The immunologic advantage of recurrent nasopharyngeal carcinoma from the viewpoint of Galectin-9/Tim-3-related changes in the tumour microenvironment

Given salvage treatment for recurrent nasopharyngeal carcinoma (NPC) remains a clinical dilemma, immunotherapy targeting NPC-specific immunosuppression may bring new hope. We analyzed the expression of CD8, CD4, Foxp3 and Tim-3 in lymphocytes, and of Galectin-9 in tumour cells between paired primary and recurrent NPC from 95 patients and we noted that there was significant increase in the expression of Galectin-9+ tumour cells (p < 0.001) and Foxp3+ lymphocytes (p < 0.001) but a significant decrease in the expression of CD8+ lymphocytes (p = 0.01) between paired primary and recurrent NPC. Of all patients, 53 patients (55.79%) and 57 patients (60%) had increased percentages of Galectin-9+ tumour cells and of Foxp3+ lymphocytes, respectively. Conversely, 42 patients (44.21%) had decreased percentages of CD8+ lymphocytes. The patients with high Galectin-9 expression in recurrent NPC frequently also had high Tim-3 (p = 0.04) and Foxp3 (p = 0.01), and low CD8 (p = 0.04) expression in lymphocytes. After multivariate analyses, low CD8 expression in lymphocytes was an independent risk factor for relapse-free survival (p = 0.002) and overall survival (p = 0.02). Our data suggests that recurrent NPC may had more immunologic advantage than primary NPC, especially the Galectin-9/Tim-3 pathway. The immunotherapies targeting Galectin-9/Tim-3/Foxp3 interaction may serve as a potential salvage treatment for recurrent NPC.

In recent years, the progress in immunotherapy has opened new fields in cancer treatment. For example, multiple immunosuppressive mechanisms in NPC have been proposed, including interleukin-10, transforming growth factor β1, Fas ligand, Foxp3+ regulatory T cells and Galectin-9 [9][10][11][12][13][14] . The Galectin-9/Tim-3 axis had been reported to be one of the NPC-specific immunosuppression pathway 13 . The immunotherapeutic drug targeting this axis are also available now 15 . Therefore, it is interesting and important to discover the changes of Galectin-9/Tim-3 pathway and generalized immunologic status (CD8+ , CD4+ and Foxp3+ lymphocytes in tumour microenvironment) between primary and recurrent NPC. The change in immunologic status between primary and recurrent NPC may have clinical implication in understanding the possibility of immunotherapy for the recurrent NPC patients. Therefore, we evaluated the paired immunologic expression of CD8+ lymphocytes, CD4+ lymphocytes, Foxp3+ lymphocytes, and the Galectin-9/Tim-3 pathway in the tumour microenvironment between primary and recurrent NPC from 95 patients.
Immunotherapy targeting NPC-specific immunosuppression may bring new hope for patients with recurrent NPC. For clinicians, the question of whether immunotherapy could be the choice of salvage treatment is imperative to answer. To answer this question, we analyzed the immunologic changes in the tumour microenvironment between primary and recurrent NPC. The scientific aim of our study is to evaluate the consequential changes in tumour immunology from chemotherapy and radiotherapy of primary NPC and understand the possible clinical implication of immunotherapeutic potential for patients with recurrent NPC in the future.

Results
Patient demographics. A total of 95 eligible patients with recurrent NPC were enrolled in this study, including 70 male and 25 female patients. Their ages ranged from 27 to 77 years, with a mean age of 50 ± 11 years at diagnosis. The length of time between when primary treatment for NPC ended and diagnosis of recurrent NPC occurred ranged from 1 to 163 months, with a mean of 26 ± 26 months. The overall follow-up period was from 12 to 174 months, with a mean of 59 ± 27 months. The recurrent NPC profile of the 97 patients are as follows: 43 patients had local recurrence in nasopharynx, 12 patients had local and regional neck recurrence, 11 patients had regional neck recurrence, 11  Furthermore, the differences in the expression of these immunologic markers between primary and recurrent NPC were verified by Wilcoxon signed-rank test. All changes in the expression of the tested immunologic markers in the paired samples are shown in Fig. 2. Again, there were significant differences in the percentages of Galectin-9+ tumour cells, Foxp3+ lymphocytes and CD8+ lymphocytes between the paired primary and recurrent NPC. The percentages of Galectin-9+ tumour cells and Foxp3+ lymphocytes increased significantly (p < 0.001 and p < 0.001, respectively) in recurrent NPC. Conversely, the percentage of CD8+ lymphocytes decreased significantly in recurrent NPC (p = 0.01).  The different patterns of immunologic markers between patients with low and high Galectin-9 expression levels in recurrent NPC. The basic characteristics and clinicopathological data of the patients with high and low Galectin-9 expression levels in recurrent NPC are shown in Table 2. There were no other significant differences between the groups with low and high Galectin-9 expression. However, the expression patterns of specific immunologic markers including CD8, Tim-3 and Foxp3 in lymphocytes differed in the group with high Galectin-9 expression (Table 3) from those with low Galectin-9 expression. In summary, the patients with high Galectin-9 expression in recurrent NPC was associated with a higher percentage of Tim-3+ lymphocytes and a more immunocompromised status (a lower percentage of CD8+ and higher percentage of Foxp3+ lymphocytes) in the tumour microenvironment.
Immunologic expression had a significant association with relapse-free survival and overall survival. In the univariate analysis (Table 4), a high percentage of Galectin-9+ tumour cells (p = 0.004), a low percentage of CD8+ lymphocytes (p = 0.005) and a high percentage of Foxp3+ lymphocytes (p = 0.004) in the recurrent tumour microenvironment as well as in the advanced stages of recurrent NPC (p < 0.001) were identified as significant risk factors for relapse-free survival. For overall survival, only an advanced stage of recurrent NPC (p < 0.001) was a significant risk factor. The relapse-free survival and overall survival curves with significant difference (p < 0.05) are shown in Fig. 3. All significant risk factors for both relapse-free survival and overall survival were further examined by multivariate analysis (Table 5). This analysis revealed that an advanced stage of recurrent disease (hazard ratio (HR): 2.58, p = 0.001) and a low percentage of CD8+ lymphocytes (HR: 2.40, p = 0.002) were independent risk factors for worse relapse-free survival. For overall survival, the analysis also revealed that an advanced stage of recurrent disease (HR: 3.92, p < 0.001) and a low percentage of CD8+ lymphocytes (HR: 2.16, p = 0.02) were independent risk factors. In short, the patients with recurrent NPC who had an advanced stage of disease or a low percentage

Discussion
After we compared immunologic marker expression in the tumour microenvironment between primary and recurrent NPC, we report two interesting findings from our study. The first finding primarily focuses on the immunologic changes between primary and recurrent NPC. In contrast to primary NPC, more than half of the patients with recurrent NPC had increased percentages of Galectin-9+ tumour cells and of Foxp3+ lymphocytes. More than one-third of the patients with recurrent NPC had increased percentages of Tim-3+ and CD4+ lymphocytes. Conversely, nearly half of the patients with recurrent NPC had a decrease in the percentage of CD8+ lymphocytes. Furthermore, the increased percentages of Galectin-9+ tumour cells and Foxp3+ lymphocytes and the decreased percentage of CD8+ lymphocytes significantly differed between primary and recurrent NPC. Finally, the patients with recurrent NPC had a high percentage of Galectin-9+ tumour cells accompanied by a low percentage of CD8+ lymphocytes and high percentages of Foxp3+ and Tim-3+ lymphocytes.
Theoretically, NPC is an EBV-related malignancy. During the process of carcinogenesis, NPC cells still retain functional antigen-presenting machinery. Therefore, the immune escape has been regarded as a specific feature of the tumour microenvironment in NPC 16,17 . Among several possible immune escape mechanisms, the immune checkpoint Galectin-9/Tim-3 interaction has been reported as one of the most common types in NPC 13,14,18,19 . This immunologic interaction could promote the formation of Foxp3+ lymphocytes and then inhibit the population of CD8+ lymphocytes 20,21 . Moreover, it has been shown that Galectin-9 was extremely abundant in NPC cells in either xenografted tumours or clinical specimens 19 . Because patients with recurrent NPC in our series had a significantly higher percentage of Galectin-9+ tumour cells, we speculate that the initial treatment may have an important effect on tumour selection. It has been reported that radiation induces dying cells to express significantly more tumour-specific antigens on their surface, resulting in CD8+ lymphocytes being recruited to perform tumour-specific killing 22 . This finding suggests that Galectin-9+ tumour cells in primary NPC may be more resistant to this radiation-induced cytotoxic killing by CD8+ lymphocytes. Subsequently, these Galectin-9+ tumour cells had selective growth advantage in recurrent NPC. In recurrent NPC, to maintain immune escape, these high Galectin-9 expressing NPC cells maintains their ability to inactivate the survival and growth of tumour-infiltrating CD8+ lymphocytes. In addition, this inhibitory effect of CD8+ lymphocytes may be related to the presence of more Foxp3+ lymphocytes in the tumour microenvironment of recurrent NPC 23 .
Our second finding primarily focuses on the impact of immunologic markers in the tumour microenvironment, specifically how a low percentage of CD8+ lymphocytes, high percentage of Galectin-9+ NPC cells and high percentage of Foxp3+ lymphocytes affect the outcome of recurrent NPC. Although a high percentage of Galectin-9+ tumour cells and a high percentage of Foxp3+ lymphocytes in the tumour microenvironment were only marginally significant, a low percentage of CD8+ lymphocytes was a significant risk factor for relapse-free survival and overall survival. Therefore, it is reasonable to speculate that the Galectin-9/Tim-3/Foxp3+ immunologic interaction and its subsequent suppression in CD8+ lymphocytes may play an important role in determining the outcome of salvage treatment. In fact, the association between CD8+ lymphocytes in the tumour microenvironment and treatment outcomes have been reported in NPC and other malignancies [24][25][26][27] . Recurrent NPC cells with high Galectin-9 expression had the ability to maintain immune escape by interacting with Tim-3+ lymphocytes 13,18,28 . The low percentage of CD8+ lymphocytes in the tumour microenvironment, which is a sign of an immunocompromised status, could serve as an effective biomarker to predict the response to salvage treatment in recurrent NPC.
In recurrent NPC, previous radiation could alter the tumour microenvironment and induce tissue fibrosis and microvasculature damage. These changes in physical barriers might play a role in tumour-mediated immune clearance because cells of the adaptive immune system must be able to infiltrate or invade into the tumour to   eradicate it. In our series, we frequently identified fibrotic tissue changes (Fig. 4) in the previous radiation field. This radiation-associated physical barrier may partially account for the decreased percentage of CD8+ lymphocytes in recurrent NPC. However, this radiation-associated fibrosis did not prevent the infiltration of Tim-3+ and Foxp3+ lymphocytes in the tumour microenvironment. A recent breakthrough in the treatment of melanoma and other cancers with immune checkpoint inhibitors has rejuvenated enthusiasm in immunotherapy for cancer treatment [29][30][31][32] . Currently, the choice of salvage treatment is limited for recurrent NPC. Immunotherapeutic strategies, especially those targeting the Galectin-9/Tim-3/Foxp3 interaction, may serve as the potential salvage treatments for recurrent NPC 33 in recurrent NPC. There were some limitations in our study. First, this retrospective study may be vulnerable to selection bias. Although we try to statistically control for this bias using Cox regression model, residual bias may remain. Second, the immunologic characteristics in different histological types, local nasopharynx, regional neck and distant sites, may have originally had different immunologic profiles 34 . These differences may reflect underlying mechanisms in the effect of immunologic changes in recurrent NPC. The strength of this study is that we succeeded in comparing all tissue samples between primary and recurrent NPC in 95 patients. Well-designed prospective case series studies with more patients are necessary in the future to clarify the changes in and the roles of immunologic profiles for recurrent NPC.

Patients and Methods
Ethics Statement. This study was approved by the Institutional Review Board of National Taiwan University Hospital. As this was a retrospective analysis of routine data, we requested and were granted a waiver of individual informed consent from the ethics committee. All methods were performed in accordance with the relevant guidelines and regulations. Patient records/information were anonymized and de-identified prior to analyses.

Statistical analysis. All immunologic factors were tested for normal distribution by Kolmogorov-Smirnov
Test. The Wilcoxon signed-rank test was used to compare the changes in percentages of all immunologic markers between primary and recurrent NPC. The median percentages of all immunologic markers were used as the cut-off points for grouping the patients into high or low level of expression groups. Chi-square tests and Fisher's exact tests (only in cell number less than five) were used to determine differences in the clinical and immunological characteristics (low or high expression) between the patients with low and high Galectin 9 expression as appropriate. The starting point of the follow-up period was defined as the completion of the curative treatment of primary NPC for each patient. The end point of the follow-up period was either June 2016 or the time when the patient died. The primary outcomes were relapse-free survival and overall survival. All sites of persistent, residual or recurrent tumours after salvage treatment were recorded as failure in terms of the relapse-free survival metric, and all deaths were recorded against the overall survival parameter. The 5-year survival rates of relapse-free and overall survival were calculated by using the Kaplan-Meier product limit method. Significance levels among the curves were determined using the log-rank test. Potential risk factors and pathological characteristics were further analyzed using a multivariate Cox regression model. All statistical analyses were performed using the SPSS software package, version 18.0 (SPSS Inc., Chicago, IL, USA). All statistical tests were two-tailed and corresponding p values < 0.05 were interpreted as statistically significant.