Dynamic changes of exhaustion features in T cells during oral carcinogenesis

Abstract Objectives This study aimed to clarify the dynamic changes of exhaustion features in T cells during oral carcinogenesis. Materials and Methods Mice were randomly divided into 4NQO group and control group. The exhaustion features of CD4+ and CD8+ T cells of both groups were detected by flow cytometry. Furthermore, multiplex immunohistochemistry was used to evaluate the expression of inhibitory receptors in human normal, dysplastic, and carcinogenesis tissues. Finally, anti‐PD‐1 antibody treatment was performed at the early premalignant phase of oral carcinogenesis. Results The proportion of naive T cells in 4NQO group was lower than those in control group, while the proportion of effector memory T cells was higher in 4NQO group. The expression of inhibitory receptors on CD4+ and CD8+ T cells increased gradually during carcinogenesis. In contrast, the secretion of cytokines by CD4+ and CD8+ T cells decreased gradually with the progression stage. Strikingly, those changes occurred before the onset of oral carcinogenesis. The expression of inhibitory receptors on T cells increased gradually as the human tissues progressed from normal, dysplasia to carcinoma. Interestingly, PD‐1 blockade at the early premalignant phase could reverse carcinogenesis progression by restoring T cell function. Conclusions T‐cell dysfunction was established at the early premalignant phase of oral carcinogenesis; PD‐1 blockade at the early premalignant phase can effectively reverse T‐cell exhaustion features and then prevent carcinogenesis progression.

is associated with reduced susceptibility to OSCC. 5,6 Nonetheless, despite the presence of antigen-specific T cells which are able to sense dysplastic epithelial cells in the oral premalignant lesions, the dysplastic cells cannot be eliminated. 7 It has been proposed that the premalignant phase of oral carcinogenesis may equivalent to the equilibrium phase of cancer immunoediting and can be shifted toward OSCC as the adaptive immunity was impaired. 8 However, whether and how the malignant transformation initiates and the dynamic changes in T-cell states during oral carcinogenesis remain largely unknown.
Accumulating evidence indicates that T cells acquire an "exhaustion" state characterized by sustained expression of inhibitory receptors, progressive loss of effector functions, and distinct transcriptional and epigenetic profiles during the progression of cancer. 9 Intriguingly, the dysfunctional tumor-specific T cells found in the tumor microenvironment are already "imprinted" at the premalignant stage preceded tumor establishment due to continuous antigen encounter. 10 It is reported that tumor-specific neoantigens may arise at the premalignant phase and can be presented to the T cells. 11 Similarly, in the context of oral carcinogenesis, oral premalignant lesions share several tumor antigens including EGFR, RAGE, and MUC1 with OSCC to stimulate T-cell-mediated immune response. 12 Besides, PD-L1, a coinhibitory receptor exerting its immunosuppressive effect via binding to PD-1, was expressed on dysplastic epithelial cells and infiltrating cells in oral premalignant lesions and showed elevated expression in oral premalignant lesions that further progressed to cancer. 13,14 Therefore, we hypothesized that T-cell dysfunction and immunosuppression occurred at the early premalignant phase of oral carcinogenesis. In addition, restoring T-cell function at the early premalignant phase could prevent carcinogenesis progression.
In this study, we used a classical 4-nitroquinoline-1-oxide (4NQO)-induced murine OSCC model to mimic human oral cavity carcinogenesis. The oral epithelial lesions in mice were evaluated by hematoxylin and eosin (H&E) staining, and the dynamic changes of exhaustion features in T cells during the development of murine OSCC were monitored by flow cytometry. Our study has revealed that systemic T-cell exhaustion emerges at the early stage of oral carcinogenesis and restoring T-cell function at this stage could prevent oral carcinogenesis progression. This discovery may help in better understanding of the underlying mechanisms of OSCC pathogenesis and developing therapeutic strategies through remodeling T-cell function at the early stage of oral carcinogenesis to prevent malignant transformation toward OSCC.

| Mice
Six-week-old female C57BL/6 mice were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. All mice were maintained in a specific pathogen-free environment. All animal study protocols were performed under the guidelines of the Institutional Animal Care and Use Committee of Sun Yat-Sen University.

| 4NQO-induced oral tumorigenesis
The 4NQO group was given sterile water containing 100 μg/ml 4-Nitroquinoline-1-oxide (4NQO, Sigma-Aldrich) for 16 weeks, while the control group was given sterile water to drink. 15,16 At week 16, the 4NQO water was changed into sterile water until the end of week 20.

| Antibody treatment
Anti-mouse PD-1 monoclonal antibody was a kind gift from Lieping Chen (Yale University School of Medicine). After 4NQO treatment for 12 weeks, the mice were randomly divided into anti-PD-1 group (αPD-1 group) and control IgG group (Ctrl IgG group). The mice of αPD-1 group were treated with 200 μg anti-PD-1 antibody intraperitoneally every week from 12 weeks to 16 weeks, while equivalent amounts of control IgG were administered intraperitoneally to the mice of Ctrl IgG group. All mice were sacrificed at week 20.

| Histology and pathological analysis
Oral lesions were identified and photographed every 4 weeks. Next, oral lesions were harvested and fixed in 10% formalin for 24 h after euthanizing mice, then sectioned into 4μm slices. The slices were then stained with hematoxylin and eosin (H&E). After that, these slices were classified into multiple histopathological grades (normal, hyperplasia, dysplasia, and carcinoma) by 2 certified pathologists.

| Flow cytometry
The Draining lymph node (Dln) and spleen were ground into a single cell suspension. The tongue tissue was cut into 1-mm 3 pieces and digested with DNase I (0.1 mg/ml) and collagenase IV (1 mg/ ml) for 1 h. After red blood cells in the single cell suspension were lysed, single cell suspension was blocked by anti-mouse FcR blocker for 30 min. The cells were further stained with antibodies in PBS with 2% FBS for 30 min at 4°C. Flow cytometry was performed according to a standardized protocol. 17 Before intracellular cytokine staining, cells were stimulated in vitro with cell stimulation cocktail (TNB-4975-UL100, Tonbo Biosciences, USA) for 5 h at 37°C with 5% CO 2 . Antibodies for flow cytometry are listed in Table S1. Flow cytometric analyses were performed using BD LSRFFortessa (BD Biosciences), and data were analyzed using FlowJo 10.5.3 software.
The fluorescence minus one (FMO) was performed as the flow cytometric controls for establishing gates. The DownSample plugin of FlowJo 10.5.3 software was used to extract 3000 events in CD3 + subgroup of each sample. 18 After concatenating the extracted data of each group, t-SNE plugin was used to obtain t-SNE graph.

| Statistical analysis
All statistical analyses were implemented with GraphPad Prism 7.0.
Fisher's exact test was used to compare the proportion of oral lesions of Ctrl IgG group and αPD-1 group. Differences in flow cytometric results were evaluated with Student's t test. The results of mIHC were evaluated with one-way analysis of variance (ANOVA).
All analyses were two-tailed, and p < 0.05 was considered significant.

| 4NQO induced the development of OSCC in mice
To investigate T-cell exhaustion during the process of OSCC, we induced precancerous and cancerous lesions in mice with 4NQO ( Figure 1A). Six mice in the 4NQO group and the control group were sacrificed every 4 weeks during the induction process. Then, the oral mucosal lesions of mice were detected. With the increase in time of 4NQO induction, hyperplasia, dysplasia, and carcinoma in the oral mucosa of mice were observed at week 8, week 12, and week 20, respectively ( Figure 1B-D). Therefore, in this study, the oral mucosa of mice at week 4, week8, and week20 was in normal, hyperplasia, and carcinoma stage, respectively, while the oral mucosa of mice at week 12 and week16 was in dysplasia stage.

| The differentiation states of T cells changed during the development of OSCC
The differentiation states of CD4 + and CD8 + T cells in the spleen, draining lymph node (Dln), and tongue were detected with flow cy-   From the above results, it can be found that the up-regulation of inhibitory receptors on CD4 + T cells appears earlier and more obvious than that on CD8 + T cells, indicating that CD4 + T cells are more vulnerable to develop an exhaustion phenotype during the development of OSCC than CD8 + T cells.

| The cytokine secretion of T cells decreased gradually during the development of OSCC
The exhaustion of T cells is also characterized by the progressive reduction in cytokine secretion. Our results showed that the secretion of IL-2 by CD4 + T cells in Dln and tongue of 4NQO group was significantly lower than that of control group from week 8 ( Figure 4A), but there was no significant difference in spleen ( Figure S4A). The production of TNFα by CD4 + T cells in Dln and tongue of 4NQO group was significantly lower than that of control group at week 16 and week 4, respectively ( Figure 4B), while there was no significant difference about the secretion of TNFα on CD4 + T cells in spleen ( Figure S4B).

In CD8 + T cells, the secretion of IL2 in Dln and tongue of 4NQO
group was significantly lower than that of control group at week 12 and week 20, respectively ( Figure 4A), while there was no significant difference in spleen ( Figure S4A). The secretion of TNFα in Dln, tongue, and spleen of 4NQO group was significantly lower than that of control group from week 16, week 4, and week 12, respectively ( Figure 4B and Figure S4B). The secretion of IFNγ in Dln, tongue, and spleen of 4NQO group was significantly lower than that of control

| The proportion of CD4 + T-cell subsets changed and MDSCs were enriched during oral carcinogenesis
It is known that CD4 + T cells comprise various subtypes, including Th1, Th2, Th17, and Treg, etc. Among them, Th1, Th17, and Treg are the most important cell subsets in carcinogenesis. Therefore, we detected the proportion of Th1, Treg, and Th17 in spleen, Dln, and tongue with flow cytometry. The results showed that the proportion of Treg in spleen, Dln, and tongue of 4NQO group was higher than that of control group at the early stage of induction, and the difference was more obvious with the extension of induction ( Figure 5A and Figure S5A). The proportion of Th17 in spleen, Dln, and tongue of 4NQO group was lower than that of control group ( Figure 5B and Figure S5B). The proportion of Th1 in spleen and Dln of 4NQO group was higher than that of control group, but no significant difference was observed between two groups in tongue ( Figure 5C and Figure   S5C). The ratio of Th17 to Treg of 4NQO group decreased from week 4 in spleen, Dln and tongue ( Figure 5D and Figure S5D). showed that the proportion of MDSCs of 4NQO group (especially in spleen) was higher than that of control group ( Figure S6).

| T-cell exhaustion was also associated with the development of human OSCC
To further validate the changes in exhausted features of T cells during the development of OSCC, we performed multiple

| PD-1 blockade at the early premalignant phase can effectively prevent oral malignant lesions progression by restoring T-cell function
Since T-cell dysfunction was established at the early premalignant phase of oral carcinogenesis, we continued to explore whether Flow cytometric analysis showed that the secretion of IL2 and TNFα by CD4 + or CD8 + T cells in tongue of αPD-1 group was significantly higher than that of Ctrl IgG group ( Figure 7D,E). However, there was no significant difference about the secretion of IL2 and TNFα by CD4 + or CD8 + T cells in Dln or spleen ( Figure 7D,E and Figure S7A). Furthermore, the secretion of IFNγ by CD8 + T cells in Dln and tongue of αPD-1 group was significantly higher than that of Ctrl IgG group, while there was no difference in spleen ( Figure 7F and Figure S7A). The expression of CTLA-4 on CD4 + or CD8 + T cells in tongue up-regulated after αPD-1 treatment, while the expression of other inhibitory receptors on CD4 + or CD8 + T cells did not significantly change after αPD-1 treatment ( Figure 7G and Figure S7B).
Then we found that the proportion of Treg in tongue increased significantly after αPD-1 treatment. The proportion of Th17 and Th1 in Dln and tongue also increased significantly after αPD-1 treatment.
However, T-cell subsets except Th17 did not change significantly in the spleen αPD-1 treatment ( Figure 7H and Figure S7C). No significant changes were found in differentiation states of CD4 + or CD8 + T cells in Dln, tongue, and spleen between αPD-1 group and Ctrl IgG group ( Figure S7D). T-cell exhaustion is a states of T cell dysfunction. We demonstrated that CD4 + and CD8 + T cells acquire their exhausted features at the early stage of oral carcinogenesis prior to tumor establishment, which could in turn accelerate the development of oral carcinogenesis. Furthermore, PD-1 blockade at the early stage of oral carcinogenesis could effectively prevent oral malignant progression.

| DISCUSS ION
We and other researchers previously reported that although PD-1 blockade can effectively prevent the occurrence of OSCC, the phenomenon of resistance against anti-PD-1 therapy still exists and therefore some mice still progress to OSCC. 17,20,[24][25][26][27] In this study, most of the oral lesions of αPD-1 group stayed at hyperplasia stage, and no oral lesions in αPD-1 group progressed to OSCC. The reason for this difference might be that the dysfunctional T cells are still in "pre-exhausted" states. These "pre-exhausted" T cells were more sensitive to PD-1 blockade therapy than terminal exhausted T cells. 28 These results provided a reference for checkpoint inhibitors blockade in treatment of oral premalignant lesions. An ongoing phase II trial (NCT04504552) aimed to treat high-risk oral premalignant lesions with checkpoint inhibitors.
Moreover, previous studies have revealed that the anti-tumor function of CD4 + T cells is not weaker than that of CD8 + T cells. 29 Here, we found that even in the premalignant lesions, CD4 + T cells were more likely to be activated by antigen stimulation than CD8 + T cells demonstrated by an increased proportion of effector T cells and Th1 cells (Figures 2 and 5). However, CD4 + T cells also showed up-regulated inhibitory receptors earlier than CD8 + T cells. These results indicated that reversal of CD4 + T cells exhaustion at the premalignant stage was also expected to be an effective strategy for OSCC prevention.
Tregs infiltration within premalignant microenvironment was significantly associated with the increased risk of cancerization to OSCC. 30 The Th17/Treg ratio was reported to be a potential diagnostic indicator for malignant transformation to OSCC. 31 Besides, promoting Th17 phenotype and inhibiting its dynamic shift to Treg phenotype in premalignant oral lesions might slow disease progression. 32,33 However, 4NQO group showed significantly lower Th17/ Treg ratio compared with control group. MDSCs were also found to be enriched in 4NQO group. It was reported that MDSCs were capable of inducing Treg, 34 which might result in a positive feedback loop that augmented immunosuppression. These findings were consistent with a recent study. 35 In addition, we found that the expression of CTLA-4 on T cells and the proportion of Treg increased significantly after anti-PD-1 therapy, which may offset the effect of therapy. These results suggested that PD-1 blockade combined with CTLA-4 blockade and Treg clearance may display a better therapeutic effect.

| CON CLUS IONS
Our study characterized the dynamic changes in T cell exhausted features during OSCC development, which provided direct evidence that T cell exhaustion occurred at the early stage of oral carcinogenesis. Interestingly, anti-PD-1 antibody treatment at the early premalignant phase could more effectively prevent the development of oral carcinogenesis than treatment after tumor formation.

ACK N OWLED G M ENT
We thank Lieping Chen for the kind gift of anti-mouse PD-1 monoclonal antibody.

CO N FLI C T O F I NTE R E S T
The authors have declared that no conflict of interest.

AUTH O R CO NTR I B UTI O N S
Zhi Wang conceptualized the study. Wenqiang Xie, Jie Shen, Dikan

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors declare that all data supporting the findings of this study are available on reasonable request from the corresponding author.