A total of 177 patients with gliomas undergoing surgery in Sanbo Brain Hospital Capital Medical University were included in this study. The diagnosis of gliomas for every patient was authenticated with laboratory examination, clinical features, and macroscopic and histological examinations. The characteristics of glioma samples, including grade, subtype, and IDH expression, were estimated by two pathologists specializing in brain tumor disorders. The follow-up information of 177 glioma patients was acquired. The endpoint of this study was defined as overall survival (OS), which is the period from the surgery date to death date without a specified cause of death. OS is a useful index to estimate the prognosis of tumor patients. The follow-up period of all patients ended in April 2019. The informed consent of biomedical research about tissue usage has been signed by every patient with glioma, with the project approved by the ethics committee of Sanbo Brain Hospital Capital Medical University (no. SBNK-2018-003-01). We ranked the 177 patients from low to high, and then we identified the first 25 patients as belonging to the low-SIGLEC9-expression group and the last 26 patients as belonging to the high-SIGLEC9-expression group based on the significant difference in survival between the high-SIGLEC9-expression group and the low-SIGLEC9-expression group. The remaining patients were categorized into the medium- SIGLEC9-expression group. The clinical data of 177 patients are shown in Supplementary Table S1 .

The protein levels of SIGLEC9 were examined by immunohistochemical staining. Formalin-fixed paraffin-embedded slides of glioma were baked for 4 h. Then, the slides were deparaffinated with dimethylbenzene and dehydrated with gradient ethanol. Furthermore, the slides were subjected to antigen retrieval with boiled citrate buffer, cyclooxygenase block with 3% hydrogen peroxide, and nonspecific antigen block with 10% goat serum. The primary antibody rabbit anti-human SIGLEC9 (Abcam, Cambridge, MA, USA) was diluted to 1:100 and incubated at 4°C overnight. On day 2, the second antibody, horseradish peroxidase-labeled goat anti-rabbit or mouse, was incubated for 1 h. Then, the slide was stained with 3′-diaminobenzidine reagent and counterstained by hematoxylin. Lastly, the total area of positive expression of SIGLEC9 was evaluated with ImageJ software by two researchers in the list of authors. The best cutoff value was determined by the survminer package of R.

The normalized values of fragments per kilobase per million mapped reads of gliomas were obtained from TCGA dataset (https://portal.gdc.cancer.gov) before April 2019. The batch effect of low-grade glioma (LGG) and GBM was removed by “sva” package. The normalized datasets of RNA-Seq were conducted as input. TCGA is a useful dataset to catalog genetic mutations responsible for cancer with the method of bioinformatics and genome sequence. The comparison of SIGLEC9 expressions in normal tissue and glioma was conducted in GEPIA website (http://gepia.cancer-pku.cn/). GEPIA is a web server about cancer and normal gene expression. In addition, we downloaded the mRNA-seq data from the CGGA dataset before October 2019. CGGA is a powerful dataset that stores the data of about more than 2,000 samples from Chinese brain tumor patients. The data of CGGA included mRNA sequencing, mRNA microarray, microRNA microarray, whole-exome sequencing, and the matched patient clinical data. The expression of SIGLEC9 was compared between different grades and subtypes of gliomas. SPSS20.0 was used to evaluate the statistical significance between different groups.

In this study, we conducted GSEA to investigate the underlying mechanisms of SIGLEC9 in gliomas. GSEA is a method that provides insights into discovering the biological mechanisms of the genes. The correlation between leukocyte infiltration and SIGLEC9 expression was conducted by single-sample GSEA (ssGSEA). The correlation between immune-related gene sets and SIGLEC9 was also conducted by ssGSEA. The ssGSEA is the extension of GSEA to measure the separate enrichment scores of each gene set. The ssGSEA could transform to the profile of gene set enrichment, which allows featuring of the cell state of biological process activities and pathways. The correlation between immune-related gene sets and SIGLEC9 was visualized in the bioinformatics website (https://www.immport.org/).

In this study, we calculated the correlation of genes and SIGLEC9 by Spearman method. We filtered genes with correlation >0.6. Gene Ontology (GO) analysis was used to conduct the functional enrichment analysis. The Database for Annotation, “clusterProfiler”, and “enrichplot” packages were used to analyze and visualize the results. GO analysis is a useful tool to investigate the biological processes, cellular components, and molecular functions. We selected signaling pathways with false discovery rate (FDR) <0.01 and count >10 from the GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment results.

The single-cell sequencing data of tumor tissues and adjacent tissues of glioma were downloaded from the GSE162631 dataset of the GEO database. We selected 4 tumor samples and 1 paracancerous tissue sample for single-cell transcriptome analysis. Raw gene expression matrices were imported and processed using the Seurat R package. The cells and genes with poor quality were filtered out. The genes expressed in at least 3 cells and high-quality cells with more than 200 genes and less than 5,000 genes were selected for the subsequent analysis. Low-quality cells containing more than 10% mitochondrial genes were excluded. Principal component analysis (PCA) with “FindNeighbors” and “FindClusters” functions was used to perform Uniform Manifold Approximation and Projection (UMAP) to screen the significant top 20 principal components. Then, we clustered the cells with a resolution of dim = 30 and visualized the clustering results using a UMAP scatterplot. Then, the different clusters were annotated to cell types based on the typical marker genes of the cells. Furthermore, we analyzed the expression of SIGLEC9 genes in tumor tissue and adjacent tissue. The ggplot2 package was used to identify the ratio of cells in tumors with adjacent tissue. We re-analyzed the macrophages separately to assess the subtypes of specific cell populations. We used Seurat standard procedures, including PCA dimensionality reduction and tSNE cell construction of clusters to extract cell subsets. Finally, we divided the macrophages into the high-SIGLEC9-expression group and the low-SIGLEC9-expression group according to the median of SIGLEC9 expression. The regulatory role of SIGLEC9 in macrophages was further analyzed.

Statistical analysis was conducted by t-test and Spearman χ ² test. Kaplan–Meier analysis was used to evaluate the survival rates. Cox proportional hazard model analysis was conducted to measure the hazard ratio and 95% confidence interval associated with SIGLEC9 expression. Statistical significance was considered when P <0.05.

We investigated the SIGLEC9 expression in different grades and subtypes of glioma patients with TCGA and CGGA databases. In TCGA database, SIGLEC9 expression was higher in tumor tissue than in adjacent normal tissue in LGG and GBM ( Figure 1A ). The expression of SIGLEC9 in grade III and grade IV glioma patients was higher than in grade II glioma patients ( Figure 1B ). As for the four subtypes, including classic, mesenchymal, and proneural subtypes, SIGLEC9 expression was higher in the mesenchymal subtype than in the other two subtypes of gliomas in TCGA database ( Figure 1D ). IDH is regarded as an index for the survival prognosis of gliomas. The results of the TCGA database showed that SIGLEC9 expression was higher in IDH wild-type glioma patients than in IDH mutated glioma patients ( Figure 1F ). The expression patterns of SIGLEC9 in different grades and subtypes of glioma patients from the CGGA database were similar to that of TCGA database. In the CGGA database, the expression of SIGLEC9 in grade IV glioma patients was higher than in grade II patients ( Figure 1C ). The SIGLEC9 level was higher in the mesenchymal subtype than in the other two subtypes of gliomas ( Figure 1E ). The SIGLEC9 expression was higher in IDH wild-type gliomas than in IDH mutated gliomas ( Figure 1G ).

Furthermore, we used TCGA and CGGA databases to analyze the survival probability of glioma patients with different expression levels of SIGLEC9. In TCGA dataset, patients with a higher SIGLEC9 expression had less survival probability than patients with a lower SIGLEC9 expression (P < 0.001) ( Figure 2A ). In the CGGA dataset, patients with a higher SIGLEC9 expression also had less survival probability than patients with a lower SIGLEC9 expression (P < 0.001) ( Figure 2B ). Furthermore, Survival analysis in different subgroups with glioma between the high-SIGLEC9 and low-SIGLEC9 groups was performed, the result was shown in Figure S1 .

We analyzed the expression of SIGLEC9 in 177 glioma patients with immunohistochemical staining. Based on the level of SIGLEC9 expression, the glioma patients were divided into three groups, and the clinical data of 177 patients are shown in Supplementary Table S1 . As shown in Figure 3A , patients with a high SIGLEC9 expression manifested a significantly shorter survival probability than those patients with a low SIGLEC9 expression (P = 0.019). Then, we analyzed the relationship between clinical characteristics and SIGLEC9 expression in glioma patients. We found that, in glioma patients with age ≧60 years, high-SIGLEC9-expression patients presented a shorter survival probability than low-SIGLEC9-expression patients (P = 0.007). Similar phenomena were also observed in patients with grade IV or glioblastoma (P = 0.032), patients with ATRX loss glioma (P = 0.002), patients without radiotherapy (P = 0.015), or patients without chemotherapy (P = 0.040). The detailed information on the clinical forest is listed in Figure 3B . The immunohistochemical map of glioma patients is shown in Figure 3C , and it can be seen intuitively that there are apparent differences in the expression of SIGLEC9 between glioma patients.

In this study, we investigated the correlation between the immune microenvironment and the SIGLEC9 levels in gliomas. Firstly, we analyzed the correlation between immune cell infiltration and SIGLEC9 expression in gliomas with TCGA database. As shown in Figure 4A , the SIGLEC9 expression was mostly positively correlated with myeloid-derived suppressor cell (MDSC) infiltration (Cor = 0.84, P < 2.2e−16), effector memory CD8⁺ T cell infiltration (Cor = 0.71, P < 2.2e−16), T follicular helper cell infiltration (Cor = 0.67, P < 2.2e−16), regulatory T cell infiltration (Cor = 0.63, P < 2.2e−16), mast cell infiltration (Cor = 0.58, P < 2.2e−16), and neutrophil infiltration (Cor = 0.27, P < 9.6e−13) in gliomas. The CGGA database was also used to explore the correlation between immune status and SIGLEC9 expression, and similar results were observed. The SIGLEC9 expression was positively correlated with MDSC infiltration (Cor = 0.88, P < 2.2e−16), effector memory CD8⁺ T cell infiltration (Cor = 0.79, P < 2.2e−16), macrophage infiltration (Cor = 0.79, P < 2.2e−16), regulatory T cell infiltration (Cor = 0.77, P < 2.2e−16), natural killer T cell infiltration (Cor = 0.77, P < 2.2e−16), and neutrophil infiltration (Cor = 0.55, P < 2.2e−16) ( Figure 4B ).

As shown in Figure 4C , glioma patients with high SIGLEC9 expression have significantly enhanced infiltration of immune cells, such as T cells, NK cells, neutrophils, aDC, B cells, DC cells, iDC cells, macrophages, mast cells, Th1, Th2, Th17, and so on. The proportion of these immune cells infiltrated was significantly reduced (pDC and Treg). We also calculated the difference in microenvironment scores between the high-SIGLEC9-expression group and the low-SIGLEC9-expression group by the ESTIMATE algorithm, and the results showed that the high-SIGLEC9-expression group had a higher immune score in StromalScore, ImmuneScore, and ESTIMATEScore (P < 0.001) ( Figure 4D ).

Then, we evaluated the expression correlation between SIGLEC9 and checkpoint members in tumor-induced immune response using Pearson correlation analysis with TCGA dataset. As shown in Figure 4E and Supplementary Table S3 , SIGLEC9 had a high concordance with LAIR1 (Cor = 0.919, P < 0.001), HAVCR2 (Cor = 0.892, P < 0.001), CD86 (Cor = 0.870, P < 0.001), and LGALS9 (Cor = 0.810, P < 0.001).

GO analysis was conducted to investigate the functions of SIGLEC9-correlated genes in data obtained from TCGA and CGGA databases. Based on the data of TCGA database, GO analysis showed that the SIGLEC9-correlated genes were mostly enriched in neutrophil activation, neutrophil degranulation, and neutrophil-mediated immunity ( Figure 5A ). The GO analysis based on the data of CGGA database also showed similar results ( Figure 5B ).

Moreover, GSEA was conducted to explore the biological functions of SIGLEC9 in gliomas. Based on the data of TCGA database, SIGLEC9 was positively correlated with the IL6-JAK-STAT3 signaling pathway (NES = 2.814, FDR = 0), KRAS signaling pathway (NES = 2.474, FDR = 0), reactive oxygen species pathway (NES = 1.896, FDR = 0), and TGF-β signaling pathway (NES = 1.928, FDR = 0) ( Figure 5C ). Similar results were obtained from the GSEA analysis based on the data of the CGGA database ( Figure 5D ), such as IL6-JAK-STAT3 signaling pathway (NES = 2.073, FDR = 0), KRAS signaling pathway (NES = 1.622, FDR = 0.001), reactive oxygen species pathway (NES = 1.668, FDR = 4.40E-04), and TGF-β signaling pathway (NES = 1.700, FDR = 2.91E-04).

Firstly, we normalized and pooled single-cell data from all samples and filtered the low-quality cells ( Figures 6A, B ). Then, we merged the tumor and adjacent tissue sample to perform unsupervised clustering to identify distinguished cell populations. Seurat v3.0 with default parameters was conducted in this study. We classified different cell subsets according to the related typical marker genes in cells. We mainly identified 10 types of cells ( Figures 6C, D ), such as glial/neuronal cells (PTPRZ1 and FABP7), DCs (HLA-DQA1, HLA-DPB1, FCER1A, and CD1C), T cells (CD3E, CD3D, CD3G, GZMK, and GZMA), mural cell (RGS5, BGN, TAGLN, NOTCH3, and PDGFRB), neutrophil (IL1R2, CSF3R, FPR2, and CXCL1), endothelial (CLDN5, VWF, ABCG2, and CAVIN2), macrophage (proliferating) (MKI67 and TOP2A), macrophage (APOC1), microglia/microphage (CD163 and F13A1), microglia (P2RY13 and SLC1A3), and microglia (P2RY12 and CX3CR1). We calculated the proportion of various types of cells in each sample and found that the proportion of microglia in normal samples was significantly higher than that in tumor samples, but the proportion of macrophages was significantly lower than that in tumor samples ( Figure 6E ). Therefore, our further analysis also focused on macrophages in tumors. Finally, we calculated the expression of SIGLEC9 in these cell subsets between glioma tissue and adjacent tissue ( Figure 6F ). In macrophages (proliferating) and macrophages subsets, the expression of SIGLEC9 in cancer tissues was higher than that in adjacent tissues. Macrophages play an essential role in glioma, and SIGLEC9 may be an important regulator of macrophages.

We extracted a subset of macrophages (macrophage) for analysis ( Figure 7A ), and we found that SIGLEC9 was positively correlated with genes (NCF2, PSMA7, VAMP3, and NCF4) (P < 0.05) that are involved in antigen processing and presentation of antigen ( Figure 7B ). Based on the median of SIGLEC9 expression, we divided the macrophages into two groups with high and low SIGLEC9 expression, and we further identified differential genes between these two groups. The GSEA analysis found that differential genes were associated with macrophage activation (16) ( Figure 7C ). The GO enrichment analysis found that upregulated differential genes were significantly enriched in pathways such as antigen processing and presentation of peptide antigen via MHC class II, antigen processing and presentation of peptide antigen, etc. ( Figure 7D ). In addition, the KEGG enrichment analysis found that upregulated differentially genes were enriched in Fc gamma R-mediated phagocytosis and antigen processing and presentation pathways ( Figure 7E ). Therefore, the differential genes of the two groups are involved in the pathway of antigen presentation, which also indicates that the expression of SIGLEC9 gene is related to the ability of macrophages to process antigens.

We also extracted macrophages (proliferating) for analysis and, as shown in Supplementary Figure S2 , we found that SIGLEC9 was positively correlated with the proliferation genes (ANAPC11, CCNB1, and PLK1) (P < 0.05) of macrophages. According to the median of SIGLEC9 expression, we further divided the macrophages (proliferating) into two groups with a high expression and a low expression of SIGLEC9, respectively. We found the differential genes between the high-SIGLEC9-expression group and the low-SIGLEC9-expression group. The GSEA analysis found that upregulated genes were associated with macrophage maturation and M2 polarization (17). The GO enrichment found that upregulated genes were significantly enriched in the regulation of cell cycle phase transition, DNA replication, T cell activation, and other pathways, which are related to cell proliferation and replication. In addition, the KEGG enrichment found that upregulated genes were significantly enriched in pathways, such as DNA replication and cell cycle, which are also related to cell proliferation and replication. It indicates that the high expression of SIGLEC9 is related to the M2 macrophage polarization and proliferation of macrophages.