3.1. WNT ligands expression signatures are tumor-specific
To investigate whether different WNT ligands function uniquely across multiple cancer types, we first compared 19 WNT ligands mRNA expression profiling across distinctive tumors with matched paratumor tissues, as well as normal tissues. The tumor datasets were downloaded from Expression Atlas (https://www.ebi.ac.uk/gxa/experiments/E-MTAB-5200)13. The expression of WNTs showed a tissue-dependent and tumor-dependent pattern (Fig. 1A). According to their expression feature, WNTs could be divided into 3 categories, tumor positive, negative, and irrelevant. WNT3, WNT4, WNT9A, WNT2B, WNT6, WNT5B and WNT11 were tumor negative WNTs, which were abundant in normal tissues, while declining in many tumors including melanoma, sarcoma, breast adenocarcinoma and colorectal adenocarcinoma. Conversely, WNT5A, WNT7B, WNT7A and WNT10A are tumor-positive WNTs, which significantly increased in many carcinomas. The remnants were tumor irrelevant WNTs. A panel of WNTs was suggested to contribute to tissue-specific tumorigenicity. Fig. 1B listed several sub-clusters. Among them, WNT3 and WNT4 were implied as melanoma tumor suppressors; WNT7A and WNT10A displayed contributing to the female reproductive system adenocarcinomas. WNT3 and WNT16 specifically increased in chronic lymphocytic leukemia (CLL), indicating the specificity of WNTs in leukemia. And the opposite effects of WNT3 in solid tumors and leukemia accentuated the complexity of WNT signaling in tumors. Surprisingly, WNT5A and WNT7B were significantly upregulated in CSCC, HNSC, LUSC and BTCC, suggesting universal property of being postulated as pan-SCCs biomarkers.
3.2. WNT receptors and co-receptors exhibit histology-dependent tumor specificity
WNTs employ more than 15 receptors and co-receptors of different families. These receptors are simultaneously regulated by a variety of antagonists or agonists such as R-spondins3. Using similar analytical methods of WNT ligands, we examined the expression profiling of WNT receptors and co-receptors, as well as four syndecans, R-spondins interacting transmembrane proteoglycans, in the above tumors. Almost all tissues exhibited a high abundance of syndecans, especially for SDC1 and SDC4, indicating a weak influence on cancer development (Fig. 2A). As for WNT receptors and co-receptors, half of them were altered in multiple types of tumor (Fig. 2A). Among them, FZD4 and ROR2 were negatively correlated with most malignant tissues, while others contributed to tumors in a histology-dependent manner (Fig. 2A). LRP5 and LRP6 act as co-receptors in β-catenin-dependent signaling. LRP6 was negligible in all above tissues compared with LRP5 (Fig. 2A). However, LRP5 preferred adenocarcinomas of digestive organs, including the pancreas, colon, esophagus and stomach (Fig. 2B). SDC3 was specifically increased in melanoma and two brain tumors, glioma and glioblastoma multiforme. SDC2 and GPC3 were up-regulated in HCC. Intriguingly, similar to WNT ligands, the SCCs demonstrated the enrichment of a set of WNT receptors and co-receptors, including GPC1, FZD6, PTK7 and FZD7 (Fig. 2B).
Taken together, these results suggested that specified WNT ligands and receptors/co-receptors could be attributed to distinctive tumors according to their tissue-of-origin.
3.3. Validation of WNTs specificity in oral squamous carcinoma
Given the specific expression pattern of WNT components in SCCs, we initially focused on quantifying WNTs expression in OSCC. Our previous work has compared genome profiling of adjacent normal, premalignant and OSCC tissues collected from two recruited OSCC patients (GSE70666)14. Here, we revisited the WNT ligands expression between paired tumors and adjacent normal tissues. Our data annotated 17 WNTs and found that WNT2B, WNT10A, WNT5A and WNT7B showed a similar cancer-positive pattern, and WNT7B was the most significant one (Fig. 3A). Subsequently, we confirmed this result by qRT-PCR in 21 OSCC clinical samples and paired adjacent nontumorous tissues. Consistently, we detected increased mRNA expression of WNT7B (P = 0.0003), as well as WNT10A (P = 0.0023) and WNT5A (P = 0.088), in most cancer samples (Fig. 3B). However, WNT2B showed no statistically significant difference (Fig. 3B). Of note, WNT10A mRNA was statistically different between tumor and adjacent normal tissue, but its abundance was low in these samples. Furthermore, WNT5A expression exhibited obvious heterogeneity (Fig. 3B, bottom panel).
3.4. Validation of LUSC specific WNT ligands and receptor/co-receptor reveals a correlation between mortality and expression pattern
To further evaluate the universalism of altered WNT5A and WNT7B in a broad range of SCCs, we examined them by IHC in LUSC tissue chips (Fig. 4A). Simultaneously, we also detected the expression of two SCCs specific WNT receptor and co-receptor, FZD7 and GPC1, in the same samples (Fig. 4A). As shown in Fig. 4B, a statistically significant elevation of WNT7B, WNT5A, FZD7 and GPC1 was observed between 72 LUSC tumors and adjacent normal tissues, consistent with the aforementioned transcriptomic analysis (Fig. 1 and Fig. 2). Similar to OSCC, WNT5A expression in LUSC was also heterogeneous as the CV (coefficient of variation) of WNT5A IHC score was 0.79. Instead, for WNT7B, FZD7 and GPC1, the CV was 0.38, 0.48 and 0.47, respectively. While positive staining for WNT7B was observed obviously in the cytoplasm of the majority of malignant tissues, WNT5A, FZD7 and GPC1 signals were presented on both the cytoplasm and membrane of cancer cells (Fig. 4A).
Next, we examined whether these proteins expression impacts clinical output. Using IHC score in individual LUSC tumor tissue as a reference, we found that the expression profiling of WNT7B, WNT5A, FZD7 and GPC1 could be divided into four sub-clusters (Fig. 5A). We further compared the death rate within each sub-cluster. Sub-cluster 1 showed high expression of all four proteins and exhibited the highest death rate, 38.89% (Fig. 5A). In sub-cluster 2, all proteins were expressed in medium abundance, of which the death rate was 36.36% (Fig. 5A). The lowest death rate was 14.29%, which belonged to sub-cluster 3 characterized by a low abundance of all proteins (Fig. 5A). As for sub-cluster 4, it had highly expressed WNT7B, GPC1 and FZD7, except for WNT5A. And it showed a low death rate, 20% (Fig. 5A). These results showed that the combined expression of WNT7B, WNT5A, GPC1 and FZD7 contributed to LUSC malignancy, and the clinical output partly relied on the balance between WNT5A and the other proteins.
3.5. WNT7B is involved in the malignant development of oral inflammation and carcinoma
Generally, there existed a prolonged inflammation course in OSCC and LUSC development15-17. Moreover, WNT7B was suggested to participate in chronic inflammations18. We found a high expression of WNT7B in the above OSCC and LUSC patients as well (Fig. 1, Fig. 3 and Fig. 4). Therefore, we assumed that WNT7B may affect oral inflammation and cancer progression. We revisited the mRNA expression of WNT7B, WNT5A, and WNT10A in our prior data (GSE70666)14. Among these OSCC positive WNTs, only WNT7B was progressively increased in normal, OLP, and OSCC tissues in both patients (Fig. 6A). We further confirmed WNT7B expression using 34 FFPE samples by IHC. Compared with adjacent nontumorous tissues, WNT7B positive signals were gradually increased in OLP, then to OSCC (Fig. 6B). Moreover, positive signals predominantly existed in OLP pathological epithelium and OSCC cancer cells (Fig. 6B). Notably, a strong positive WNT7B signal also existed in macrophages (Fig. 6B). Thus, these results suggested the feasibility of using WNT7B as an indicator of the transition from oral inflammation to tumor progression.
3.6. WNT7B promote tumor invasion by its downstream genes including MMP1
WNT signaling initiates proliferation and EMTs in multiple carcinomas19. Considering the gradual up-regulation of WNT7B in OLP and OSCC, we sought to investigate the biological functions of WNT7B in the OSCC progression. We constructed WNT7B stable knockdown OSCC cell lines (SCC9 and FaDu) (Fig. 7A and 7B) and performed a proliferation assay and cell cycle detection to evaluate the contribution of WNT7B to cell growth. Results showed that the knockdown of WNT7B did not inhibit the proliferation of tumor cells (Supplementary Fig. 1A). However, WNT7B was suggested to improve the tolerance of cells to nutritional deficiency, as WNT7B knockdown cells recovery from cell cycle synchronization more slowly than the control (Supplementary Fig. 1B and 1C).
We subsequently examined the effects of WNT7B on the migration of OSCC cells and found that WNT7B knockdown significantly inhibited the migration ability of OSCC cells, suggesting that WNT7B promoted the motility of these cells (Fig. 7C and 7D). Given WNT7B is typically involved in canonical WNT signaling, we then investigate whether this invasion capability was attributed to canonical signaling downstream genes. According to the probable WNT target genes (http://web.stanford.edu/group/nusselab/cgi-bin/WNT/target_genes), we chose CDH1 (cadherin1) and matrix metalloproteinases (MMPs) as candidates because they show a high correlation with tumor invasion and metastasis20,21. We compared the mRNA abundance of all WNT downstream MMP members and CDH1 in normal, OLP, and cancer tissues using our previous data (GSE70666)14. MMP1, MMP3, and MMP12 were up-regulated in both inflammation and cancer (Fig. 7E). However, MMP1 was the most significant increased one and showed well consistency in two OSCC patients. Further verification of MMP1 expression in the above 34 FFPE samples showed that, contrary to the weak expression in normal tissues, strong cytoplasmic positive signals were accumulated in cancer cells (Fig. 7F). In OLP tissues, the epithelium was loosely organized and a positive signal was localized in epidermal basal cells. Interestingly, FFPE samples showed a good correlation between WNT7B and MMP1 proteins expression (R2 = 0.66) (Fig. 7F). Furthermore, compared with the control, knockdown of WNT7B was accompanied by reduced expression of MMP1 in both SCC9 and FaDu cells (Fig. 7G). Therefore, our results revealed the involvement of WNT7B and MMP1 in oral inflammation and tumor progression.