SLA2 is a prognostic marker in HNSCC and correlates with immune cell infiltration in the tumor microenvironment

Data acquisition, preprocessing, and ethics statement Pan-cancer ribonucleic acid sequencing (RNA-seq) data were obtained from UCSC XENA database (https:// xenabrowser.net/ datap ages/) [8]. Level 3 RNA RNA-seq expression data and clinical data for HNSCC (502 HNSCC tissues vs. 44 normal adjacent cancer tissues) were downloaded from TCGA portal (https:// portal. gdc. cancer. gov/) [9]. Before analysis, transcripts per million reads (TPM) RNA-seq data were log transformed (Log2(TPM + 1)). We use TCGA to analyze the SLA2 expression difference between HNSCC and normal head and neck tissues. As TCGA are open publicly available databases, the data collection from the databases was compliant with all applicable laws, regulations, and policies for the protection of human subjects, and all written informed consents were obtained from all subjects involved.

We performed univariate and multivariate Cox regression analyses to determine the appropriate terms for building nomograms. A forest was applied to display the p value, hazard ratio (HR), and 95% confidence intervals (95% CI) for each variable using the “forestplot” R package through R software. A nomogram was created according to the results of the multivariate Cox proportional hazards analysis to predict the total recurrence rate in 1, 3, and 5 years. The “rms” R package was used to evaluate the risk of recurrence for individual patients by the points associated with each risk factor. We extracted survival information for each sample in TCGA. Indicators such as overall survival (OS) and disease-specific survival (DSS) were used to elucidate the relationship betweens SLA2 and the prognosis of HNSCC patients. Kaplan–Meier (KM) and log-rank tests were used for the survival analysis of HNSCC (p < 0.05), and survival curves were analyzed by the “survminer” and “survivor” R packages.

GEPIA is a Newly developed interactive web server. Its web site is http://​gepia.​cancer-pku.​cn/​index.​html [10]. This database provides online analysis based on data from TCGA and the Genotype-Tissue Expression (GTeX) projects. In our study, we used the given TCGA expression data set to assess the difference of SLA2 between HNSCC and normal head and neck tissues. In addition, we used GEPIA to analyze the prognostic significance of SLA2 mRNA expression in HNSCC and draw the survival curve with log rank p value.

Single-sample Gene Set Enrichment Analysis (ssGSEA) was used to assess the infiltration of individual immune cell populations, e.g., T helper 2 cells (Th2 cells), cluster of differentiation 8 T cells (CD8 T cells), natural killer CD56dim cells (NK CD56dim cells), and others. The analysis was performed using Gene Set Variation Analysis (GSVA) package [11, 12]. The TIMER (https://​cistrome.​shinyapps.​io/​timer/​) database was applied to explore the correlations between HNSCC and TIICs (Li et al. 2017) [13]. GSVA package of R was used to analyze the correlations between SLA2 expression levels and the infiltrations of T cells, B cells, regulatory T cell (TReg), CD8 T cells, dendritic cell (DC), activated dendritic cell (aDC), Cytotoxic cells, effective memory T cell (Tem), Eosinophils, and NK CD56dim cells.

Sialic acid binding Ig-like lectin 15 (SIGLEC15), T cell immunoreceptor with Ig and ITIM domains (TIGIT), cytotoxic T lymphocyte-associated antigen-4 (CTLA4), cluster of differentiation 274 (CD274), hepatitis A virus cellular receptor 2 (HAVCR2), lymphocyte-activation gene 3 (LAG3), programmed cell death 1 (PDCD1), and programmed cell death 1 ligand 2 (PDCD1LG2) were selected to be immune-checkpoint-relevant transcripts, and the expression data of these eight genes were extracted. R package “ggplot2” “pheatmap” and “immuneeconv” were used to assess the expression of the immune-checkpoints and co-expression of SLA2 with these immune-checkpoints. Potential immune checkpoint blockade response was predicted with the TIDE algorithm (Jiang et al. 2018) [14].

RNA-sequencing expression profiles and corresponding clinical information for HNSCC were downloaded from the TCGA data set (https://​portal.​gdc.​com). Predicting the chemotherapeutic response for each sample based on the largest publicly available pharmacogenomics database [the Genomics of Drug Sensitivity in Cancer (GDSC), https://​www.​cancerrxgene.​org/​] [15‐17]. The prediction process was implemented by R package “pRRophetic”. The samples' half-maximal inhibitory concentration (IC50) was estimated by ridge regression. All the parameters were set as the default values. Using the batch effect of the combat and tissue type of all tissue, the duplicate gene expression was summarized as the mean value using the batch effect of the combat and tissue type of all tissue. Predicting the relative IC50 of cisplatin and methotrexate in patients with HNSCC.

LinkedOmics is a public website on http://​www.​linkedomics.​org/​login.​php, which contains 32 TCGA cancer types of multiple omics data, including federation generated from the Clinical Proteomics Tumor Analysis Consortium [18]. We applied this online tool to identify and analyze SLA2 co-expression genes in head and neck cancer cohort from TCGA, which contains 502 samples. The data from Linkfinder are signed and sorted; after that, GO analyses, including molecular function (MF), biological process (BP), cellular component (CC), and KEGG analysis, are performed by the GSEA method.

Gene expression data were normalized by log2-transformation. Statistical analyses were conducted using R software (version4.0.2). The analysis of SLA2 expression in TIMER and TCGA databases showed p value. The survival curve generated by GEPIA analysis showed p value. The Cox proportional hazards model, KM analyses, and log-rank test were conducted for all survival analyses. Spearman’s test or Pearson’s test was applied to analyze the correlation between two variables; p value < 0.05 was considered significant. R software (version 4.0.2) was used for statistical analysis.

Using the data from TCGA, we found that SLA2 expression was significantly elevated in 19 of 33 cancer types, including kidney renal papillary cell carcinoma (KIRP), kidney renal clear cell carcinoma (KIRC), stomach adenocarcinoma (STAD), among others (Fig. 1A). Similarly, SLA2 expression was significantly increased in HNSCC tissues compared with normal samples (Fig. 1B). It has the same result as in Fig. 1A, with a p value less than 0.001. For the 43 paired samples, SLA2 expression was also significantly higher in cancer tissues than in matched normal adjacent tissue samples (Fig. 1C).

The survival data of SLA2 expression in HNSCC patients were analyzed. KM survival curves were used to evaluate the relationship between the SLA2 expression and survival outcomes. The cut-off value of the high and low SLA2 expression group was set as the median. The results showed that patients with higher SLA2 mRNA expression had longer OS (p < 0.019) (Fig. 2A), DSS (p < 0.0277) (Fig. 2B). Consequently, the expression of SLA2 mRNA in HNSCC is associated with survival. The result from GEPIA is also consistent with our results (Fig. 2C).

We then performed univariate and multivariate Cox regression analyses using several clinical features and SLA2 expression level. Furthermore, univariate and multivariate Cox regression analyses illustrated that SLA2 expression (p < 0.01), pN-stage (p < 0.05), and pM-stage (p < 0.01) were important independent factors to the prognosis of HNSCC (Fig. 3A, B). We further constructed a nomogram that combined only three independent prognostic factors (including SLA2, pN-stage, and pM-stage) to provide a quantitative guideline for clinicians to predict the probability of 1-, 3-, and 5-year OS in HNSCC patients (Fig. 3C). Each patient is given a total score through adding each prognostic parameter point, with a higher total score meaning a worse outcome for that patient. In addition, the calibration curves showed that the nomogram performed well in estimating 1-, 3-, and 5-year OS (Fig. 3D). Therefore, SLA2 may be a potential diagnostic marker for HNSCC.

To understand the effects of SLA2 expression on tumor microenvironment, immune infiltration analysis was performed using ssGSEA method. The correlations between enrichments of immune cell in Head and neck tissues and SLA2 expression levels were calculated using Spearman correlation analyses. As shown in Fig. 4A, SLA2 expression levels were positively correlated with numerous types of immune cells, e.g., T cells, Cytotoxic cells, TReg, aDC, NK CD56dim cells, follicular helper T cell (TFH), CD8 T cells, T helper cells, T helper 1 cells (Th1 cells), B cells, DC, plasmacytoid dendritic cells (pDC), Tem cells, Macrophages, Th2 cells, Eosinophils, interdigitating dendritic cell (iDC), Th17 cells, NK cells, central memory T cells (Tcm), Mast cells, and Neutrophils, all of which play an important role in anti-tumor immunity. Next, we further used the CIBERSORT algorithm to investigate the tumor-infiltrating immune cells, we found that the expression of SLA2 was positively correlated with the infiltration of B cells, cluster of differentiation 8-positive T cells (CD8 + T cells), cluster of differentiation 4-positive T cells (CD4 + T cells), Macrophages cells, Neutrophils cells, and Dendritic cells (Fig. 4B). Similarly, the infiltrations of other immune cells in Head and neck tissues are observed, GSVA package of R was used to analyze the correlations between SLA2 expression levels and the infiltrations of T cells, B cells, TReg, CD8 T cells, DC, aDC, Cytotoxic cells, Tem, Eosinophils, and NK CD56dim cells. As shown in Fig. 4C, SLA2 expression levels were significantly positively correlated with the infiltrations of those immune cells (p < 0.01).

We further performed gene co-expression analysis to evaluate the mechanisms that SLA2 was correlated with the infiltration of immune cells in HNSCC. Major histocompatibility complex gene (MHC genes), immunosuppressive, immune activation genes, chemokine and chemokine receptor-related genes were assessed. The results showed that SLA2 positively co-expressed with all listed MHC genes (Fig. 5A). Notably, almost all immunosuppressive genes positively co-expressed with SLA2 (Fig. 5B). SLA2 had a positive relationship with most of the immune activation genes, such as genes that encoding cluster of differentiation 48 T cells (CD48), cluster of differentiation 28 T cells (CD28), Interleukin-6 (IL6), C–X–C motif chemokine receptor 4 (CXCR4) (Fig. 5C). Similarly, SLA2 had a positive correlation with some chemokine and chemokine receptors (Fig. 5D, E). The expression of immune-checkpoints, including CD274, CTLA4, HAVCR2, LAG3, PDCD1, PDCD1LG2, TIGIT, and SIGLEC15 (Fig. 6A), were further investigated in the World Health Organization (WHO) grade II and III of HNSCC. The results illustrated that LAG3 (p < 3.35e−05), PDCD1 (p < 1.06e−04), CTLA4 (p < 2.55e−03), HAVCR2 (p < 1.72e−03), TIGIT (p < 1.22e−03), and SIGLEC15 (p < 9.39e−02) were upregulated in WHO grade III compared with WHO grade II of HNSCC (Fig. 6B). In addition, we revealed that HNSCC patients in WHO grade III had a better response to immune checkpoint blockade compared with HNSCC patients in WHO grade I (Fig. 6C). Moreover, the results demonstrated that SLA2 positively co-expressed with these immune-checkpoints (Table 1), indicating that SLA2 may be a potential immunotherapy target.

RNA-sequencing expression profiles and corresponding clinical information for HNSCC were downloaded from the TCGA data set (https://​portal.​gdc.​com). The results showed that IC50 of cisplatin and methotrexate was negatively correlated with SLA2 expression (Fig. 6D, E).

To study the functional network of SLA2 neighborhood genes in HNSCC, we first identified SLA2 neighborhood genes with LinkedOmics. The results are shown in a volcanic plot (Fig. 7A). In addition, the top 50 positively and negatively related genes were shown in the heat map, respectively (Fig. 7B, C). Significant GO term analysis by GSEA indicated that genes correlated with SLA2 were located mainly in the side of membrane, receptor complex, secretory granule membrane, endocytic vesicle, membrane region, and endosome membrane, where they were involved in leukocyte cell–cell adhesion, response to interferon-gamma, and regulation of immune effector process. These related genes also served as antigen binding, cytokine receptor activity, phosphatidylinositol 3-kinase activity, peptide receptor activity, Src homology domain 3 (SH3) domain binding, and cytokine receptor binding (Fig. 7D–F). KEGG pathway analysis demonstrated that these genes related to SLA2 were mainly enriched in signal pathways, such as hematopoietic cell lineage, cell adhesion molecules (CAMs), natural killer cell mediated cytotoxicity, measles, and chemokine signaling pathway (Fig. 7G).