Introduction

TP53, located at 17p13.1, encodes the p53 protein, which guards cells against cancerous transition by promoting cell cycle arrest, apoptosis and DNA damage repair [1, 2]. Ordinarily, p53 acts as a transcription factor to govern the expression of target genes, such as CCNE1 and ATM [3]. Furthermore, p53 can exert transcription-independent protective roles through interaction with other transcription factors and cellular compartments pathways [4,5,6,7]. In addition to its cell-intrinsic efficacy, p53 can also wheel tumor immune microenvironment (TME) [8]. Large studies have revealed that the p53 controls tumor-immune system crosstalk by modulating cytokines, MHC-I antigen processing pathway, and immune checkpoints [9].

TP53 mutations occur ~50% in gastric cancer (GC), up to 70% in metastasis, and predominantly in tumors with chromosomal instability (CIN) [10, 11]. However, TP53 mutations in GC have controversial impact on survival and chemotherapy sensitivity [10]. The initial studies showed that TP53 mutational status was independent indicator to poor survival and response to cisplatin-based chemotherapy [12,13,14], which were challenged by later research [15, 16]. Furthermore, mutated TP53 was found to inhibit gastric cancer immunity [17], and predict poor response to immunotherapy in patients with metastatic solid tumors including GC [18]. Moreover, TP53 loss of heterozygosity (LOH) indicating a severe p53 dysfunction have been reported in GC, revealing an association with tumor progression and poor patient survival [19,20,21]. These preceding studies with heterogeneous definition of TP53 alteration presented limited explanation of mechanism, and failed to reach a consensus on how TP53 alteration influence clinical outcomes in GC.

Previous studies have confirmed the efficacy of the VAF/TP clinical classifier in characterizing gene mutation statuses and differentiating between genomic changes and clinically neutral events [22]. We divided TP53 status into three groups according to genome sequencing, namely, clonal mutations with LOH (C-LOH), clonal diploid or subclonal mutations (CD-SC), and wild type (WT). Likewise, according to immunohistochemical (IHC) staining, p53 expression was divided into over-expression (OE), Null and WT. Our study, including the Cancer Genome Atlas (TCGA) cohort, the Samsung Medical Center (SMC) cohort, the Zhongshan Hospital (ZSHS) cohort and the FUSCC cohort, studied the association between TP53 status and clinical outcomes (survival, chemotherapy and immunotherapy sensitivity), and evaluated the underlying mechanism (genomes and TME).

Patients and methods

Study population

Overall 890 patients from four cohorts were enrolled in this study, including the ZSHS cohort based on tissue samples, two cohorts based on sequencing data (namely the TCGA and SMC cohorts), and the FUSCC cohort, which possesses both IHC staining data and targeted-genes sequencing data. The ZSHS Cohort originally enrolled 496 GC patients with full informed consent and the approval of the Clinical Research Ethics Committee of Zhongshan Hospital affiliated with Fudan University (Shanghai, China; approval number: Y2015-054). All these patients underwent partial or total gastrectomy in Zhongshan Hospital between August 2007 and December 2008. However, 23 patients were excluded due to dot loss or incomplete clinical information, and we ultimately included 473 patients from Zhongshan Hospital in this study. These patients were followed until April 2014 with a 42-month median follow-up time. The FUSCC cohort included 27 GC patients with full informed consent and the study was approved by the Institutional Review Board (IRB) of FUSCC (IRB number 050432-4-2108). All the patients underwent gastrectomy and targeted-genes sequencing in Fudan University Shanghai Cancer Center during January 2022 to August 2023. Specifically, we included TCGA cohort (n = 410) as the discovery cohort to measure TP53 mutation status; similarly, ZSHS cohort (n = 473) was included as the validation cohort to assess p53 protein expression patterns. FUSCC cohort (n = 27) was included to confirm the consistency of TP53 mutation status with p53 protein expression. SMC cohort (n = 43) was included to explore the predictive value of TP53 mutation status for responsiveness to anti-PD-1 immunotherapy in GC patients.

Immunohistochemistry (IHC)

Firstly, the tissue microarray slides were baked at 60 °C for 6 h, placed in xylene and gradient alcohol to dewax, and then washed with phosphate-buffered saline for 3 times. To inhibit endogenous peroxidase, we treated the slides with 3% H2O2 for 30 min at 37 °C. Next, the slides were performed heat mediated antigen retrieval with 0.01 M citrate buffer (pH 6.0). After blocked with 10% goat serum (ZSGB-BIO) at 37 °C for 2 h for eliminating non-specific reactions, the slides were subsequently incubated with prediluted primary anti-p53 antibody (Leica, NCL-L-p53-DO7, 1: 800) at 4 °C overnight. Subsequent to reaction with HRP-labeled secondary antibody (ZSGB-BIO) for 30 min at 37 °C, diaminobenzidine (DAB)-H2O2 and hematoxylin were applied to stain the reaction products and nucleus, respectively. Finally, the slides were fixed with mounting medium for analysis.

Evaluation of IHC score

IHC images were evaluated by two experienced pathologists under a microscope at high power field (HPF, ×200), both of whom were blinded to the patients’ clinical data. Representative images of IHC staining and corresponding Hematoxylin and Eosin staining were shown in Supplementary Fig. S1. Cases were assigned to one of three categories: (1) WT: 1–80% of tumor cell nuclei staining positive, usually with variable intensity; (2) Null: no tumor cell nuclear staining with a positive internal control; and (3) OE: uniform and intense nuclear staining in at least 80% of tumor cell nuclei (estimated). As for tumors with contradictory disposition in p53 expression pattern from the two pathologists, a second-round evaluation would be conducted.

Computational analysis

Statistical analyses included in this study were performed by SPSS (version 26.0) and R software (version 4.1.2). All conducted analyses were two sided, and P < 0.05 was considered statistically significant. Continuous variables were analyzed by student’s t test or one-way ANOVA followed by Tukey’s multiple comparisons. For categorical variables, Chi-square test was adopted. The cut-off value of variant allele frequency (VAF) / tumor purity (TP) for the classification of TP53 C-LOH and TP53 CD-SC subgroups was set to 0.9. For the SMC cohort, 40 differentially expressed genes were utilized to generate the classification of TP53 C-LOH and TP53 CD-SC based on transcriptomic profiling, with the relevant genes being listed in Supplementary Table S1. The genomic characteristics of the patients from TCGA, containing VAF and TP data, fraction of genome altered (FGA), aneuploidy score and whole-genome doubling (WGD), were directly downloaded from cBioPortal (http://www.cbioportal.org). The cut-off value of Homologous Recombination Deficiency (HRD) score for the classification of HRD and Homologous Recombination Proficiency (HRP) subgroups was set to 42. Associations between TP53 mutation status and enrichment of immune-related signatures (IFN-α Response, IFN-γ Response, Antigen Processing and Presentation) in the TCGA cohort were assessed with gene sets enrichment analysis (GSEA) using the GSEA software. The immune-related transcription factors activities including STAT5B, STAT5A, STAT2, IRF1, STAT3, REL, NFKB1, RELA, and STAT1 were assessed with DoRothEA. Other immune-related features were directly acquired from GDC pan-immune data portal (https://gdc.cancer.gov/about-data/publications/panimmune).

Results

TP53 CD-SC associates with superior overall survival in gastric cancer

To explore the impact of TP53 mutation status on prognosis of gastric cancer patients, we conceived variant allele VAF / TP value to classify TP53 mutations as TP53 C-LOH, TP53 CD-SC, and TP53 WT in the TCGA cohort (Discovery Cohort). The Kaplan-Meier curve showed that patients with TP53 CD-SC had superior overall survival (OS) compared with TP53 C-LOH patients (log rank P = 0.014, Fig. 1A). A similar trend was also found between the OS of TP53 CD-SC and TP53 C-LOH subgroup (log rank P = 0.071, Fig. 1A), despite no statistical significance. Thus, we combined TP53 C-LOH with TP53 WT subgroup for further analysis and found a significant longer OS in patients with TP53 CD-SC compared with other patients (log rank P = 0.029, Fig. 1B). To understand the prognostic effects of p53 proteins in GC, we divided patients into p53 WT, p53 OE or p53 Null subgroups according to IHC staining in ZSHS cohort (Validation Cohort). However, although a similar trend to that in the TCGA cohort was observed between p53 null vs p53 OE (log rank P = 0.053, Fig. 1C), there was no significant difference in OS among p53 OE vs p53 Null/WT (log rank P = 0.120, Fig. 1D). To unify the results of the TCGA and ZSHS cohorts, we evaluated the correlation between TP53 status and p53 expression in the FUSCC cohort. As showed in Supplementary Fig. S2, TP53 C-LOH and CD-SC consistently exhibited high correlation with p53 Null and OE, respectively. Furthermore, 11/13 patients with TP53 WT displayed p53 WT expression, while 2/13 showed p53 OE, indicating a consistent relationship between TP53 status and p53 expression patterns.

Fig. 1: Clinical outcomes of TP53/p53-based subgroups in the TCGA and ZSHS cohort.
figure 1

Kaplan-Meier survival curves for the three subgroups of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT) in the TCGA cohort (a). Kaplan-Meier survival curves for comparing TP53 C-LOH with other tumors in the TCGA cohort (b). Kaplan-Meier survival curves of the three subgroups of interest (p53 OE, p53 Null and p53 WT) in the ZSHS cohort (c). Kaplan-Meier survival curves for comparing p53 OE with other tumors in the ZSHS cohort (d). CD-SC, clonal diploid-subclonal; C-LOH, clonal with loss of heterozygosity; WT wild type; OE overexpression. Log-rank test was conducted for Kaplan-Meier curves. P ≤ 0.05 was considered statistical significance.

Full size image

TP53 C-LOH predicts worse adjuvant chemotherapy response in gastric cancer

It has been widely recognized that treatment with fluorouracil-based adjuvant chemotherapy (ACT) could improve the OS for patients with stage II and III gastric cancer [23, 24], which was also identified in the TCGA and ZSHS cohort (log rank P < 0.001 and log rank P < 0.001, respectively; Fig. 2a, e). Considering the important role of TP53 gene in chemotherapy response [25], we further explored the association between ACT and OS among patients with stage II/III disease belonging to different TP53-based subgroups. Interestingly, fluorouracil-based ACT could only improve overall survival in TP53 CD-SC (log rank P = 0.008, Fig. 2c) and TP53 WT (log rank P = 0.043, Fig. 2d) subgroup, but no longer improved overall survival in patients with TP53 C-LOH (log rank P = 0.321, Fig. 2b). While in ZSHS cohort, we found that ACT had a positive prognostic effect only in p53 OE (log rank P = 0.001, Fig. 2g) and p53 WT (log rank P = 0.001, Fig. 2h) subgroups, but not in p53 Null subgroup (log rank P = 0.206, Fig. 2f).

Fig. 2: Benefits from adjuvant chemotherapy in patients with different TP53/p53 status.
figure 2

Kaplan-Meier analyses of overall survival according to ACT chemotherapy treatment for all stage II/III gastric cancer patients (a), TP53 C-LOH subgroup (b), TP53 CD-SC subgroup (c) and TP53 WT subgroup (d) in the TCGA cohort. Kaplan-Meier analyses of overall survival according to ACT chemotherapy treatment for all stage II/III gastric cancer patients (e), p53 Null subgroup (f), p53 OE subgroup (g) and p53 WT subgroup (h) in the ZSHS cohort. CD-SC, clonal diploid-subclonal; C-LOH, clonal with loss of heterozygosity; WT wild type; OE overexpression. Log-rank test was conducted for Kaplan-Meier curves. P ≤ 0.05 was considered statistical significance.

Full size image

TP53 C-LOH predicts worse anti-PD1 immunotherapy response in gastric cancer

Recently, immunotherapy have been playing more and more important roles in GC treatment. We further enrolled another cohort from the Samsung Medical Center (the SMC cohort) consisting of patients treated with anti-PD1 immunotherapy to evaluate the predictive value of TP53 status for immunotherapy. The baseline clinical and molecular characteristics of the SMC cohort were displayed in Fig. 3a. We found that patients in TP53 C-LOH subgroup demonstrated a significantly decreased response rate compared with those in TP53 CD-SC and TP53 WT subgroup (P < 0.001, Fig. 3b). Meanwhile, patients in TP53 C-LOH subgroup demonstrated the worst OS among all patients treated with anti-PD1 immunotherapy (log rank P < 0.001, Fig. 3c), and there was no significant overlap between TP53 mutation status and existing immunotherapy response predictive markers [26], such as microsatellite instability status and PD-L1 expression (Supplementary Fig. S3).

Fig. 3: TP53 status predicts responsiveness to anti-PD-1 immunotherapy in gastric cancer.
figure 3

Heatmap demonstrated responsiveness to pembrolizumab and molecular parameters based on TP53 status in the SMC cohort (n = 43) (a). Association of the BOR with the TP53 status in the ICB cohort (b). Kaplan-Meier curves of overall survival based on TP53 status in ICB cohort (n = 43) (c). BOR best of response, CIN chromosomal instability, EBV EBV-positive, GS genomically stable, MSI microsatellite instability, TMB tumor mutation burden. Log-rank test was conducted for Kaplan-Meier curves. P ≤ 0.05 was considered statistical significance.

Full size image

Correlation between genomic feature and TP53 status in gastric cancer

Given the comprehensive impact on therapeutic responsiveness, we next tried to clarify molecular characteristic in patients with different TP53 mutation status in GC. In order to explore the genomic mechanism underling the association between TP53 status and clinical outcomes, TCGA cohort was used for the analysis of the correlations between genomic features and TP53 status. Although tumor mutation burden (TMB) has been identified a positive predictive biomarker for better response to immunotherapy [27], there was no significant difference across subgroups based on TP53 status (Fig. 4a), which indicated that the predictive value of TP53 status might be independent of TMB. However, we discovered more significant instability at the chromosomal level, characterized by elevated FGA [28] (P < 0.001, Fig. 4b), aneuploidy score [29] (P < 0.001, Fig. 4c) and WGD [30] (P < 0.001, Fig. 4d), in TP53 C-LOH subgroup. Moreover, patients with TP53 C-LOH harbored a higher degree of HRD, which is one of the mechanisms of chromosome instability (P < 0.001, Fig. 4e, f). Consistently, TP53 C-LOH GC were mainly enriched in CIN subtype of GC (P < 0.001, Fig. 4g). Together, all the above findings indicated that TP53 C-LOH patients presented higher level of chromosomal instability.

Fig. 4: Genomic characteristics associated with different TP53 status in gastric cancer.
figure 4

Box plots for TMB (a), FGA (b), aneuploidy score (c) in the three subgroups of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT). Stacked bar chart displaying the distribution of WGD across the three subsets of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT) (d). Box plots for HRD score in the three subgroups of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT) (e). Stacked bar chart displaying the association of HRR status (f) and TCGA subtypes (g) with TP53 status. Analyses of ERBB2 and CCNE1 copy number alteration in the three subgroups of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT) (h). Different genomic alterations accompanied with ERBB2 amplification in TP53 C-LOH and TP53 CD-SC subgroups (i). TMB tumor mutation burden, FGA fragment genome alteration, WGD whole genome duplication. CD-SC clonal diploid-subclonal, C-LOH clonal with loss of heterozygosity, WT wild type.

Full size image

In view of the attenuated benefit from chemotherapy and immunotherapy, we further assessed the association between TP53 status and known targetable alterations. It was shown that ERBB2 and CCNE1 amplification were more significant in TP53 C-LOH subgroup (ERBB2 29.21%; CCNE1 29.21%) compared with TP53 CD-SC (ERBB2 13.83%; CCNE1 11.70%) and TP53 WT subgroups (ERBB2 6.61%; CCNE1 3.08%) (Fig. 4h). Moreover, the amplification of ERBB2 was more likely to be accompanied by CCNE1 amplification and HRD in TP53 C-LOH subgroup (Fig. 4i), which indicated possibility for combination of anti-HER2 therapy with PARP inhibitor or cell cycle inhibitors in these patients.

TP53 C-LOH promotes immune suppression in gastric cancer

To uncover the possible mechanisms for poor anti-PD1 immunotherapy response in TP53 C-LOH patients, we further analyzed the immune characteristics of patients with different TP53 mutation status in TCGA cohort and ZSHS cohort. Based on immunogenomic differences, a previous extensive analysis has identified six cancer immune subtypes: Wound Healing, IFN-γ Dominant, Inflammatory, Lymphocyte Depleted, Immunologically Quiet, and TGF-β Dominant [31]. We discovered that IFN-γ dominant phenotype was less distributed in TP53 C-LOH subgroup (P < 0.001, Fig. 5a). Consistently, patients with TP53 C-LOH showed the lowest IFN-γ Response signature (Fig. 5b). By gene set enrichment analysis, we found down-regulation of IFN-α Response, IFN-γ Response and Antigen Processing and Presentation pathways (P = 0.028, P = 0.011, P = 0.006; respectively; Fig. 5c). In addition, inflammatory associated transcription factors were also down-regulated in TP53 C-LOH subgroup (Fig. 5d). Neoantigens are the basis of effective anti-tumor immune response, thus we further analyzed the association between TP53 mutation status and tumor neoantigens predicted by computed method [32]. It was shown that patients with TP53 C-LOH harbored significantly decreased neoantigens compared with TP53 CD-SC and TP53 WT patients (Fig. 5e). Interestingly, CD8+ T cells was not positively correlated with neoantigens in TP53 C-LOH subgroup (Fig. 5f), and further analysis revealed that the infiltration of CD8+ T cells was independent of p53 status (Supplementary Fig. S4), which indicated that neoantigens might not trigger anti-tumor immune response in this subgroup of GC. Taken together, all these findings revealed that TP53 C-LOH might promote immune suppression in GC.

Fig. 5: Immunophenotypic characteristics of patients with different TP53 status.
figure 5

Distribution of immune subtypes across the three subgroups of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT) (a). Quantification of immune subtype signatures in the three subgroups of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT). b Gene set enrichment analysis indicated down-regulation of IFN-α Response, IFN-γ Response and Antigen Processing and Presentation pathways in TP53 C-LOH subgroups (c). Enrichment analysis for immune associated transcription factor activities in TP53 C-LOH subgroup (d). Boxplot for predicted neoantigens (NeoAgs) level in the three subgroups of interest (TP53 C-LOH, TP53 CD-SC and TP53 WT) (e). The correlations of predicted neoantigens with CD8+ T cells in TP53 C-LOH subgroups and TP53 CD-SC subgroups (f). CD-SC clonal diploid-subclonal, C-LOH clonal with loss of heterozygosity, WT wild type.

Full size image

Discussion

In clinical setting, rapid advances in sequencing technology and the advent of deep genomic analysis have broadened the horizon of insight into intrinsic tumor characteristics and led to the identification of novel therapeutic targets and biomarkers [3]. Genomic alterations have been shown to be strongly associated with therapeutic responsiveness and tumor progression in various tumors, including GC [33,34,35,36]. TP53 is the guardian of the genome, and there is a recognized association between its loss and genomic instability [37, 38]. Moreover, TP53 mutation has been shown to be associated with patient survival risk and treatment response in several cancer types [12, 14]. However, the clinical significance of TP53 alterations in GC has not been fully elucidated. In this study, we identified different types of TP53 mutation based on clonality/LOH status and revealed, for the first time, that different TP53 mutations matched with different genomic and immune microenvironment features and patient clinical outcomes.

TP53 mutation is among the most frequent mutations in GC, affecting more than 40% of GC patients [39]. Previous studies tended to be limited to focus on the impact of overall TP53 mutation on patient survival status, resulting in its contribution to the clinical course of the disease often being overlooked and largely underestimated. Our study in multiple independent cohorts found that compared to TP53 WT, only TP53 CD-SC was able to suggest better OS in GC patients, while patients with TP53 C-LOH had similar survival outcomes as TP53 WT patients. These results suggest that the inclusion of clonality/LOH factors in the assessment of mutational status can help to provide a more refined interpretation of high-dimensional sequencing data and mitigate the impact of background noise from synonymous mutations at the clinical level [40], thus locating the true clinically significant mutational subtypes to the extent possible.

We found that TP53 clonal status affects the response of GC patients with TP53 mutation to first-line treatment strategies. Patients with TP53 CD-SC had the best responsiveness to 5-fluoropyrimidine-based ACT compared to TP53 WT and TP53 C-LOH patients. After receiving pembrolizumab monotherapy, TP53 WT patients showed the optimal response, while patients with TP53 C-LOH barely responded to ICI, suggesting that in future clinical practice, it may be possible to precisely stage GC patients based on TP53 mutation status. In patients with TP53 C-LOH, existing first-line treatment strategies may not be effective, and combination therapy or exploration of novel treatment strategies were required for this population, while patients with TP53 CD-SC may receive standard ACT or ICI. Although this decision must take into account other clinicopathological factors (age, tumor status, disease burden, etc.), our study suggests that TP53 status may help us in clinical decision making and promote individualized treatment of GC patients.

In view of the refractory nature of TP53 C-LOH GC, we sought to further explore their genomic and immunological features in order to find the underlying mechanisms and potential therapeutic targets for their treatment insensitivity. We found that TP53 C-LOH GC exhibited higher levels of chromosomal instability, as evidenced by elevated FGA, whole ploidy scores, whole genome duplication and HRD levels. Interestingly, TP53 C-LOH was also strongly associated with lower neoantigen levels and inflammatory responses, suggesting a ‘cold’ immune phenotype, providing a possible explanation for the poor responsiveness to ACT and immunotherapy. In addition, the concomitant amplification of ERBB2 and CCNE1 observed in GC with TP53 C-LOH, combined with its chromosomal instability, suggests that anti-HER2 therapy combined with PARP inhibitors or cell cycle inhibitors may be applied to improve the clinical outcome in this population.

Even though our results were reproducible in independent cohorts, we are aware that there still existed some limitations in our study. Inter-cohort heterogeneity must be emphasized because of the differences in epidemiology, clinicopathological characteristics, and therapeutic strategies of GC patients from the East and the West [41]. Some of the conclusions in our study were based on retrospective observations of subgroups and partially did not achieve significant consistency across all cohorts. Furthermore, the efficacy and safety of combinatory therapy for TP53 C-LOH GC has not been assessed. Thus, further validation is required to confirm our results within the framework of more extensive, multi-centered clinical trials. Additionally, owing to data source constraints, it is challenging to evaluate the association between TP53 mutation status and metastatic burden. Future studies focusing on cohorts of metastatic GC patients are needed to explore this relationship further.