Comparison of expression
Based on the TCGA data set, we performed a differential analysis between 501 tumor tissues and 44 paracancerous cases. We found elevated NAT10 expression in tumor tissues relative to normal tissues (Fig. 1A); paired comparison of 44 HNSCC patients showed significantly higher NAT10 expression levels in tumor tissues than in adjacent normal tissues (Fig. 1B). The results indicate that the high expression level of NAT10 may be closely related to the occurrence of HNSCC.
Relationship between NAT10 and clinical characteristics
As the results show (Fig. 2A-F), the expression level of NAT10 was significantly associated with T typing, AJCC stage and histological grade, but showed no significant correlation with variables such as age and sex. NAT10 was significantly increased in the late stage, late T grade and late grade, indicating that the high expression of NAT10 will affect the prognosis of HSNCC. Patients with lower NAT10 levels had significantly reduced OS relative to those with lower NAT10 levels (p=0.002, Fig. 2H); similar results were obtained for the progression-free interval aspect (Fig. 2G). The above results could indicate that the NAT10 gene is an oncogene of HNSCC. Clinical correlation heat maps showed that NAT10 gene expression levels were correlated with the grade in HNSCC patients (Fig. 2I). The univariate COX regression analysis demonstrated that NAT10 expression level, AJCC stage, T typing, and N typing were significantly associated with the overall survival of HHSCC patients (Fig. 2J). The results of the multivariate COX regression analysis suggested that the high expression of NAT10 was an independent risk factor for poor prognosis in HNSCC patients (Fig. 2K).
NAT10 participates in multiple related pathways in HNSCC samples
To explore the relevant pathways of NAT10 up-regulation in HNSCC, TCGA tumor sample datasets were collected by GSEA. We found that NAT10 was involved in Base excision repair, homologous recombination, mismatch repair, DNA replication, cell cycle, nucleotide excision repair, RNA degradation, spliceosome, one carbon pool by folate, glyoxylate and dicarboxylate metabolism, purine metabolism. pyrimidine metabolism, RNA polymerase, cysteine and methionine metabolism, aminoacyl TRNA biosynthesis, valine leucine and isoleucine biosynthesis, basal transcription factors, ubiquitin mediated proteolysis, non-homologous end joining, lysine degradation signal pathways (Fig. 3). The left, yellow portion is the high corresponding pathway enrichment score, indicating that these pathways are upregulated in the NAT10 high expression group, also indicating that the poor prognosis of patients with high NAT10 expression state may be due to differences in these pathways.
Relationship between NAT10 and the immune checkpoints
Immune checkpoints between the high and low expression groups were HHLA2, TNFSF4, CD244, BTNL2, CD44, CD40, CD276, TNFSF15, VTCN1, and NRP1. The analysis results (Fig. 4A) showed that, except for CD244 being slightly lower, the NAT10 gene high expression group at the immune checkpoint was higher than the low expression group. In patients with high NAT10 gene expression, CD44, CD40, CD276 and other immune checkpoints were expressed significantly higher, indicating that high expression of the NAT10 gene leads to poor prognosis, perhaps also through upregulation of these checkpoints. Additionally, we found different patterns of correlation between immune cells (Fig. 4B).
Relationship between NAT10 expression and immune cells
The proportion of immune-infiltrating cells between the high and low groups was calculated using the CIBERSORT method (Fig. 5A). In the high expression group, B cells were naive, T cells CD4 memory resting, T cell CD4 memory activated, NK cells resting, macrophages (M0, M1, M2), dendritic cells activated, and levels of NAT10 activation were higher than in the low expression group. There were no significant differences in NAT10 expression levels between B cell memory, CD4 naive cells, T cells, monocytes, eosinophils, and neutrophils. The expression of NAT10 and the infiltration of eight immune cells in the HNSCC group are indicated (Fig. 5B-I): NK cell resting (P=0.0035) and T cell CD4 memory resting (P=2.9e-08) were positively correlated with the proportion of NAT10 infiltration in HNSCC. In contrast, NAT10 expression was negatively associated with dendritic cell resting (P=0.0022), neutrophils (P=0.02), NK cell activation (P=0.018), T cell CD8 (P=0.000688), T cell follicular assist (P=2e-04), and T cell regulation (Treg) (P=2e-07). The results of the immune infiltration analysis showed that the proportion of immune cells infiltration was significantly changed in patients with high NAT10 expression, which could clarify the reason that the poor prognosis of patients with high NAT10 expression may be due to the increased proportion of these immune cells infiltration.
Differential expression of immune efficacy and drug efficacy
The difference in immune efficacy between the two groups was calculated according to the TIDE algorithm. The result indicates (Fig. 6A): there was no significant difference between the high and low expression groups, indicating that the expression of the NAT10 gene does not affect the efficacy of immunotherapy. Drug sensitivity results of the GDSC database for two groups with common chemotherapy and targeted therapy (Fig. 6B): Six drugs, crezotinib, dasatinib, imatinib, paclitaxel, sorafenib, and sunitinib, showed high semi-inhibitory concentration values in patients with low NAT10 expression. The high IC50 values of the patients in the low expression group indicate poor sensitivity to drugs in this group, indicating that the high expression patients are suitable for these drugs.
Immune pathway activity and TMB
By comparing the activities of 13 immune-related pathways and the infiltration fraction of 16 immune cells between high and low expression of the two groups (Fig. 7A): except for T cells regulatory (Tregs), Type I IFN Response and Type II IFN Response activities, the low expression group had higher scores than the high expression group. The TMB results show (Fig. 7B) that NAT10 expression was positively correlated with the tumor mutation burden expression level.
Prognostic significance of the immunomodulators associated with NAT10
Some immune regulators of HNSCC were positively or negatively correlated with NAT10 expression (Fig. 8A). A total of 17 related prognostic genes were selected by univariate cox (Fig. 8B): BTLA, CD244, CD27, CD96, CD28, CD40LG, CTLA4, TIGIT, ICOS, LTA, PVR, TNFRSF13B, TNFRSF13C, TNFRSF17, TNFRSF18, TNFRSF25 and TNFRSF4. Based on these selected regulator genes, 12 related immune genes were finally selected, which are: BTLA, CD244, CD96, CD40LG, ICOS, LTA, PVR, TNFRSF13C, TNFRSF17, TNFRSF18, TNFRSF25 and TNFRSF4. These 12 genes were included in the model as variables (Fig. 9A-B). Risk score = (-0.6569019*expression of BTLA)+(-0.1569486* expression of CD244)+ (0.2309998*expression of CD96)+(-0.1402918 * expression of CD40LG)+(-0.0146550*expression of ICOS)+(0.0.4044274 *expression of LTA)+(0.2462431*expression of PVR)+(0.0293222*expression of TNFRSF13C)+(-0.0177497*expression of TNFRSF17)+(-0.0646251*expression of TNFRSF18)+(-0.2106798*expression of TNFRSF25)+(-0.2371697 *expression of TNFRSF4), the risk cutoff value of the model is-0.51. The distribution status of risk values, risk score, and survival status of HNSCC are shown in Fig. 9C, E, G, and I. All the results were validated in the test group (Fig. 9D, F, H, J). The Kaplan-Meier survival analysis showed that the survival rate was significantly higher in the low-risk group than in the high-risk group (Fig. 9K). The ROC curve shows the AUC of the model prediction at 1-, 3- and 5-years as 0.696, 0.691 and 0.639 (Fig. 9M); the model also showed good prediction accuracy in the test group (Fig. 9L. N). Combining the above results demonstrates that the risk score can effectively predict the prognosis of HNSCC patients.
Construction and evaluation of the nomogram
Risk scores for patients with TCGA-HNSC were related to advanced pathological parameters (Fig. 10A). A nomogram was constructed to predict survival status in HNSCC patients, with gender, N classification, risk, clinical stage, and T classification as variables, with a specific score. According to the reality of each sample, the score of each prognostic factor was added to obtain the total score (Fig. 10B). A higher total score represents a higher chance of an event. This approach made it possible to individually estimate the probability of relapse, death, or drug adherence [18]. We were able to calculate the survival status of 1-,3- and 5-years and draw the ROC curve (Fig. 10C), with the resulting distribution of 0.716, 0.749, 0.668, respectively, which suggests good predictive accuracy. The resulting calibration curve compares well with the ideal model (Fig. 10D).