Higher TOX Genes Expression Is Associated With Poor Overall Survival for Patients With Acute Myeloid Leukemia

Thymocyte selection-associated HMG box (TOX) is a transcription factor that belongs to the high mobility group box (HMG-box) superfamily, which includes four subfamily members: TOX, TOX2, TOX3, and TOX4. TOX is related to the formation of multiple malignancies and contributes to CD8+ T cell exhaustion in solid tumors. However, little is known about the role of TOX genes in hematological malignancies. In this study, we explored the prognostic value of TOX genes from 40 patients with de novo acute myeloid leukemia (AML) by quantitative real-time PCR (qRT-PCR) in a training cohort and validated the results using transcriptome data from 167 de novo AML patients from the Cancer Genome Atlas (TCGA) database. In the training cohort, higher expression of TOX and TOX4 was detected in the AML samples, whereas lower TOX3 expression was found. Moreover, both the training and validation results indicated that higher TOX2, TOX3, and TOX4 expression of AML patients (3-year OS: 0% vs. 37%, P = 0.036; 3-year OS: 4% vs. 61%, P < 0.001; 3-year OS: 0% vs. 32%, P = 0.010) and the AML patients with highly co-expressed TOX, TOX2, TOX4 genes (3-year OS: 0% vs. 25% vs. 75%, P = 0.001) were associated with poor overall survival (OS). Interestingly, TOX2 was positively correlated with CTLA-4, PD-1, TIGIT, and PDL-2 (rs = 0.43, P = 0.006; rs = 0.43, P = 0.006; rs = 0.56, P < 0.001; rs = 0.54, P < 0.001). In conclusion, higher expression of TOX genes was associated with poor OS for AML patients, which was related to the up-regulation of immune checkpoint genes. These data might provide novel predictors for AML outcome and direction for further investigation of the possibility of using TOX genes in novel targeted therapies for AML.


INTRODUCTION
In recent years with the improvement of chemotherapy regimens and the development of hematopoietic stem cell transplantation technology, acute myeloid leukemia (AML) patient treatment has achieved certain curative effects. However, there is still a high risk of relapse and a low disease-free survival rate (1,2). The immune escape of tumor cells is a crucial cause of relapse and refractory AML (3). It has been shown that in the tumor microenvironment, tumor cells induce the expression of immune checkpoint (IC) genes, such as programmed cell death protein 1 (PD-1), cytotoxic T lymphocyte-associated molecule-4 (CTLA-4), and lymphocyte-activation gene 3 (LAG-3), leading to T cell exhaustion and immune escape (4)(5)(6)(7)(8)(9). Clinical trials of targeted inhibitory antibodies, such as anti-PD-1 and anti-CTLA-4, in solid tumors have demonstrated their significant effects (10). In contrast, the clinical effectiveness of such immune therapies appears to be relatively different for different AML cases and clinical trials with different outcomes (11)(12)(13). Therefore, it is worth exploring the immune biomarkers that may be related to the effects of immune checkpoint blockade and revision of T cell exhaustion as well as their association with clinical outcome in AML (14).
Thymocyte selection-associated HMG box (TOX), a transcription factor that can bind to DNA, belongs to the high mobility group box (HMG-box) superfamily. TOX includes four subfamily members (TOX1-4, TOX1 is also known as TOX) (15). TOX is a crucial transcription factor related to the development of malignancies and contributing to CD8+ T cell exhaustion in patients with solid tumors (16)(17)(18). For example, TOX is positively correlated with larger tumor size, lower differentiation, later tumor node metastasis (TNM) stage, and facilitating endocytic recycling of PD-1 (17). In tumor-infiltrating CD8+ T cells from human melanoma and non-small cell lung cancer (NSCLC), increased expression of TOX in CD8+ T cells is associated with high expression of PD-1 (19). In contrast, there are few studies on TOX genes in hematological malignancies. TOX is highly expressed in acute lymphoblastic leukemia (ALL), particularly in T cell -ALL (T-ALL). High expression of TOX inhibits the function of the repair factors KU70/KU80 causing abnormal non-homologous end joining (NHEJ) repair (17). Although TOX is positively expressed in almost all ALL cases, TOX deletion has also been detected in ALL patients (20). Therefore, the mechanism by which TOX plays a role in ALL remains to be investigated.
In our previous study, we found higher TOX expression concurrent with PD-1, Tim-3, or CD244 in T cells from patients with B cell non-Hodgkin's lymphoma (B-NHL), which suggested that TOX may be involved in inducing CD8+ T cell exhaustion by co-regulation with immune checkpoint proteins (21).
In this study, we investigated the expression characteristics and prognostic value of the TOX genes and analyzed the correlation between TOX and IC genes in peripheral blood (PB) samples from AML patients in our clinical center. The results were further validated with high-throughput sequencing data from The Cancer Genome Atlas (TCGA) database in a more significant number of patients.

PB Sample Information
In this study, we collected PB mononuclear cells from 40 de novo AML patients with informed consent (15 males and 25 females) who ranged in age from 12 to 83 years and provided informed consent from March 2016 to March 2021. We also included 17 AML-complete response (CR) patients (11 males and 6 females) whose ages ranged from 12 to 62 years ( Figure 1). In addition, we collected PB white blood cells (WBCs) from 25 healthy individuals (HIs), including 12 males and 23 females, whose ages ranged from 19 to 70 years as a control population. Overall survival (OS) was defined as the time from diagnosis to death or last follow-up. The clinical information of the patients in the training cohort was listed in Table 1

TCGA Dataset
The gene expression data and the clinical information of 167 de novo AML patients were obtained from the TCGA (https:// cancergenome.nih.gov/) database by UCSC XENA (https:// xenabrowser.net/datapages/) (6). The gene expression data from the TCGA database comprised the validation cohort for OS analysis and were used to validate the results of the training cohort.

Quantitative Real-Time PCR
RNA isolation was performed using peripheral blood mononuclear cells (PBMCs) samples. Reverse transcription of RNA into cDNA was performed according to the manufacturer's instructions for the Reverse Transcription Kit (ABI, USA). The gene expression levels were quantified according to the manufacturer's instructions in the qRT-PCR kit (TIANGEN, China) (6), and b2M was used as an internal control. The sequences of the primers used for qRT-PCR are listed in Supplementary Table 1. The gene expression results are presented as the fold change lg (2^-DDCT*100).

Optimal Prognostic Cutoff Values
The Optimal prognostic cutoff values for TOX, TOX2, TOX3, and TOX4 were determined using the maximally selected rank statistics from the 'maxstat' R package, which was provided to the 'survminer' R package (22). This is an outcome-oriented method providing a value of a cut-point that corresponds to the most significant relationship with survival. According to the optimal cut-points of TOX genes, AML patients were divided into low-and high-expression groups to plot and compare Kaplan-Meier curves.  Figure S1) for continuous variables, which were obtained using the "Survminer" package, and the log-rank test was used for comparison. A Correlation heatmap was generated using the "ggcorrplot" package, and it was analyzed using Spearman's coefficient. Univariate and multivariate COX regression analyses were used to identify independent prognostic factors. The Mann-Whitney test was used for comparison between two groups, and the Kruskal-Wallis test was used to compare multiple gene expression groups. A two-tailed p-value < 0.05 was considered statistically significant.

Higher Expression of TOX Genes Is Associated With Poor OS in AML Patients
To investigate the role of altered TOX expression in the clinical outcome of AML patients, we collected the clinical information of FIGURE 1 | Workflow of study. A total of 40 AML patients from our clinical center were designated as the training cohort. Peripheral blood was collected from these patients to obtain PBMCs, which were used to translate RNA into cDNA. qRT-PCR was used to detect the expression levels of the TOX genes and ICs. After patient follow-up, the data were analyzed by expression characteristics, overall survival analysis, and correlation analysis. The gene expression data and clinical information of 167 de novo AML patients obtained from the TCGA were designated as a validation cohort. PBMCs, peripheral blood mononuclear cells; ICs, immune checkpoint genes; qRT-PCR, Quantitative Real-Time PCR; TCGA, The Cancer Genome Atlas.
Considering an additive effect on the outcome if multiple TOX genes are aberrantly elevated, we characterize the predictive value of co-expression of TOX genes in AML. Using the coexpression of TOX genes to evaluate the OS, we found that lower OS was observed in TOX2 high TOX4 high AML patients in comparison with TOX2 high TOX4 low or TOX2 low TOX4 high AML patients and TOX2 low TOX4 low AML patients (3-year OS: 0% vs. 25% vs. 56%, P = 0.002, Figure 4A). In addition, TOX high TOX2 high TOX4 high AML patients are also related to the poor prognosis of patients (3-year OS: 0% vs. 25% vs. 75%, P = 0.001, Figure 4B). The same results were also confirmed in the validation cohort (3-year OS: 21% vs. 38% vs. 100%, P < 0.001; 3year OS: 22% vs. 36% vs. 100%, P = 0.008, Figures 4C, D).

Correlation of TOX and IC Genes Expression in AML
Based on the previous finding of TOX expression concurrent with that of PD-1 and Tim-3 in T cells from patients with lymphoma (21), we analyzed the correlation of the gene expression level of the TOX genes and IC genes in AML patients ( Figures 5A-C). Significantly, TOX2 has a positive correlation with TIGIT, PD-1, CTLA-4, and PDL2 (r s = 0.43, P = 0.006; r s = 0.43, P = 0.006; r s = 0.56, P < 0.001; r s = 0.54, P < 0.001). Moreover, the expression levels of TOX and TOX4 had a positive correlation (r s = 0.41, P = 0.008), while the expression level of TOX and TOX2 demonstrated a trend toward a negative correlation (r s = -0.133, P = 0.412). Interestingly, there was a significantly negative correlation between TOX and TOX2 expression in the TCGA dataset (r s = -0.23, P = 0.003). These results were confirmed in the validation cohort (r s = 0. 34, P < 0.001; r s = 0.29, P < 0.001; r s = 0. 44, P < 0.001; r s = 0. 26, P < 0.001, Figure 5B).

DISCUSSION
The biomarkers for AML outcome, particularly those related to immune suppression, which often occurs in cancer immune escape, are far from clear. Recent studies have shown that TOX is a crucial transcription factor that contributes to T cell exhaustion and is involved in tumor development (16,23). However, how TOX is altered in AML remains unclear. In this  study, we explored the expression characteristics of the four TOX family members in AML samples. Interestingly, the expression patterns of the TOX genes were different with the expression of TOX and TOX4 significantly increased, TOX3 was decreased, and TOX2 demonstrated an increasing expression trend. These differences may be due to the functional differences of TOXs in the AML subtypes and may be involved in clinical outcome. It is well known that AML includes eight subtypes and is a heterogeneous disease (24). For instance, AML-M3 is a particular subtype with a favorable outcome. Indeed, the expression of TOX genes in patients of different AML subtypes i.e., AML-M2, M3, and M5, was different. Overall, TOX and TOX4 had low expression in the M3 group and high expression in the M2 and M5 groups, while TOX3 was low in AML-M2 and high in AML-M5. Therefore, the study of the TOX family as a biomarker may provide particular predictive value for the study of patients with different types. It has been reported that both TOX and TOX2 are correlated with CD8+ T cell exhaustion (25). In this study, we also found that TOX2 is positively correlated with CTLA-4, PD-1, TIGIT, and PDL-2 in AML samples from either our center or from the TCGA dataset. Previous studies have indicated that higher PD-1, PD-L1, or CTLA-4 expression is associated with poor OS in AML (6). Thus, TOX2 may be one more biomarker for predicting the T cell immune suppression related to the clinical outcome of AML. However, we did not find any association between the expression of TOX and the immune checkpoint genes. The reason for this discrepancy may be the level of TOX expression in different cells.
It is possible that the association between TOX and PD-1 or CTLA-4 co-expression only occurs in T cells (21) and not PBMCs, which include a high percentage of AML cells. We found that TOX genes are also expressed in AML cell lines and primary AML cells (data not shown), which suggests that there are different patterns of expression for TOX genes in T cells and AML cells, which may play a different role. TOX3 is an essential protective transactivator in neurons (26) and plays different roles in various tumors (27)(28)(29). Moreover, TOX4 regulates the cell cycle and fate (30,31), but there is no information regarding the expression characteristics of TOX4 in cancer or hematological malignancies. In this study, we characterized the expression patterns of TOX3 and TOX4 in AML. Interestingly, the expression of TOX3 was significantly decreased and different from the other TOX genes, while TOX4 was highly expressed. Whether these genes play different regulatory roles in AML requires further investigation. We also considered whether there is any correlation or complementation between the expression of the TOX genes. From our results, we could find a positive correlation between TOX and TOX4 in the training and validation groups, while a negative correlation or correlation trend between TOX and TOX2 expression was found. Whether this is complementary regulation remains an open question.
To further discuss the role of the altered expression of the TOX genes in AML, we explored the association between the expression of the TOX genes and the OS of AML patients. Our results demonstrated that AML patients with high TOX2 expression have an inferior OS. Combined with the finding that TOX2 positively correlates with IC genes and is highly expressed in AML-CR patients, overexpression of TOX2 together with T cell exhaustion may be a reason why AML patients have poor OS. It is worth further investigating the value in predicting OS for AML by evaluating the co-expression of TOX with immune checkpoint genes (32). Although the expression of TOX3 is decreased in AML, patients with higher TOX3 expression have an inferior 3-year OS, the function of TOX3 is needed to further characterize. Moreover, we found that patients with increased TOX3 expression are primarily AML-M5 patients. These findings may provide a precise and valuable predictor of OS for AML. Similarly, higher expression of TOX4 is also related to poor prognosis. An interesting finding is that when the TOX, TOX2, TOX4 genes co-expressed highly in AML patients, the prognosis of these patients is significantly poor. This finding indicates that TOX genes play a negative role in AML patients to a large extent. Overall, TOX genes may be potential biomarkers for predicting clinical outcomes in AML, and their blockade may be considered a new direction for the treatment of AML patients.
In summary, in this study, we characterized the altered expression of TOX genes in AML and defined their different roles. We also demonstrated that TOX2 is positively correlated with the CTLA-4, PD-1, TIGIT, and PDL-2 genes. Moreover, higher expression of TOX2, TOX3, and TOX4 of AML patients and the AML patients with highly co-expressed TOX, TOX2, TOX4 genes were associated with poor OS for AML patients, which may be related to the upregulation of immune checkpoint genes. These data indicate that TOX genes may be novel predictors for clinical outcomes in AML. Moreover, TOX, as the upstream molecule of immune checkpoint proteins, is not only expressed in AML cells but also associated with T cell exhaustion, which might provide direction for future investigations of the possibility of the dual effect of TOX targeted inhibition, inhibiting proliferation of AML cells and revising T cell exhaustion and restoring anti-AML T cell function.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Ethics Committee of the School of Medicine of Jinan University. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

AUTHOR CONTRIBUTIONS
SC and YL contributed to the concept development and study design and edited the manuscript. CL performed the experiments, interpreted the data, and wrote the manuscript. YZ and CC helped write the manuscript. SH, TD, and XBZ supported reviewing the references and preparing figures. JT and XFZ collected the clinical information and provided clinical peripheral blood samples. All authors contributed to the article and approved the submitted version.