The RNA-sequencing (RNA-seq) data and relevant clinical data across 33 tumor types and normal tissues of 15,776 samples were downloaded from The Cancer Genome Atlas (http://cancergenome.nih.gov) and the Genotype-Tissue Expression (GTEx) database (https://www.gtexportal.org/home/-index.html), which contained the extraction of ESCA: GTEx normal (n=653); TCGA paraneoplastic (n=13); and TCGA tumor (n=182). Furthermore, the GSE23400 (n=208) (20), GSE33426 (n=71) (21), GSE53625 (n=358) (22), and GSE45670 (n=38) (23) datasets were used for the validation of the expression difference analysis. The samples with missing expression data were excluded from the study. In addition, the downloaded data were used to explore the relationship of LAMP2 with various clinicopathological parameters (including the pathologic stage, N stage, M stage, and residual tumor), diagnosis [receiver operating characteristic (ROC)], prognosis [overall survival (OS), progression-free interval (PFI), and disease-specific survival (DSS)], and gene coexpression and bioenrichment [GO, KEGG, and gene set enrichment analysis (GSEA)] in ESCA (ESCC and EAC).

The following tumor types were included: bladder urothelial carcinoma (BLCA); breast-invasive carcinoma (BRCA); cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC); cholangiocarcinoma (CHOL); colon adenocarcinoma (COAD); rectum adenocarcinoma (READ); lymphoid neoplasm diffuse large B-cell lymphoma (DLBC); glioblastoma multiforme (GBM); glioma (GBMLGG); head and neck squamous cell carcinoma (HNSC); kidney chromophobe (KICH); kidney renal clear cell carcinoma (KIRC); kidney renal papillary cell carcinoma (KIRP); acute myeloid leukemia (LAML); brain lower-grade glioma (LGG); liver hepatocellular carcinoma (LIHC); lung adenocarcinoma (LUAD); lung squamous cell carcinoma (LUSC); mesothelioma (MESO); ovarian serous cystadenocarcinoma (OV); pancreatic adenocarcinoma (PAAD); prostate adenocarcinoma (PRAD); rectal adenocarcinoma (READ); sarcoma (SARC); skin cutaneous melanoma (SKCM); stomach adenocarcinoma (STAD); testicular germ cell tumors (TGCTs); thyroid carcinoma (THCA); thymoma (THYM); uterine corpus endometrial carcinoma (UCEC); and oral squamous cell carcinoma (OSCC).

Tumor Immune Estimation Resource 2.0 (TIMER 2.0; http://timer.cistrome.org/) is an updated interactive web server that allows the investigation and visualization of tumor immunity. TIMER 2.0 assists in finding the associations between gene expression, mutations, immune infiltration, and survival characteristics in the TCGA cohort (24). In this study, the expression of LAMP2 in multiple cancer types was assessed using the “Exploration-Gene_DE” model. The “Immune-Gene” module was employed to explore the relationship between LAMP2 expression and the infiltration levels of immune cells (neutrophils, macrophages, dendritic cells, B cells, CD8+ T cells, and CD4+ T cells) on a pan-cancer basis.

Gene Expression Profiling Interactive Analysis 2 (GEPIA2; http://gepia2.cancer-pku.cn/#index) is a free web portal for the differential gene expression analysis of TCGA and GTEx data (25). In the present study, LAMP2 expression was analyzed using the TCGA-ESCA dataset. LAMP2 expression in ESCA and paraneoplastic tissue samples was studied using the “Expression DIY” module in GEPIA2.

The UALCAN database (http://ualcan.path.uab.edu/index.html) can be used to analyze online data (TCGA, The MET500 metastatic cancer cohort, and Clinical Proteomic Tumor Analysis Consortium) and clinical data with respect to the differential gene expression between tumor and normal tissues (26). UALCAN was used here to study LAMP2 expression and its pan-cancer relationships with protein expression in the following 10 subtypes: k1 overexpression of proteasome complex proteins, glycolysis proteins, and pentose phosphate pathway proteins), k2 (adaptive immune system related; associated with T-cell activation; expression of major histocompatibility complex proteins), k3 (innate immune system related; overexpression of complement system proteins; involvement of eosinophils, neutrophils, mast cells, and macrophages; hypoxia signature), k4 (represents basal-like breast cancer; overexpression of YAP1 and MYC targets), k5 (epithelial signature; normoxia signature; overexpression of YAP1 and MYC targets; overexpression of oxidative phosphorylation and the tricarboxylic acid (TCA) cycle proteins), k6 (stromal related; overexpression of matrix metallopeptidases; Wnt and Notch pathway signatures; hypoxia signature), k7 (stromal related; overexpression of collagen VI proteins; Wnt and Notch pathway signatures), k8 (overexpression of Golgi apparatus-related proteins; Ras pathway signature), k9 (found in KIRC cases only; overexpression of hemoglobin complex proteins), and k10 (overexpression of endoplasmic reticulum (ER)-related proteins and steroid biosynthesis pathway proteins).

PrognoScan (http://dna00.bio.kyutech.ac.jp/PrognoScan/index.html) is a database for the meta-analysis of the prognostic value of genes (27). It was used here to validate the use of LAMP2 in prognosis in cancers using the GEO dataset.

TISIDB (http://cis.hku.hk/TISIDB/) is a website for exploring tumor-immune system interactions that integrates numerous data types (28). In this study, TISIDB was used to construct a heat map showing the correlations among immunomodulators, lymphocytes, chemokines (or receptors), and gene expression. In addition, TISIDB was used to explain the correlations of LAMP2 expression with immune subtypes and molecular subtypes of tumors.

The tumor immune dysfunction and exclusion (TIDE) algorithm (http://tide.dfci.harvard.edu/query/) provides data to support the studies of T-cell dysfunction and immunotherapy resistance in cancer based on large clinical datasets (29). In this study, the general predictive ability of LAMP2 with respect to treatment response in different cancer types was compared with those of nine standardized biomarkers for tumor immune response using the “Biomarker Evaluation” model, including TIDE, the microsatellite instability (MSI) score, tumor mutational burden (TMB), cluster of differentiation 274 (CD274), cluster of differentiation 8 (CD8), interferon-γ (IFNG), T-cell clonality (T.Clonality), B-cell clonality (B.Clonality), and Merck18 (29, 30). p < 0.05 was defined as statistically significant.

OncoLnc (http://www.oncolnc.org/) is a tool that allows the interactive exploration of survival correlations based on the survival data of 8,647 patients from 21 cancer studies conducted in TCGA (31). In this work, the survival analysis for LAMP2 in ESCA was performed using OncoLnc.

The LAMP2 interaction network was constructed using the GeneMANIA database (32) (http://www.genemania.org). The STRING online database (33) (https://string-db.org/) was used to construct the LAMP2 protein–protein interaction network.

To predict the function of LAMP2 and its associated pathways, we performed a correlation analysis between LAMP2 and other genes in ESCA using TCGA data. GO analysis is an efficient bioinformatics tool for identifying the biological processes, cellular components, and molecular functions linked to the gene of interest. GSEA is a computational method for determining statistical differences in the expression status of a set of genes between two organisms (34). GSEA was used to investigate the potential mechanism of LAMP2. We further explored potential functional pathways based on the top 250 genes using the clusterProfiler R package (35) (Version 3.14.3). Adjusted P < 0.05 was considered to indicate the meaningful enrichment of a pathway.

Tumor specimens were obtained from 40 consecutive patients undergoing a single surgical resection in the First Affiliated Hospital of Shantou University from 2020 to 2021. None of the patients had received preoperative chemotherapy or radiotherapy. Ethical approval was granted by the clinical research ethics committee of the First Affiliated Hospital of Shantou University. Tumor tissues were fixed in 10% formalin and embedded in paraffin. Paraffin-embedded 4 µm thick tissue sections were automatically immunohistochemically stained using a Ventana BenchMark XT immunostainer (Ventana Medical Systems, Tucson, AZ, United States) with a basic Diaminobenzidine (DAB) Kit (Ventana Cat, Tucson, USA). The specimens were diluted to 1:400 with a prediluted polyclonal anti-lamp2 Rabbit polyclonal antibody (pAb) (Wanleibio, Shenyang, China) and incubated for 24 min. Specimens were restained with hematoxylin.

Immunohistochemical sections were visualized and analyzed after full-slide digitization using the Panoramic Scan and Image Pro-Plus (IPP) software. The density means and integrated optical density (IOD) of IPP are representative parameters for assessing immunostaining quantification, allowing for an increased sensitivity of scoring and enabling a more reliable and reproducible protein expression analysis. To further compare the expression of LAMP2 in other cancers, the expression of LAMP2 in five tumor types and the corresponding normal tissues was validated using The Human Protein Atlas (THPA) (https://www.proteinatlas.org/) database (36). Three pairs of samples (cancer and normal tissue) were downloaded for each type of cancer. The density means and IOD of IPP were calculated. The GraphPad Prism software (version 5.0) was used to perform unpaired t-tests (Student’s t-test) on the average IOD values obtained from the cancer and normal tissues. p < 0.05 was defined as statistically significant.

The R package (version 3.6.3) was used for performing all statistical tests, and the ggplot2 package (3.3.3 version) was used for visualization. Kaplan–Meier survival analyses were performed with the “survival R” and “survminer R” packages in the R software. The ROC analysis was performed with the qROC package (version 1.17.0.1). The immune infiltration algorithm (ssGSVA) in the GSVA package (version 1.34.0) was used to calculate immune scores (37). LAMP2 differential expression analysis was performed using the DESeq2 package (38) (version 1.26.0). t-test or Wilcoxon rank sum test was used for continuous variables and Pearson’s chi-square test for categorical variables. p < 0.05 was considered to indicate statistical significance (ns, p ≥ 0.05; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001).

LAMP2 expression was identified using the data from different GEO datasets. The expression of LAMP2 was higher in ESCC tissues than in adjacent normal tissues in the GSE33426 ( Figure 1A ), GSE45670 ( Figure 1B ), GSE53625 ( Figure 1C ), and GSE23400 ( Figure 1D ) datasets. Furthermore, based on TCGA data, LAMP2 protein expression was higher in EAC ( Figure 1E ) and ESCA ( Figures 1F, G ) compared with adjacent normal tissues. The statistical analyses performed on ESCA samples and adjacent normal tissue samples from TCGA showed that LAMP2 expression had an effective predictive value (area under the curve [AUC] = 0.939) ( Figure 1H ). These findings indicate that LAMP2 expression is upregulated in ESCA and suggest an important regulatory role of LAMP2 in ESCA progression.

We investigated the correlations between LAMP2 expression and clinicopathological features in ESCA including the pathologic stages, N stage, M stage, residual tumor, primary therapy outcome, gender, race, age, body mass index (BMI), histological type, radiation therapy, tumor central location, Barrett’s esophagus, columnar metaplasia, and T stage. A high expression of LAMP2 was significantly associated with pathologic stages II and III ( Figure 2A ), N0 and N1 stages ( Figure 2B ), M0 and M1 stages ( Figure 2C ), residual tumor (R0, R1, and R2) ( Figure 2D ), complete response ( Figure 2E ), gender (female, male) ( Figure 2F ), race (Asian, White, Black, or African-American) ( Figure 2G ), age (≤60, >60) ( Figure 2H ), BMI (≤25, >25) ( Figure 2I ), histological type (squamous cell carcinoma) ( Figure 2J ), radiation therapy ( Figure 2K ), tumor central location (distal, mid, proximal) ( Figure 2L ), Barrett’s esophagus (no) ( Figure 2M ), columnar metaplasis (no) ( Figure 2N ) and T2 stage ( Figure 2O ). These correlations of LAMP2 expression with various clinicopathological features highlight that more attention should be paid to the patients with certain concomitant clinical traits, such as an age over 60 years, the absence of Barrett’s esophagus, or distal central location of the tumor.

Next, we downloaded prognostic information for patients with EAC and ESCA from TCGA to analyze the prognostic value of LAMP2. A high expression of LAMP2 in patients with EAC was significantly associated with worse OS ( Figure 3A ), PFI ( Figure 3B ), and DSS ( Figure 3C ). Subsequently, the prognostic information of patients with ESCA was obtained from TCGA, GEPIA2, and Oncolnc to analyze the prognostic value of LAMP2. We investigated the relationship between LAMP2 expression and prognosis in patients with ESCA. The results showed that a high expression of LAMP2 was significantly related to worse DSS ( Figure 3D ) in patients with ESCA from TCGA. In the GEPIA2 and Oncolnc data, we found that a high expression of LAMP2 was associated with worse OS in patients with ESCA ( Figures 3E, F ). These results suggest that LAMP2 is significantly associated with the prognosis of ESCA.

Kaplan–Meier analysis was used to better understand and explore the correlations of LAMP2 expression with various clinical characteristics in patients with ESCA. As shown in Figure 3G , high LAMP2 expression, adenocarcinoma, and a history of alcohol use were significantly correlated with worse OS and DSS ( Figure 3G ). Moreover, we found that LAMP2 expression was significantly associated with poor DSS in ESCA patients >60 years of age or without Barrett’s esophagus or a distal tumor central location ( Figure 3G ). The upregulation of LAMP2 expression was only associated with poorer OS in patients with histological grade 2 and those without columnar metaplasia ( Figure 3G ). These results suggest that LAMP2 expression has a prognostic value in ESCA.

GeneMANIA was used to construct a gene–gene interaction network for LAMP2. As shown in Figure 4A , the 20 genes closely related to LAMP2 were identified, including LAMP1, CD60, IMPDH1, LAMP3, and LAMP5 ( Figure 4A ). A protein–protein interaction network for LAMP2 was established using the STRING database ( Figure 5B ). There were 27 edges and 11 nodes, including SLC40A1, TFR2, and HFE (P<0.001) ( Figure 4B ).

Median LAMP2 mRNA expression was used to divide samples into high-expression and low- expression groups. A total of 708 DEGs were identified, including 485 upregulated and 223 downregulated genes, between the LAMP2 low- and high-expression groups (|log fold change| > 2, P < 0.05) ( Figure 4C ). Heat maps were used to illustrate the associations between the top 50 related genes and LAMP2 expression levels. As shown in the heat map of positive correlations ( Figure 5A ), the expression level of LAMP2 was positively associated with those of LINC02457 (r = 0.555) ( Figure 5B ), AC005865.1 (r = 0.543) ( Figure 5C ), LRRC38 (r = 0.529) ( Figure 5D ), cyclic nucleotide-gated channel subunit beta 1 (CNGB1; r = 0.524) ( Figure 5E ), EFCAB1 (r = 0.523) ( Figure 5F ), KREMEN2 (r = 0.514) ( Figure 5G ), AC099066.2 (r = 0.510) ( Figure 5H ), PHF24 (r = 0.508) ( Figure 5I ), LAMA1 (r = 0.505) ( Figure 5J ), and SLC6A2 (r = 0.502) ( Figure 5K ). Among the negatively correlated genes ( Figure 6A ), the top 10 were AGR3 (r = −0.576) ( Figure 6B ), CDC42EP5 (r = −0.562) ( Figure 6C ), CLRN3 (r = −0.534) ( Figure 6D ), SMIM24 (r = −0.532) ( Figure 6E ), the polymeric immunoglobulin receptor (PIGR; r = −0.530) ( Figure 6F ), IHH (r = −0.517) ( Figure 6G ), DMBT1 (r = −0.510) ( Figure 6H ), MIR3131 (r = −0.503) ( Figure 6I ), MS4A8 (r = −0.503) ( Figure 6J ), and TM4SF5 (r = −0.503) ( Figure 6K ).

To better comprehend the characteristics of LAMP2 and its related pathways, we performed a correlation analysis on LAMP2 and the DEGs in ESCA using TCGA data. The top 250 linked genes most positively correlated with LAMP2 were picked for enrichment analysis. We further studied the potential functional pathway based on the top 250 genes using the clusterProfiler R package. According to the GO enrichment analysis, the primary biological processes included a humoral immune response, an antimicrobial humoral response, epidermis development, digestion, epidermal cell differentiation, and keratinocyte differentiation. The main cellular components were the immunoglobulin complex, blood microparticle, and cornified envelope. The molecular functions were principally related to carbohydrate binding, glucuronosyltransferase activity, and peptidoglycan binding ( Figures 7A–C ). In addition, KEGG pathway analysis identified the enrichment and crosstalk of the top 250 genes in protein digestion and absorption and retinol metabolism ( Figure 7D ). GSEA was used to search for reactome and KEGG pathways: the ABC transporter in lipid homeostasis; acetylcholine binding and downstream events; activation of the mRNA upon the binding of the cap-binding complex and eukaryotic initiation factors (eIFs), and subsequent binding to 43S; and activation of the TFAP2 (AP-2) family of transcription factors were extremely enriched ( Figure 7E ). Alanine aspartate and glutamate metabolism, aldosterone-regulated sodium reabsorption, allograft rejection, and the allograft rejection pathway were significantly enriched in the KEGG analysis ( Figure 7F ). These results strongly suggest that LAMP2 is involved in the regulation of the immune response in ESCA.

Spearman correlation analysis showed that the expression level of LAMP2 in ESCA was related to immune cell infiltration levels as quantified by ssGSVA. The size of the dot in Figure 8A shows the absolute value of Spearman’s r, indicating that infiltrating cells were correlated with LAMP2 in the differential expression analysis. Specifically, LAMP2 was positively correlated with TCM (T central memory) cells, Th2 cells, and natural killer (NK) cells ( Figure 8A ) and negatively correlated with Th17 cells, pDC, NK CD56 bright cells, and eosinophils ( Figure 8A ). The infiltration of TCM and Th17 cells in the LAMP2 differential expression analysis was statistically significant (P<0.001) ( Figures 8B, C ). The relative enrichment scores of TCM and Th17 cells showed significantly positive and negative correlations with LAMP2 expression levels, respectively (P<0.001) ( Figure 8D ). Next, we further analyzed the relationship between LAMP2 expression and the abundance of 28 TILs using TISIDB. Figure 9A shows the relationships between TPM4 expression and LAMP2 in different types of cancer. In particular, LAMP2 expression was significantly associated with multiple types of TILs ( Figures 9B–I ) in ESCA. LAMP2 expression was meaningfully negatively associated with infiltrating levels of monocytes (rho = −0.348, p = 1.4e-06) ( Figure 9B ), Th17 cells (rho = −0.354, p = 8.98e-07) ( Figure 9C ), CD56dim cells (rho = −0.284, p = 9.73e-05) ( Figure 9D ), Act_B cells (rho = −0.152, p = 0.0385) ( Figure 9E ), Act_CD8 cells (rho = −0.15, p = 0.0417) ( Figure 9F ), Act_DC cells (rho = −0.183, p = 0.0127) ( Figure 9G ), and eosinophils (rho = −0.24, p = 0.00104) ( Figure 9H ) and positively correlated with the infiltrating levels of Th2 cells (rho = 0.146, p = 0.048) ( Figure 9I ) in ESCA.

The TIMER database was used to analyze LAMP2 mRNA expression in 25 commonly occurring types of human cancer. LAMP2 expression was significantly upregulated in 24 cancer types, BRCA, CESC, CHOL, COAD, GBM, HNSC, KICH, KIRC, KIRP, LAML, LGG, LIHC, LUAD, LUSC, OV, PAAD, PRAD, READ, SKCM, STAD, TGCT, THCA, THYM, and UCEC, whereas it was downregulated only in DLBC ( Figure 10A ). Then, LAMP2 expression levels were determined using the GEPIA2 database. We found that LAMP2 was significantly more highly expressed in eight human tumor types (GBM, LGG, LIHC, PAAD, PRAD, SKCM, STAD, and READ) compared with normal tumor tissues, a result consistent with those shown in Figure 10A ( Figure 10B ). LAMP2 expression was significantly upregulated in 11 cancer types (BLCA, BRCA, CHOL, COAD, HNSC, KIRC, LIHC, LUAD, LUSC, READ, and STAD ( Figure 10C ). In addition, we analyzed the protein expression of LAMP2 across 10 cancer subtypes in CPTAC samples based on UALCAN data ( Figure 10D ). In CPTAC samples, high LAMP2 expression was correlated strongly with subtypes k3 and k10. These findings suggest that LAMP2 may have an important regulatory role in the progression of various cancers.

ROC curves were constructed to assess the diagnostic value of LAMP2 in other cancers. The results showed that LAMP2 had a certain accuracy (AUC > 0.7) in predicting 20 cancer types: PAAD (AUC = 0.970) ( Figure 11A ), READ (AUC = 0.938) ( Figure 11B ), CHOL (AUC = 0.938) ( Figure 11C ), GBM (AUC = 0.927) ( Figure 11D ), STAD (AUC = 0.913) ( Figure 11E ), GBMLGG (AUC = 0.889) ( Figure 11F ), LGG (AUC = 0.879) ( Figure 11G ), COAD (AUC = 0.868) ( Figure 11H ), LAML (AUC = 0.862) ( Figure 11I ), LUSC (AUC = 0.842) ( Figure 11J ), THCA (AUC = 0.816) ( Figure 11K ), LIHC (AUC = 0.815) ( Figure 11L ), SKCM (AUC = 0.803) ( Figure 11M ), BRCA (AUC = 0.791) ( Figure 11N ), HNSC (AUC = 0.765) ( Figure 11O ), OV (AUC = 0.765) ( Figure 11P ), OSCC (AUC = 0.757) ( Figure 11Q ), PRAD (AUC = 0.756) ( Figure 11R ), TGCT (AUC = 0.734) ( Figure 11S ), and LUAD (AUC=0.702) ( Figure 11T ). LAMP2 had high accuracy in predicting PAAD, READ, CHOL, GBM, and STAD (AUC > 0.9). In addition, LAMP2 had a better reliability in the remaining cancers (0.7 < AUC < 0.9). These results suggest that LAMP2 has valid pan-cancer diagnostic value.

The expression levels of LAMP2 were particularly correlated with the PFI, DSS, and OS of patients with GBMLGG, LGG, SARC, and BLCA. Higher LAMP2 expression was associated with worse PFI, DSS, and OS (P < 0.001) in GBMLGG ( Figures 12A, F, K ); worse PFI (P < 0.01), DSS (P < 0.01) and OS (P < 0.001) in LGG ( Figures 12B, G, L ); worse PFI (P < 0.05), DSS (P < 0.01), and OS (P < 0.01) in SARC ( Figures 12C, H, M ); worse PFI (P < 0.01), DSS (P < 0.01), and OS (P < 0.05) in BLCA ( Figures 12D, I, N ); worse PFI (P < 0.05) in KIRP ( Figure 12E ); worse DSS (P < 0.05) ( Figure 12J ) and OS (P = 0.01) ( Figure 12O ) in MESO; and worse OS (P < 0.05) in BRCA and OV ( Figures 12P, Q ). In addition, we validated the prognostic value of LAMP2 in a variety of cancers including LUAD, GBMLGG, breast cancer, colorectal cancer, blood cancer (follicular lymphoma), and soft tissue cancer (liposarcoma) using PrognoScan based on the GEO dataset ( Supplementary Table 1 ). The results verified the prognostic value of LAMP2 on a pan-cancer basis.

To evaluate the relevance of LAMP2 to immunity in cancer progression, we explored the relationships between LAMP2 expression and immunoinhibitors, immunostimulators, chemokines, and receptors in human heterogeneous carcinomas based on TISIDB. As shown in Figure 13 , LAMP2 expression levels were positively correlated with various immunoinhibitors (including CD274, the colony-stimulating factor 1 receptor [CSF1R], and TGFBR1) ( Figure 13A ), several immune stimulators (including CD80, CD86, and IL6R) ( Figure 13C ), various chemokines (including CXCL8, CXCL10, and CXCL11) ( Figure 13E ), and multiple receptors (including CCR1, CCR2, and CX3CR1) ( Figure 13F ). In contrast, the immunoinhibitors ADORA2A and CD160, the immunostimulator TNFRSF25, and the receptor CCR10 were significantly negatively correlated with LAMP2 expression ( Figures 13A, C, F ). Overall, as shown in Figure 13B , there were significant pan-cancer positive correlations between LAMP2 expression and infiltration of six immune cell types (B cells, CD4+ T cells, CD8+ T cells, macrophages, neutrophils, and dendritic cells). Only six tumor types (BRCA, BRCA-luminal, COAD, KIRC, LGG, and LUAD) were consistently associated with the six immune cell types. The strongest correlations were with B cells in KTCH (r = 0.57, p < 0.01) and THCA (r = 0.52, p < 0.01); with CD4+ T cells in THCA (r = 0.47, p < 0.01) and HNSC-HPVpos (r = 0.44, p < 0.01); with CD8+ T cells in PAAD (r = 0.65, p < 0.01), KICH (r = 0.65, p < 0.01); and PRAD (r = 0.59, p < 0.01); with dendritic cells in PAAD (r = 0.58, p < 0.01) and DLBC (r = 0.55, p < 0.01); with macrophages in PAAD (r = 0.66, p < 0.01) and THCA (r = 0.51, p < 0.01); and with neutrophils in DLBC (r = 0.69, p < 0.01) and SKCM-Primary (r = 0.54, p < 0.01) (all data are from the TIMER2.0 database, and values are given to two decimal places). By contrast, five tumor types (ESCA, KICH, KIRP, THYM, and UCEC) were significantly negatively associated with CD4+ T cells, and only THCA was negatively associated with CD8+ T cells. None of the other cancer types showed any significant correlation (p > 0.05) between LAMP2 expression levels and tumor infiltration by immune cells ( Figure 13B ).

We assessed the biomarker relatedness of LAMP2 by comparing it using standardized biomarkers based on predictive power response outcomes and OS in the immune checkpoint blockade (ICB) subcohort. These results showed that LAMP2 alone had an AUC value >0.5 in 12 of the 25 ICB subcohorts ( Figure 13D ). LAMP2 showed a higher predictive value than TMB, T. Clonality, and B. Clonality, which respectively had AUC values >0.5 in seven, nine, and seven ICB subcohorts. However, the predictive value of LAMP2 was lower than those of TIDE, MSI score, CD274, CD8, IFNG, and Merck 18. In summary, LAMP2 is related to immune modulatory factors in a variety of tumor types.

An investigation of the correlations between LAMP2 expression and tumor molecular subtypes and immune subtypes based on TISIDB showed that LAMP2 expression was meaningfully linked with several different immune subtypes (C1: wound healing, C2: IFN-gamma dominant, C3: inflammatory, C4: lymphocyte depleted, C5: immunologically quiet, and C6: TGF-b dominant) in five cancer types ( Figure 14A ): BLCA ( Figure 14B ), BRCA ( Figure 14C ), LIHC ( Figure 14D ), STAD ( Figure 14E ), and TGCT ( Figure 14F ). We also found that LAMP2 was differentially expressed in the diverse molecular subtypes of various tumors ( Figure 14G ), including BRCA, COAD, STAD, LIHC, and ESCA. Specifically, LAMP2 expression was higher in the Her2 molecular subtype of BRCA ( Figure 14H ), the CIN molecular subtype of COAD ( Figure 14I ), HM-SNV molecular subtype of STAD ( Figure 14J ), iCluster:2 molecular subtype of LIHC ( Figure 14K ), and ESCC subtype of ESCA compared with the other molecular subtypes of the respective tumor types ( Figure 14L ). In conclusion, LAMP2 is highly expressed in the immune and molecular subtypes of some cancer types.

We performed immunohistochemical staining and found that the expression levels of LAMP2 protein in ESCC tissues were significantly higher than those in corresponding normal tissues ( Figures 15A, C ). The subcellular localization of LAMP2 in cancer cells indicated that it was predominantly expressed in cytoplasmic lipid droplets ( Figure 15B ). Finally, the results further confirmed the differences in LAMP2 expression between the tissues of five tumor types and the corresponding normal tissues from THPA and showed that LAMP2 was significantly highly expressed in gliomas ( Figure 15D ), BLCA ( Figure 15E ), BRCA ( Figure 15F ), OV ( Figure 15G ), and KIRP ( Figure 15H ).