TGIF2-Induced HMGB3 Promotes Esophageal Squamous Cell Carcinoma Progression Through TGFβ Signaling

Background: Proliferation and metastasis are the major malignant phenotypes of esophageal squamous cell carcinoma (ESCC) and the main causes for poor survival in patients with ESCC. Nevertheless, the underlying mechanisms of ESCC proliferation and metastasis remains unclear. The high mobility group box protein family 3 (HMGB3) is one of the HMGB family members. It is critically involved in the occurrence and development of various carcinomas. However, the knowledge of HMGB3 in ESCC remains limited. In this study, we elucidated the role of HMGB3 in ESCC proliferation and metastasis, and the concrete mechanism. Methods: Expression level of HMGB3 and TGF-β interacting factor 2 (TGIF2) in ESCC cell lines and tissues was quantied by qRT-PCR, Western Blot, and immunohistochemistry. In vitro and in vivo assays revealed the functions of TGIF2 and HMGB3 in ESCC. RNA-seq was performed to search for the downstream signaling of HMGB3. ChIP assay and were performed to explore the relationship of HMGB3 and TGIF2. HMGB3-interacting protein was validated by immunoprecipitation. Results: Higher expression of TGIF2 and HMGB3 was observed in ESCC cell lines and tissues and was associated with worse prognosis of ESCC patients. TGIF2 and HMGB3 upregulation could promote ESCC proliferation and metastasis, and vice versa. TGIF2 and HMGB3 upregulation can activate Smad-dependent TGF-β signaling. TGIF2 can transcriptionally regulate HMGB3, and its TGF-β inducing capability and oncogenic role are at least partly HMGB3-dependent. Additionally, TLR3 was identied as a client protein of HMGB3, and their combination might be the reason of TGF-β activation. Conclusions: Collectively, HMGB3-dependent TGIF2 overexpression activates TGF-β signaling and promotes the proliferation and metastasis of ESCC via TLR3 regulation. These ndings revealed that TGIF2 and HMGB3 could be prognostic mutated HMGB3 promoter which was cloned into reporter effect on Different analysis shows the positive regulating effect of on


Background
Esophageal squamous cell carcinoma (ESCC) is the seventh most prevalent cancer and has the sixthhighest cancer-associated mortality rate wordwide [1,2]. Although improvement in earlier diagnosis and effective treatment have enhanced the overall survival (OS) rate of ESCC patients, long-term survival of ESCC patients remains poor because of unlimited proliferation and distant metastasis [3]. ESCC patients' ve-year survival rates are less than 20% [4]. Multiple studies were devoted to disclose the underlying mechanisms of ESCC but their results were far from satisfactory [4]. Therefore, continuous and in-depth investigation of the mechanisms of ESCC and establishing novel therapeutic strategies are still in urgent need.
Thus, it can assist the malignant transformation of CRC [22], melanoma [23], and lung adenocarcinoma (LUAD) [24]. The expression of TGIF2 and HMGB3 are positively associated in pro-B cells, and they both donate to the self-renewal of fetal-B cells [25]. However, the characteristic of TGIF2 and the relationship between HMGB3 and TGIF2 in carcinoma have not been investigated.
This study highlighted the oncogenic role of TGIF2 and HMGB3 in the proliferation and metastasis of ESCC. They are both highly expressed in ESCC and their overexpression is closely related to the poor prognosis of ESCC patients. In addition, both TGIF2 and HMGB3 lead to the activation of TGF-β signaling. Mechanistically, TGIF2 transcriptionally regulate HMGB3 and its malignancy inducing capability, and TGF-β activating capability is partly HMGB3-dependent. Collectively, this study rstly con rmed the role of HMGB3, TGIF2 in ESCC progression and the concrete mechanism involved in this.
This study improves the understanding of function of HMGB3, TGIF2 as well as the relationship between them. Targeting TGIF2/HMGB3/TGF-β axis may be promising for the treatment of ESCC. ESCC cells were infected with the MOI set as 40. The cells were then selected using 1-4 ug/ml puromycin for 14 days.

Materials And Methods
2.6 CCK-8 assay 1× 10 3 cells in 100μl were seeded into 96-well plates per well. We replaced the original medium with CCK-8 solution (TransDetect Cell Counting Kit, Transgene, Beijing, China) with complete medium mixing at a 1:9 ratio at the same time in 5 days. We subsequently incubated cells at 37 °C for three hours. Each sample was inspected at 450 nm, and a microplate reader (Bio-Rad, CA, USA)was used to obtain the absorbance.

In vitro migration and invasion assays
In vitro migration and invasion abilities of infected cell lines were measured using 24-well Transwells (8μm pore size, Corning, Inc., NY, USA). In the migration experiment, 4 × 10 5 cells were seeded in the top chamber. While in the invasion experiment, every top chamber was coated with 200 mg/ml Matrigel and dried for one hour at 37 ℃. Then, 8 × 10 5 cells were plated in the chamber, and the invading and migrating cells at the lower layer were counted 36 h later.

Colony formation assay
The ESCC cells suspension (800 cells) was seeded in 6-well plates. They were then cultured for two weeks in a 5% CO 2 incubator at 37•C. Subsequently, the cells were xed with 10% formalin for 15 min, after which they were stained via 0.1% crystal violet for 15 min.

Cell cycle assay
The seeded stable infected ESCC cells attaining the log phase via trypsinization were harvested to 1ml centrifuge tube. Subsequently, we rinsed cells in a phosphate-buffered saline (PBS) buffer. We then xed the samples in 75% ethanol for 1 hour at -20℃. Fixed cells were rinsed again by PBS, followed by 400ul propidium iodide (Servicebio,) staining within 100ul RNase at 37℃ for 30 minutes in the dark. We nally implemented cell cycle analysis by the ow cytometer (CytoFLEX, Beckman Coulter, Brea, CA, United States).

Xenograft experiment and in vivo metastasis experiment
Balb/c mice (6 weeks old) were obtained from Shanghai Experimental Animal Center of the Chinese Academy of Sciences (Shanghai, China). The mice were raised in speci c pathogen-free conditions in the Animal Research and Care Committee of the Fourth Force Military Medical University. We randomly assigned these mice into experimental or control groups. We created the xenograft tumor models via subcutaneous injection of 5 × 10 6 indicated cells. We measured the tumor size every three days, and calculated the volume using the following formula: volume = (length × width2)/2. Mice were sacri ced when the volume of the largest tumor in the group is close to 1000 . We conducted tail vein injection assays in BALB/C nude mice (5 weeks old) via 4 × 106 indicated cells. Mice were sacri ced after 6 weeks when they were in poor condition, and the lungs and livers were extracted for histological staining and examination. The survival time of mice was also recorded. The use of live animals in this research was approved by the Committee inTeaching and Research (CULATR) at Fourth Military Medical University.

Immunohistochemistry
The human ESCC tissue microarray (TMA) (Lot No. HEsoS180Su08) was purchased from Shanghai Outdo Biotech Co., LTD (Shanghai, China). It contained 114 ESCC and 66 para-tumor tissues. These tissues were obtained from patients who underwent radical esophagectomy from April 2006 to December 2008. This research was rati ed by the ethics committee of Xijing Hospital, Fourth Military Medical University. Xylene and then graded alcohol were utilized to depara nize the TMA slides. The endogenous peroxidase activity was blocked using 3% H 2 O 2 , and 10% goat plasma was used to block the slices.
Pathologists were invited to view the TMA utilizing the histochemical score (H-score) to assess the staining intensity and the percentage of stained cells. The staining intensity was scored on a range of 0 to 3: 0 (no staining), 1 (weakly staining), 2 (moderate staining), and 3 (strong staining). The percentage of stained cells was scored as follows: 0 (no staining, 0%), 1 (staining range, 1-25%), 2 (staining range, 26-50%), 3 (staining range, 51-75%), or 4 (staining range, 76-100% We performed a log-rank assessment to inspect the connection between HMGB3, TGIF2 expression and the prognosis of ESCC. two-tailed Student t-test was employed to inspect differences between variables in the variable groups. p<0.05 was set as signi cantly different.

HMGB3 is upregulated in ESCC and predict poor prognosis
To investigate the role of HMGB3 in ESCC. We initially analyzed the gene expression level of HMGB3 in The Cancer Genome Atlas (TCGA) dataset (https://cancergenome.nih.gov/abouttcga/overview). The result showed that the expression of HMGB3 was signi cantly upregulated in ESCC tissues than that in adjacent tissues (Fig. 1A). Speci cally, both HMGB3 mRNA and protein levels in the ESCC cell lines were critically higher than those in normal het-1A cell (Fig 1C, D). Besides, the expression of HMGB3 was analyzed in 21 pairs of ESCC tissues and adjacent nontumor tissues using qRT-PCR. As shown in Fig. 1E, mRNA expression of HMGB3 was higher in ESCC tissues than the adjacent normal control (p=0.007).
Additionally, the ESCC patients in the TCGA database were divided into high and low HMGB3 expression groups. The higher expression of HMGB3 was correlated with a shorter OS rate (Fig. 1B).
Tissue microarray was also employed to further evaluate the correlation between HMGB3 expression and clinical features. The HMGB3 protein level was critically higher in ESCC tissues than in adjacent nontumorous tissues (p<0.001) (Fig. 1F, G). Higher expression of HMGB3 protein is associated with a higher possibility of tumor invasion grade (p=0.05) and higher American Joint Committee on Cancer (AJCC) stage (p=0.007). Moreover, males tended to have higher HMGB3 expression (p=0.046) ( Table 1).
The Kaplan-Meier analysis showed that patients with HMGB3 overexpression have shorter OS rates than patients with lower HMGB3 expression (Fig. 1H).

Overexpression of HMGB3 promotes ESCC cell proliferation, migration, and invasion in vitro
Based on the expression level of HMGB3 in ESCC cells, we infected EC9706 with lentivirus to upregulate HMGB3 expression. We also downregulated HMGB3 expression by si-RNA in EC109 and ECA109. The expression level was con rmed by qRT-PCR ( Fig. 2A) and WB (Fig. 2B). Cell Counting Kit 8 (CCK-8) assay (Fig. 2C) and colony formation assay (Fig. 2D) indicated that HMGB3 upregulation could promote the ESCC cells' proliferation, while HMGB3 downregulation attenuated the proliferative ability. Further, we conducted a cell cycle assay to disclose whether the change of cell proliferation was associated with alteration in the cell cycle pro le. Consequently, the results showed that ectopic expression of HMGB3 in EC9706 attenuated G1/S arrest, while HMGB3 downregulation in EC109 and ECA109 lowered the rate of cells in the S phase ( Figure 2E). Transwell assay con rmed that the upregulation of HMGB3 could promote the invasion and metastasis ability of EC9706, while HMGB3 knockdown repressed this ability (Fig. 2F).

Overexpression of HMGB3 promotes ESCC cell proliferation, migration, and invasion in vivo.
We established another stable cell lines by lentivirus, which suppresses HMGB3 expression in EC109 and ECA109. The result of qRT-PCR and WB showed that infection of LV-shHMGB3-1 could reduce the expression of HMGB3 mRNA and protein level in indicated cells (Fig. 3A, B). Subcutaneous xenograft nude mice models were established by stable cell lines of EC9706 with HMGB3 overexpression, ECA109 with HMGB3 knockdown and their negative control (NC) (Fig. 3C). The tumor volume and weight assessment showed that HMGB3 overexpression stimulated tumor growth, while HMGB3 downregulation suppressed tumor proliferation in vivo (Fig. 3D, E). Immunohistochemistry (IHC) staining for Ki67 and HMGB3 showed signi cant increases in the growth of tumor cells with high HMGB3 expression (Fig. 3F). We further performed in vivo lung metastasis experiments. The data revealed that HMGB3 upregulation shortened the OS of mice in the HMGB3 overexpression group (Fig. 3G) and enhanced ESCC cells lung metastasis incidence, while HMGB3 knockdown resulted in the opposite effect (Fig. 3H). Hematoxylin and eosin (H&E) staining showed an increased number of metastatic lung nodules on HMGB3 overexpression, whereas HMGB3 knockdown reduced the metastatic capability (Fig. 3I). In conclusion, HMGB3 plays an oncogenic role in ESCC cell proliferation and metastasis in vivo.

TGIF2 transcriptionally regulates HMGB3
To investigate the concrete mechanism of HMGB3 in the tumorigenesis of ESCC, we predicted the transcriptional factors of HMGB3 in the JASPAR database (http://jaspar.genereg.net/) (Fig. 4A). The data revealed that TGIF2 might regulate HMGB3 as a transcriptional factor. A recent study also highlighted that their expression level was positively associated, and they both donate to initiate self-renewing fetal pro-B cells [25]. The correlation analysis of the TCGA database also indicated that the expression level of TGIF2 and HMGB3 might be positively associated (R=0.18, p=0.018) (Fig. 4B). To verify this hypothesis, we initially analyzed the expression of TGIF2 in ESCC tumor cell lines and normal cell het-1A by qRT-PCR (Fig. 4C) and WB (Fig. 1D). The results showed that TGIF2 expression was lower in EC9706, whereas it was higher in EC109 and ECA109. We then established stable cell lines to upregulate TGIF2 in EC9706 and suppress TGIF2 in EC109 and ECA109 by lentivirus. The expression level of TGIF2 and HMGB3 was con rmed using qRT-PCR (Fig. 4D) and WB (Fig. 4E).
Subsequently, we determined to con rm the direct relation between TGIF2 and HMGB3. Through sequence analysis, we identi ed two possible TGIF2-binding sites in the HMGB3 promoter (Fig. 4F). The chromatin immunoprecipitation (ChIP) showed that overexpression of TGIF2 enhanced binding of TGIF2 to binding site 1 of HMGB3 promoter, while the binding of TGIF2 to binding site 2 of HMGB3 promoter was not changed (Fig. 4F, G). We also developed and transferred plasmid to mutate the site between -2000bp to 500bp of HMGB3 promoter. In this progression, binding site 1 is mutated and binding site 2 remains unchanged (Fig. 4I). Luciferase activity further veri ed that TGIF2 overexpression could enhance the luciferase activity when binding to HMGB3 promoter. However, the mutation of the HMGB3 promoter binding site 1 inhibited this effect, indicating the binding of TGIF2 to the binding site 1 of HMGB3 promoter (Fig. 4I).

Overexpression of TGIF2 promotes ESCC cell proliferation, migration, and invasion in vitro and in vivo.
Using the established stable cell lines with TGIF2 overexpression or knock down as indicated, we performed CCK-8 (Fig. S1A) and colony formation assay to uncover the role of TGIF2 in ESCC cell proliferation in vitro (Fig. S1B). The results showed that TGIF2 upregulation could promote the proliferation of ESCC cells, while TGIF2 suppression delayed these malignant phenotypes. Subsequently, cell cycle assay was conducted to verify whether its expression can regulate the cell cycle progression (Fig. S1C). The assay highlighted that the role of TGIF2 in promoting proliferation may result from its ability to decrease G1/S arrest of cancer cells. The subcutaneous xenograft nude mice model was also conducted to show the role of TGIF in inducing proliferation in vivo (Fig. 5A). In the xenograft nude mice, the tumor volumes and weight of the TGIF2 overexpression group were substantially larger than those in the NC group. Reversely, tumors formed by the TGIF2 knockdown group had critically smaller volumes and lower weight than those of the NC group (Fig. 5B, C). IHC for Ki67 and TGIF2 revealed higher TGIF2 expression signi cantly increases the proliferation of tumor cells (Fig. 5D).
Transwell assay (Fig. S1D) showed that TGIF2 overexpression could promote the invasive and metastatic ability of ESCC cells, and vice versa. The in vivo metastasis experiment was also performed to disclose the roles of TGIF2 in in uencing ESCC cells' metastatic capability in vivo. TGIF2 upregulation leads to a shorter OS rate of mice (Fig. 5E) and contributes to a higher rate of metastatic lung modules (Fig. 5F, G), and vice versa. Collectively, TGIF2 functions as an oncogene that promotes ESCC cell proliferation and metastasis in vitro and in vivo.
3.6 HMGB3 is essential for TGIF2-mediated ESCC cell proliferation and metastasis.
We further veri ed the role of HMGB3 in TGIF2-mediated ESCC proliferation and metastasis. We deleted HMGB3 in TGIF2-overexpressing EC9706 cells and upregulated HMGB3 in TGIF2-knockdown ECA109 cells by infection of lentivirus. The expression level of HMGB3 and TGIF2 was veri ed using WB (Fig. 6A).
In vitro tumorigenesis experiments including CCK-8 and colony formation assay indicated that ESCC cells' growth ( Fig. S2A, B) critically decreased upon silencing of HMGB3 in EC9706 cells overexpressing TGIF2. In contrast, HMGB3 overexpression reversed the TGIF2 knock down-induced ESCC cell proliferation. Speci cally, cell cycle assay showed that the critical role of HMGB3 in TGIF2-mediated proliferation resulted from its capability in inducing the increasement of S phase rate of ESCC cells (Fig.   S2C). In vivo experiments (Fig. 6B) showed signi cant downregulation in tumor growth and tumor weight upon silencing of HMGB3 in EC9706-TGIF2 cells (Fig. 6C, D). Additionally, IHC effect for Ki67 staining was also attenuated upon silencing of HMGB3 in EC9706-TGIF2 cells (Fig. 6E).
Additionally, the Transwell assay veri ed the promoting role of HMGB3 in TGIF2 induced ESCC metastasis (Fig. S2D). In vivo metastatic experiment showed that HMGB3 suppression increased the OS (Fig. 6F), and reduced lung metastasis rates and the numbers of metastatic lung nodules (Fig. 6G, H) of mice in the EC9706-TGIF2 group. Thus, HMGB3 is essential for TGIF2-mediated ESCC cell proliferation and metastasis.

Clinical signi cance of TGIF2 and correlation between TGIF2 and HMGB3
To investigate the role of TGIF2 and the correlation between TGIF2 and HMGB3 in ESCC tissues, we further analyzed the mRNA expression in the 21 pairs of ESCC and normal control tissues. As revealed in Fig. 7A, TGIF2 was higher expressed in ESCC than in adjacent nontumor (p=0.0457). Additionally, IHC was also performed to analyze the protein expression level on the TMA (Fig. 7B). Consistently, ESCC tissues showed higher TGIF2 protein expression levels than the normal (p=0.0087) (Fig. 7C). Correlation analysis con rmed the positive correlation between TGIF2 and HMGB3 (r=0.4924, p=0.023) in 21 pairs samples and the TMA (r=0.4331, p<0.001) (Fig. 7D, E). The Chi-square test indicated that HMGB3 overexpression positively correlated with its master regulator TGIF2 (Fig. 7F). Speci cally, elevated expression of TGIF2 was closely related to higher pathological grade (Table 1). According to the K-M plotter, TGIF2 overexpression was associated with poor prognosis of ESCC patients (Fig. 7G). Besides, cooverexpression of TGIF2 and HMGB3 showed the lowest OS rate, and both of their low expression indicated the highest OS rate in the cohort of ESCC patients, suggesting that TGIF2/HMGB3 axis is critical in the prognosis of ESCC patients (Fig. 7H).

TGIF2 and HMGB3 can positively regulate TGF-β signaling in ESCC
To further identify the possible downstream target of HMGB3, we performed RNA-seq. The volcano plot and heatmap showed 168 upregulated genes and 141 downregulated ones in ECA109 cells compared with the ECA109 infected with LV-shHMGB1-1 (Fig. 8A, B). The speci c expression of each gene is presented in Table S1. The Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis indicated that TGF-β signaling might be regulated by HMGB3 (Fig. 8C). To con rm the relation between HMGB3 and TGF-β signaling, we performed WB. The data indicated that HMGB3 overexpression could enhance the expression of TGF-β, SMAD2, SMAD3, p-SMAD2, p-SMAD3, and SMAD2/3, and low expression of HMGB3 can downregulate these molecules (Fig. 8D). TGIF2 can also positively regulate TGF-β signaling pathway in a Smad-dependent way (Fig. 8E), and this is partly HMGB3-dependent (Fig.  8F). HMGB3 knockdown attenuated the inducing effect of TGIF2 overexpression on TGF-β signaling in EC9706, while HMGB3 upregulation reversed the inhibiting function of TGIF2 knockdown in ECA109.
Recent studies highlighted the relationship between HMGB and TLR family. Besides, TLR3 was reported to be involved in TGF-β activation in various carcinomas. According to RNA-seq and recent articles, we subsequently investigated the possible relationship between HMGB3, TLR3, and TGF-β. we initially activated TLR3 using poly (I:C). Consistent with previous studies, activation of TLR3 can induce the expression of TGF-β signaling in ESCC cell line (Fig. 8G). Moreover, we detected TLR3 in the indicated cell lines, and found that overexpression of TGIF2 and HMGB3 could enhance the expression of TLR3, and vice versa (Fig. 8D, E). Besides, HMGB3 knockdown in the EC9706-TGIF2 cells attenuated the upregulation of TLR3 and TGF-β signaling. While upregulation of HMGB3 in the ECA109 with TGIF2 knockdown cells enhanced the TLR3 and TGF-β signaling, indicating the role of HMGB3 in positively regulating TLR3 (Fig. 8F).
Previous study fully con rmed HMGB family and TLR family could be combined as donors and receptors to induce in ammation and carcinogenesis [12]. To further investigate the interaction between HMGB3 and TLR3, we employed a reciprocal coimmunoprecipitation (Co-IP) assay. Co-IP of the endogenous protein veri ed that HMGB3 could directly combine with TLR3 compare to IgG in EC9706 and ECA109.

Discussion
Current diagnostic and therapeutic methods for ESCC have improved, and ESCC patients' OS has been prolonged [3]. However, ESCC patients' prognosis still cannot meet expectations, and unlimited proliferation and metastasis are still the main reasons for ESCC patients' poor prognosis [4]. Thus, investigation of the mechanisms underlying ESCC proliferation and metastasis is urgently needed to develop therapies. Herein, we identi ed that TGIF2 transcriptionally regulated HMGB3, and their expression was positively associated with ESCC proliferation and metastasis. TGIF2 and HMGB3 expression levels were relatively higher in ESCC patients, and ESCC patients with higher HMGB3 or TGIF2 levels tended to have a worse prognosis. We also found that TGIF2/HMGB3 axis can positively regulate TGF-β signaling, and this may be related to the combination of HMGB3 and TLR3.
Similar to its family members (HMGB1 and HMGB2), HMGB3 critically participate in building nucleoprotein complexes by altering the chromatin structures, leading to the combination of multiple factors [5] and modi cation of these DNAs [37]. HMGB3 was associated with poor prognosis of GC [8], HCC [9] and ESCC [10]. Researches also highlighted the role of HMGB3 in proliferation [11], metastasis [12], hypoxia [13] and drug resistance [14]. However, the concrete mechanism of HMGB3 in inducing ESCC tumorigenesis was still less explored. Herein, we con rmed that HMGB3 could induce proliferation and metastasis of ESCC in vitro and in vivo. IHC also indicated that high expression of HMGB3 is associated with poor prognosis of ESCC patients, and overexpression of HMGB3 may lead to higher tumor invasion grade and higher AJCC stage. We further made an in-depth inquiry into the potential mechanism.
In previous research, HMGB3 was reported to be regulated by miRNAs including miR-200b [12], miR-205-5p [11], miR-532-5p [38], and proteins including EGb 761 [39] and MMP7 [17]. We predicted possible transcriptional factors of HMGB3 employing the JASPAR database. After researching recent studies, we selected TGIF2 as an ideal target because previous studies showed that they were both overexpressed to promote the self-renewing capability in the pro-B cells [25]. Besides, sequence analysis also predicted 2 possible binding sites of HMGB3 promoter for TGIF2 (binding site 1: -1138bp ~ -894bp and binding site 2: -4566bp ~ -4638bp). TGIF2, like TGIF1, functions by binding with R-SMAD and recruiting histone deacetylases (HADCs) or directly binding to the promoter of DNAs independent of SMADs and HADC3 [24]. TGIF2 was reported as a transcriptional activator of CDH1, and thus it could lead to EMT of ESCC cells [22]. While Fructosyltransferase 8 (FUT8) can be transcriptionally regulated by TGIF2, leading to distant metastasis of melanoma [23]. ERK/MAPK signaling participates in the phosphorylation of TGIF2, and phosphorylated TGIF2 can transcriptionally enhance the expression of OCT4 and promote stemness of lung adenocarcinoma (LUAD) [24]. In our study, we con rmed that TGIF2 could bind to the promoter of HMGB3 and positively regulate its expression via qRT-PCR, WB, ChIP, and luciferase report. TGIF2 was reported to be involved in the progression of various carcinomas, however, the role of TGIF2 in ESCC was not investigated in previous study [23][24][25]. This study also rstly veri ed that TGIF2 can act as one oncogene in ESCC carcinogenesis. Its expression level is higher in ESCC tissues, and its high expression leads to poor prognosis of ESCC patients. Higher TGIF2 expression level may be related to higher pathological grade. Additionally, its overexpression can lead to proliferation and metastasis of ESCC cells, analyzed by in vitro and in vivo analysis. Speci cally, the oncogenic role of TGIF2 is at least partly HMGB3-dependent.
Downstream target of HMGB3 in ESCC was also not disclosed in previous study [11,17]. We further employed RNA-seq to underline the downstream target of HMGB3. KEGG enrichment analysis highlighted the possible relationship between HMGB3 and TGF-β signaling. TGF-β signaling can function through Smad-dependent or Smad-independent ways, and it plays a dual role in tumorigenesis [40]. As is generally accepted, it inhibits the malignant transformation of normal cells while inducing tumorigenesis during tumor progression [19]. The reasons for its duality are multifaceted, and the mechanisms of its tumor-inducing ability are partly disclosed, including but not limited to evasion of immune surveillance [41], myo broblast mobilization [42], and osteoclast mobilization [43]. TGF-β was reported to be involved in ESCC carcinogenesis [43,45]. Recent studies veri ed that TGIF2 and HMGB1 were closely related to TGF-β signaling [19,40]. TGIF2 was universally accepted as a classical binding protein of R-Smad [19].
Herein, we discovered that TGIF2 and HMGB3 could positively regulate Smad-dependent TGF-β signaling in ESCC, and TGIF2's function in inducing TGF-β signaling is partly HMGB3 dependent, employing WB analysis.
Given that HMGB3 could positively regulate TGF-β signaling, we wanted to nd a HMGB3 downstream target, which is also involved in the regulation of TGF-β signaling. We observed the different expressed genes in ECA109-shHMGB3-1 and NC groups and searching through multiple studies. We then selected TLR3 for the following reasons. First, in many cases, the TLR family is essential for the in ammation and carcinoma-inducing function of HMGB1 [34,44], which shares an identical biochemical structure with HMGB3 [2]. Consequently, TLR3, as a signi cant member of the TLR family, may be vital for HMGB3's participation in carcinogenesis. Second, TLR3 was certi ed to positively regulate TGF-β signaling in many carcinomas, including neuroblastoma [45], breast cancer [46], and lung cancer [35]. TGF-β is upregulated in ESCC patients and can induce tumorigenesis of ESCC [32,33]. TLR3 may also positively regulate TGFβ in ESCC, and thus TLR3 may be a connection point between HMGB3 and TGF-β. Investigation of the relationship between HMGB3 and TLR3 can further disclose the mechanism of HMGB3 and the HMGB family. Based on the direct combination between HMGB1 and TLR4 [34,35,44,47], we speculated that HMGB3 might regulate TLR3 by directly combining with it. WB indicated that TLR3 activation using poly (I:C) could induce the expression of Smad-dependent TGF-β signaling. Additionally, TGIF2 and HMGB3 could positively regulate TLR3, and the inducing role of TGIF2 is partly HMGB3-dependent. Speci cally, the Co-IP test con rmed the direct combination of HMGB3 and TLR3.
In conclusion, we veri ed TGIF2's role in positively regulating HMGB3 as a transcriptional factor. TGIF2 and HMGB3 were both upregulated in ESCC patients, and their overexpression indicated a poor prognosis. TGIF2 and HMGB3 could promote proliferation and metastasis of ESCC and induce the expression of TLR3 and TGF-β signaling. The inducing capability of TGIF2 is at least partly HMGB3dependent. Additionally, TLR3 can activate TGF-β signaling in ESCC, and this can be positively regulated by HMGB3. The direct combination between HMGB3 and TLR3 may be the concrete mechanism for their TGF-β signaling-inducing capability.
Herein, we rstly investigated the oncogenic role of TGIF2 and HMGB3 in ESCC proliferation and metastasis. The relationship between HMGB3, TGIF2, TLR3, and TGF-β signaling was rstly disclosed, and we also further perfected the understanding of these molecules and their families. Targeting the TGIF2/HMGB3/TLR3 axis may be helpful for the diagnosis and treatment of ESCC. However, limitations in our study must also be discussed. Firstly, the number of ESCC patients in our cohort was relatively insu cient, which may explain TGIF2's insigni cant role in the multicox regression analysis. Secondly, because some mice were in relatively poor condition before the xenograft experiment, we excluded them for ethical reasons. The number in different groups was not the same (4 or 5 as indicated). Thirdly, the reason why the combination of HMGB3 and TGIF2 could lead to the activation of TGF-β signaling was not clearly veri ed. Consequently, further studies with a larger ESCC cohort and in-depth molecule mechanisms of TGIF2/HMGB3/TLR3 and in ESCC should be conducted in the future.

Conclusion
In conclusion, we rstly reported that TGIF2 could promote ESCC proliferation and metastasis via transcriptional regulation of HMGB. The inducing capability of TGIF2 is partly HMGB3-dependent. Additionally, TLR3 was veri ed as a novel binding partner of HMGB3. Combination of HMGB3 and TLR3 might be the reason for HMGB3 and TGIF2 to activate the Smad-dependent TGF-β signaling. Comprehension and investigation of the TGIF2/HMGB3/TLR3 axis in ESCC will provide novel perspective of the TGIF, HMGB and TLR family as well as the mechanism of ESCC progression. In addition, TGIF2/HMGB3/TLR3 axis may be promising target for treating ESCC.