The EnVision + HPR/DAB kit was purchased from Shanghai Gene Technology Co., Ltd. (Shanghai, China), and the cytokine CXCL9 from PeproTech Inc. (Rocky Hill, NJ, USA). CXCR3 neutralizing antibodies (2.0 µg/ml; mouse, monoclonal; cat. no. 49801) and IgG (2.0 µg/ml; mouse; cat. no. A7028) were purchased from R&D Systems (Minneapolis, MN, USA) and Beyotime Biotechnology Co., Ltd. (Shanghai, China), respectively. Anti-E-cadherin (1:200; rabbit, polyclonal; cat. no. ab53226) and anti-vimentin (1:100; rabbit, monoclonal; cat. no. ab16700) were both purchased from Abcam (Cambridge, UK). The phalloidin-labeled cytoskeleton staining kit was purchased from Sigma-Aldrich (St. Louis, MO, USA), the CCK-8 kit from Dojindo Molecular Technologies, Inc. (Kumamoto, Japan), the plasmid extraction kit from Qiagen GmbH (Hilden, Germany), the high efficiency transfection (HET) kit from Shenzhen Biowit Technologies Co., Ltd. (Shenzhen, China), and the RNA reverse transcription kit from Thermo Fisher Scientific (Waltham, MA, USA).

Ten normal tongue mucosa specimens were used as negative controls. TSCC specimens (provided by Professor Tiejun Li from Peking University Stomatological Hospital) with pathologically definitive diagnosis were collected from 51 cases without preoperative treatment, averaging 22 years old, and including 30 males and 21 females. Among them, 28 cases had lymph node (LN) metastasis while the remaining 23 had no LN metastasis. CXCL9 and CXCR3 expression levels were detected by Envision immunohistochemistry and intracellular brown particles reflected positive signals. The semi-quantitative additive score method was used to quantify positive staining by counting cells in 10 high power fields randomly in every slide. Specific criteria were: i) score based on the percentage of positive cells (0, ≤5%; 1, 6–25%; 2, 26–50%; 3, 51–75%; 4, >75%); ii) score based on signal intensity in most cells (1, slight yellow; 2, yellow; 3, brown). The product of the score based on the percentage of positive cells and the signal intensity score were used as the final score: negative, score of 0 (−); slightly positive, score of 1–4 (+); positive, score of 5–8 (++); strongly positive, score of 9–12 (+++). Double blinded quantification of the slides by 2 people was applied and a difference of 3 points between values required re-examination.

The human TSCC cell line Cal-27 (ATCC; Manassas, VA, USA) was cultured in high-glucose DMEM with 10% FBS, at 37°C and 5% CO₂. Cells at the logarithmic growth phase were used for experiments.

The hCXCR3 expression vector pLVX-hCXCR3-mCMV-ZsGreen-PGK-Puro was constructed by DNA synthesis (1107 bp) with the following primers pLVX-hCXCR3-EcoRI-F, 5′-CCGGAATTCGCCACCATGGTCCTTGAGGTGAGTGACCACCAAG-3′; pLVX-hCXCR3-BamHI-R, 5′-CGCGGATCCTCACAAGCCCGAGTAGGAGGCCTCTGAG-3′. Prior to large volume plasmid extraction, the vector was characterized by enzymatic digestion and sequencing. The lentivirus was packaged by infecting 293T cells and concentrated. Following purification, Cal-27 cells were infected with the lentivirus under selection by puromycin.

The shCXCR3 (pLVX-ShRNA-puro-hCXCR3) expression vector was constructed by DNA synthesis with the following primers: hCXCR3-F, 5′-GATCCGCCTACTGCTATGCCCACATCTTCAAGAGAGATGTGGGCATAGCAGTAGGCTTTTTTCTCGAGG-3′; hCXCR3-R, 5′-AATTCCTCGAGAAAAAAGCCTACTGCTATGCCCACATCTCTCTTGAAGATGTGGGCATAGCAGTAGGCG-3′.

Cells at the log-phase were trypsinized, seeded in 96-well plates at a density of 1.2×10⁴ cells/ml in 100 µl, and cultured overnight at 37°C. Then, the culture medium was aspirated, and supplemented with 100 µl of CXCL9 protein, CXCR3 neutralizing antibodies (2.0 µg/ml), and IgG (2.0 µg/ml) for 24 h, respectively, followed by the addition of 10 µl of CCK-8 reagent for 3 h. The absorbance at 450 nm was assessed on a microplate reader. Each experiment was repeated 3 times.

Cells at the log-phase were seeded in 6-well plates at a density of 5×10⁵ cells/well. At a confluency of 95%, 0.5 ml of CXCL9, CXCR3 neutralizing antibodies (2.0 µg/ml) and IgG (2.0 µg/ml) were added for 3 h. Then, the cell monolayer was scratched with a 200-µl pipette tip. Detached or dead cells were washed out with PBS, and serum-free culture media were added. A total of 3 fields/group were photographed under a fluorescence microscope at 12, 24, and 48 h, respectively. Cell migration rates were calculated with ImageJ software 6.0 and each experiment was repeated three times.

Transwell chambers were used for cell invasion assays. The upper chambers were seeded with 0.5 ml of Cal-27 cells in high-glucose complete medium, with the lower chambers filled with 0.7 ml of high-glucose complete culture medium. For each group, three different interventions were applied: i) the lower chamber was supplemented with 100 nM CXCL9; ii) the upper chamber was supplemented with 2 µg/ml CXCR3 neutralizing antibodies in combination with 100 nM CXCL9 protein in the lower chamber; and iii) the upper chamber was supplemented with 2 µg/ml IgG combined with 100 nM CXCL9 in the lower chamber. After 24 h, the membranes were separated, stained, sealed, and assessed for invasive cells under a microscope. Every experiment was repeated 3 times.

Cells under routine culture were supplemented with 1 ml of the CXCL9 protein, CXCR3 neutralizing antibodies (2.0 µg/ml), and IgG (2.0 µg/ml), followed by 4% paraformaldehyde fixation for 48 h. Then, 0.5 ml of 40 µg/ml phalloidin-rhodamine solution was added for 60 min in the dark in a humid box at room temperature, and sealed before analysis by fluorescence microscopy. Edge aggregation of F-actin was assessed in Cal-27 cells.

Cultured cells were fixed in paraformaldehyde at 4°C, permeabilized with 0.1% Triton X-100, blocked with 10% goat serum, and incubated with anti-E-cadherin (1:200; rabbit, polyclonal; cat. no. ab53226) or anti-vimentin (1:100; rabbit, monoclonal; cat. no. ab16700, both from Abcam) primary antibodies overnight at 4°C in the dark, in a humid box. This was followed by incubation with fluorescence-labeled secondary antibody (2 µg/ml, goat anti-rabbit IgG (H+L); antibody, cat. no. A11012; Thermo Fisher Scientific) in a shaker incubator. Counterstaining was performed with 100 µg/ml DAPI before analysis by fluorescence microscopy.

Cal-27 cells were lysed on ice with lysis buffer. Proteins in whole cell lysates were separated by SDS-PAGE, electrically transferred onto Immobilon-P membranes, incubated with primary antibodies anti-Akt2 (1:500; rabbit, polyclonal; cat. no. ab8805), anti-p-Akt (1:500; rabbit, polyclonal; cat. no. ab38513), anti-eIF4E (1:500; rabbit, monoclonal; cat. no. ab33766), anti-p-eIF4E (1:1,000; rabbit, monoclonal; cat. no. ab76256; all from Abcam) overnight at 4°C, followed by 1 h of incubation with a horseradish peroxidase (HRP)-labeled secondary antibody (1:3,000; goat anti-rabbit IgG-HRP; cat. no. sc2004; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) at room temperature before chemiluminescence detection (Western Chemiluminescent HRP substrate; Millipore, Billerica, MA, USA).

The Wilcoxon non-parametric test was applied to assess intergroup differences for enumeration data. One-way ANOVA was used for measurement data; intergroup differences were analyzed using the LSD-t method. P<0.05 indicated a statistically significant difference.

In tongue mucosa epithelial tissue specimens from the 10 normal cases, no CXCL9-positive cells were detected, i.e., CXCL9 (−). In addition, CXCR3-positive cells were observed in individual specimens, and positive cells were limited and primarily located in the basal layer, i.e., CXCR3 (−) (Fig. 1A and B). Of the 51 TSCC specimens, 24 (47.06%) were CXCL9 (−), 10 (19.60%) CXCL9 (+), 10 (19.60%) CXCL9 (++), and 7 (13.73%) CXCL9 (+++). The 23 TSCC specimens without cervical LN metastasis were mostly CXCL9 (−)/(+), while the 20 out of 28 with cervical LN metastasis were CXCL9 (+) to (+++). Statistical analysis revealed significantly higher CXCL9 expression in TSCC with cervical LN metastasis compared with those without metastasis (P<0.01) as assessed by the Wilcoxon non-parametric test (Z= −2.68, p=0.0074) (Fig. 1C).

Of the 51 TSCC specimens, 10 (19.61%) were CXCR3 (−), 8 (15.69%) CXCR3 (+), 15 (29.41%) CXCR3 (++), and 18 (35.29%) CXCR3 (+++). The 23 TSCC without cervical LN metastasis were mostly CXCR3 (−) to (+), while the 28 with cervical LN metastasis were mostly CXCR3 (++) to (+++); a statistically significant difference (P<0.01) in CXCR3 expression in the LN metastasis group was reflected in the Wilcoxon nonparametric test (Z= −2.66, P=0.0079) (Fig. 1C).

The hCXCR3 and shCXCR3 expression vectors were constructed (Fig. 2A and B). Cal-27 cells were seeded in 96-well plates and incubated with CXCL9 at 0, 50, 100, and 150 nM, respectively, for 24 h, followed by 3 h of incubation with CCK-8 reagent at room temperature before absorbance assessment at 450 nm. Compared with the control group, no statistical difference (P>0.05) was detected in terms of cell proliferation (Fig. 2C).

CXCR3 neutralizing antibodies did not promote Cal-27 cell proliferation. Cells treated for 24 h with 100 nM CXCL9 and CXCR3 neutralizing antibodies (2.0 µg/ml) or IgG (2.0 µg/ml) exhibited no statistically significant differences (P>0.05) in proliferation (Fig. 2C).

CXCR3 overexpression did not affect Cal-27 cell proliferation. Cal-27-GFP and Cal-27-CXCR3 cells were incubated for 24 h with 100 nM CXCL9, followed by 3 h of incubation with CCK-8 reagent at room temperature before absorbance assessment at 450 nm. Compared with the control group, cell proliferation exhibited no statistically significant difference (P>0.05) (Fig. 2C).

Low CXCR3 expression was not related to Cal-27 cell proliferation. Cal-27-shRNA-scramble and Cal-27-shRNA-hCXCR3 transfected cells were treated for 24 h with 100 nM CXCL9, followed by 3 h of incubation with CCK-8 reagent at room temperature before absorbance assessment at 450 nm. Compared with the control group, cell proliferation exhibited no statistically significant difference (P>0.05) (Fig. 2C). The aforementioned data indicated that CXCL9 did not promote Cal-27 cell proliferation.

Cal-27 cells seeded in 6-well plates were cultured for 3 h with CXCL9 at 0, 50, 100 and 150 nM, respectively; then the cell monolayer was scratched. After 0, 12, 24 and 48 h of incubation, respectively, the samples were imaged, and migration areas were determined. Compared with the control group, a higher migration rate of Cal-27 cells was observed after treatment with various concentrations of CXCL9 (except for 0 and 50 nM after 12 h), with statistical significance (P<0.05) (Fig. 3A). Compared with the CXCL9 group, 100 nM CXCL9 administered in combination with CXCR3 neutralizing antibodies (2.0 µg/ml) and IgG (2.0 µg/ml) resulted in significantly decreased migration at 48 h, from 93.7±6.13% to 83.8±3.40% (P<0.05) (Fig. 3B).

CXCR3 overexpression enhanced CXCL9-stimulating effects on Cal-27 cell migration. Cal-27-GFP and Cal-27-CXCR3 cells were treated for 3 h with 100 nM CXCL9, followed by scratching. The samples were imaged at 0, 12, 24 and 48 h, respectively, to evaluate the migration areas. Compared with the control group, treatment resulted in significantly increased migration areas (P<0.05) (Fig. 3C).

Low CXCR3 expression attenuated the effect of CXCL9 on Cal-27 cell migration. Cal-27-shRNA-Scramble and Cal-27-shRNA-hCXCR3 cells were scratched and imaged at 0, 12, 24, and 48 h of culture, respectively. Compared with the control group, silencing of CXCR3 resulted in significantly reduced migration areas (P<0.05) (Fig. 3D). These results revealed that CXCL9 promoted Cal-27 migration, an effect closely related to its receptor CXCR3.

CXCR3 expressing Cal-27 cells seeded in invasion chambers were cultured for 24 h with 100 nM CXCL9 in combination with CXCR3 neutralizing antibodies (2.0 µg/ml) or IgG (2.0 µg/ml). After staining with 0.1% crystal violet, the specimens were imaged. Compared with the control group, a significant increase in the number of invasive cells was observed (P<0.05). Compared with the CXCL9 group, treatment with CXCL9 and CXCR3 neutralizing antibodies resulted in significantly reduced cell invasion (P<0.05) (Fig. 4A).

CXCR3 overexpression positively regulated the inducive effect of CXCL9 on invasion in Cal-27 cells. Cal-27-GFP and Cal-27-CXCR3 cells were seeded in invasion chambers and treated for 24 h with 100 nM CXCL9, followed by staining with 0.1% crystal violet. Compared with the control group, a significant increase in the number of invasive cells was observed (P<0.05) (Fig. 4B).

Low CXCR3 expression inhibited the inducive effect of CXCL9 on invasion in Cal-27 cells. Cal-27-shRNA-Scramble and Cal-27-shRNA-hCXCR3 cells were seeded in invasion chambers, and treated for 24 h with 100 nM CXCL9, followed by staining with 0.1% crystal violet and imaging. Compared with the control group, a significant decrease in the number of invasive cells was observed (P<0.05) (Fig. 4C). In summary, CXCL9 enhanced the invasive capability of Cal-27 cells, an effect closely related to its receptor CXCR3.

Routinely cultured Cal-27 cells were coated on slides, and incubated for 48 h with 100 nM CXCL9, CXCR3 neutralizing antibodies (2.0 µg/ml) and IgG (2.0 µg/ml), and incubated with phalloidin-rhodamine for 60 min before imaging. Compared with the control group, edge aggregation of F-actin, specifically binding to phalloidin-rhodamine, was evident. Compared with the CXCL9 group, addition of CXCR3 neutralizing antibodies led to significantly decreased edge aggregation of F-actin (Fig. 5A).

CXCR3 overexpression enhanced the inducive effect of CXCL9 on edge aggregation of F-actin. Cal-27-GFP and Cal-27-CXCR3 cells were coated on slides and cultured routinely for 48 h with CXCL9. After blocking with 1% BSA, incubation was performed with rhodamine-labeled phalloidin for 60 min before imaging. Compared with the Cal-27-GFP group, edge aggregation of F-actin was improved in Cal-27-CXCR3 cells (Fig. 5B).

Reduced CXCR3 expression attenuated the inducive effect of CXCL9 on edge aggregation of F-actin in Cal-27 cells. Cal-27-shRNA-Scramble and Cal-27-shRNA-hCXCR3 cells were seeded on slides and cultured routinely for 48 h with CXCL9. After blocking with 1% BSA, incubation was performed with 40 µg/ml rhodamine-labeled phalloidin for 60 min before imaging. Compared with the shRNA-Scramble group, edge aggregation of F-actin was reduced in Cal-27-shRNA-hCXCR3 cells (Fig. 5C). Overall, CXCL9 enhanced the edge aggregation of F-actin in Cal-27 cells, an effect closely related to its receptor.

Cal-27 cells from the test and control groups were coated on slides and cultured routinely. Both groups underwent the same process, except for 24-h treatment with 100 nM CXCL9 in the test group. Cells were fixed with paraformaldehyde, permeabilized for 20 min with 0.1% Triton X-100, blocked for 2 h in goat serum, and incubated with anti-E-cadherin or anti-vimentin primary antibodies, followed by rhodamine-labeled secondary antibodies and DAPI staining before imaging. Compared with the control group, E-cadherin expression in Cal-27 cells was significantly reduced in the CXCL9 group while vimentin levels were increased. Compared with the CXCL9 group, addition of CXCR3 neutralizing antibodies attenuated the reduction of E-cadherin expression as well as vimentin level increase. For western blotting, Cal-27 cells from the CXCL9 and control groups were lysed, and proteins in the resulting lysates were separated by SDS-PAGE and electrophoretically transferred onto Immobilon-P membranes. The membranes were then incubated with primary antibodies (anti-E-cadherin and anti-vimentin) overnight at 4°C, followed by 1 h of incubation with HRP-labeled secondary antibodies at room temperature before chemiluminescent detection. The results were consistent with the aforementioned immunofluorescence findings (Fig. 6A and D).

CXCR3 overexpression enhanced the inducive effect of CXCL9 on expression increase of vimentin and E-cadherin level decrease in Cal-27 cells. Cal-27-GFP and Cal-27-CXCR3 cells were treated for 24 h with 100 nM CXCL9, fixed with paraformaldehyde and permeabilized for 20 min with 0.1% Triton X-100. After blocking for 2 h in goat serum, the samples were sequentially incubated with primary antibodies (anti-E-cadherin or anti-vimentin) and rhodamine-labeled secondary antibodies, followed by DAPI counterstaining before imaging. Compared with the control group, E-cadherin expression in Cal-27 cells was reduced significantly whereas vimentin levels were enhanced. For western blotting, Cal-27-GFP and Cal-27-CXCR3 cells from the CXCL9 and control groups were lysed. Proteins in whole cell lysates were separated by SDS-PAGE, and electrophoretically transferred onto Immobilon-P membranes. The latter were incubated with primary antibodies (anti-E-cadherin and anti-vimentin) overnight at 4°C, followed by 1 h of incubation with HRP-labeled secondary antibodies at room temperature before chemiluminescent detection. The results were consistent with the immunofluorescence findings (Fig. 6B and D).

Silencing of CXCR3 attenuated the inducive effect of CXCL9 on E-cadherin level reduction and vimentin expression increase in Cal-27 cells. Cal-27-shRNA-Scramble and Cal-27-shRNA-hCXCR3 cells were treated for 24 h with 100 nM CXCL9. After sequential incubation with primary antibodies (anti-E-cadherin or anti-vimentin) and rhodamine-labeled secondary antibodies, the samples were treated with DAPI before imaging. Compared with the Scramble group, CXCR3 silencing did not result in E-cadherin reduction in Cal-27 cells but it did decrease the expression level of vimentin. For western blotting, Cal-27-shRNA-Scramble and Cal-27-shRNA-hCXCR3 cells of the CXCL9 and control groups were lysed. Proteins in whole cell lysates were then separated by SDS-PAGE and electrophoretically transferred onto Immobilon-P membranes. This was followed by sequential incubation with primary antibodies (anti-E-cadherin or vimentin) overnight at 4°C, and HRP-labeled secondary antibodies at room temperature 1 h before chemiluminescent detection. The results were consistent with the immunofluorescence findings (Fig. 6C and D). Hence, in Cal-27 cells, CXCL9 downregulated E-cadherin, an epithelial cell marker, while concomitantly upregulating vimentin, a mesenchymal marker, an activity closely related to its receptor CXCR3.

The CXCL9/CXCR3 axis activates Akt signaling in Cal-27 cells. To further explore how the CXCL9/CXCR3 axis induces EMT in Cal-27, we treated Cal-27 cells with 100 nM CXCL9, and whole cell total protein was extracted at 10, 30, and 60 min, respectively, for western blotting. The results indicated an induction of phosphorylated Akt2 and eIF4E by CXCL9, which peaked at 30 min, with statistically significant differences (P<0.05) (Fig. 7A). According to the aforementioned findings, the cells were incubated with 100 nM CXCL9, CXCR3 neutralizing antibodies (2.0 µg/ml) and IgG (2.0 µg/ml). Compared with the CXCL9 group, addition of CXCR3 neutralizing antibodies resulted in decreased relative amounts of p-Akt2 (from 6.28±0.39 to 5.01±0.07) and p-eIF4E (from 2.44±0.21 to 1.42±0.16), with statistically significant differences (P<0.05) (Fig. 7B).

Moreover, CXCR3 overexpression enhanced the inducive effect of CXCL9 on p-Akt2 and p-eIF4E in Cal-27 cells. Cal-27-GFP and Cal-27-CXCR3 cells were stimulated for 30 min with 100 nM CXCL9, and total cell protein was extracted for western blotting. Compared with the control group, relative amounts of p-Akt2 in Cal-27-CXCR3 cells were significantly increased, from 1.11±0.12 to 1.46±0.06, as well as p-eIF4E, from 0.73±0.11 to 1.11±0.09 (P<0.05) (Fig. 7C).

Silencing of CXCR3 attenuated the inducive effect of CXCL9 on p-Akt2 and p-eIF4E in Cal-27 cells. Cal-27-shRNA-Scramble and Cal-27-shRNA-hCXCR3 cells were stimulated for 30 min with 100 nM CXCL9, and total cell protein was extracted for western blotting. Relative amounts of p-Akt2 were decreased from 2.48±0.14 in the control group to 1.55±0.10 in the interference group, as well as relative p-eIF4E, from 0.32±0.03 to 0.12±0.02, with statistically significant intergroup differences (P<0.05) (Fig. 7D). In summary, the CXCL9/CXCR3 axis upregulated the phosphorylated forms of Akt2 and eIF4E in Cal-27 cells.