To explore the expressions of GBE1 and CD163 in LUAD tissues and paired adjacent lung tissues, tissue microarray (HLugA180Su04) containing 92 LUAD tissues and 88 adjacent lung tissues were obtained from Shanghai Outdo Biotech Co, Ltd. All included patients did not receive neoadjuvant therapy and received operation between January 2008 and December 2013. Eliminating 11 ineffective LUAD tissues, a total of 81 LUAD cases included in these two studies. The sections were mounted onto slides coated with 3‐aminopropyltriethoxysilane. After drying and dewaxing, the sections underwent high-pressure antigen retrieval and reacted with antibodies against GBE1 (ab223799, dilution 1:300; Abcam, San Francisco, CA, USA) and CD163 (ab182422, dilution 1:500; Abcam, San Francisco, CA, USA). Furthermore, the slide was incubated with HRP-labeled anti-rabbit IgG (Dako, Glostrup, Denmark) for 45 min at room temperature. Finally, the slide was stained with 3,3-diaminobenzidine (DAB) and was counterstained with hematoxylin. The EnVision+ detection system (Dako) was used according to the manufacturer’s instructions. The score of IHC was calculated by multiplying the intensity (0–3) and extent (0%–100%) of staining for each tissue point. These tissue samples were collected with bioethics approvals and informed consents. The diagnosis of LUAD was validated by pathological evidence, and the tumor grades and clinical stages were classified using the 7th American Joint Committee on Cancer (AJCC) TNM criteria. The experiments were approved by the Ethics Committee of National Cancer Center/Cancer Hospital) Chinese Academy of Medical Sciences and Peking Union Medical College (IRB Approval No. NCC2019C-167).

The correlation between GBE1 expression and prognosis in LUAD was analyzed by the PrognoScan database (http://www.abren.net/PrognoScan/), which collects plenty of publicly available cancer microarray datasets, and consequently, could be utilized to explore the association between a gene and clinical outcome in cancer research (24). The hazard ratio (HR) and adjusted p-value were assessed. Furthermore, we explored the correlation between GBE1 expression and prognosis by using TCGA cohort and TMA cohort. The patients with other malignancies were excluded in TCGA cohort. The cutoff values of high and low expression were calculated by R package “OptimalCutpoints.” The Kaplan-Meier method and Log-rank test were used to compare the prognosis of patients with different GBE1 expression. The multivariable Cox regression analysis was used to confirm whether GBE1 is an independent prognostic factor.

We further explored the relationship between clinical characteristics and GBE1 expression by using TCGA database and TMA cohort. Level 3 RNA-Seq V2 data for 20,530 genes of lung adenocarcinoma patients (TCGA-LUAD Cohort) were downloaded from the UCSC Xena browser (http://xena.ucsc.edu/) TCGA hub. Updated clinical and survival information for TCGA patients were also obtained from the UCSC Xena browser. The differential expression of GBE1 between paired tumor tissues and normal tissues were analyzed. The age, gender, T stage, N stage, M stage, and TNM stage of LUAD patients were extracted from the TCGA cohort to analyze the GBE1 expression in these different subgroups. In TMA cohort, gender, age, T, N, TNM, and grade were also collected and analyzed. Furthermore, we evaluated the expression of CD163 in the TMA cohort.

We evaluated the proportions of immune cells in LUAD by using the CIBERSORT method. The CIBERSORT algorithm is an immune cell infiltration estimation analysis tool, which can be used to estimate the immune cell infiltration by deconvoluting the expression matrix of immune cell subtypes based on linear support vector regression (25). Immune cells fraction data downloaded from TCGA Data Portal (https://gdc.cancer.gov/about-data/publications/panimmune, filename: TCGA.Kallisto.fullIDs.cibersort.relative.tsv) (26). Correlation analyses between the expression of GBE1 and immune cell filtration were performed with Pearson’s correlation by using the “corrplot” function with R package corrplot. Furthermore, we also validated the associations between GBE1 expression and correlated immune infiltrates via the TIMER database, including B cells, CD4⁺ T cells, CD8⁺ T cells, macrophages, and dendritic cells. TIMER (https://cistrome.shinyapps.io/timer/) is a web server designed for exploring the associations between immune infiltrates and different factors, including gene expression, clinical outcomes, somatic mutations, etc. In total, 10,897 tumors from 32 cancer types were collected to estimate the abundance of immune infiltrates (27). For further investigation, the correlations of GBE1 expression with gene markers of M1 macrophage (NOS2, IRF5, PTGS2), M2 macrophage (CD163, VSIG4, MS4A4A), CD8⁺ T cells (CD8A, CD8B), dendritic cells (CD1C, ITGAX), and CD4⁺ T cells (CD4) were also explored in LUAD. All these gene markers have been reported in previous published studies (28–35). The correlations in TIMER were adjusted by tumor purity with the left-most panel (36). The correlation modules were presented as the expression scatterplot for a pair of specified genes, as well as Spearman correlation and statistical significance. The y-axis represents the expression of GBE1, and the x-axis represents the expression of corresponding gene markers of immune cells. Gene expression levels were demonstrated by LOG2 RSEM.

The differences between categorical variables were analyzed by Chi-square test. The counting variables were expressed as mean ± standard deviation (SD), and the differences between counting variables were analyzed by Student’s t-test (number of variable = 2) or variance analysis (number of variable ≥2). The prognosis was estimated by Kaplan-Meier analysis with Log-rank test and Cox regression analysis. The strength of the correlation was determined using absolute values in TIMER, as shown below: very weak (0.00–0.19), weak (0.20–0.39), moderate (0.40–0.59), strong (0.60–0.79), and very strong (0.80–1.0) (30). Statistical analyses were processed using SPSS 23.0 software (SPSS, Inc., Chicago, IL, USA), and p < 0.05 was considered significant.

A total of 512 LUAD patients in TCGA database were selected, and the optimal cutoff value of GBE1 expression was 9.5. We firstly evaluated the differential GBE1 expression between 38 paired LUAD tissues and adjacent tissues. The results showed that GBE1 expression in LUAD tissues was higher than the adjacent tissues ( Table 1 ). In TMA cohort, we analyzed the differential GBE1 expression between 81 paired LUAD tissues and adjacent tissues by IHC. The IHC results suggested that GBE1 was mainly expressed in the cytoplasm and overexpressed in LUAD tissues ( Table 1 ). The different staining intensities of GBE1 are displayed in Figure 1 .

Firstly, the correlation between GBE1 expression and survival was evaluated using the PrognoScan database. The results of three cohorts (HARVARD-LC, GSE31210, jacob-00182-UM) in PrognoScan showed that high expression of GBE1 had a poor prognosis in LUAD ( Figures 2A–D ). Furthermore, we also estimated the correlation between GBE1 expression and survival in TCGA cohort. The results revealed that 5-year survival rate of high-expression group of GBE1 in LUAD was significantly lower than that of low-expression group (low vs. high = 48.4% vs. 25.7%, Figure 3A ). In addition, we also validated the prognostic value of GBE1 expression by using TMA-based IHC in 81 paired LUAD tissues and corresponding adjacent nontumor tissues. The result was consistent with that of PrognoScan database and TCGA cohort (low vs. high = 64.0% vs. 22.9%, Figure 3B ).

To confirm whether GBE1 is an independent prognostic factor in LUAD patients, we performed Cox multivariable regression by using TCGA cohort and TMA cohort. The results showed that GBE1, age, gender, and TNM stage were independent prognostic factors ( Figures 3C, D ). Therefore, GBE1 has a potential to serve as a poor prognostic biomarker in LUAD patients.

In TCGA cohort, we found that higher mRNA expression of GBE1 was correlated with advanced T classification (p < 0.01), N classification (p = 0.05), M classification (p = 0.04), and TNM stage (p = 0.02). However, there was no significant difference between GBE1 expression and age (p = 0.50) or gender (p = 0.10) ( Figures 4A–F ). We further explore the protein expression of GBE1 by utilizing the TMA cohort. The results showed that GBE1 was related with T classification (p = 0.004), N classification (p = 0.011), grade (p < 0.001), and TNM stage (p = 0.001) ( Table 2 ). These results suggested that GBE1 may be associated with LUAD progression.

To investigate the relationship between GBE1 expression and immune-infiltrating cells, we used CIBERSORT algorithm to calculate the correlations between GBE1 mRNA expression and immune marker sets of various immune cells, including T cells and B cells, M1 and M2 macrophages, and dendritic cells ( Figure 5A ). The results revealed that GBE1 expression was positively correlated with infiltration of M2 macrophages, CD4⁺ memory-activated T cells, and resting dendritic cells. To further validate the relationships between GBE1 and infiltration of immune cells, we also analyzed the correlations between GBE1 expression and infiltration level of various immune cells by TIMER ( Figure 5B ). The results showed that GBE1 expression was positively correlated with macrophage infiltration, CD8⁺ T-cell infiltration, and dendritic cell infiltration. We also estimated the relationships between GBE1 and immune marker genes of immune cells, included M1 (NOS2, IRF5, PTGS2) macrophages, M2 (CD163, VSIG4, MS4A4A) macrophages, CD8⁺ T cells (CD8A, CD8B), dendritic cells (CD1C, ITGAX), and CD4⁺ T cells (CD4) in TIMER. The results showed that GBE1 exhibited positive association with CD163, VSIG4, and MS4A4A in LUAD (r = 0.414, p < 0.001; r = 0.275, p < 0.001; r = 0.275, p < 0.001). Moreover, GBE1 showed positive associations with markers of M1 macrophages and CD4, IRF5, PTGS2, and ITGAX in LUAD ( Figure 6 ). Interestingly, CD163 mRNA expression was also positively correlated with GBE1 mRNA expression in TCGA cohort ( Figure 7A ). We further confirmed the relationship between GBE1 expression and CD163⁺ tumor-associated macrophage infiltration by using TMA-based IHC, and the IHC staining of the same tissues are shown in Figures 8A–H . The result revealed that GBE1 was strongly positively related to the CD163⁺ tumor-associated macrophage infiltration, which was consistent with the results of CIBERSORT algorithm, TIMER database, and TCGA cohort ( Figure 7B ). We dichotomized the infiltration level of CD163⁺ tumor-associated macrophages into low-infiltration group and high-infiltration group based on the median value. The result showed that the expression level of GBE1 was significantly lower in the low infiltration CD163 group when compared with high-infiltration group both in mRNA and protein levels ( Figures 7C, D ). These findings suggested that GBE1 may regulate macrophage polarization in LUAD. High expression of GBE1 may promote high infiltration level of CD163⁺ tumor-associated macrophages in LUAD.