Clinical implications and genetical insights of SOX6 expression in acute myeloid leukemia

Transcription factor SOX6 belongs to Sry‐related high‐mobility‐group box (SOX) family, has been reported to be downregulated and acts as a tumor-suppressor gene in various solid tumors, but in acute myeloid leukemia (AML) is incompletely understood. The SOX6 expression was analyzed between AML patients and normal controls from public data and our research cohort. Correlations between SOX6 expression and clinical, genetic features together with survival were further analyzed. In both public and our present datasets, we demonstrated that SOX6 expression is notably downregulated in AML patients compared with normal controls. Moreover, the expression level of SOX6 was dynamic, along with the disease status. SOX6 was significantly decreased in relapsed/refractory AML compared with complete remission AML. Clinically, SOX6 underexpression was significantly correlated with bone marrow blasts, and WBC counts. Furthermore, decreased expression of SOX6 was more common in core binding factor AML (CBF-AML), rarely found in complex karyotype AML (CK-AML), and correlated with FLT3 mutations. By survival analyses, low-expression of SOX6 was associated with shorter overall survival (OS) and event-free survival (EFS) among cytogenetic normal AML (CN-AML) patients. Moreover, both univariate and multivariate analyses showed that low SOX6 expression was an independent unfavorable prognostic biomarker for CN-AML. Our findings indicated that SOX6 underexpression, as a frequent event in AML, was associated with genetic abnormalities and prognosis in AML. SOX6 might be a valuable biomarker for risk stratification, predicting prognosis and relapse of AML.


Introduction
Acute myeloid leukemia (AML) is an aggressive hematopoietic neoplasm with high heterogeneous (Dohner et al. 2015;Short et al. 2018). Although AML treatment has significantly improved, with options including chemotherapy, hematopoietic stem cell transplantation (HSCT), cellular and target therapy, AML is generally still incurable. Molecular and cytogenetic biomarkers are the most important for 1 3 treatment strategy and risk stratification. For example, the treatment of PML-RARA positive AML with ATRA represents an important advancement in AML therapy (Huang et al. 1988). With the development of molecular genetic diagnostic technology, a number of molecular markers have already been used to optimize the risk stratification. For instance, NPM1 and biallelic CEBPA mutations are associated with favorable outcome, and mutations in TP53, FLT3-ITD, RUNX1, and ASXL1 may indicate poor outcomes in AML patients (Prada-Arismendy et al. 2017;Short et al. 2018). In addition, the aberrant expression of some genes (MN1, BAALC, ERG, etc.) could also influence the prognosis (Jentzsch et al. 2020;Langer et al. 2009;Schwind et al. 2010).
Sry-related high-mobility-group box (SOX) family contains a highly conserved DNA-binding high-mobility-group (HMG) domain and plays essential roles in embryogenesis, disease development and tumorigenes (Castillo and Sanchez-Cespedes, 2012;Dong et al. 2004;Grimm et al. 2020). Alteration and dysfunction of SOX family members may cause various diseases, especially cancer (Castillo and Sanchez-Cespedes, 2012;Dong et al. 2004;Grimm et al. 2020;Luo et al. 2022). SOX6 is a member of the SOXD subcategory. Substantial evidence reveals that SOX6 acts as a tumor suppressor and is downregulated in various human malignancies including osteosarcoma, ovarian carcinoma, esophageal squamous cell carcinoma, pancreatic cancer, clear cell renal cell carcinoma, lung adenocarcinoma, and hepatocellular carcinoma (Chen et al. 2020;Guo et al. 2013;Jiang et al. 2018;Li et al. 2017;Lv et al. 2020;Qin et al. 2011;Wang et al. 2018).
In hematopoiesis and leukemogenesis, researches mainly focus on the role of SOX6 in erythropoiesis (Cantu et al. 2011;Dumitriu et al. 2010Dumitriu et al. , 2006, but not on leukemia. Rarely, only one study showed that SOX6 overexpression arrested the proliferation of BCR-ABL1 + and JAK2V617F + leukemic cells (Barbarani et al. 2019). The biological function and clinical significance of SOX6 in AML are still largely unclear. In the present study, we attempted to investigate the expression of SOX6 in AML, and the effects of SOX6 on AML survival.

Patients
A cohort of 118 newly diagnosed AML patients (ND-AML), 16 relapsed or refractory AML patients (RR-AML), and 63 complete remission AML patients (CR-AML) were enrolled in this study with approval of the ethics committee of our hospital from February 2017 to February 2022. The diagnosis and classifications of the patients were based on the French-American-British (FAB) classification and 2016 revised World Health Organization (WHO) criteria. Control samples (n = 17) were obtained from donors without any malignant BM disorder, containing 11 healthy donors.

Online source
The online publicly available datasets from TCGA (The Cancer Genome Atlas), GEO (Gene Expression Omnibus) and GEPIA (Gene Expression Profiling Interactive Analysis) were analyzed in this study. The expression of SOX6 in AML patients and normal controls was analyzed in GEPIA dataset (http:// gepia. cancer-pku. cn/). The data of TCGA-LAML (NEJM 2013) were obtained from TCGA databases (http:// www. cbiop ortal. org/). In TCGA-LAML cohort, a total of 173 adult de novo AML patients with complete clinical information and SOX6 expression data were included in the study.
In addition, two independent cohorts from GEO database (GSE13159 and GSE12417) were also included. GSE13159 was used to compare the difference of SOX6 expression between AML (n = 542) and healthy donors (n = 74). GSE12417 (78 and 162 CN-AML patients) was used to validate the prognostic value of SOX6 expression in CN-AML through the online tool Genomicscape (www. genom icsca pe. com). The online tool GeneMANIA (http:// genem ania. org/) was used to construct the SOX6-centered gene functional interaction network.

Statistical analyses
The differences between continuous variables were using unpaired t test or the Mann-Whitney U test. Comparisons of categorical variables were analyzed using the χ 2 test or Fisher's exact test. In all cohorts, overall survival (OS) was defined as the day from diagnosis to last follow-up or death. Event-free survival (EFS) was defined as the day from diagnosis to relapse or death. EFS and OS were analyzed though Kaplan-Meier analysis using Log-rank test. Univariate and multivariate Cox regression models were used to analyze the prognostic significance of SOX6 expression. Statistical analyses of this study were performed using R software 4.0.2 and GraphPad Prism 8. For all tests, P value < 0.05 was considered statistically significant.

Low SOX6 expression associated with AML
By using the online web GEPIA, we found that SOX6 expression in AML patients from TCGA-LAML was significantly decreased as compared with GTEx normal samples (P < 0.01, Fig. 1a). Furthermore, result from GEO dataset (GSE13159) indicated that SOX6 expression was lower in AML cases than that in healthy controls (P < 0.0001, Fig. 1b).
To validate the public database results, we further analyzed SOX6 expression in a cohort of 118 ND-AML patients from our hospital. Consistently, SOX6 expression was significantly downregulated in AML compared with normal controls (P < 0.0001, Fig. 1c). Moreover, we observed that SOX6 expression in CR-AML patients was significantly increased compared with ND-AML patients (P < 0.0001, Fig. 1d). The relative SOX6 expression was markedly downregulated in RR-AML cases compared with CR-AML patients (P < 0.0001, Fig. 1d).

Association between SOX6 expression and clinical features
To explore the association of SOX6 expression with clinical features in AML, the median level of SOX6 expression was used to split AML patients into low SOX6 expression (SOX6 low ) and high SOX6 expression (SOX6 high ) groups. In our cohort, 118 AML patients were divided into SOX6 high (n = 59) and SOX6 low (n = 59) groups. The comparison of clinical characteristics between SOX6 high and SOX6 low groups was summarized in Table 1. The SOX6 low group had higher BM blasts (P = 0.0001) and WBC counts (P = 0.0112) compared with the SOX6 high group. No remarkable differences were found in sex, age, hemoglobin, platelet and FAB subtype between two groups.
In TCGA-LAML cohort, patients with SOX6 level ≥ median were defined as SOX6 high (n = 87), others were defined as SOX6 low (n = 86). The comparison of clinical characteristics between SOX6 high and SOX6 low groups was summarized in Table 2. As expectedly, the results for TCGA-LAML cohort were similar to our cohort. In TCGA-LAML, the SOX6 low group was also found to be associated with higher BM blasts (P = 0.0054) and WBC counts (P < 0.0001) than SOX6 high group.

Association between SOX6 expression and genetic abnormalities
We further determined the correlation of SOX6 expression with genetic abnormalities in AML. In both TCGA-LAML and our datasets, remarkable difference was found in the distribution of karyotypes between SOX6 low and SOX6 high subgroups (our data: P = 0.0438, Table 1; TCGA-LAML data: P = 0.0005, Table 2). In detail, we showed SOX6 expression among groups with core binding factor AML (CBF-AML), complex karyotype AML (CK-AML) or neither (Fig. 2). CBF-AML corresponds to two distinct cytogenetic subtypes of AML, namely t(8;21) and inv(16)/t(16;16) (Jahn et al. 2020). In both public (TCGA-LAML and GSE13159) and our cohorts, we found significantly different SOX6 expression among the three categories. AML patients with low SOX6 expression more commonly exhibited CBF-AML (TCGA-LAML: P = 0.0010, Fig Fig. 2c).

Prognostic value of SOX6 expression in AML
We first compared EFS and OS in 173 AML patients (TCGA-LAML) with low and high SOX6 expression groups. The result showed that low expression of SOX6 predicted significantly worse EFS (P = 0.015, Fig. 3a), but not OS (P = 0.091, Fig. 3b). We also analyzed the relationship between SOX6 expression level and survival in 79 CN-AML patients (TCGA-LAML). Significant differences in survival were observed between SOX6 low and SOX6 high groups among CN-AML patients (P = 0.020 for EFS, Fig. 3c; P = 0.026 for OS, Fig. 3d). To validate the prognostic value of SOX6 expression, we then obtained data from GEO dataset (GSE12417), including two cohorts of 78 and 162 CN-AML patients. We reached the same conclusion that CN-AML patients with low SOX6 expression suffered significantly shorter OS (cohort1: n = 78, P = 0.012, Fig. 3e; cohort2: n = 162, P = 0.024, Fig. 3f).
In addition, we also performed univariate and multivariate analyses on OS in CN-AML patients (TCGA-LAML), including the expression level of SOX6 and some common prognostic factors (age, WBC, gene mutations and treatment regimen). As shown in Table 3, high SOX6 expression was significantly and independently associated with a better OS both in univariate (P = 0.028) and multivariate analysis (P = 0.032).

Discussion
SOX6 has been extensively studied in various solid tumor but remains largely unknown in hematological disorders. In most solid cancer, SOX6 seems to act as a tumor-suppressor gene and remarkably inhibit tumor cell growth, proliferation, migration and invasion (Chen et al. 2020;Li et al. 2017;Lv et al. 2020;Z. Wang et al. 2018). Consistently, Gloria Barbarani et al. demonstrated that SOX6 display an anti-proliferative effect on a variety of leukemic cells (Barbarani et al. 2019). As a tumor suppressor, SOX6 is frequently downregulated in different human malignancies (Chen et al. 2020;   Guo et al. 2013;W. Jiang et al. 2018;Li et al. 2017;Lv et al. 2020;Qin et al. 2011;Wang et al. 2018) (except some brain tumor (Ueda et al. 2004) and Ewing sarcoma (Marchetto & Grunewald, 2020)). In our study, similar finding was also observed in AML. By using three cohorts of AML patients (GEPIA, GSE13159 and our datasets), we confirmed that SOX6 expression was notably downregulated in AML. In addition, the expression level of SOX6 was dynamic, along with the disease status. SOX6 was significantly decreased in RR-AML compared with CR-AML. This result indicated that SOX6 might be a potential biomarker to predict relapse of AML. Besides, the abnormal expression of SOX6 might be a promising prognostic marker. In some solid tumor, SOX6 has been found to associate negatively with tumor size, stage, metastasis (Chen et al. 2020;Qin et al. 2011), and patients with downregulation of SOX6 exhibited adverse prognosis (Chen et al. 2020;Guo et al. 2013;Lv et al. 2020;Qin et al. 2011;Z. Wang et al. 2018). Consistent with this, a significant negative effect of low SOX6 expression on OS and/or EFS was observed among CN-AML patients in our study, by analysis of multi-datasets from TCGA-LAML and GEO. Both univariate and multivariate analyses showed that low SOX6 expression was related to inferior OS in CN-AML. Then, we further analyzed the correlation between SOX6 expression and clinical features of AML patients. In both public and our present datasets, we found that SOX6 expression was negatively correlated with some traditional AML risk factors (including BM blasts, WBC counts and FLT3 mutation). Interestingly, a markedly correlation of SOX6 expression with some specific AML subtypes (CBF-AML and CK-AML) was noticed. According to the 2017 European LeukemiaNet (ELN) genetic risk stratification (Dohner et al. 2017), CBF-AML is categorized as favorable-risk AML and CK-AML is adverse-risk AML. Low SOX6 expression was commonly exhibited in CBF-AML and rarely in CK-AML. This may explain that downregulation of SOX6 didn't correspond to worse OS in whole-cohort AML. Thus, we consider that SOX6 expression might be a valuable factor that affecting prognosis in CN-AML, but not in whole-cohort AML. Overall survival

Days from diagnosis Days from diagnosis
To understand potential biological role of SOX6 in AML, we built a SOX6-centered gene-gene interaction network. Among the interacted genes with SOX6, some have been confirmed to play critical roles in AML, such as TOX3, SOX2, IRF9, IL1RL1 and TSLP. A number of investigations showed the expression of TOX3, SOX2 and IRF9 were declined in AML (Liang et al. 2021;Tian et al. 2018;Tosic et al. 2018), and the increased expression of SOX2 and TOX3 related to poor prognosis of AML (Liang et al. 2021;Tosic et al. 2018). In addition, a series of researches showed that IRF9, IL1RL1, and TSLP participated in the biological function of leukemogenesis, including cell survival, cell growth and/or cell apoptosis (Quentmeier et al. 2001;Tian et al. 2018;Wang et al. 2019) [36-39]. In human AML, IRF9 knockdown could promote leukemic cell proliferation, colony formation and survival (Tian et al. 2018). IL1RL1 has been reported to be dynamically expressed in CBF-MYH11 + leukemic cells and promoted their survival (Y. Wang et al. 2019). Similarly, TSLP could prevent apoptosis and stimulate growth of the human AML cell line MUTZ-3 (Quentmeier et al. 2001). Therefore, we hypothesized that SOX6 downregulation might have multiple effects on leukemogenesis, by interacting with SOX6-related genes mentioned above. The interaction between SOX6 and its associated genes need to been verified experimentally in future.
In summary, our discoveries confirmed that SOX6 was decreased expression in AML, and associated with distinct genetic abnormalities and prognosis of AML. SOX6 might be used as a potential biomarker for risk stratification and predicting relapse of AML. Fig. 3 Survival analysis of AML patients according to SOX6 expression. a EFS of whole-cohort AML from TCGA-LAML. b OS of whole-cohort AML from TCGA-LAML. c EFS of CN-AML from TCGA-LAML. d OS of CN-AML from TCGA-LAML. e OS of CN-AML from GSE12417 (cohort1: n = 78, SOX6 probe: 227498_at) by using Genomicscape. f OS of CN-AML from GSE12417 (cohort2: n = 162, SOX6 probe: 227498_at) by using Genomicscape ◂ Author contributions HS and YL designed the project; DJ and HS performed experiments; HS analyzed data and wrote the manuscript; QZ, EL, DJ and YL performed revised the manuscript. All authors reviewed the manuscript. All authors read and approved the final manuscript.
Funding None.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Conflict of interest
The authors have declared that no competing interest exists.
Ethical approval This study was approved by the Ethics Committee of The Third Xiangya Hospital of the Central South University. The study was conducted in accordance with the principles of the Declaration of Helsinki.
Consent to participate All participants in the study provided written informed consent.