Lymphocyte activating gene 3 protein expression in nasopharyngeal carcinoma is correlated with programmed cell death-1 and programmed cell death ligand-1, tumor-infiltrating lymphocytes

Background Immunotherapy has shown promising efficacy in patients with nasopharyngeal carcinoma (NPC). Lymphocyte activating 3 gene (LAG-3) represents a significant immune target, however, its relationship with NPC remains unclear. This study aimed to evaluate LAG-3 expression in NPC and its association with CD3+ tumor-infiltrating lymphocytes (TILs), Granzyme B (GZMB), programmed death ligand 1 (PD-L1), and programmed death 1 (PD-1) expression. Methods A total of 182 patients with NPC from Sun Yat-sen University Cancer Center, China, were included in this retrospective study. LAG-3 expression in 15 NPC cell lines and LAG-3, CD3+ TILs, GZMB, PD-L1 and PD-1 in clinical samples were estimated using immunohistochemistry. The Chi-square test was used to estimate the association between LAG-3, other biomarkers, and clinical characteristics. Survival analysis was performed using the Kaplan–Meier method and the Cox regression model. Results LAG-3 was negatively expressed in all of the 15 NPC cell lines, whereas, 147 patients with NPC (80.8%) exhibited high LAG-3 expression on TILs from tumor tissues. Male patients and those who were EBV-positive presented higher LAG-3 expression. Correlation analyses showed that LAG-3 expression was related to PD-1 expression on TILs, as well as, PD-L1 expression on tumor cells (TCs) and TILs. Both the univariate and multivariate Cox models indicated that pathological type III (P = 0.036), higher LAG-3 on TILs (P < 0.001), higher PD-L1 on TCs (P = 0.027), and higher PD-1 on TILs (P < 0.001) were associated with poorer disease-free survival (DFS). However, lower PD-L1 expression on TILs was related to superior DFS only in the univariate Cox analyses (P = 0.002). Conclusion Higher LAG-3 and PD-1 on TILs, and higher PD-L1 expression on TCs, and pathological type III were identified as independent risk factors for poorer DFS in NPC patients. Our data demonstrate that LAG-3 is a promising inhibitory receptor that may play an important role in anti-NPC therapy. Supplementary Information The online version contains supplementary material available at 10.1186/s12935-021-02162-w.


Background
Nasopharyngeal carcinoma (NPC) is a common malignancy of the upper or side wall of the nasopharyngeal chamber. Somewhat unexplainably, NPC has distinct disparities in its geographical distribution, with a particularly high occurrence in Guangdong province, China [1]. NPC is closely related to the Epstein-Barr virus (EBV), with the EBV proven as its main etiologic cause [2]. With recent developments in radiotherapy and combined chemoradiotherapy, the survival of NPC patients has been substantially prolonged [3,4]. Despite advances in treatment, local relapse and distant metastasis continue to represent major causes of cancer progression following anti-NPC therapy [5]. Due to limited advances made regarding chemotherapy regimens, it is critical to explore novel approaches for the treatment of metastatic NPC that alleviate toxicity and promote survival benefits.
Immune checkpoints expressed on tumor cells and immune cells play a crucial role in inhibiting or enhancing anti-tumor immunity. Blocking immune checkpoints has become a promising anti-neoplastic strategy [6,7]. Recently, immunotherapeutic strategies targeting programmed cell death-1 (PD-1), and programmed cell death ligand-1 (PD-L1) checkpoints have been associated with a remarkable anti-tumor response among various solid tumors [8]. An increasing number of clinical trials involving immunotherapy have shown promising outcomes for NPC patients [6,9]. PD-L1 has been shown to adjust type 1 T helper (Th1) autoimmune reactions, and is expressed on both tumor cells (TCs) and tumorinfiltrating lymphocytes (TILs) [10]. PD-L1 promotes tumor cell apoptosis by activating PD-1 expressed on T lymphocytes [11]. PD-1 is a suppressive receptor located on activated T lymphocytes that regulates immunological suppression and immune escape [12]. Inhibiting PD-1 or PD-L1 signaling is a potential therapeutic strategy to strengthen the immune response towards tumor cells. Numerous promising predictive biomarkers for immunotherapy have been suggested for various cancer types, including PD-L1 [13,14], TILs [15], level of microsatellite instability (MSI) [16], and tumor mutational burden (TMB) [17].
Recently, anti-PD-1 and PD-L1 treatments have been associated with potential clinical effects in some NPC patients [18][19][20][21][22]. However, other related studies have found that immuno-monotherapy was ineffective in NPC patients. For instance, a phase I study showed that the overall response of patients with advanced NPC treated with camrelizumab therapy was 34% (95% CI  and the median progression-free survival (PFS) was only 5.6 months (95% CI 3.3-7.9) [23]. Another study reported that the median PFS of pembrolizumab monotherapy was only 3.7 months, (95% CI 2.1-13.4) and only 2.8 months (95% CI 1. 8-7.4) following treatment with nivolumab monotherapy as palliative treatment for NPC [6,24]. It has also been reported that only 25% of NPC patients will benefit from anti-PD-1/PD-L1 immunotherapy, highlighting the critical need for further research into novel therapeutic regimens [25].
Lymphocyte activating gene 3 (LAG-3, also termed CD223), is a 51-KD transmembrane protein and a member of the immune globulin superfamily [26]. It represents another potential therapeutic target. It is mainly expressed on natural killer cells [26], B cells [27], TILs [28], and dendritic cells [29,30]. LAG-3 was first identified in the 1990s, its structure is similar to CD4 as both have four extracellular domains [26]. The LAG-3 gene is located near the CD4 gene on chromosome 12, and about 20% of the amino acid sequence of LAG-3 and CD4 are identical. Hence, LAG-3 acts as a ligand to bind to the major histocompatibility complex (MHC) class II, and possesses an even higher affinity than CD4 [31][32][33]. Pre-clinical studies indicate that LAG-3 inhibition activates the effector capabilities of T cells and synergizes with other immune checkpoint inhibitors (i.e., anti-PD-1/PD-L1) [34][35][36][37][38], which provides a strong rationale for simultaneously targeting LAG-3 and PD-1/PD-L1 to enhance anti-tumor T cell immunity. The upregulation of LAG-3 in TILs and in MHC II+ tumors that are resistant to anti-PD-1 antibodies supports this idea [39]. An increasing number of basic and clinical studies have begun to adopt a LAG-3 blockade strategy. As of March 2021, 14 anti-cancer drugs target LAG-3 (data source: https:// www. clini caltr ials. gov). Moreover, a number of pharmaceutical companies in China have distributed LAG-3 fusion proteins, antibodies, and bispecific antibodies targeting LAG-3 and other immune checkpoints. These companies have also applied for clinical applications to enhance the efficacy of immunotherapy (data source: https:// www. cde. org. cn/). A previous study indicated that LAG-3 expression is closely related to a worse survival of patients with non-small cell lung cancer (NSCLC) [25]. Conversely, several studies indicate that LAG-3 expression is related to a better survival for gastric cancer [40] and breast cancer patients [41]. However, it remains unclear whether LAG-3 has a significant influence on the prognosis of NPC patients.
It has been reported that LAG-3 can induce T cell dysfunction in the tumor microenvironment (TME) [42][43][44]. Previous studies have shown that the TME interaction with LAG-3 on TILs can modulate an anti-cancer immunoreaction [45]. The extent of TILs infiltration in the TME is related to the treatment effects of PD-1/PD-L1 inhibition [18,46]. A recent study indicated that lower CD3+ TIL infiltration was related to a poorer DFS for NPC patients [47]. It is well established that the immune system can eradicate infected or transformed cells, which is largely mediated through the activities of natural killer (NK) and cytotoxic T lymphocyte (CTL) cells [48]. The main mechanism of cellular apoptosis induced by NK cells, and CTLs is through the release of the granzyme B protein (GZMB) [49]. GZMB is localized inside endosomes as a zymogen and is subsequently activated by cathepsins to produce the fully active form of GZMB [50]. A meta-analysis found that GZMB+ lymphocytes were significantly associated with a better overall survival (OS) for patients with hepatocellular carcinoma (HCC) [51] and colorectal cancer [52]. However, the clinical relevance of PD-1, TILs, GZMB, PD-L1, and LAG-3 in patients with NPC remains unclear.
Here, we explore the association between LAG-3 expression and clinical characteristics at the cellular level and in tumor samples from NPC patients. We also evaluate the relationship between PD-1, GZMB, CD3+ TILs, PD-L1 expression and the prognosis of NPC patients.

Patients
This retrospective study included 182 patients who were pathologically diagnosed with NPC between January 1, 2006 and December 30, 2018 at Sun Yat-sen University Cancer Center (SYSUCC), China. Malignancy stages were determined according to the tumor node metastasis (TNM) staging method (eighth version) of the American Joint Committee on Cancer (AJCC). Qualified patients were 18-80 years of age with pathologically diagnosed NPC, no second primary cancer, and no distant metastases. We collected patient information regarding age, gender, smoking history, EBV status, family history, TNM stage, pathologic types, and treatment.

Cell lines
A tissue microarray (TMA) (2 mm) involving 15 NPC cell lines was generated in the laboratory, and all specimens were evaluated in triplicate. Cells were collected, settled for one night, compounded with 0.9% Sepharose, and allowed to curdle at room temperature for at least 5 min. Each solidified agar pellet was lightly positioned in a cassette and placed in 70% alcohol. The solidified agar pellets were disposed of and implanted in paraffin blocks. The cores were then taken from each encased block to produce the TMA, from which 4-mm sections were sliced.

Immunohistochemical analysis for LAG-3, CD3, GZMB, PD-L1, and PD-1 expression
Pathologically identified, formalin fixed, paraffin-embedded NPC specimens from patients who were biopsied at SYSUCC were retrospectively tested. Archived hematoxylin-eosin staining sections were assessed by two independent pathologists. Immunohistochemical (IHC) staining for LAG-3, CD3, GZMB, PD-L1, and PD-1 expression was conducted using sections obtained from the formalin-fixed diagnostic specimens. Briefly, 4-µm sections were deparaffinized in xylene, rehydrated, and then treated with a citrate antigen restore buffer (pH 9.0) to expose the antigen in the sections. After processing following the conventional steps, the slides were incu- . After washing them three times with a phosphate-buffered saline (PBS), 5 min per wash, the sections were sequentially incubated with a Horseradish peroxidase (HRP)-conjugated goat anti-human secondary antibody (PV6000, ZSGB-BIO, Beijing, China). An evaluation was performed using 3, 3′-Diaminobenzidine (DAB) substrate kits (ZLI-9017, ZSGB-BIO, Beijing, China). The sections were stained with hematoxylin for 4 min and counterstained with bluing reagent for 4 min. The slides were washed and then dehydrated in 70% to 100% alcohol baths followed by xylene baths before coverslip application.

Pathological evaluation of LAG-3, CD3+ TILs, GZMB, and PD-1/PD-L1 expression
The evaluation of LAG-3, CD3+ TILs, and GZMB, PD-1, and PD-L1 expression in the smears were performed by two pathologists who were blinded to the results of the whole tumor sections. Five randomly selected highpower sites (400×) in every sample were chosen to estimate the number of positive cells. The expression of LAG-3, PD-1, GZMB, and CD3+ TILs was assessed in the tumor stroma only, and PD-L1 staining was assessed in both the tumor and stromal cells [53].

Follow-up
Patients were followed up every 3 months for the first 3 years, and every 6 months over the next 2 years, and finally once each year thereafter. A semiannual follow-up was conducted until the end of the study or the death of the patient, whichever occurred first. The last follow-up time for all living patients was October 2020.

Statistical analysis
The statistical analyses were conducted using SPSS version 22 (IBM, Armonk, NY, USA) and GraphPad Prism 8 (GraphPad Software, La Jolla, CA, USA). Optimal cut-off values for the biomarkers were obtained using X-tile Software. Chi-square tests and Fisher's exact tests were used to estimate the relationship between LAG-3 and clinical characteristics, CD3, GZMB, PD-L1, and PD-1. The odds ratios (ORs) for LAG-3 expression were estimated for the variables: age, EBV status, family history, gender, smoking status, pathological pattern, TNM staging, CD3, GZMB, PD-L1, and PD-1. Survival was evaluated using the Kaplan-Meier approach. A Cox regression analysis was used to explore the correlation between the clinicopathological variables, the above biomarkers, and DFS. A P-value less than 0.15 in the univariate analysis was used to screen the values eligible for the multivariate analysis with the Cox proportional hazard model, together with 95% confidence intervals (CI). A P < 0.05 for all statistics was considered significant.

Patient clinical characteristics
In this study, 297 patients originally diagnosed with NPC at SYSUCC were screened for eligibility, 115 patients did not meet the inclusion criteria, of which 50 patients had no tumor staging or pathological type, four patients had other primary tumors, and 16 patients were lost to follow-up, and 45 patients had insufficient paraffin sections. Therefore, 182 NPC patients were evaluated using a series of screening steps (Fig. 2). The median followup time was 23.2 months in this patient population, and 50 (27.5%) patients were females and 132 (72.5%) were males. A total of 115 (63.2%) patients were EBV positive, 67 (36.8%) were EBV-negative. The patients' age ranged from 24 to 76 years old, with a median age of 49 years old. A total of 26 (14.3%) patients had a family history of NPC. Most patients were pathological type III (69.8%), only 14 (7.7%) and 41 (22.5%) patients were pathological type I and II, respectively. There were 31 (17%) smokers and 151 (83%) nonsmokers. The cancer stages were as follows: stage I, nine (4.9%); stage II, 35 (19.2%); stage III, 96 (52.7%); and stage IV, 42 (23.2%). Forty (22%) patients were treated with induction chemotherapy, 11 (6%) with radiotherapy, and 130 (71.4%) with radiochemotherapy ( Table 1).

Logistic regression model analysis to predict LAG-3 expression
The calculated ORs for LAG-3 expression were 2.535 (95% CI 1.157-5.551) and 0.513 (95% CI 0.225-1.170) when low PD-1 expression was compared with high PD-1 expression on TILs and low PD-L1 expression was compared with high PD-L1 expression on TILs in the logistic regression model univariate analysis. The calculated ORs for LAG-3 expression were 0.271 (95% CI 0.105-0.695) and 3.439 (95% CI 1.280-9.237) when low PD-L1 expression on TCs was compared with high expression, and low PD-L1 expression on TILs was compared with high expression in the logistic regression model multivariate analysis (Table 5).

The impact of LAG-3 and other checkpoints on DFS
We found that lower LAG-3, PD-1, and PD-L1 expression was associated with a more favorable survival prognosis. The Kaplan-Meier analysis estimated that higher LAG-3 expression on TILs ( (Fig. 5C), as well as, PD-L1  (Fig. 5G).

Cox regression analysis for DFS
All univariate and multivariate analyses of prognostic factors are summarized in  The optimal cut-off value for LAG-3 expression on TILs was 14, patients with LAG-3 expression on more than 14 cells were associated with poorer disease-free survival (DFS) than patients with LAG-3 expression on fewer than 14 cells (A-C). The optimal cut-off value for PD-L1 expression on TC was nine, patients with PD-L1 on TCs with a score lower than nine had better survival than those with PD-L1 on TCs with a score higher than nine (D, E). The optimal cut-off value for PD-1 expression on TILs was two cells, patients with PD-1 expression on fewer than two cells was related to superior DFS than those with PD-1 expression on more than two cells (G, H)

Discussion
LAG-3 represents a promising immune checkpoint inhibitor and has been investigated as a target for the treatment of solid tumors in many studies [57]. One study found that LAG-3 suppresses antitumor immunity in Hodgkin's lymphoma [58]. Anti-LAG-3 treatment was also found to restrict breast carcinoma growth in an animal model [35]. Moreover, simultaneously inhibiting LAG-3 and PD-1 signaling can strengthen the T lymphocyte response in ovarian carcinoma [59]. An increasing number of basic research and clinical studies have begun to adopt a LAG-3 blockade strategy in the treatment of solid tumors [60,61]. However, data regarding LAG-3 expression in NPC and its correlation with TILs, GZMB, PD-1, and PD-L1 remains unclear in NPC patients.  A limited number of studies have investigated LAG-3 expression in NPC and its relationship with TILs, PD-1, PD-L1, and GZMB. In our study, we first examined LAG-3 expression in NPC cell lines as well as clinical specimens and found that LAG-3 was negatively expressed on NPC cell lines regardless of EBV status, but that it was highly expressed on TILs in NPC cancer specimens. Male patients and those who were EBV-positive displayed higher LAG-3 expression. We also discovered that LAG-3 was closely related to PD-1 and PD-L1 expression. A survival analysis demonstrated that NPC patients with lower LAG-3, PD-1, and PD-L1 expression had a longer DFS. Importantly, higher LAG-3, PD-1, and PD-L1 expression on TCs, and pathological type III were confirmed to be independent prognostic factors for poorer DFS in NPC patients.
In regions where NPC is prevalent, it is primarily related to EBV infection status [62]. The World Health Organization (WHO) has identified histological categories of NPC, among which, keratinizing squamous cell malignancy is defined as type I, differentiated non-keratinizing malignancy is defined as type II, and undifferentiated non-keratinizing malignancy is defined as type III, which is closely linked with EBV infection status [63][64][65].
Our findings indicate that patients who are EBV positive exhibit higher LAG-3 expression. In addition, higher expression of LAG-3 and pathological type III were identified as independent prognostic risk factors for poorer DFS. The correlation between LAG-3 and EBV status may explain the poor prognosis of NPC patients with pathological type III.
Our findings demonstrate that LAG-3 is positively related to PD-L1 expression on TCs and PD-1 expression on TILs, which is in line with the results of previous studies. One study showed that LAG-3 was closely related to PD-1 and PD-L1 expression in NSCLC [25]. LAG-3 and PD-1 can regulate immune activation and synchronously increase immunity [66]. A study investigating ovarian cancer found that LAG-3 and PD-1 can down-regulate TILs [67]. Moreover, it was shown that LAG-3 and PD-1 can synchronously regulate the behavior and anticancer response of T lymphocytes [68]. An in vivo study found that targeting the PD-1 or LAG-3 signaling pathways could stimulate T lymphocytes, and that the combined inhibition of these pathways had a greater effect than the inhibition of each pathway alone [34]. Previous studies  have found that LAG-3 protein expression may act synergistically with PD-1 or PD-L1 monoclonal antibodies [68][69][70]. Anti-LAG-3 therapy has also been shown to effectively modulate regulatory T lymphocytes, [71] whereas other immune checkpoints (e.g., PD-1, PD-L1, and CTLA-4 (cytotoxic T lymphocyte associated antigen-4)) have not. A total of 14 anti-cancer drugs targeting LAG-3 have been developed as of March 2021 (Table 7) (data source: https:// www. clini caltr ials. gov). IMP321 was the first anti-cancer drug targeting LAG-3 to enter clinical trials. Clinical research has identified that a double blockade of immune checkpoint molecules results in enhanced clinical survival in various cancers, including renal cell carcinoma [72], melanoma [73], NSCLC [74], and small cell lung cancer [75]. It is important to note that in the reported camrelizumab antibody therapy NPC study, six out of eight patients who had formerly received ipilimumab (anti-CTLA-4) treatment exhibited a clinical response [23]. This finding indicates that a combination of immunotherapeutic strategies warrants further research. The findings of our study indicate that the inhibition of both LAG-3 and PD-1/PD-L1 can enhance the anti-cancer response as part of a synergy. Bispecific antibodies (BsAbs) (e.g., anti-LAG3, PD-1/PD-L1) have been exploited for extensive clinical use. MGD013 is a BsAb therapy that simultaneously targets both LAG-3 and PD-1 to suppress immune checkpoint inhibition, promote T cell activation, and improve anti-cancer immunity. Similarly, F-star exploited a BsAb termed FS118, which simultaneously targets LAG-3 and PD-L1. In addition, numerous pharmaceutical companies in China have created LAG-3 fusion proteins, antibodies, and bispecific antibodies, which are undergoing clinical applications (Table 8) (data source: https:// www. cde. org. cn/). Since LAG-3 is closely related to PD-1 and PD-L1, our study provides a novel insight and a theoretical foundation for the future development of LAG-3 and PD-1/PD-L1 bispecific antibodies to enhance the efficacy of immunotherapy for NPC.
We also observed that LAG-3 expression was associated with poor survival, which is in accordance with the results of other studies. One study indicated that high LAG-3 expression was related to worse survival in patients with NSCLC [25]. In chronic lymphocytic leukemia, LAG-3 also serves as a new predictive marker: Table 5 Univariate and multivariate analysis for prediction of LAG-3 expression in all patients   5 The survival curves of LAG-3, PD-1, and PD-L1 on TILs, PDL1 on TCs layered by the cut-off values estimated by X-tile. A Disease-free survival (DFS) differed significantly between patients with LAG-3 expression on fewer than 14 cells and those with LAG-3 expression on more than 14 cells (P < 0.001). B DFS differed significantly between patients with PD-1 expression on fewer than two cells and those with PD-1 expression on more than two cells (P < 0.001). C DFS differed significantly between patients with PD-L1 on TCs with a score lower than nine and those with PD-L1 on TCs with a score higher than nine (P = 0.027). D DFS differed significantly between patients with PD-L1 expression on fewer than 1% of TILs and those with PD-L1 expression on more than 1% of TILs (P = 0.002). E DFS differed significantly between patients with lower LAG-3 and PD-L1 expression on TILs and those with higher LAG-3 and PD-L1 expression on TILs (P < 0.001). F DFS differed significantly between patients with lower LAG-3 and PD-1 expression and those with higher LAG-3 and PD-1 expression (P < 0.001). G DFS differed significantly between patients with lower LAG-3 and PD-L1 expression on TCs and those with higher LAG-3 and PD-L1 expression on TCs (P = 0.037) higher expression of LAG-3 was associated with shorter survival [76]. Yet, in contrast, several studies have found that high LAG-3 was correlated with better survival in patients with gastric carcinoma [40] and breast carcinoma [41]. Thus, LAG-3 represents a potential immune checkpoint target. Despite the numerous ongoing anti-LAG-3 studies, the literature related to LAG-3 and NPC remains insufficient. Consequently, it is necessary to further explore the prognostic value of LAG-3 in NPC.
The findings of our present study indicate that LAG-3 may participate in the tumor immune escape of NPC as an interpretation of the observed poor survival in NPC patients. In addition, the association of LAG-3 and GZMB, and CD3+ TIL expression was also analyzed. However, no correlation between LAG-3 and GZMB or CD3+ TIL expression was found. A previous study indicated that lower CD3+ TIL infiltration was related to a worse DFS in patients with NPC [47] and HCC [77], which is inconsistent with our findings. Our finding that CD3 was unrelated to NPC prognosis may be attributed to our comparatively small case size. Further investigation with a larger sample size and an independent cohort of patients is required. There are some limitations of our study. First, this was a retrospective study, and we only collected information from one institute, and the overall survival data was insufficient for rigorous analysis. Secondly, since previous related reports have not identified the optimal cut-off value for LAG-3, X-tile was used to determine the cut-off values for predicting DFS. Finally, the sample size of this population was small. Further research involving a larger number of NPC patients is required.

Conclusions
Immune checkpoints play a critical role in immune regulation. Yet, the synergistic effects between multiple immune targets remains unknown. CTLA-4, PD-1, and PD-L1 antibody immunotherapy have demonstrated significant effectiveness for the treatment of some cancers. LAG-3 represents another potential therapeutic target whose synergistic effect requires further investigation. In our present study, we found that LAG-3 was closely associated with PD-1/PD-L1 expression. Positive LAG-3  expression or the expression of both LAG-3 and PD-L1 has been associated with early cancer relapse. Male and EBV-positive patients were associated with higher LAG-3 expression in our study. Lower LAG-3, PD-1, and PD-L1 expression were also associated with a longer DFS. Importantly, high LAG-3, PD-1, and PD-L1 expression on TCs, and Pathological type III were confirmed to be independent risk factors for poorer DFS in NPC patients. Based on the findings of our study and observations supporting its potential synergistic function when administered in conjunction with anti-PD-1/PD-L1, the inhibition LAG-3 is a promising inhibitory receptor and anti-LAG3 will likely play a critical role in anti-neoplastic therapy. Our study provides a theoretical foundation for the exploitation of BsAbs against LAG-3 or PD-1 on TILs, and PD-L1 on TCs. However, to date, there has been minimal research into the synergistic interactions between LAG-3 and other promising immune checkpoint molecules, such as T cell immunoglobulin-3. Future investigations are warranted.