PTPRD/PTPRT mutation correlates to treatment outcomes of immunotherapy and immune landscape in pan-cancers

Abstract

Objective

PTPRD and PTPRT are phosphatases of the JAK-STAT pathway related to immunotherapy. However, the role and mechanism of PTPRD and PTPRT mutations in multiple cancers remains unclear.

Methods

Clinical data and PTPRD/PTPRT mutation information from 12 cohorts were collected and classified as a discovery cohort and three validation cohorts. The association between PTPRD/PTPRT mutations and immunotherapeutic efficacy was analyzed. Then, the association between PTPRD/PTPRT mutation and immune profiles was analyzed using The Cancer Genome Atlas (TCGA) cohort.

Results

A total of 2,392 patients across 20 cancer types were included in this study. Our results showed that patients harboring PTPRD/PTPRT mutation, especially co-mutations, had a significantly elevated response rate to immunotherapy in multiple cancers. Patients with PTPRD/PTPRT mutation had a higher objective response rate (ORR) (P=0.002), longer overall survival (OS) (P=0.005) and progression-free survival (PFS) (P=0.038). Importantly, the above findings were further verified in validation cohorts. In addition, we found that the PTPRD/PTPRT co-mutations (co-mut) subgroup exhibited an immune-activated phenotype, the wild-type subgroup tended to have an immune-desert phenotype, and the uni-mutation (uni-mut) subgroup might have an immune-mixed phenotype. Our further analyses suggested that combining programmed cell death ligand 1 (PD-L1) expression and PTPRD/PTPRT mutation can be used to screen patients who may benefit from immunotherapy.

Conclusions

PTPRD/PTPRT mutation could serve as a potential predictive biomarker for cancer immunotherapy.

Introduction

Immunotherapy has emerged as an important therapeutic strategy to inhibit tumor growth by activating anti-tumor immune response and curbing the immune escape of tumor cells (1). Nivolumab combined with chemotherapy increased the overall survival (OS) of patients with gastric cancer from 11±1 months to 13±1 months (2). Besides, among patients with resectable non-small cell lung cancer (NSCLC), neoadjuvant nivolumab plus chemotherapy resulted in a significantly longer event-free survival of 10.8 months and a significantly higher proportion of patients with a pathological complete response rate than chemotherapy alone (3). Furthermore, programmed cell death 1 (PD-1) blockade plus chemotherapy has been recommended as the most important strategy in multiple cancers (4-7). However, a subset of cancer patients respond poor to immunotherapy, so reliable biomarkers must be explored to identify the potential population who will benefit from immunotherapy (8-10).

Although some biomarkers have been used as potential predictors of immunotherapy, such as programmed cell death ligand 1 (PD-L1) expression (CPS), tumor mutation burden (TMB) and tumor immune dysfunction and exclusion (TIDE) score (11-14). These markers have some limitations (8,15). PD-L1 is considered to be the most commonly used indicator for PD-1 inhibitors however, there are still some inconsistencies regarding to the different antibodies and scoring criteria (10,16). TIDE has only been applied to gene expression profile and cannot yet be used for individual prediction. Owing to its high stability, genomic analysis is an effective approach for predicting immunotherapy, the most representative marker of which is TMB. While each mutation was given the same weight, confounding factors might be incorporated. Moreover, TMB requires the implementation of whole-exome sequencing or next-generation sequencing panels, which are more expensive and require higher DNA sample quality. Therefore, identifying key mutations that predict immunotherapy may be an ideal approach for screening patients who are sensitive to immunotherapy.

JAK-STAT pathways have been reported to play important roles in T-cell immunity (17). Additionally, JAK-STAT pathways are related to the upregulated expression of PD-L1 (18,19). PTPRD and PTPRT are two members of the receptor-protein tyrosine phosphatases family, which have been reported as mediators of the JAK-STAT signaling pathway (20,21). In addition, patients with PTPRD/PTPRT mutation had a better survival advantage in OS/progression-free survival (PFS) than the wild-type group in NSCLC (22). Several studies have also reported the clinical significance of PTPRD/PTPRT alterations in cancers treated with immunotherapy, however, with limited sample size, study cohort, efficacy evaluation or the comprehensive analysis. Few studies have yet explored the significance of the co-mutation (co-mut) of PTPRD and PTPRT in immunotherapy.

Therefore, in this study, we systematically collected and consolidated a large amount of genomic and clinical data of pan-cancers to evaluate the predictive efficiency of PTPRD/PTPRT mutation for immunotherapy, including up to 12 immunotherapy cohorts across 20 tumor types, which also included a validation cohort 3 from Peking University Cancer Hospital. Our study explored differences in co-mut, uni-mutation (uni-mut), and wild-type groups across these cohorts from multiple perspectives such as OS, PFS, response, and clinical benefit. We found that PTPRD/PTPRT mutation, particularly co-mutant groups, predicted better clinical outcomes in patients with multiple cancers who were treated with immunotherapy and were associated with enhanced anti-tumor immunity.

Materials and methods

Sources of clinical cohort

First, we identified and integrated a discovery cohort based on six cohorts receiving immunotherapy in the cBioPortal (https://www.cbioportal.org/) (23-28). Among them, the cohort from Miao et al. integrated and filtered seven previously published cohorts (29-35). We also integrated four published immunotherapy cohorts as the validation cohort 1 (36-39), a cohort sequenced using MSK-IMPACT panels as the validation cohort 2 (40) and a cohort (PUCH cohort) from Peking University Cancer Hospital as the validation cohort 3 (41). After strict selection criteria were applied, a total of 2,392 patients receiving immunotherapy were enrolled in the present study (Supplementary Figure S1A). In addition, two non-immunotherapy cohorts were integrated and included in this study: the cohort from Zehir et al. and The Cancer Genome Atlas (TCGA) cohort containing 20 cancer types (https://portal.gdc.cancer.gov/) (Supplementary Figure S1B). For PTPRD/PTPRT mutations, all non-synonymous mutation types were considered for inclusion in the study, including nonsense, frameshift, missense, translation start site, splice site and nonstop (42). Mutations in both PTPRD and PTPRT are considered co-mut, mutations in only one of them are defined as uni-mut, and no mutations in either are considered wild-type.

Figure S1.

Flowchart of process used to screen patients included in immunotherapy cohort (A) and non-immunotherapy cohort (B). OS, overall survival; PFS, progression-free survival; TCGA, The Cancer Genome Atlas.

Clinical outcomes

The primary clinical outcomes used in this study included response [objective response rate (ORR) and non-ORR], clinical benefit [durable clinical benefit (DCB) and no clinical benefit (NCB)], PFS, and OS. Responses were assessed using the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 (24). DCB was classified as complete response (CR)/partial response (PR), or stable disease (SD) lasting >6 months, while NCB was classified as SD lasting ≤6 months or progression of disease (PD). In the immunotherapy cohort, OS was measured from the start date of immunotherapy therapy, and patients who did not die were censored at the date of last contact. PFS was assessed from the date the patient began immunotherapy to the date of progression, recurrence, or death of any cause. In the non-immunotherapy cohort, OS was calculated from the date of the first chemotherapy.

Tumor immunogenicity analysis

TMB was defined as the total number of non-synonymous somatic mutations per megabase (Mb) of genome examined. In this study, we mean-normalized the TMB in each cohort. MANTIS was used to estimate microsatellite instability (MSI) scores across 20 cancers from TCGA (43). In addition, non-synonymous single-nucleotide variants (SNVs) neoantigens, indel neoantigens, intratumor heterogeneity, and copy number alterations (CNA) in the TCGA dataset were obtained from Thorsson et al (44). The quantitative protein data of PD-L1 (PD-L1 RPPA) were obtained from The Cancer Proteome Atlas (TCPA) database (https://www.tcpaportal.org/tcpa/).

Tumor immune microenvironment (TIME) analysis

The leukocyte fraction, lymphocytes based on the CIBERSORT algorithm, and tumor-infiltrating lymphocytes (TIL) were also obtained from Thorsson et al. In addition, 18 immune-associated signatures (Supplementary Table S1) with clear anti-tumor/pro-tumor functions adopted from Bagaev et al. (45) and the “GSVA” package were used to determine the single sample gene set enrichment (ssGSEA) scores of each immune signature (46). In addition, the expression of cytokines, chemokines, immunoinhibitors, immunostimulators, and MHC molecules was summarized to assess TIME.

Table S1. Description of immune-associated signatures with anti-tumor/pro-tumor functions.

Term	Function	Genes
CAFs,cancer-associated fibroblasts; MDSCs, myeloid-derived suppressor cells; DC,dendritic cell; NK, natural killer.
Anti-tumor cytokines	Anti-tumor	TNF, IFNB1, IFNA2, CCL3, TNF, SF10, IL21
B cells	Anti-tumor	CD19, MS4A1, TNFRSF13C, CR2, TNFRSF17, TNFRSF13B, CD22, CD79A, CD79B, BLK, FCRL5, PAX5, STAP1
CAFs	Pro-tumor	COL1A1, COL1A2, COL5A1, ACTA2, FGF2, FAP, LRP1, CD248, COL6A1, COL6A2, COL6A3, CXCL12, FBLN1, LUM, MFAP5, MMP3, MMP2, PDGFRB, PDGFRA
Effector cell traffic	Anti-tumor	CXCL9, CXCL10, CXCL11, CX3CL1, CCL3, CCL4, CX3CR1, CCL5, CXCR3
Effector cells	Anti-tumor	IFNG, GZMA, GZMB, PRF1, GZMK, ZAP70, GNLY, FASLG, TBX21, EOMES, CD8A, CD8B
Endothelium	Pro-tumor	NOS3, KDR, FLT1, VCAM1, VWF, CDH5, MMRN1, ENG, CLEC14A, MMRN2
Immune suppression by myeloid cells (MDSCs)	Pro-tumor	IDO1, ARG1, IL10, CYBB, PTGS2, IL4I1, IL6
M1 signature	Anti-tumor	NOS2, TNF, IL1B, SOCS3, CMKLR1, IRF5, IL12A, IL12B, IL23A
Macrophage and DC traffic	Anti-tumor	CCL2, CCL7, CCL8, XCL1, CCR2, XCR1, CSF1R, CSF1
Myeloid cells traffic	Pro-tumor	CSF2, CSF3, CXCL12, CCL26, IL6, CXCL8, CXCL5, CSF1R, CSF2RA, CSF3R, CXCR4, IL6R, CXCR2, CCL15, CSF1
NK cells	Anti-tumor	NKG7, CD160, CD244, NCR1, KLRC2, KLRK1, CD226, GZMH, GNLY, IFNG, KIR2DL4, EOMES, GZMB, FGFBP2, KLRF1, SH2D1B, NCR3
Pro-tumor cytokines	Pro-tumor	IL10, TGFB1, TGFB2, TGFB3, IL22, MIF, IL6
T cells	Anti-tumor	TBX21, ITK, CD3D, CD3E, CD3G, TRAC, TRBC1, TRBC2, CD28, CD5, TRAT1
Th1 signature	Anti-tumor	IFNG, IL2, CD40LG, IL21, TBX21, STAT4, IL12RB2
Th2 signature	Pro-tumor	IL4, IL5, IL13, IL10, GATA3, CCR4
Treg	Pro-tumor	FOXP3, CTLA4, IL10, TNFRSF18, CCR8, IKZF4, IKZF2
Treg and Th2 traffic	Pro-tumor	CCL17, CCL22, CCL1, CCL28, CCR4, CCR8, CCR10
Tumor-associated macrophages	Pro-tumor	IL10, MRC1, MSR1, CD163, CSF1R, IL4I1, SIGLEC1, CD68

Pathway enrichment analysis

Gene set enrichment analysis (GSEA) was used for pathway enrichment analysis (47). The Molecular Signatures Database (MSigDB) of HALLMARK gene sets, Kyoto Encyclopedia of Genes and Genomes (KEGG) gene sets, PID gene sets, Reactome gene sets, WP gene sets and BioCarta were used for enrichment analysis.

Statistical analysis

Fisher’s exact test was performed to compare differences among categorical variables, and multivariable logistic regression analysis was used to adjust for confounding factors including age, gender, therapy, and calculate the odds ratio (OR) value. For continuous variables, the Wilcoxon rank-sum test was performed to compare differences between the two groups, and the Kruskal-Wallis test was conducted to compare differences between three or more groups. For survival analysis, the Kaplan-Meier method and log-rank test were used to compare the survival distribution, and multivariable Cox regression analysis was applied to adjust for confounding factors including age, gender, therapy, and calculate hazard ratio (HR) values. All statistical analyses were done using SPSS software (Version 22.0; IBM Corp., New York, USA) and the R package and P<0.05 was considered statistically significant.

Results

PTPRD/PTPRT mutation predicted favorable clinical outcomes to immunotherapy in discovery cohort

The baseline patient characteristics of the discovery and validation cohorts are summarized in Table 1. In the discovery cohort, six independent cohorts from cBioPortal, including 782 patients with clinical information, were included in our study, with 37 (4.7%) patients with PTPRD/PTPRD co-mut, 155 (19.8%) with uni-mut, and 590 without mutations. Following immunotherapy, the PTPRD/PTPRD co-mut group had the best response, followed by the uni-mut group (P=0.002) (Figure 1A), and the mutant group showed a better clinical benefit than the wild-type group (P=0.002) (Figure 1B). In addition, survival analysis showed that the PTPRD/PTPRD co-mut group had a OS advantage over the uni-mut group and even more over the wild-type group (P=0.005) (Figure 1C). Expectedly, PFS analysis revealed survival differences among the three groups, consistent with OS trends (P=0.038) (Figure 1D).

Table 1. Summary of clinical characteristics of discovery and validation cohorts.

Characteristics	n (%)
	Discovery cohort (N=782)	Validation cohort 1 (N=245)	Validation cohort 2 (N=1,291)	Validation cohort 3 (N=75)
NA, not available; OS, overall survival; PFS, progression-free survival.
Age [median (range)] (year)	65 (18−92)	56 (27−86)	62 (15−90)	59 (15−76)
Gender
Female	331 (42.3)	46 (18.8)	462 (35.8)	21 (28.0)
Male	451 (57.7)	26 (10.6)	829 (64.2)	54 (72.0)
NA	0 (0)	173 (70.6)	0 (0)	0 (0)
Therapy
Combination	132 (16.9)	20 (8.2)	209 (16.2)	0 (0)
Monotherapy	650 (83.1)	225 (91.8)	1,082 (83.8)	75 (100)
NA	0 (0)	0 (0)	0 (0)	0 (0)
Treatment response
Response	226 (28.9)	39 (15.9)	0 (0)	16 (21.3)
Non-response	541 (69.2)	133 (54.3)	0 (0)	59 (78.7)
NA	15 (1.9)	73 (29.8)	1,291 (100)	0 (0)
Treatment benefit
Benefit	315 (40.3)	29 (11.8)	0 (0)	NA
Non-benefit	430 (55.0)	103 (42.0)	0 (0)	NA
NA	37 (4.7)	113 (46.1)	1,291 (100)	NA
OS
Recorded	664 (84.9)	139 (56.7)	1,291 (100)	75 (100)
Not recorded	118 (15.1)	106 (43.3)	0 (0)	0 (0)
PFS
Recorded	730 (93.4)	132 (53.9)	0 (0)	74 (98.7)
Not recorded	52 (6.6)	113 (46.1)	1,291 (100)	1 (1.3)
Mutation status
PTPRT uni-mut	78 (10.0)	12 (4.9)	96 (7.4)	4 (5.3)
PTPRD uni-mut	77 (9.9)	22 (9.0)	67 (5.2)	8 (10.7)
Co-mut	37 (4.7)	6 (2.4)	48 (3.7)	1 (1.3)
Wild	590 (75.4)	205 (83.7)	1,080 (83.7)	62 (82.7)

Figure 1.

Association between PTPRD/PTPRT mutation and clinical outcomes in discovery cohort. (A) Histogram depicting proportions of ORR in co-mut, uni-mut and wild-type groups (Fisher’s test P=0.002, multivariate adjust P=0.006; OR=1.64, 95% CI: 1.15−2.33); (B) Histogram depicting proportions of clinical benefit in the co-mut, uni-mut and wild-type groups (Fisher’s test P=0.002, multivariate adjust P=0.005; OR=1.62, 95% CI: 1.15−2.27); (C) Kaplan-Meier and Cox analysis comparing OS among the co-mut, uni-mut and wild-type groups (Log rank test P=0.005, multivariate adjust P=0.019; OR=0.73, 95% CI: 0.55−0.95); (D) Kaplan-Meier and Cox analysis comparing PFS among the co-mut, uni-mut and wild-type groups (Log rank test P=0.038, multivariate adjust P=0.002; OR=0.68, 95% CI: 0.53−0.87). ORR, objective response rate; OR, odds ratio; 95% CI, 95% confidence interval; NCB, no clinical benefit; DCB, durable clinical benefit; HR, hazard ratio; OS, overall survival; PFS, progression-free survival.

PTPRD/PTPRT mutation predicted favorable clinical outcomes in validation cohorts

To further investigate the efficacy of immunotherapy in patients with PTPRD/PTPRT mutations, we collected four additional independent cohorts receiving immunotherapy and merged them into validation cohort 1. Likewise, in the four clinical outcomes of response rate, clinical benefit rate, OS, and PFS, the co-mut group showed superiority over the uni-mut group and even more so over the wild-type group (Figure 2A−D). Unfortunately, the multivariable analysis could not be performed because most of the patient age, gender, and therapy information were lacking in validation cohort 1. Validation cohort 2 was an immunotherapy cohort including 1,291 patients with OS information. Survival analysis revealed significant differences among the three groups (P<0.001) (Figure 2E). Similarly, in the PUCH cohort, co-mut were detected in only one patient, who achieved a clinical response, survived, and experienced no progression events by the end of the follow-up period. Besides, the mutant group showed superiority in terms of OS, PFS, and response rate compared with the wild-type group (Figure 2F−H). To further illustrate that the benefit from the PTPRD/PTPRD mutation is derived from immunotherapy and not from other causes, we explored a large cohort of non-immunotherapies (Zehir et al. cohort), and retained only the same cancer subtypes in the above immunotherapy cohorts. Survival analysis failed to reveal significant differences among the three groups, and surprisingly, the uni-mut group appeared to have worse OS (Supplementary Figure S2), implying that immunotherapy may have reversed the poor prognostic characteristics of the uni-mut group.

Figure 2.

Association between PTPRD/PTPRT mutation and clinical outcomes in validation cohorts. (A) Histogram depicting proportions of ORR in co-mut, uni-mut and wild-type groups in validation cohort 1 (Fish’s test P=0.043); (B) Histogram depicting proportions of clinical benefit in co-mut, uni-mut and wild-type groups in validation cohort 1 (Fish’s test P=0.036); (C) Kaplan-Meier analysis comparing OS among co-mut, uni-mut and wild-type groups in validation cohort 1 (P=0.036); (D) Kaplan-Meier analysis comparing PFS among co-mut, uni-mut and wild-type groups in validation cohort 1 (P=0.098); (E) Kaplan-Meier and Cox analysis comparing OS among co-mut, uni-mut and wild-type groups in validation cohort 2 (Log rank test P<0.001, multivariate adjust P<0.001; HR=0.62, 95% CI: 0.51−0.75); (F) Histogram depicting proportions of ORR between mutation and wild-type groups in PUCH cohort (Fisher’s test P=0.460, multivariate adjust P=0.300; OR=2.08, 95% CI: 0.52−8.29); (G,H) Kaplan-Meier and Cox analysis comparing PFS (Log rank test P=0.100, multivariate adjust P=0.084; HR=0.46, 95% CI: 0.19−1.11) (G) and OS (Log rank test P=0.422, multivariate adjust P=0.348; HR=0.63, 95% CI: 0.24−1.65) (H) between mutation and wild-type groups in PUCH cohort. ORR, objective response rate; NCB, no clinical benefit; DCB, durable clinical benefit; OS, overall survival; PFS, progression-free survival; HR, hazard ratio; 95% CI, 95% confidence interval; OR, odds ratio.

Figure S2.

Kaplan-Meier analysis comparing OS among co-mut, uni-mut and wild groups in Zehir cohort (P=0.183). OS, overall survival.

Predictive value of PTPRD/PTPRT mutation in different therapy subgroups

In the therapy subgroup analysis, we first selected subgroups of patients receiving combination therapy or monotherapy among all patients receiving immunotherapy. In the monotherapy subgroup, the PTPRD/PTPRD co-mut group had the best efficacy, including a long OS and PFS and superior ORR and benefits, followed by the uni-mut group, while the wild-type group had the worst efficacy (Supplementary Figure S3A−D). However, in the combination subgroup, only similar trends were observed (Supplementary Figure S3E−H). Among the various monotherapy options, our focus was on anti-PD-1/anti-PD-L1 and anti CTLA4, which are currently at the forefront of immunotherapy research (48). We evaluated four clinical outcomes: response rate, clinical benefit rate, OS, and PFS. In the anti-PD-1/anti-PD-L1 subgroups, the co-mut group exhibited superior performance compared to that of the uni-mut group, and even greater performance compared to that of the wild-type group (Supplementary Figure S3I−L). However, we did not observe a similar trend in the anti-CTLA4 subgroup (Supplementary Figure S3M−P).

Figure S3.

Predictive value of PTPRD/PTPRT mutation in different therapy subgroups. (A,B) Histogram depicting proportions of ORR (Fish’s test P<0.001) (A) and benefit (Fish’s test P<0.001) (B) in co-mut, uni-mut and wild-type groups in monotherapy subgroup; (C,D) Kaplan-Meier analysis comparing OS (P<0.001) (C) and PFS (P=0.016) (D) among co-mut, uni-mut and wild-type groups in monotherapy subgroup; (E,F) Histogram depicting proportions of ORR (Fish’s test P=0.052) (E) and benefit (Fish’s test P=0.146) (F) in co-mut, uni-mut and wild-type groups in combination therapy subgroup; (G,H) Kaplan-Meier analysis comparing OS (P=0.618) (G) and PFS (P=0.076) (H) among co-mut, uni-mut and wild-type groups in combination therapy subgroup; (I,J) Histogram depicting proportions of ORR (Fish’s test P<0.001) (I) and benefit (Fish’s test P<0.001) (J) in co-mut, uni-mut and wild-type groups in anti-PD-1/anti-PD-L1 subgroup; (K,L) Kaplan-Meier analysis comparing OS (P<0.001) (K) and PFS (P=0.020) (L) among co-mut, uni-mut and wild-type groups in anti-PD-1/anti-PD-L1 subgroup; (M,N) Histogram depicting proportions of ORR (Fish’s test P=0.858) (M) and benefit (Fish’s test P=0.659) (N) in co-mut, uni-mut and wild-type groups in anti-CTLA4 subgroup; (O,P) Kaplan-Meier analysis comparing OS (P=0.006) (O) and PFS (P=0.927) (P) among co-mut, uni-mut and wild-type groups in anti-CTLA4 subgroup. ORR, objective response rate; OS, overall survival; PFS, progression-free survival.

Association of PTPRD/PTPRT mutation and TIME

To further investigate the mechanism of the extrinsic immune response in tumors with PTPRD/PTPRT mutations, multi-omics data from the TCGA cohort, including 20 types of cancers, were analyzed. Based on the results of previous calculations (44), we found that the leukocyte fraction, lymphocytes based on the CIBERSORT algorithm and TILs, were higher in the uni-mut and co-mut groups than in the wild-type group, revealing that the wild-type group has “cold tumor” characteristics (Figure 3A−C). Subsequent analyses of cell signatures also indicated the characteristics of immune-deserts in the wild-type group, as evidenced by an overall downregulation of pro-tumor and anti-tumor immune cells, as well as the significant upregulation of pro-tumor endothelium (Figure 3D). The co-mut and uni-mut groups also showed significantly different immune infiltration characteristics, as evidenced by the relative downregulation of pro-tumor immune cells and cytokines, in addition to the upregulation of anti-tumor immune cells and cytokines in the co-mut group compared to the uni-mut group (Figure 3D). Immune-related protein analysis revealed that activated cytokines and chemokines, including CCL5, CD8A, CXCL10, CXCL11, CXCL9, GZMA, GZMB, IFNG, PRF1, TBX2, and TNF, were upregulated in both uni-mut and co-mut groups, and inhibited cytokines and chemokines, including CCL22, CXCL12, and IL1B, were downregulated in the co-mut group (Figure 3E). In addition, the majority of immunomodulators and antigen presentation-related proteins were significantly upregulated in the mutant group (Figure 3E). Based on the above analyses, we found that these three groups had significantly distinct TIME infiltration characteristics: the wild-type group was classified as the immune-desert phenotype, the co-mut group was classified as the immune-activated phenotype, andthe uni-mut group was classified as the immune-mixed phenotype.

Figure 3.

Association of PTPRD/PTPRT mutation and TIME in TCGA cohort. (A) Comparison of leukocyte fractions among co-mut, uni-mut and wild-type groups; (B) Comparison of lymphocyte fractions estimated by the CIBERSORT method among co-mut, uni-mut and wild-type groups; (C) Comparison of TIL fractions among co-mut, uni-mut and wild-type groups; (D) Comparison of 18 immune signatures estimated by the ssGSEA method based on RNA-sequencing data among the co-mut, uni-mut and wild-type groups; (E) Comparison of immune-related gene expression among co-mut, uni-mut and wild-type groups. TIME, tumor immune microenvironment; TCGA, The Cancer Genome Atlas; TIL, tumor infiltrating lymphocyte; ssGSEA, single sample gene set enrichment. *, kruskal. test P<0.05.

Association of PTPRD/PTPRT mutation and immunotherapy markers

To further compare PTPRD/PTPRT mutations and other immunotherapy predictors, we collected and compiled predictive markers for immunotherapy, including TMB (49), MSI (50), neoantigen (51), and PD-L1 (protein) (38) which were positively correlated with immunotherapy efficacy, CNA (52), which was negatively correlated with immunotherapy efficacy, and some signatures (14,39,49,53-59). We found that TMB and neoantigens (both SNV neoantigens and indel neoantigens) in the wild-type group, uni-mut group, and co-mut group showed a progressively increasing trend (Figure 4A−C and 4G−J). Additionally, the MSI score was significantly higher in the mutant group (Figure 4D) than in the wild-type group. CAN load was significantly lower in the co-mut group than in the uni-mut group (P=0.033) (Figure 4E). It is worth mentioning that we did not find a correlation between PTPRD/PTPRT mutation and PD-L1 (protein), revealing that PTPRD/PTPRT mutation is a PD-L1-independent predictor (Figure 4F). Notably, the uni-mut and co-mut groups exhibited a significant enrichment in signatures positively associated with immunotherapy; however, only the co-mut group showed a significant downregulation in signatures negatively associated with immunotherapy (Figure 4K). These findings may partially explain the correlation between PTPRD/PTPRT mutations and the benefits of immunotherapy.

Figure 4.

Association of PTPRD/PTPRT mutation and immunotherapy markers. (A) Comparison of TMB among the co-mut, uni-mut and wild-type groups in TCGA cohort; (B,C) Comparison of SNV-neoantigen (B) and Indel-neoantigen (C) among co-mut, uni-mut and wild-type groups in TCGA cohort; (D−F) Comparison of MSI score (D), CNA load (E), PD-L1 (protein) (F) among co-mut, uni-mut and wild-type groups in TCGA cohort; (G−J) Comparison of TMB among the co-mut, uni-mut and wild-type groups in discovery cohort (G), validation cohort 1 (H), validation cohort 2 (I) and Zehir cohort (J); (K) Comparison of immunotherapy signatures based on RNA-sequencing data among co-mut, uni-mut and wild-type groups. TMB, tumor mutation burden; TCGA, The Cancer Genome Atlas; MSI, microsatellite instability; CNA, copy number alteration; PD-L1, programmed cell death ligand 1. *, kruskal. test P<0.05.

PTPRD/PTPRT mutation combined with PD-L1 (protein) to predict immunotherapy efficacy

Based on the results of the analysis presented above (Figure 4F), PD-L1 expression is a potential combined predictor of efficacy. We obtained the expression data of PD-L1 in the immunotherapy cohort including 150 patients with NSCLC from Rizvi et al. and Hallmann et al. (25,27) and 41 patients with gastric cancer (GC) from Kim et al (38). PD-L1 positive was defined as a PD-L1 score greater than 0. We could not identify patients with co-mutations in the GC cohort or patients with both co-mutation and positive PD-L1 in the NSCLS cohort, possibly due to the small sample size. We also found no correlation between PTPRD/PTPRT mutations and PD-L1 expression in both GC (Supplementary Figure S4A) and NSCLC (Supplementary Figure S4B) cohorts. Further analysis revealed that the introduction of PD-L1 could better differentiate the immunotherapy efficacy in both the GC (Response: P=0.069 → P<0.001) (Supplementary Figure S4C,D) and NSCLC cohorts (Response: P=0.084 → P<0.001; Benefit: P=0.350 → P<0.001) (Supplementary Figure S4E−H). It could also better differentiate patient survival in the NSCLC cohort (OS: P=0.353 and P=0.047; PFS: P=0.054 and P<0.001) (Supplementary Figure S4I−L). These results suggest that the combination of PTPRD/PTPRT mutation with PD-L1 expression further improves the predictive power of immunotherapy.

Figure S4.

PTPRD/PTPRT mutation combined with PD-L1 (protein) to predict immunotherapy efficacy. (A,B) Histogram depicting proportions of positive PD-L1 among co-mut, uni-mut and wild groups in GC cohort (Fish’s test P=0.586) (A) and in NSCLC cohort (Fish’s test P=0.120) (B); (C,D) Histogram depicting proportions of ORR between uni-mut and wild groups (Fish’s test P=0.069) (C) and among Uni-mut&Pos, Uni-mut/Pos and Wild&Neg groups (Fish’s test P<0.001) (D) in GC cohort; (E,F) Histogram depicting proportions of ORR among co-mut, uni-mut and wild groups (Fish’s test P=0.084) (E) and among co-mut, uni-mut&Pos, uni-mut/pos and wild&neg groups (Fish’s test P<0.001) (F) in NSCLS cohort; (G,H) Histogram depicting proportions of clinical benefit among co-mut, uni-mut and wild groups (Fish’s test P=0.350) (G) and among co-mut, uni-mut&pos, uni-mut/pos and wild&neg groups (Fish’s test P<0.001) (H) in NSCLS cohort; (I,J) Kaplan-Meier analysis comparing OS among co-mut, uni-mut and wild groups (P=0.353) (I) and among co-mut, uni-mut&pos, uni-mut/pos and wild&neg groups (P=0.047) (J) in NSCLC cohort; (K,L) Kaplan-Meier analysis comparing PFS among co-mut, uni-mut and wild groups (P=0.054) (K) and among co-mut, uni-mut&pos, uni-mut/pos and wild&neg groups (P<0.001) (L) in NSCLC cohort. uni-mut&pos, defined as uni-mut occurring simultaneously with positive PD-L1; uni-mut/pos, defined as uni-mut or positive PD-L1; wild&neg, defined as wild occurring simultaneously with negative PD-L1. PD-L1, programmed cell death ligand 1; GC, gastric cancer; NSCLC, non-small cell lung cancer; ORR, objective response rate; OS, overall survival; NCB, no clinical benefit; DCB, durable clinical benefit; PFS, progression-free survival.

Pathway enrichment analysis of PTPRD/PTPRT mutation in TCGA cohort

GSEA enrichment analysis indicated that several pathways varied significantly between the mutant and wild-type groups; the mutant group was mainly enriched in immune-related pathways, the JAK-STAT signaling pathway, cell cycle, and glycolysis, and the wild-type group was mainly enriched in the PI3K/AKT signaling pathway, focal adhesion, MAPK signaling pathway, gap junction, and other cancerous pathways (Figure 5A−C). Our data revealed that the mutant group had higher levels of glycolysis, as evidenced by the fact that almost all related enzymes were upregulated in the mutant group (Figure 5D). Using the mean gene expression of all glycolytic enzymes to measure glycolytic activity, we found a statistically significant difference between the mutant and wild-type groups (P<0.001) (Figure 5E). Lactate is a key product of glycolysis involve with the Warburg effect, and is a potent inhibitor of CD8 T cells (60), glycolysis may therefore be a potential immunosuppressive factor in the mutant group. The use of glycolytic inhibitors in combination with immune checkpoint inhibitors in the mutant group may further improve patient outcomes and survival.

Figure 5.

Pathway enrichment analysis between mutant and wild-type groups in TCGA cohort. (A) Differences in pathway activities scored by GSEA between mutant and miscode wild-type groups; (B,C) GSEA plot depicting representative pathways identified by GSEA between mutant (B) and wild-type (C) groups; (D) Alteration of key enzymes involved in the glycolysis in mutant group; (E) Comparison of glycolytic activity between mutant and wild-type groups (****, P<0.001). TCGA, The Cancer Genome Atlas; GSEA, gene set enrichment analysis.

Discussion

In this study, we collected and collated mutational information and clinical data from pan-cancer cohorts to systematically assess differences in the clinical outcomes of immunotherapy in different PTPRD/PTPRT mutation groups. We found that PTPRD/PTPRT mutations were significantly associated with enhanced anti-tumor immunity and immunogenicity. In addition, PD-L1 expression combined with PTPRD/PTPRT mutation can better differentiate the groups that benefit from immunotherapy than others.

TIME analysis showed that the three mutant groups had significantly distinct TIME infiltration characteristics: the wild-type group was classified as the immune-desert phenotype, the co-mut group was classified as the immune-activated phenotype, and the uni-mut group was classified as the immune-mixed phenotype. These results explain why the three mutant groups had different clinical outcomes. Several types of cells exert crucial effects during the formation of the immune phenotype. For instance, previous studies have shown that endothelial cells can drive T-cell exclusion and thus induce an immune desert phenotype to promote immune escape (61). Therefore, targeting endothelial-cadherin can promote T-cell-mediated immunotherapy (62). Interestingly, we found that the PTPRD/PTPRT mutation increased the abundance of endothelial cells, suggesting a pro-tumor TIME. Collectively, these findings suggest that endothelial cells may be another mechanism underlying immunotherapy resistance mediated by the PTPRD/PTPRT mutation.

The enrichment analysis of RNA-seq showed some meaningful biological changes, including immune-related pathways, the JAK-STAT signaling pathway, cell cycle, and glycolysis. PTPRD/PTPRT, two members involved in the JAK-STAT signaling pathway, cause dysregulation of this pathway when mutated. In addition, several studies have shown that cell cycle and proliferation may benefit immunotherapy. For example, Pabla et al. revealed a significant survival advantage for moderately proliferative tumors compared to their combined highly/poorly proliferative counterparts (63) and Bagaev et al. also revealed that the most suitable subtype for immunotherapy in pancreatic cancer shows proliferative activity (45). We observed elevated glycolytic activity in the mutant group. Increased tumor glycolysis has been shown to suppress anti-tumor immunity by inhibiting T-cell function and trafficking to TIME (64). Therefore, glycolysis inhibition may be combined with checkpoint inhibitors to further liberate anti-tumor immunity in the mutant group.

This result is partially consistent with the results of a previous studies. An early study found that mutations in PTPRD/PTPRT were associated with improved immunotherapy efficacy only in NSCLC (22). Subsequent studies also found that mutations in PTPRD or PTPRT suggest better immunotherapy efficacy in both NSCLC and SKCM (65,66). Recently, with the widespread application of immunotherapy in pan-cancer, a large amount of data information has been generated. Shang et al. found that the PTPRD/PTPRT mutation group had better OS compared with the wild group in three cohorts including 9 cancer types (67). Our study explored differences in co-mut, uni-mut, and wild-type groups across 12 immunotherapy cohorts including 20 cancer types from multiple perspectives such as OS, PFS, response, and clinical benefit. Furthermore, we highlight that the co-mut group exhibited better clinical outcome and immune characteristics. Moreover, our results suggest that the combination of PTPRD/PTPRT mutation with PD-L1 expression further improves the predictive power of immunotherapy. Therefore, the present study is a better supplement and deeper continuation of previous research results.

This study has several limitations. First, the immunotherapy cohorts included in this study mainly received PD-1, PD-L1 and CTLA4 antibodies, but we did not discuss further immunotherapy strategies. Second, the data information utilized in this study was obtained from already published studies, which did not disclose sufficiently comprehensive clinical information, such as tumor stage for our further exploration, Third, the results of this study are based on the analysis of retrospective data, and there is still a lack of prospective studies to further confirm these findings.

Conclusions

Our study demonstrated a better clinical benefit of immunotherapy in patients with cancer with PTPRD/PTPRT mutation, especially the co-mut group, compared to the wild-type group. In addition, we also found that the different mutation status presented distinct TIME and immunogenicity profiles. Remarkably, PD-L1 combined with PTPRD/PTPRT mutation can better differentiate patients who may benefit from immunotherapy. Therefore, PTPRD/PTPRT mutation could serve as a potential predictive biomarker for immunotherapy.