oxygen perfusion among the tumor cells, resulting in hypoxia and consequently radiation resistance. However, it is also believed that the amount of newly formed vessels in the tumor stroma does not necessarily lead to increased blood flow [20]. Furthermore, not only the tumor cells but also endothelial cells require oxygen. The newly formed vessels are inefficient and may compromise oxygen perfusion and consumption [21]. According to this hypothesis, normal vessels will reorganize to increase the oxygen perfusion of tumor cells and decrease the oxygen supply to the endothelial cells of inefficient vessels. Under such circumstances, more oxygenic tumor cells will be sensitive to the radiation and killed. However, this proposed mechanism needs to be confirmed by additional studies.

There may also be other mechanisms that underlie CXCL10-mediated chemoradiosensitivity. One laboratory in China elucidated the molecular mechanism behind the anti-tumor activity of CXCL10 using HeLa cells. They demonstrated that CXCL10 upregulation after irradiation may prolong the G1 phase and delay the S phase of the cell cycle, as evidenced by upregulated p27 protein and downregulated cyclin E. Therefore, more cells stayed in the G1 phase, which resulted in higher sensitivity to radiation treatment [22].

Recently, Rentoftet al. has found that high CXCL10 mRNA expression was associated with a poor response to radiotherapy in patients with squamous cell carcinoma of the tongue. Besides, patients with low CXCL10 mRNA expression had a better survival [23]. Their conclusions are opposite to ours, which means that the effect of CXCL10 to radiation may be various in different cancers.

Through the comparison of clinical parameters and CXCL10 expression, we found that there was a significant difference between post-operative T stage and CXCL10 expression level. In the high CXCL10 expression group, there were more patients with a low post-operative T stage. This finding is similar to the previous result demonstrating an association between CXCL10 and tumor response. In addition, we found an association between low-sited rectal cancer and higher CXCL10 expression, although the basis for this relationship is currently unclear.

It has long been a question of what the correlation between mRNA and protein expression is. In the present study, we have not found any correlation between CXCL10 mRNA and protein expressions. Likewise, some other researchers have discovered a lack of mRNA–protein correlation when comparing the results of RT-PCR and immunohistochemistry [24, 25]. Different regulation mechanisms may play a role in influencing mRNA–protein correlation, such as synthesis and degradation rates, which will affect the amount of the two molecules differently. Specifically, transcription, mRNA degradation, post-transcription, translation, and protein degradation may all contribute to the variation in mRNA and protein concentrations [26]. These possible mechanisms in the regulation process of mRNA and protein expression inspired us to have a further investigation in the future.

There were certain limitations to our study, such as the small sample size. Therefore, additional studies with larger sample sizes and more paired samples may be necessary in order to further validate CXCL10 as a predictive biomarker. Moreover, because it has not been clarified through which mechanism CXCL10 may increase the sensitivity to NCRT in rectal cancer patients, further studies are necessary to further elucidate this association.

In conclusion, our results indicated that CXCL10 mRNA expression, rather than CXCL10 protein expression, has the potential to be a biomarker for predicting chemoradiosensitivity in rectal cancer patients who receive NCRT. We found that high expression of CXCL10 before treatment was associated with sensitivity to NCRT. However, further studies investigating the mechanism linking chemoradiosensitivity with CXCL10 are warranted.