Based on univariate Cox analysis for survival analysis，we got one hundred and sixty-three genes (p<0.05). Then, a total of two subtypes were identified by the ConsensusClusterPlus. Then we divided the patients into the subtype1 (a total of 951 SCC patients) and subtype2 (a total of 417 SCC patients) based on the one hundred and sixty-three genes of survival analysis (Figure1A). 252 CESCs, 95 ESCAs 520 HNSCs, and 501 LUSCs were included in this study. A total of 90 Grade I, 461 Grade II, 248 Grade III, and 8 Grade IV SCC were analyzed (Table1).
Genomic correlations with angiogenesis subtypes
Studying the effects of genomic alterations on the generation of angiogenesis subtypes. As result, the cancer DNA fraction, subclonal genome fraction, and LOH fraction altered were significantly higher in the subtype1 (Figure1B). Additionally, the tumor purity was significantly lower and the ploidy was significantly higher in the subtype1 (Figure1C). However, the HRD-LOH presented no significant difference between the two subtypes. As for the HRD in the subtype1, were significantly higher than that in the subtype2 (Figure1D). The more prime LOH was found in the subtype1 (Figure1E). Moreover, the first 20 highly mutated genes were shown in oncoplot (Figure2A~2B). And a landscape of mutations in terms of number of mutations and mutation types (Figure2C~2D). Ten significantly enriched oncogenic pathways, with RTK-RAS, WNT, Hippo, PI3K, Cell_Cycle, MYC, TGF-Beta, TP53, and NRF2 having the greatest impact in subtype 1(Figure2E). The TGF-Beta and NRF2 pathways had the greatest effect in subtype 2(Figure2F).
Differentially expressed genes and network regulation of angiogenesis-subtype
A total of 1368 SCC samples showed differential expression of genes. We identified 163 differential genes by t-test, including 124 genes with up-regulated expression and 18 genes with down-regulated expression in the subtype1 (P<0.05, Figure3A). In the alternative, we obtained a score for each sample's association with angiogenesis genes. Next, we analyzed the differently expressed microRNA between the two subtypes. As result, a total of 34 differently expressed miRNA were found between two subtypes, including 11 down-regulated miRNAs and 23 up-regulated miRNAs (Figrue3B). To reveal whether the regulatory relationship between TF and miRNA has an impact on cancer, we established the regulation network for TF-miRNA-target (Figure3C). Finally, most of the drug targets were shown to be also altered by the angiogenesis (Figure3D).
Enrichment Pathways for angiogenesis-subtype
As result, twenty pathways were significantly enriched (P<0.05) by performing KEGG pathway analysis associated with DEGs. These included the PI3K/Akt signaling pathway and ECM-receptor interactions (Figure4A). Notably, twenty GO pathways with DEGs were also found to be enriched, extracellular matrix structural constituent had the highest -log10 P value, with the maximum difference (Figure4B). The correlation between different pathways and angiogenesis subtype (Figure4C). After GSEA analysis, these pathways were strongly associated with angiogenesis in tumor progression (Figure4D). As for subtype1, ten KEGG pathways were found to be upregulated including focal adhesion, pathways in cancer, adherens junction, renal cell carcinoma, regulation of actin cytoskeleton, neurotrophin signaling pathway, ECM receptor interaction, TGF beta signaling pathway, ERBB signaling pathway, GAP junction. On the other hand, another ten KEGG pathways were up-regulated in subtype2 such as spliceosome, base excision repair DNA replication. Notably, ten hallmarks pathways were also found to be enriched in the subtype1, including epithelial-mesenchymal transition, inflammatory response, hypoxia, and angiogenesis As for subtype2, oxidative phosphorylation DNA repair, E2F targets, MCY targets v1, G2M checkpoint, MYC targets v2 were found to be the enriched hallmarks pathways. Thus we defined subtype1 as the angiogenesis subtype and subtype 2 as the non-angiogenesis subtype.
The immune microenvironment in the angiogenesis subtypes
By applying CIBERSORT, a computational method capable of inferring leukocyte subtypes in bulk tumor transcriptomes, we discovered complex associations between 22 different leukocyte subsets and angiogenesis subtype (Figure5A). Next, we also explored the distribution of the immune cells between the angiogenesis subtype and non-angiogenesis subtype (Figure5B). Additionally, we also evaluated the immune score by using the ESTIMATE algorithm. As result, the ESTIMATE score immune score and stromal score were significantly higher in the angiogenesis subtype and the tumor purity was significantly lower in the non-angiogenesis subtype (Figure5C).
Next, we aimed to explore the microenvironment in these two angiogenesis subtypes. We analyzed the expression of 15 immune checkpoints between two angiogenesis subtypes. The results showed that 8 of the 15 immune checkpoints (ADORA2A, BTLA, 276, CYBB, HAVCR2, SIGLEC7, SIGLEC9, and VTCN1) were significantly up-regulated while C10orf54 were significantly down-regulated in the subtype1 (Figure5D).
The clinical implication of angiogenesis subtypes
To evaluate the survival significance of angiogenesis subtype for SCC patients. The KM curves revealed that the patients in subtype1 have a lower survival DFI than that in subtype2 (p=0.017, Figure6A). Additionally, the patients of the angiogenesis subtype were revealed to have a poor OS outcome than that in non-angiogenesis subtype patients (P=0.00013, Figure6B). However, there was no statistical significance was found for the DSS and PFI (Figure6C~D).