Fig 2: PTK7 was the most significantly deregulated kinase in metastatic HCC and was enriched in advanced HCC tumors. (A) A total of 509 commonly deregulated kinases in cancer were extracted from the nCounter Human Kinase Panel Gene List, and their expression pattern was examined in the TCGA-LIHC patient cohort. Thirty-five candidate kinases were overexpressed in tumor and advanced HCC, and ranked (from top to bottom) according to their extent of up-regulation and evaluated against the GSE45114 data set, which described transcriptomic profiles of nontumor and tumor tissues harvested from metastatic HCC samples and their nonmetastatic counterparts. The heatmap was constructed to represent the fold-change in the expression of each candidate kinase in HCC tumors compared with nontumor samples under a metastatic or nonmetastatic background. Only 3 kinases (TTK, PTK7, and CDC7, marked with an asterisk) were significantly overexpressed in metastatic HCC, and PTK7 was the most significantly deregulated. (B) Statistical evaluation of the fold change in PTK7 expression (tumor vs nontumor) for 228 pairs of HCC samples evenly grouped by their predicted risk of metastasis as described in the GSE14520 cohort by paired t test between each tumor/nontumor comparison. The predicted risk of metastasis was defined by the integral signature score of 161 metastatic genes that robustly correlated with recurrence and adverse clinical outcome.25 (C) The distribution of early and advanced HCC cases in the high-PTK7 and low-PTK7 samples of the TCGA-LIHC data set compared using the chi-squared test. (D) The overall survival distribution of patients with high or low PTK7 expression in the TCGA-LIHC data set compared using the Gehan–Breslow–Wilcoxon test. (E) Representative images of high-PTK7 staining from IHC on TMA containing tumor and nontumor samples of HCC patients. Scale bar: 100 µm. (F) The distribution of high- and low-PTK7 cases in tumor and nontumor samples of TMA compared using the chi-squared test. (G) The overall survival distribution of TMA patients with high or low PTK7 expression compared using the log-rank test. P values (*P < .05, **P < .01, ***P < .001).

Fig 3: The expression level and transcriptomic activity of SOX9 are clinically and functionally correlated with PTK7. (A) A total of 133 TFs potentially binding to the promoter region of PTK7, proximal to the transcriptional start site (between -1000 and +1), were compiled from the online Gene Transcription Regulation Database. Fourteen candidate TFs that highly correlated with PTK7 (r > 0.3, determined by Pearson correlation) and were overexpressed in tumor HCC at the messenger RNA (mRNA) level (analyzed using TCGA-LIHC data as a normalized count) were examined further for their specific enrichment in metastatic HCC using GSE45114 data. Only SOX9, nuclear respiratory factor 1 (NRF1), and TEA domain transcription factor (TEAD4) satisfied all of the tested criteria. The correlation of SOX9 and PTK7 expression at the mRNA level is shown for reference. (B) IHC was performed on TMA to investigate the extent of SOX9 nuclear localization in tumor HCC samples in which cases were designated as SOX9 high (>75% nuclei), medium (25%–75% nuclei), or low (<25% nuclei), depending on the amount of tumor cells showing apparent SOX9 nuclear staining. Graded expression of SOX9 was compared with PTK7 in the same TMA. Representative images of IHC are shown. Scale bar: 100 µm. (C) The distribution of cases with various nuclear SOX9 expression in the high-PTK7 and low-PTK7 samples of TMA was compared using the chi-squared test. (D) Two truncated versions of the PTK7 promoter region, one containing the predicted SOX9 binding site (T1) and one without (T2), were expressed in HCC cell lines via transient transfection. The ability of the promoter elements to drive protein expression was quantified using the Dual-Glo luciferase reporter assay and compared using the Student t test. (E) A ChIP assay was performed in HCC cell lines to examine the physical binding of SOX9 to the PTK7 promoter region. Targeted enrichment of the predicted binding sequence was quantified using reverse-transcription quantitative PCR and compared with mouse IgG immunoprecipitation control (IgG). (F) The messenger RNA (mRNA) levels of PTK7 and SOX9 were quantified and validated in HCC cells after successful knockdown of SOX9 by short hairpin RNA (shRNA). (G) Western blot validation of PTK7 suppression after SOX9 depletion by shRNA knockdown. (H) Two truncated PTK7 promoter elements, one harboring the predicted SOX9-binding motif (T1) and the other lacking the same motif (T2), were transfected into SOX9-knockdown (KD) cells and nontarget control (NTC) cells. The resultant luciferase signal induced by the 2 elements in the different cell lines was compared using the Student t test. P values (*P < .05, **P < .01, ***P < .001).

Fig 4: The enrichment of PTK7 in HCC is responsive to the activation of TGF-ß signaling and potentially mediated by stimulated SOX9 activity. (A) Patients in the TCGA-LIHC cohort were ranked according to the transcriptomic level of PTK7 in their tumors. Publicly available RNA sequencing data of the 25% highest-ranking patients were subjected to GSEA to identify molecular signatures and pathways deregulated under enriched PTK7 conditions compared with the 25% lowest-ranking patients. TGF-ß signaling, which is highly relevant to metastasis, was identified as one of the top pathways significantly enriched under high-PTK7 conditions. The top 10 deregulated pathways under the high-PTK7 condition are arranged in a bubble diagram for reference. (B) Heatmap showing the correlation of the tumor PTK7 level with 18 TGF-ß signaling signature genes in all of the paired tumor/nontumor (T/NT) cases of the TCGA-LIHC cohort. The T/NT fold change of all signature genes was calculated, and their activation status for each HCC case was based on a numeric scoring system, where a score of 2 means the most activated and a score of -2 is the most inactivated. A TGF-ß score was assigned to each patient by integrating the scores of all 18 signature genes. An overall zero or negative TGF-ß score was representative of a TGF-ß–suppressive environment, and a positive overall score was representative of a TGF-ß–activated environment. PTK7 expression was compared between HCC samples with a positive or negative TGF-ß score using the Student t test. (C) Western blot validation of dose-dependent up-regulation of PTK7 and SOX9 expression in PLC/PRF/5 cells after a single treatment of recombinant TGF-ß1 for 24 hours at various doses. Quantification of PTK7 and SOX9 expression was performed in ImageJ (National Institutes of Health, Bethesda, MD). (D) Western blot validation of PTK7 and SOX9 expression in PLC/PRF/5 cells after TGF-ß1 single treatment or in combination with a TGF-ßR1–specific inhibitor, LY-364947, for 24 hours. (E) Representative images of immunocytochemistry targeting SOX9 in PLC/PRF/5 cells after TGF-ß1 single treatment or in combination with LY-364947 (Combo) for 24 hours. 4',6-Diamidino-2-phenylindole (DAPI) staining was used to locate the nuclei of stained cells. The average nuclear localization of SOX9 was quantified in ImageJ and compared between treatment groups using the Student t test. (F) Western blot validation of PTK7 levels in SOX9-knockdown PLC/PRF/5 cells after TGF-ß1 single treatment. DMSO, dimethyl sulfoxide; DN, down-regulated; FDR, false discovery rate; IF, immunofluorescence; mRNA, messenger RNA; NTC, nontarget control. P values (*P < .05, **P < .01, ***P < .001).

Fig 5: In vitro analyses of TGF-ß1–treated HCC cells. (A) PTK7 correlation with 6 exhausted T-cell signature markers (HAVCR2, TIGIT, LAG3, PDCD1, CXCL13, and LAYN) was examined by Pearson correlation using TCGA-LIHC data set and aided by the publicly available GEPIA2 analysis platform with all expression in terms of transcript per million. (B) Statistical comparison of the expression of TGFB1/2/3 in metastatic and nonmetastatic HCC by GSE45114 and the Student t test. (C) Same statistical evaluation of the GSE14520 data set as described in Figure 1B for the fold change in TGFB1 expression of paired tumor/nontumor HCC samples with different predicted risk of metastasis, by paired t test. (D) Representative images and the quantification of migrating cells by Transwell assay after TGF-ß1 single treatment of 24 hours’ duration. Statistical comparisons by the Student t test. (E) Western blot validation of dosage-dependent up-regulation of PTK7 and SOX9 expression in Hep3B cells after a single treatment of recombinant TGF-ß1 for a duration of 24 hours at various dosages. Quantification of PTK7 and SOX9 expression was performed by ImageJ. (F) Western blot validation of PTK7 and SOX9 expression in Hep3B cells after TGF-ß1 single treatment or in combination with LY-364947, a TGF-ßR1–specific inhibitor, for a duration of 24 hours. (G) Representative images of immunocytochemistry targeting SOX9 in Hep3B cells after TGF-ß1 single treatment or in combination with LY-364947 (Combo) for a duration of 24 hours. 4',6-Diamidino-2-phenylindole (DAPI) staining was used to locate the nuclei of stained cells. The average nuclear localization of the SOX9 was quantified by ImageJ and compared between treatment groups by the Student t test. DMSO, dimethyl sulfoxide. P values (*P < .05, **P < .01, ***P < .001).