Gene expression analysis
To determine the expression differences of PTPRD in tumor tissues and normal tissues, PTPRD mRNA levels in more than 10,000 tumor and normal tissue samples across 23 cancer types from the TCGA database were evaluated using the TIMER database (Supplementary Table 1, Supplementary Table 2). As shown in Fig. 1A, PTPRD was downregulated in lung adenocarcinoma (LUAD), uterine corpus endometrial carcinoma (UCEC), kidney renal clear cell carcinoma (KIRC), bladder carcinoma (BLCA), head and neck carcinoma (HNSC) HPV+, liver hepatocellular carcinoma (LIHC), lung squamous carcinoma (LUSC), glioblastoma multiforme (GBM), thyroid carcinoma (THCA), prostate adenocarcinoma (PRAD), cervical squamous carcinoma (CESC), and stomach adenocarcinoma (STAD) compared with the adjacent normal tissues. Besides, PTPRD was upregulated in head and neck carcinoma (HNSC), and kidney renal papillary cell carcinoma (KIRP) compared to the normal tissues. Moreover, we further explored PTPRD expression across 33 cancer types using the GEPIA2 database, containing the TCGA and GTEx data. As shown in Fig. 1B, PTPRD expression was lower in BLCA, CESC, KIRC, LUAD, LUSC, ovarian serous cystadenocarcinoma (OV), THCA, UCEC, and testicular germ cell tumors (TGCT) compared with the normal tissues, and PTPRD expression was higher in rectum adenocarcinoma (READ) and thymoma (THYM). Moreover, a significant correlation between PTPRD expression and the pathological stages was observed using the GEPIA2 database in multiple cancers, including BLCA, KIRC, and skin cutaneous melanoma (SKCM) (Fig. 1C). These findings suggested that PTPRD might perform diverse functions in cancer progression. Also, the UALCAN database was applied to analyze PTPRD protein expression across tumor and normal cases in the CPTAC datasets, and the results showed that lower protein expression of PTPRD was observed in UCEC, LUAD, clear cell renal cell carcinoma (ccRCC), ovarian cancer, and breast cancer compared with adjacent normal tissues (Fig. 1D).
Survival analysis
To investigate the prognostic value of PTPRD across multiple cancers from the TCGA project, the impact of PTPRD expression on survival was evaluated using the GEPIA2 database. As shown in Fig. 2A, low PTPRD expression was associated with a better OS (overall survival) for STAD, uveal melanoma (UVM), SKCM, and BLCA, while low PTPRD expression was related to the poor OS for READ and brain lower-grade glioma (LGG). Besides, low PTPRD expression was linked to a better DFS (disease-free survival) for UVM and KIRP, while low PTPRD expression was related to poor DFS for KIRP. To further investigate the prognostic potential of PTPRD in different tumors, the PrognoScan database was used to assess the relationship between PTPRD expression and prognosis of patients with different cancers based on the data retrieved from Gene Expression Omnibus (GEO) datasets (Supplementary Table 3). The results exhibited that PTPRD expression affected the survival of breast cancer, ovarian cancer, lung cancer, and liposarcoma. As shown in Fig. 2B, the cohort (GSE8894, N = 138, P = 0.029) of lung cancer patients with low PTPRD expression exhibited a poorer relapse-free survival. Besides, the cohort (GSE30929, N = 140, P = 0.003) containing 140 samples of liposarcoma showed that low PTPRD expression was observably associated with improved distant recurrence-free survival (Fig. 2C). Similarly, the GSE26712 (N = 185, P = 0.002) and GSE9891 (N = 278, P = 0.001) cohorts containing 185 and 278 ovarian cancer cases, respectively, showed that low PTPRD expression was significantly associated with improved overall survival (Fig. 2D and Fig. 2E). Additionally, patients with low PTPRD expression had poorer disease-specific survival in the GSE3494-GPL96 cohort (N = 236, P = 0.003) (Fig. 2F), and patients with low PTPRD expression was correlated with improved distant metastasis-free survival in the GSE2034 cohort (N = 286, P = 0.030) (Fig. 2G). Therefore, it is conceivable that PTPRD expression has an impact on the prognosis of multiple cancers.
Genetic alteration analysis
It has been broadly recognized that genomic alteration is closely linked to tumorigenesis. To investigate the genomic aberration of PTPRD in pan-cancer, a combined study containing 10967 samples from 32 studies was extracted from the TCGA database for the genetic alteration analysis (Supplementary Table 4). The mutation frequency and genetic alteration type of PTPRD were explored in various cancers (Fig. 3A). The mutation frequency of PTPRD was 9.5% (1045/10950) in the TCGA pan-cancer cohort with stomach adenocarcinoma (25.2%) ranking first followed by skin cutaneous melanoma (25.0%), lung adenocarcinoma (23.1%), and head and neck squamous cell carcinoma (19.1%). Besides, the genetic alteration profiling of PTPRD exhibited that its deep deletion was one of the most significant factors for alteration in stomach adenocarcinoma, cholangiocarcinoma, testicular germ cell tumors, head and neck squamous cell carcinoma, lung squamous cell carcinoma, bladder urothelial carcinoma, and lung adenocarcinoma. As shown in Fig. 3B, PTPRD mutations have abundant types of mutation alteration, including missense, nonsense, silent mutation, insertion or deletion, duplication, and frameshift mutation, and G203E, S431L, R705Q, and L1053I sites were observed many times. Moreover, the association between genetic aberrations of PTPRD and the survival prognosis of cases with various types of tumors has been investigated, and the results exhibited that kidney renal clear cell carcinoma cases with altered PTPRD mutations had worse prognosis in disease-specific (P = 1.780e-5), progression-free (P = 3.899e-4) and overall (P = 6.366e-4) survival, but not disease-free (P = 0.595) survival, compared with cases with unaltered PTPRD (Fig. 3C, 3D, 3E).
The relationship between PTPRD mutation and response to immunotherapy in pan-cancer
A combined cohort of 2185 immune checkpoint blockade (ICB)-treated patients (249 from Dana Farber Cancer Institute (DFCI) and 1936 from Memorial Sloan Kettering Cancer Center (MSKCC)) were further used to investigate the relationship between PTPRD mutation and response to ICB therapy (Supplementary Table 5). A comparatively high proportion of PTPRD mutation cases occurred in patients with melanoma (20.59% of 471 cases) and non-small cell lung cancer (12.48% of 681 cases) (Fig. 4A). Moreover, a significant difference in the number of coexisting mutations between patients with PTPRD mutation type and PTPRD wild type. For instance, PTPRD mutation coexisted more mutated NF1 (36.84% VS. 8.89%), PTPRT (34.65% VS. 8.02%), GRIN2A (28.95 VS. 5.88%), PAK5 (25.55% VS. 4.70%) and FLT1 (21.49% VS. 3.88%) (Fig. 4B). Besides, TP53, NF1, PTPRT, TERT, PREX2, KMT2D, GRIN2A, KMT2C, FAT1, PAK5 mutations are high-frequency mutations that occur in both PTPRD mutation and PTPRD wild group (Fig. 4C).
In the cohort of ICB-treated non-small cell lung cancer (N = 406), the OS of the PTPRD-mutant patients (N = 49) was better than that of those without the PTPRD mutation (P = 0.011) (Fig. 4D). Significant differences were found in the distribution of TMB Score (N = 350, P < 10e-10), mutation count (N = 673, P < 10e-10), mutation rate (N = 240, P = 3.78e-10), smoker (N = 296, P = 6.74e-8), and RECIST (N = 56, P = 0.002) between the PTPRD-mutation and PTPRD-wildtype groups in the ICB-treated non-small cell lung cancer cohort (Fig. 4E). The ICB-treated non-small cell lung cancer patients with PTPRD mutation have higher TMB scores, more mutation counts, and increased mutation rate. Moreover, the proportion of former smokers (10.26% VS. 9.73%), ever smokers (82.05% VS. 62.65%) and current smokers (5.13% VS. 4.67%) in ICB-treated non-small cell lung cancer patients with PTPRD mutation is higher than that of patients without PTPRD mutation (Fig. 4F). Of the six non-small cell lung cancer patients with PTPRD mutation who received immunotherapy, one patient had a complete response (CR), four patients achieved partial response (PR), and one obtained stable disease (SD). Among the 50 non-small cell lung cancer patients without PTPRD mutation who received immunotherapy, 12 patients (24%) achieved partial response (PR) and 20 patients (40%) had stable disease (SD) (Fig. 4G).
For the melanoma patients (N = 471) who received immune checkpoint inhibitors, better OS could be observed in the PTPRD-mutation group (N = 97, P = 0.005) (Fig. 4H). In a subgroup of 471 melanoma patients from the ICB-treated cohort, patients with PTPRD mutation had more mutation counts, and TMB score was significantly higher in PTPRD-mutated patients compared with that in PTPRD-wildtype patients (Fig. 4I).
Immune infiltration analysis
Immune infiltration was closely associated with the occurrence and development of cancer. Therefore, the relationship between PTPRD expression and the level of immune cell infiltration in various cancers of TCGA was evaluated by investigating the coefficient of PTPRD expression and immune infiltration level using the TIMER database. The findings revealed that PTPRD expression had significant correlations with tumor purity in multiple cancers, including UVM, LUAD, and LUSC. In UVM, the PTPRD expression level was negatively correlated with neutrophil (r=-0.263, P = 0.021), and positively correlated with CD8 + T cells (r = 0.251, P = 0.028) (Fig. 5A). In LUAD, PTPRD expression was positively correlated with CD4 + T cells (r = 0.256, P = 1.16e-08), macrophages (r = 0.362, P = 1.86e-16), neutrophils (r = 0.287, P = 1.30e-10), and dendritic cells (r = 0.302, P = 9.47e-12) (Fig. 5B). In LUSC, PTPRD expression was positively correlated with macrophages (r = 0.174, P = 1.28e-04) (Fig. 5C). Moreover, we assessed the relationship between PTPRD mutation and immune cell infiltration level in various cancers of TCGA using the TIMER database, and the results showed that the mutated PTPRD status was correlated with high infiltrating levels of CD4 + T cells (P = 0.026) in SKCM (Fig. 5D). Additionally, the TISIDB database was used to further evaluate the abundance of tumor-infiltrating lymphocytes (TILs) and PTPRD mutation, and the results showed that PTPRD mutation was positively correlated with memory B cell abundance (P = 0.015), and negatively correlated with plasmacytoid dendritic cell (P = 0.009) in LUAD (Fig. 5E). Therefore, the findings revealed that PTPRD was closely linked with immune infiltration.
Protein-protein interaction network analysis
To further investigate the potential mechanisms of PTPRD, we conducted a PPI network of PTPRD based on the top 50 PTPRD-related genes with the STRING database (Supplementary Table 6). 51 nodes and 227 edges were obtained in the PPI network (avg. local clustering coefficient: 0.801, PPI enrichment p-value: < 1.0e-16) and, the combined score of SLITRK1, IL1RAPL1, SLITRK2, IRS1, and PPFIA1 with PTPRD is > 0.5 in the PPI network (Fig. 6A). Based on the GEPIA2 database, the expression levels of IRS1 were negatively correlated with PTPRD, and the expression levels of SLITRK1, IL1RAPL1, SLITRK2, and PPFIA1 were positively correlated with PTPRD (Fig. 6B). The findings indicated that SLITRK1, IL1RAPL1, SLITRK2, IRS1, and PPFIA1 were closely related to the modulation and function of PTPRD.
Functional enrichment analysis
To further explore the underlying mechanism of PTPRD in tumorigenesis, we identified the targeting PTPRD-binding proteins and the PTPRD expression-correlated genes for functional enrichment analysis. We gathered a total of 50 PTPRD-binding proteins using the STRING database, which was validated by experimental evidence (Fig. 6A). Moreover, we used the GEPIA2 database to integrate all TCGA tumor expression data and obtained the top 100 genes related to PTPRD expression (Supplementary Table 7). As shown in Fig. 7A, PTPRD were significantly positively correlated with GATS (GATS, stromal antigen 3 opposite strand) (R = 0.58, P < 0.001), NRXN1 (neurexin 1) (R = 0.58, P < 0.001), QKI (QKI, KH domain containing, RNA binding) (R = 0.59, P < 0.001), RP11 (re-mRNA processing factor 31) (R = 0.58, P < 0.001), and AS1 (prostaglandin D2 receptor (DP)) (R = 0.75, P < 0.001). The top 50 PTPRD-binding proteins were obtained from the PTPRD -target network which was constructed by the STRING database, and the top 100 genes related to PTPRD were screened from the GEPIA2 database. As shown in Fig. 7B, the overlapping target related to PTPRD was SLITRK2 (SLIT and NTRK-like family, member 2) by drawing a Venn diagram. The corresponding heatmap data showed a positive correlation between PTPRD and SLITRK2 in various cancers (Fig. 7C). We combined the top 50 PTPRD-binding proteins and the top 100 genes related to PTPRD to perform KEGG and GO enrichment analyses using the metascape database. The GO enrichment analysis suggested that most of these genes are related to the pathways or cellular biology of RNA metabolisms, such as neuron projection morphogenesis, synapse organization, presynapse, axon, positive regulation of nervous system development, and others (Fig. 7D). The KEGG enrichment analysis indicated that cell adhesion molecules, foxo signaling pathway, AGE-RAGE signaling pathway in diabetic complications, Wnt signaling pathway, GABAergic synapse, hedgehog signaling pathway, neurotrophin signaling pathway, and adherens junction might be involved in the effect of PTPRD on tumor pathogenesis (Fig. 7E).