DCE‐MRI for early evaluation of therapeutic response in esophageal cancer after concurrent chemoradiotherapy and its values in predicting HIF‐1α expression

To examine the feasibility of quantitative dynamic contrast‐enhanced magnetic resonance imaging (DCE‐MRI) in the early assessment of the therapeutic response to concurrent chemoradiotherapy (CRT) in esophageal cancer (EC) patients and to determine its value in predicting HIF‐1α expression. EC patients underwent DCE‐MRI 1 week pre‐CRT and 3 weeks post‐CRT (3w‐CRT). According to tumor regression post‐treatment, patients were divided into sensitive group (SG) and resistant group (RG). HIF‐1α expression was assessed by immunohistochemistry (IHC). Quantitative parameters (ktrans, kep, and ve) were compared between the SG and RG groups, as well as between the HIF‐1α(+) and HIF‐1α(−) groups. Receiver operating characteristic (ROC) curve analysis was performed to detect the best predictor of the above parameters in the therapeutic response and in predicting HIF‐1α expression. Totally 34 and 5 patients were included in the SG and RG, respectively. Pre‐ktrans and pre‐kep were decreased significantly in the SG at 3w‐CRT (p < .01), whereas only pre‐kep was decreased in the RG (p = .037). Pre‐ktrans was higher in the SG compared with the RG (p < .01). Meanwhile, absolute Δktrans (post‐ktrans–pre‐ktrans) was reduced more substantially in the SG compared with the RG. Δktrans also had the highest area under the curve (AUC = 0.929) in distinguishing SG from RG. Based on IHC, 13 and 11 patients were HIF‐1α(+) and HIF‐1α(−), respectively. At 3w‐CRT, post‐ktrans was markedly lower than pre‐ktrans in the HIF‐1α(+) group (p < .01); however, both ktrans and kep in the HIF‐1α(−) group were dramatically reduced than pre‐treatment values (both p < .01). Pre‐ktrans was significantly higher in the HIF‐1α(−) group compared with the HIF‐1α(+) group (p = .002) and constituted an excellent parameter for predicting HIF‐1α expression (AUC = 0.881). DCE‐MRI is effective in the early assessment of the therapeutic response after CRT, offering a novel noninvasive method for predicting HIF‐1α expression in advanced EC patients.

group (RG). HIF-1α expression was assessed by immunohistochemistry (IHC). Quantitative parameters (ktrans, kep, and ve) were compared between the SG and RG groups, as well as between the HIF-1α(+) and HIF-1α(À) groups. Receiver operating characteristic (ROC) curve analysis was performed to detect the best predictor of the above parameters in the therapeutic response and in predicting HIF-1α expression.
Totally 34 and 5 patients were included in the SG and RG, respectively. Pre-ktrans and pre-kep were decreased significantly in the SG at 3w-CRT (p < .01), whereas only pre-kep was decreased in the RG (p = .037). Pre-ktrans was higher in the SG compared with the RG (p < .01). Meanwhile, absolute Δktrans (post-ktrans-pre-ktrans) was reduced more substantially in the SG compared with the RG. Δktrans also had the highest area under the curve (AUC = 0.929) in distinguishing SG from RG. Based on IHC, 13 and 11 patients were HIF-1α(+) and HIF-1α(À), respectively. At 3w-CRT, post-ktrans was markedly lower than pre-ktrans in the HIF-1α(+) group (p < .01); however, both ktrans and kep in the HIF-1α(À) group were dramatically reduced than pre-treatment values (both p < .01). Pre-ktrans was significantly higher in the HIF-1α (À) group compared with the HIF-1α(+) group (p = .002) and constituted an excellent Xiaodong Xie and Lingling Gu contributed equally to this study. parameter for predicting HIF-1α expression (AUC = 0.881). DCE-MRI is effective in the early assessment of the therapeutic response after CRT, offering a novel noninvasive method for predicting HIF-1α expression in advanced EC patients.

K E Y W O R D S
concurrent chemoradiotherapy, dynamic contrast-enhanced magnetic resonance imaging, early evaluation, esophageal cancer, hypoxia-inducible factor-1-alpha

| INTRODUCTION
It is estimated that there are 570 000 new diagnoses of esophageal cancer (EC) and 510 000 deaths per year, representing the sixth most common cause of cancer-related deaths worldwide. 1,2 In China, it is also a leading cause of death, and the majority of cases are diagnosed at advanced stages and not eligible for surgery. 3,4 At present, a definitive CRT has been given priority as the standard treatment for inoperable EC, and good response could increase patient survival. 5,6 However, owing to individual differences and tumor heterogeneity, not all patients could benefit from the CRT approach. 7,8 The treatment effect depends heavily on the response to CRT, and it is crucial to predict the response as early as possible to avoid side effects for timely adjustment of treatment strategies.
The main traditional imaging methods for evaluating the treatment response of EC patients administered CRT include esophagography and computed tomography (CT), which analyze pathological changes of the esophagus only for morphology, lagging behind biometric changes. 9,10 Although fluorodeoxyglucose positron emission tomography/computed tomography (FDG-PET/CT) could reflect metabolic changes and is helpful in evaluating the treatment response to CRT in EC patients, radiation exposure and high cost are fatal disadvantages, especially for long-term follow-up.
Magnetic resonance imaging (MRI) has been widely used in tumor detection and treatment evaluation thanks to excellent soft tissue resolution and nonionizing radiation. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is an advanced MRI; beside the advantages of traditional MRI, it measures the properties of tumor microvascular structure and permeability, and evaluates the functional tissue information with quantitative parameters. 11 In DCE-MRI examination, quantitative parameters such as ktrans (volume transfer constant in min À1 ), kep (rate constant in min À1 ), and ve (volume fraction of the extravascular extracellular space, which is dimensionless) could be obtained from pharmacokinetic models. 12,13 Current studies have reported the potential role of DCE-MRI parameters in assessing the treatment response in different tumors 6,13-16 as well as its value in predicting tumor-related biomarkers. 17,18 Regarding the prediction of tumor-related biomarkers, based on the advantages of DEC-MRI in determining vascular permeability, hypoxia-inducible factor-1-alpha (HIF-1α) was chosen as a research indicator in this study. As a core transcription factor under hypoxia in the microenvironment, HIF-1α is involved in mediating biological behaviors such as apoptosis, proliferation, and migration. 19,20 Current reports have further suggested that hypoxic cells could not only increase the resistance of tumor cells to CRT, but also render the tumor more invasive and prone to metastasis. 21,22 In addition, higher HIF-1α expression is associated with lower treatment response and reduced survival. 23,24 Therefore, the prediction of HIF-1α expression would also contribute to early efficacy assessment in EC patients. At present, data reported about tumor hypoxia are typically dependent on the pathological method, which is an invasive and undesirable approach for some patients. Hence, it would be beneficial for the patients to predict HIF-1α expression by DCE-MRI in a noninvasive manner.
Based on the above analysis, this study aimed at detecting the role of DCE-MRI in detecting early treatment response in EC patients administered CRT. In addition, the possible associations of DCE-MRI parameters with HIF-1α expression were assessed, to offer a novel and noninvasive approach for prediction. Moreover, the relationships among DCE-MRI parameters, therapeutic response, and HIF-1α expression were explored.

| Ethics statement
The Institutional Review Board (IRB) of our hospital approved this study, and written informed consent was obtained from each participant. All experiments were performed in accordance with the ethical standards of the World Medical Association (Declaration of Helsinki).

| Patients
Inclusion criteria were (a) EC confirmed by endoscopic biopsy;

| MR image analysis
The dynamic data were processed with the Omni-Kinetics postprocessing software (GE Healthcare), which could fit T1-weighted DCE MRI data to the Tofts linear model, and quantitative kinetic parameters (ktrans, kep and ve) were calculated. Two radiologists with 8 and 11 years of experience in digestive radiology were blinded to the treatment results and independently performed data analysis.
In this study, we artificially divided esophageal lesions into upper, middle, and lower segments, and three regions of interest (ROIs) were manually outlined randomly in each segment ( Figure 1). Areas of necrotic tissue, hemorrhage, calcification, and blood vessels were avoided while setting the ROIs. Finally, the averages of the three segments (totally nine ROIs) for various parameters were obtained in every patient. On T2WI, the thickened esophageal wall had a relatively higher signal intensity compared with normal esophageal tissue. The small FOV and HR T2WI could make the signal intensity contrast more obvious; in addition, DWI/ADC could further help determine the exact boundaries of the esophageal lesions for delineating ROIs in CE-T1WI ( Figure 2).

| Immunohistochemical analysis of HIF-1α
HIF-1α was assessed in paraffin-embedded tissue samples sectioned at 5 μm. Briefly, all sections were deparaffinized, and antigen retrieval was performed under high pressure for 2 min. Nonspecific binding was blocked with serum for 15 min at 37 C. The sections were stained with primary monoclonal rabbit anti-human HIF-1α antibody (Abcam, Cambridge, UK) in a humidified chamber for 60 min at 37 C. The specimens were next stained with goat anti-rabbit secondary antibodies, in a humidified container for 30 min at 37 C. HIF-1α expression was visualized with 3, 3-diaminobenzidine (DAB) followed by counterstaining with hematoxylin. HIF-1α expression was determined by assessing the percentage of tumor cells with cytoplasmic staining, and staining intensity was evaluated with the following classification system: 0, no staining; I, staining in less than 10% of tumor cells; II, staining in 10%-50% of cells; III, staining in over 50% of cells. 0 and I were considered a negative (À) pattern, whereas II and III were positive (+) patterns. 26

| Statistical analysis
All statistical analyses were performed with the SPSS 23.0 statistical software (SPSS Inc., Chicago, IL). Categorical data were compared by the F I G U R E 1 Diagrammatic representation of region of interest (ROI) delineation. The esophageal lesion was artificially divided into upper, middle, and lower segments, each containing three random ROIs Fisher's exact test and Kruskal-Wallis test. The Shapiro-Wilk test was performed to determine whether quantitative parameters had a normal distribution. Normally distributed data were compared by the Student's t test; the Mann-Whitney U test was adopted for non-normally distributed data. Paired comparisons were performed by the Wilcoxon test.
The diagnostic performances of these parameters in predicting treatment response or HIF-1α expression were tested by receiver operating characteristic (ROC) curve analysis. The maximal Youden index (Youden index = sensitivity + specificity -1) was calculated to obtain a reasonable threshold. p < .05 was considered statistically significant.  Table 2, pre-ktrans and pre-kep were decreased significantly in the SG after 3w-CRT (p < .01), whereas only pre-kep was significantly reduced in the RG (p = .037). Pre-ktrans was also decreased in the RG at 3w-CRT, although it showed no statistically significant difference (p = .225). Although ve was increased in the RG and decreased in the SG after 3w-CRT, no statistical significance was detected (p = .319 and .48, respectively).

| Changes in parameters between the SG and
RG at the pre-CRT and 3w-CRT time-points As shown in Table 3, pre-ktrans was higher in the SG compared with the RG (p < .01), and absolute Δktrans was reduced more substantially F I G U R E 2 Multifunctional sequences for determining the exact boundary of an esophageal lesion for ROI delineation. A middle thoracic esophageal tumor at the pre-chemotherapy (CRT) and 3w-CRT time-points with multifunctional sequences images. The small field of view (FOV) and HR T2WI increase the signal intensity of the thickened esophageal wall more obviously. In addition, DWI/ADC could further help in determining the exact boundary of the esophageal lesion for ROI delineation on CE-T1WI in the SG compared with the RG. No statistically significant differences were detected between the SG and RG in post-ktrans (p = .473), pre-kep (p = .579), post-kep (p = .226), Δkep (p = .685), pre-ve (p = .475), post-ve (p = .914), and Δve (p = .38). According to ROC analysis, Δktrans was the best parameter for early distinction of SG from RG; at a threshold of 0.4416, its sensitivity was 97.1%, with a specificity of 80.0% (AUC = 0.929; Figure 3). 3.2 | DCE-MRI-derived parameters for predicting HIF-1α(À) and HIF-1α(+) EC patients
T A B L E 6 Comparisons of parameters between the HIF-1α(À) and HIF-1α(+) groups at the pre-CRT and 3w-CRT time-points Parameters HIF-1α(À) (n = 13) HIF-1α(+) (n = 11) p Evaluating the early response to CRT, we found that pre-ktrans in the SG was higher than that of the RG (p < .01) and pre-ktrans was decreased significantly in the SG at 3w-CRT (p < .01). To obtain adequate nutrients for growing and metastasizing, malignant neoplasms have developed both structurally and functionally new vessels, which are leaky with a hazard pattern of interconnections. 33 This results in higher blood flow and endothelial permeability in the tumor, thereby increasing k-trans. Therefore, we considered the relationship between high pre-ktrans and sensitive treatment may be associated with higher blood flow and endothelial permeability, which improves accessibility to chemotherapy and sensitivity to radiation. 15,34 In addition, we found that absolute Δktrans was reduced more substantially in the SG compared with the RG. In addition, it was considered the best parameter in distinguishing SG from RG. Δktrans is believed to represent a relative reduction of vessel endothelial permeability due to fibrosis, and the relative variation is often thought to be more representative and stable.
In HIF-1α prediction research, we compared parameters between the HIF-1α(À) and HIF-1α(+) groups at pre-CRT and 3w-CRT. The data showed that pre-ktrans was significantly higher in the HIF-1α(À) compared with the HIF-1α(+), and its sensitivity in distinguishing SG from RG could reach 100% (specificity of 76.1%; AUC 0.881). These findings suggest that pre-ktrans derived from DCE-MRI could be an excellent and promising imaging biomarker for predicting the expression the HIF-1α. Moreover, higher pre-ktrans in the HIF-1α(À) may reflect a better treatment response compared with the HIF-1α(+), according to the above early response evaluation and previous studies. 35 Inevitably, there were several limitations in this study. First, the sample size was relatively small, and a larger sample is necessary in further studies. Secondly, we used biopsy samples to evaluate the expression of HIF-1α in EC. Considering tumor heterogeneity, the F I G U R E 4 ROC curve for pre-ktrans. The pre-ktrans value was a promising parameter in predicting the expression of HIF-1α in EC patients, with 100% sensitivity, 76.9% specificity, and an AUC of 0.881 results may not represent the biomarker expression of the entire tumor.
Therefore, the assessment of the entire tumor is required to further study the associations of DCE-MRI parameters with tumor molecular markers. Thirdly, other immunohistochemical biomarkers are related to EC diagnosis, treatment, and prognosis. Thus, more investigations of cancer biomarkers and MRI parameters are warranted.
In conclusion, DCE-MRI could be a promising tool for early assessment of tumor response to CRT and HIF-1α expression prediction in advanced EC patients. Δktrans was the best parameter in early distinction of SG from RG, and pre-ktrans represented an excellent parameter in predicting HIF-1α expression. Also, the prediction of HIF-1α expression could also contribute to efficacy evaluation. The core conclusions of this study are shown in Figure 5.
F I G U R E 5 Diagrammatic representation of the core conclusions. DCE-MRI could help timely assess the response to CRT and predict HIF-1α expression in advanced EC patients. In addition, the prediction of HIF-1α expression would improve response assessment