USP21 expression is aberrantly upregulated and indicates an unfavorable prognosis in ESCC

Although one recent study reported increased protein levels of USP21 in 4 cases of esophageal cancer tissues [30], USP21 expression levels in ESCC (the most predominant pathological subtype of esophageal cancer in Asian populations), its association with clinicopathological factors, and its effect on ESCC prognosis remain unknown. Here, we detected USP21 protein levels in 86 pairs of ESCC samples with complete clinical information (Supplementary Table 1). IHC scores demonstrated that ESCC tissues had significantly upregulated levels of USP21 protein compared to adjacent normal esophageal mucosa (Fig. 1A, B). Receiver operating characteristic (ROC) analysis revealed that USP21 protein levels possess the potential ability to distinguish ESCC tumors from their matched normal tissues (Fig. 1C). Further analysis based on IHC evaluation exhibited that ESCC tumor tissues at advanced statuses of primary tumor (T3-4) or with regional lymph node metastasis (N1-3) had higher levels of USP21 protein compared with those at early primary tumor statuses (T1-2) (Fig. 1D) or without positive lymph nodes (N0) (Fig. 1E). Consistently, USP21 protein expression in stage III or IV ESCC specimens was obviously higher than that in ESCC tissues with stage I and II or with stage I, II, and III (Fig. 1F). Additionally, poorly differentiated (G3) ESCC tumor samples displayed increased USP21 protein levels compared to well/moderately differentiated (G1-2) tumors (Fig. S1A). However, no significant association was found between USP21 protein levels and other clinical characteristics (age and gender) (Fig. S1B, C). To determine the effect of USP21 protein levels on ESCC prognosis, 86 patients were categorized into low or high USP21 expression groups according to IHC scores. Kaplan-Meier analyses demonstrated that ESCC patients with high USP21 levels displayed a shorter overall survival (OS) and progression-free survival (PFS) than patients with low expression of USP21 (Fig. 1G, H). To clarify whether USP21 protein expression is an independent predictor for the prognosis of ESCC patients, we performed a multivariate analysis by establishing the Cox-regression Hazard models. As shown in Supplementary Table 6, the status of the regional lymph node (N1-3 vs. N0) can independently predict OS and PFS, while USP21 protein expression (High vs. Low) is only an independent predictor for PFS. These results suggest that in terms of prognostic prediction for ESCC patients, USP21 may be no more useful than the standard staging, especially compared with the status of regional lymph nodes. Additionally, USP21 mRNA levels in 10 cases of fresh ESCC tumor samples were obviously upregulated compared with adjacent normal tissues (Fig. 1I). Consistently, increased mRNA levels of USP21 in ESCC tissues were identified based on TCGA data from the UCSC Xena platform [31] and GEO data (GSE130078) [32] (Fig. 1J, K). USP21 expression was further evaluated in normal human esophageal epithelial cells (HEEC) and five ESCC cell lines (Eca-109, KYSE-150, KYSE-30, KYSE-410, and KYSE-510). As shown in Fig. 1L, M, USP21 mRNA and protein levels in four ESCC cell lines (Eca-109, KYSE-150, KYSE-30, and KYSE-510) were remarkedly elevated compared to those in HEEC cells. Collectively, our results suggest dysregulated expression of USP21 in ESCC, which predicts a worse clinical prognosis.

Fig. 1: USP21 expression is aberrantly upregulated and indicates an unfavorable prognosis in ESCC.

A, B USP21 protein levels in ESCC tumor and adjacent normal esophageal mucosa tissues were evaluated by IHC. Representative staining images with scale bars (A) and a comparison of IHC scores (B) are present. C The diagnostic value of USP21 protein levels for ESCC was evaluated using the receiver operating characteristic (ROC) analysis. Further analysis based on IHC scores from ESCC samples with different primary tumor statuses (T1-2 and T3-4) (D), with no metastasis (N0) or metastasis (N1-3) of regional lymph nodes (E), with different stages (I, II, III, and IV) (F). Prognostic differences in patients with high or low levels of USP21 were assessed by Kaplan-Meier curves for overall survival (OS) (G) and progression-free survival (PFS) (H). I USP21 mRNA levels were detected in 10 pairs of fresh ESCC tissues through RT-qPCR. The expression of USP21 mRNA was evaluated using TCGA data from the UCSC Xena platform (J) and GEO data (GSE130078) (K). USP21 mRNA (L) and protein levels (M) were detected using RT-qPCR and western blot in HEEC, Eca-109, KYSE-150, KYSE-30, KYSE-410, and KYSE-510 cells. Statistical significance was determined by applying a Wilcoxon matched-pairs signed rank test (B), a Mann-Whitney test (D–F), log-rank test (G, H), unpaired t-test (I, J, L), or paired t-test (K). All P values and n-numbers are indicated on the corresponding graphs.

Dysregulated USP21 accelerates ESCC progression through its deubiquitinase activity in vitro and in vivo

In order to investigate the function of USP21 on ESCC progression, we knocked down endogenous USP21 expression in KYSE-150 and Eca-109 cells by transfecting siRNAs. The efficiency of USP21 depletion was verified by western blot and quantitation (Fig. 2A, Fig. S2A–C). To prove the specificity of siRNAs against USP21, we constructed the plasmid expressing siRNA-resistant USP21 (USP21SR) that translates the same sequence of amino acids (aa) as wild-type USP21 (USP21WT) but possesses mutant base sequence with resistance to si-USP21#1 and si-USP21#2. Our results revealed that si-USP21#1 or si-USP21#2 obviously reduced USP21WT protein levels but did not affect USP21SR expression (Fig. S2D). Functional experiments in vitro revealed that siRNA-regulated depletion of endogenous USP21 remarkedly suppressed plate colony formation (Fig. 2B, Fig. S2E), migration, and invasion (Fig. 2C, Fig. S2F) of KYSE-150 and Eca-109 cells. Consistently, USP21 knockdown significantly downregulated the protein levels of PCNA (a marker for cell proliferation) and N-Cadherin (a marker for mesenchymal cells) but upregulated the expression of E-Cadherin (a marker for epithelial cells) (Fig. 2A, Fig. S2B). To further demonstrate the consequence induced by USP21 depletion in vivo, we established ESCC cell lines with stably shRNA-mediated depletion of endogenous USP21 (sh-USP21) or negative control (sh-NC), in which stable USP21 knockdown efficiency was shown in Fig. 2D. In xenograft and lung metastasis mice models, we found that USP21 downregulation obviously inhibited tumor growth rate, weight and Ki-67 protein expression (Fig. 2E–G, Fig. S3A) and decreased metastatic nodules in mice lung (Fig. 2H, I, Fig. S3B). These results indicate the cancer-promoting function of USP21 in ESCC progression.

Fig. 2: Dysregulated USP21 accelerates ESCC progression through its deubiquitinase activity in vitro and in vivo.

A KYSE-150 cells were transfected with si-NC, si-USP21#1, or si-USP21#2. Western blot analysis was used to detect the protein levels of USP21, PCNA, E-Cadherin, and N-Cadherin. Plate colony formation assays (B) were applied to assess the proliferation ability of KYSE-150 cells transfected with si-NC, si-USP21#1, or si-USP21#2, while transwell assays (C) were used to evaluate cell migration and invasion capability. Representative staining and quantification as indicated. D Western blot analysis for verifying USP21 protein levels in stable KYSE-150 cells with sh-USP21 or sh-NC expression. The image of xenografts (E), tumor volume curve (F), and quantification for tumor weights (G) from BALB/c-nude mice (n = 7) subcutaneously xenografted with KYSE-150 cells expressing sh-USP21 or sh-NC for 19 days. H, I KYSE-150 cells with sh-USP21 or sh-NC expression were injected into the tail vein of BALB/c-nude mice (n = 4) for 6 weeks. Representative images (H) and statistical analysis (I) of metastatic nodules in mice lungs. J Western blot analysis was applied to determine the indicated protein levels in KYSE-150 cells with Vector, Flag-USP21WT, or Flag-USP21C221A expression. The Vector-, USP21WT-, or USP21C221A-expressing KYSE-150 cells growth, migration, and invasion were evaluated by plate colony formation (K) and transwell assays (L). Representative images and quantification as shown. M–O KYSE-150 cells with Vector-, USP21WT-, or USP21C221A-expression were subcutaneously xenografted into BALB/c-nude mice (n = 6) for 16 days. The representative xenograft image (M), volume curve for tumor growth (N), and statistical analysis for tumor weights (O) as indicated. P, Q The Vector-, USP21WT-, or USP21C221A-expressing KYSE-150 cells were injected into BALB/c-nude mice (n = 5) through a tail vein for 6 weeks. Representative images (P) and statistical analysis (Q) for pulmonary metastatic nodules as shown. Scale bars (red line) in (C) and (L) are 100 μm. The data in (B, C, F, G, I, K, L, N, O, Q) are presented as means ± SD. An unpaired t-test was performed to determine the statistical significance in (B, C, G, I, K, L, O, Q) while a two-way ANOVA was used for (F, N). The P values are displayed in the corresponding panels, respectively.

To determine whether the oncogenic role of USP21 in ESCC is dependent on its deubiquitination activity, we constructed plasmids expressing USP21WT or mutated USP21 with catalytic inactivity (USP21C221A). As shown in Fig. 2J and Fig. S4A, USP21WT, and USP21C221A expression levels were comparable in KYSE-150 and Eca-109 cells, which eliminated the effect of USP21 expression differences on subsequent functional experiments. Plate colony formation assays revealed that ectopic USP21WT expression enhanced colony formation in KYSE-150 and Eca-109 cells compared with vector control, while USP21C221A overexpression failed to promote the rates of colony formation (Fig. 2K, Fig. S4B). Similarly, in transwell assays, USP21C221A displayed no tumor-facilitating function in cell migration and invasion compared to USP21WT overexpression (Fig. 2L, Fig. S4C). Consistently, USP21WT overexpression significantly increased the expression of PCNA and N-Cadherin but reduced E-Cadherin levels while ectopic USP21C221A expression could not result in a similar alteration of these markers (Fig. 2J, Fig. S4A). The quantitation of western blots for PCNA, E-Cadherin, and N-Cadherin in KYSE-150 cells expressing Vector, USP21WT, or USP21C221A is displayed in Fig. S4D–F. Furthermore, USP21WT, but not USP21C221A mutation, significantly promoted tumor growth and upregulated Ki-67 protein levels in subcutaneous xenograft models (Fig. 2M–O, Fig. S5A) and increased the number of metastatic nodules in experimental mice models with lung metastasis (Fig. 2P, Q, Fig. S5B). Interestingly, USP21C221A displayed an inhibition on xenograft tumor growth compared with vector control in vivo (Fig. 2M–O). In summary, our results suggest the oncogenic function of USP21 in ESCC is exactly dependent on its activity of deubiquitinase.

USP21 binds to, deubiquitinates, and stabilizes the G3BP1 protein

It has been reported that USP21 acts as a deubiquitinase to stabilize certain oncoproteins, which promotes the progression of many human cancers [14, 15, 17, 19, 20, 23, 24]. Hence, to find the potential target regulated by USP21 in ESCC, we performed Co-IP assays to obtain a USP21-containing complex in KYSE-150 cells with ectopic expression of Flag-USP21WT, followed by mass spectrometry analysis (data not shown). Among all potential proteins binding to USP21, G3BP1 aroused our attention and interest, because G3BP1 displayed a relatively high enrichment in the Co-IP complex and was reported as an oncogene in ESCC progression [33]. The physical binding between USP21 and G3BP1 proteins was further verified in KYSE-150 and Eca-109 cells by using an antibody against G3BP1 or USP21 (Fig. 3A). In addition, we determined the binding between ectopic USP21 and G3BP1 proteins in HEK293T cells through IP with Flag- or HA-antibody (Fig. 3B). Subsequently, to map the essential domains mediating the binding between USP21 and G3BP1 protein, we constructed truncated USP21- or G3BP1-expressing plasmids, Flag-USP21-\(\Delta\)1 with the deletion of nuclear export signal (NES) domain (134-152 aa), Flag-USP21-\(\Delta\) 2 with the deletion of USP domain (212-558 aa), HA-G3BP1-\(\Delta\) 1 lacking the domain of nuclear transport factor 2 (NTF2) (11-133 aa), and HA-G3BP1-\(\Delta\) 2 lacking the domain of RNA recognition motif (RRM) (340-415 aa) (Fig. 3C, D). As shown in Fig. 3E, F, the deletion of the USP domain (212-558 aa) in USP21 or the domain of NTF2 (11-133 aa) in G3BP1 abolished their binding, which suggests the indispensable role of the two domains in their physical interaction.

Fig. 3: USP21 binds to, deubiquitinates, and stabilizes G3BP1.

A Co-immunoprecipitation (Co-IP) assays were applied to validate the interaction between endogenous USP21 and G3BP1 proteins in KYSE-150 and Eca-109 cells using antibodies against G3BP1 or USP21. B HEK293T cells were transfected with Flag-USP21WT- and HA-G3BP1WT-expressing plasmids. Co-IP assays with Flag or HA antibody were used to determine the binding of ectopic USP21 and G3BP1. C, D Schematic diagram displaying wild type or truncated mutation of USP21 and G3BP1 proteins. Co-IP assays with HA or Flag antibody were applied to determine the bindings between truncated USP21 and G3BP1WT (E) or between USP21WT and truncated G3BP1 proteins (F) followed by immunoblotting with indicated antibodies. Western blot analysis was performed to detect G3BP1 protein levels in KYSE-150 cells with USP21 depletion (G) or overexpression (H). I The nucleus and cytoplasmic proteins were extracted separately for western blot with G3BP1 protein in KYSE-150 cells with ectopic USP21WT expression. GAPDH was used as an internal control for cytoplasmic lysate while Lamin B1 was applied for the loading control of nucleus proteins. KYSE-150 cells were transfected with si-NC or si-USP21#1 followed by treatment with Vehicle, MG132 (20 μM) (J), or chloroquine (50 μM) (K) for 24 h. Western blot analysis was used for the detection of G3BP1 protein levels. L, M KYSE-150 cells with stable expression of sh-NC or sh-USP21 were treated with 50 μg/mL of Cycloheximide (CHX), a protein synthesis inhibitor, for 0, 8, 16, 24, and 32 h. Western blot analysis (L) and quantification of protein half-life (M) for G3BP1 protein as shown. N HEK293T cells were transfected with G3BP1WT-, Vector/USP21WT/USP21C221A-, and Ub-expressing plasmids followed by MG132 (20 μM) treatment for 6 h. Co-IP assays were applied to detect G3BP1 ubiquitination.

To determine USP21-mediated G3BP1 expression levels, we knocked down USP21 expression by applying siRNA in KYSE-150 and Eca-109, in which USP21 depletion downregulated the expression of G3BP1 protein but had no effect on G3BP1 mRNA levels (Fig. 3G, Fig. S6A, B). Furthermore, we explored how ectopic USP21WT- or USP21C221A-expression mediates G3BP1 protein levels. As expected, ectopic USP21WT significantly increased G3BP1 protein expression compared with vector control, whereas USP21C221A overexpression failed to upregulate the protein levels of G3BP1 (Fig. 3H, Fig. S6C). Neither USP21WT nor USP21C221A expression altered G3BP1 mRNA levels (Fig. S6D). In addition, we found that USP21-mediated upregulation of G3BP1 predominantly occurred in the cytoplasm (Fig. 3I, Fig. S6E).

Interestingly, reduced G3BP1 protein induced by USP21 depletion was abolished by the treatment of MG132, a proteasome inhibitor, but not chloroquine, an autophagy inhibitor (Fig. 3J, K, Fig. S7A, B). These results indicate that USP21 mediates G3BP1 protein levels via proteasome-dependent degradation. Moreover, a time-dependent decrease of G3BP1 protein was found in KYSE-150 cells treated with cycloheximide (CHX), a protein synthesis inhibitor, however, USP21 knockdown markedly shortened the half-life of G3BP1 protein (Fig. 3L, M). To further verify the effect of USP21 deubiquitinase activity on G3BP1 protein levels, the plasmids of His-Ub and HA-G3BP1WT combined with Flag-USP21WT- or Flag-USP21C221A-expressing plasmids were transfected into HEK293T cells, followed by MG132 treatment, in which USP21WT, but not USP21C221A, decreased the ubiquitination of G3BP1 protein compared with vector control (Fig. 3N). Collectively, our results strongly demonstrate that USP21 enhances the stabilization of G3BP1 protein by binding to and deubiquitinating it.

G3BP1 is essential for USP21-mediated ESCC progression

Considering that USP21 binds to and stabilizes G3BP1, we speculated that USP21 functions as an oncogene in ESCC via G3BP1. To demonstrate our hypothesis, KYSE-150 and Eca-109 cells were co-transfected with si-NC or si-G3BP1 combined with vector or USP21WT-expressing plasmids. The efficiency of ectopic USP21 expression and endogenous G3BP1 knockdown was verified in Fig. 4A and Fig. S8A. Plate formation and transwell assays revealed that USP21WT-induced enhancement of cell growth, migration, and invasion in KYSE-150 and Eca-109 cells was obviously abolished by G3BP1 depletion (Fig. 4B–D, Fig. S8B–D). To further identify the function of G3BP1 in USP21-mediated progression of ESCC, we transfected G3BP1-expressing plasmids to KYSE-150 and Eca-109 cells with siRNA-regulated USP21 knockdown, as shown in Fig. 4E and Fig. S8E. Cell function experiments in vitro demonstrated that weakened abilities of cell proliferation, migration, and invasion in KYSE-150 and Eca-109 cells with USP21 depletion were significantly restored by ectopic overexpression of G3BP1 (Fig. 4F–H, Fig. S8F–H). Additionally, IHC for mice xenograft sections revealed that G3BP1 protein levels in xenograft tumors were significantly decreased in USP21-depleted group (Fig. 4I). Consistently, G3BP1 protein was remarkedly increased in subcutaneous tumors derived from USP21WT-expressing, but not USP21C221A-expressing, KYSE-150 cells compared with the control group (Fig. 4J). Taken together, our results demonstrate that USP21 plays an oncogenic role in ESCC progression through upregulating G3BP1.

Fig. 4: G3BP1 is essential for USP21-mediated ESCC progression.

A Western blot analysis was conducted to determine indicated protein expression in Vector- or USP21WT-expressing KYSE-150 cells transfected with si-NC or si-G3BP1. Plate colony formation (B) and transwell assays (C, D) were performed to assess the growth, migration, and invasion of KYSE-150 cells expressing Vector + si-NC, USP21WT + si-NC, Vector + si-G3BP1, or USP21WT + si-G3BP1. Representative images and statistical quantification as indicated. E Western blot analysis was performed to detect indicated protein levels in Vector- or G3BP1WT-expressing KYSE-150 cells with transfection of si-NC or si-USP21#1. Plate colony formation (F) and transwell assays (G, H) were used to evaluate the growth, migration, and invasion of KYSE-150 cells expressing si-NC + Vector, si-NC + G3BP1WT, si-USP21#1 + Vector, or si-USP21#1 + G3BP1WT. Representative images and statistical quantification as shown. IHC analysis was performed to evaluate G3BP1 protein levels in the sections of mice xenograft tumor from the group of sh-NC/sh-USP21 (I) or Vector/USP21WT/USP21C221A (J). Scale bars (red line) are 100 μm in (C, D, G, H). The data for statistical analysis are presented as means ± SD and the unpaired t-test was applied to determine the statistical significance (B–D, F–H). The P values are shown in the corresponding panels, respectively.

USP21 activates Wnt/β-Catenin signaling to accelerate ESCC progression in a G3BP1-dependent manner

G3BP1 has been reported to regulate the activity of the Wnt/β-Catenin signaling pathway in esophageal cancer, colon cancer, and breast cancer [33,34,35]. In the present study, we found that knocking-down USP21 remarkedly reduced the expression level of β-Catenin, the key effector of Wnt/β-Catenin signaling, (Fig. 5A, Fig. S9A) while overexpression of USP21WT, but not USP21C221A, significantly upregulated β-Catenin levels (Fig. 5B, Fig. S9B). Given that increased accumulation of β-Catenin in the nucleus is a key indicator for the activation of Wnt signaling, we separately extracted the cytoplasmic and nuclear proteins from KYSE-150 transfected with si-NC, si-USP21#1 or si-USP21#2 or expressing Vector, USP21WT, or USP21C221A. Our results exhibited that the knockdown of USP21 obviously reduced nuclear accumulation of β-Catenin, while USP21WT, but not USP21C221A, increased β-Catenin protein levels in the cell nucleus (Fig. S9C, D). Furthermore, a dual-luciferase reporter assay revealed that the activity of luciferase from TOP plasmids was remarkedly inhibited by USP21 depletion but enhanced by the overexpression of USP21WT, but not USP21C221A, however, both USP21 knockdown and overexpression had no obvious effect on luciferase activity from FOP plasmids (Fig. 5C, D). These results solidly demonstrate the positive regulation of USP21 on Wnt/β-Catenin signaling. To figure out whether USP21-mediated activation of Wnt/β-Catenin signaling depends on G3BP1, we knocked down G3BP1 in USP21WT-expressing KYSE-150 cells and found that increased expression of β-Catenin induced by USP21 overexpression was obviously diminished by G3BP1 knockdown (Fig. 5E, S9E). Similarly, we overexpressed G3BP1 in USP21-depleted KYSE-150 cells and demonstrated that β-Catenin downregulation caused by USP21 depletion was significantly restored by ectopic G3BP1WT expression (Fig. S9F). These results indicate that USP21 mediates the Wnt/β-Catenin signaling pathway in a G3BP1-dependent manner. Then, we asked whether USP21 promotes ESCC malignant biological behaviors via activating the Wnt/β-Catenin pathway. As shown in Fig. 5F and Fig. S9G, treatment with IWR-1, an inhibitor for Wnt/β-Catenin signaling, significantly decreased β-Catenin levels in USP21WT-expressing KYSE-150 cells. Plate colony formation and transwell assays revealed that enhanced proliferation, migration, and invasion of KYSE-150 cells with USP21WT overexpression were significantly diminished by IWR-1 treatment (Fig. 5G–I). On the other hand, we found that laduviglusib, an activator for Wnt/β-Catenin signaling, obviously reversed the downregulation of β-Catenin protein (Fig. 5J, Fig. S9H) and rescued the weakened malignant phenotypes in KYSE-150 with USP21 depletion (Fig. 5K–M). Besides the experiments about the chemical modulator of the Wnt pathway, we introduced β-Catenin-expressing plasmid to genetically prove the role of the Wnt pathway in USP21-mediated ESCC progression. Our results revealed that ectopic β-Catenin expression significantly abolished si-USP21#1-induced inhibition of proliferation, migration, and invasion of KYSE-150 cells (Fig. S10A–D). In conclusion, our results demonstrate that USP21 exercises oncogenic effects in ESCC progression through regulating Wnt/β-Catenin.

Fig. 5: USP21 activates Wnt/β-Catenin signaling to accelerate ESCC progression in a G3BP1-dependent manner.

Western blot analysis was used to detect β-Catenin protein levels in KYSE-150 cells transfected with si-NC, si-USP21#1, or si-USP21#2 (A) or transfected with Vector-, USP21WT-, or USP21C221A-expressing plasmids (B). Dual-luciferase reporter assays were used to measure the luciferase activity of TOP and FOP plasmids in KYSE-150 cells with stable USP21 knockdown (C) or overexpression (D). The pRL-TK plasmids were co-transfected as an internal control. E Western blot analysis was conducted to detect β-Catenin protein levels in KYSE-150 cells transfected with Vector + si-NC, USP21WT + si-NC, Vector + si-G3BP1, or USP21WT + si-G3BP1. F–I KYSE-150 cells expressing Vector or USP21WT were treated with vehicle or 10 μM of IWR-1, an inhibitor for Wnt signaling, for 48 h. Western blot analysis for β-Catenin protein levels (F), plate colony formation for cell growth (G), and transwell assays for cell migration and invasion (H, I). J–M KYSE-150 cells were transfected with si-NC or si-USP21 followed by treatment with laduviglusib (10 μM), a potent activator for Wnt signaling, or vehicle for 48 h. Western blot to detect β-Catenin protein (J), plate colony formation to evaluate cell growth (K), and transwell assays to measure cell migration and invasion (L, M). Representative stainings and quantification as indicated. Scale bars (red line) are 100 μm in (H, I, L, M). Statistical significance was identified with an unpaired t-test, and all analyzed data are shown as means ± SD (C, D, G–I, K–M). The P values are respectively indicated in the corresponding position.

DSF inhibits USP21-drived ESCC progression

The anti-tumor effect of DSF has been demonstrated in many human cancers, such as breast cancer, lung cancer, pancreatic cancer, glioblastoma, melanoma, and esophageal cancer [36, 37]. Interestingly, it has been reported that DSF functions as a potent inhibitor against deubiquitinase activity of USP21 [38]. This prompted us to investigate whether DSF suppresses USP21-mediated G3BP1 upregulation, activation of Wnt/β-Catenin signaling, and ESCC malignant progression. We treated USP21WT-expressing KYSE-150 cells with DSF and found that USP21-induced upregulation of G3BP1 and β-Catenin protein was obviously diminished by DSF treatment according to Western blot analysis and its quantitation (Fig. 6A–C). Plate colony formation and transwell assays demonstrated that USP21-mediated enhancement of cell proliferation, migration, and invasion was remarkedly abolished in KYSE-150 cells when treated with DSF (Fig. 6D–F). Furthermore, in vivo experiments revealed that USP21WT overexpression accelerated subcutaneous tumor growth of KYSE-150 cells in mice xenograft models, however, DSF administration significantly abolished USP21WT-induced tumor proliferation (Fig. 6G–I). Consistently, IHC for xenograft tumor section indicated that the protein levels of Ki-67 and G3BP1 in USP21WT-expressing group were obviously upregulated compared with those in vector control, while USP21WT-induced increase of these proteins was greatly diminished by DSF treatment (Fig. 6J). Similarly, in experimental lung metastasis models, USP21WT-expressing KYSE-150 cells formed more pulmonary nodules than vector control, while DSF significantly decreased the number of USP21WT-induced nodules in mice lung (Fig. 6K–M). Together, DSF dramatically impaired USP21-induced activation of the G3BP1/Wnt/β-Catenin signaling axis and inhibited USP21-driving ESCC progression.

Fig. 6: DSF inhibits the USP21-drived ESCC progression.

A–F The vector- or USP21WT-expressing KYSE-150 cells were treated with 5 μM of DSF or vehicle control. Western blot analysis to detect G3BP1 and β-Catenin protein levels (A), quantitation for western blots of G3BP1 and β-Catenin from three replicate experiments as shown (B, C), plate colony formation to assess cell growth (D), and transwell assays to evaluate cell migration and invasion (E, F). Scale bars (red line) indicate 100 μm in (E, F). G–I BALB/c-nude mice (n = 7) were subcutaneously xenografted with vector- or USP21WT-expressing KYSE-150 cells for 16 days and treated with an intraperitoneal injection of DSF (200 mg/kg) or vehicle control every 2 days at the initiation of the model establishment. Tumor growth curve (G), the representative image of xenograft tumors (H), and tumor weights quantification (I) as indicated. J The protein levels of G3BP1 and Ki-67 in the sections of xenografts from (H) were evaluated using IHC. Scale bars as indicated. K–M KYSE-150 cells with vector or USP21WT expression were injected into BALB/c-nude mice (n = 5) via tail vein for 6 weeks. At the initiation of establishing metastatic lung xenograft model, mice were treated with DSF (200 mg/kg) or vehicle control through an intraperitoneal injection every 2 days. Representative images for pulmonary metastatic nodules (K), corresponding H&E staining with the scale bars (L), and statistical analysis (M) as shown. All data for statistical analysis are presented as means ± SD. Statistical significance of differences was identified using an unpaired t-test for (B–F, I, M) or the two-way ANOVA for (G), and the P values are displayed correspondingly.

The clinical significance of the USP21/G3BP1 axis in ESCC patients

To explore the potential function of the USP21/G3BP1 axis in ESCC patients, we continued to evaluate G3BP1 protein levels in 86 pairs of ESCC tissues. IHC scores exhibited that G3BP1 protein had an obvious increase in ESCC tumor samples compared to that in normal esophageal mucosa (Fig. 7A, B). ROC analysis suggested that G3BP1 protein levels have the potential ability to distinguish normal and ESCC tissues (Fig. S11A). Similar to USP21, higher protein levels of G3BP1 were found in ESCC tissues with T3-4 or N1-3 when compared to tumor samples with T1-2 (Fig. S11B) or N0 (Fig. S11C). ESCC samples with more advanced stages (III or IV) had higher G3BP1 protein levels than those with stage I or II (Fig. S11D). ESCC tissues in Stage II also exhibited higher levels of G3BP1 protein than those in Stage I (Fig. S11D). In addition, ESCC samples with poor differentiation (G3) showed upregulated G3BP1 protein levels compared to tumor tissues with well/moderate differentiation (G1-2) (Fig. S11E). Nonetheless, there was no significant correlation between other clinical factors (age and gender) and G3BP1 protein expression (Fig. S11F, G). Additionally, survival analyses revealed a strong correlation between G3BP1 protein levels and OS/PFS (Fig. S11H, I). More importantly, spearman correlation analysis showed that increased USP21 protein levels were positively correlated to upregulated protein expression of G3BP1 (Fig. 7C). ESCC patients with simultaneously high protein levels of USP21 and G3BP1 showed a worse OS and PFS than those with low expression of USP21 and G3BP1 (Fig. 7D, E), with low USP21 and high G3BP1 expression (Fig. 7F, G), or with high USP21 and low G3BP1 expression (Fig. 7H, I). Together, these results suggest aberrant activation of the USP21/G3BP1 axis in ESCC, which indicates an unfavorable clinical outcome.

Fig. 7: The clinical significance of the USP21/G3BP1 axis in ESCC patients.

A, B G3BP1 protein expression was determined using IHC in the ESCC samples (86 pairs). Representative IHC staining of G3BP1 with scale bars (A) and quantification for IHC scores (B) as displayed. C Spearman correlation analysis was performed to evaluate the association between USP21 and G3BP1 protein levels in 86 pairs of ESCC tumor tissues. Kaplan-Meier curves with log-rank test were conducted to evaluate OS and PFS rates between ESCC patients with simultaneously high USP21 and G3BP1 expression and those with low protein levels of USP21 and G3BP1 (D, E), those with low USP2