Evaluation of Tumor DNA Sequencing Results in Patients with Gastric and Gastroesophageal Junction Adenocarcinoma Stratified by TP53 Mutation Status

Abstract Background Gastric cancer (GC) and gastroesophageal junction adenocarcinomas (GEJ) are molecularly diverse. TP53 is the most frequently altered gene with approximately 50% of patients harboring mutations. This qualitative study describes the distinct genomic alterations in GCs and GEJs stratified by TP53 mutation status. Patients and Methods Tumor DNA sequencing results of 324 genes from 3741 patients with GC and GEJ were obtained from Foundation Medicine. Association between gene mutation frequency and TP53 mutation status was examined using Fisher’s exact test. Functional gene groupings representing molecular pathways suggested to be differentially mutated in TP53 wild-type (TP53WT) and TP53 mutant (TP53MUT) tumors were identified. The association of the frequency of tumors containing a gene mutation in the molecular pathways of interest and TP53 mutation status was assessed using Fisher’s exact test with a P-value of <.01 deemed statistically significant for all analyses. Results TP53 mutations were noted in 61.6% of 2946 GCs and 81.4% of 795 GEJs (P < .001). Forty-nine genes had statistically different mutation frequencies in TP53WT vs. TP53MUT patients. TP53WT tumors more likely had mutations related to DNA mismatch repair, homologous recombination repair, DNA and histone methylation, Wnt/B-catenin, PI3K/Akt/mTOR, and chromatin remodeling complexes. TP53MUT tumors more likely had mutations related to fibroblast growth factor, epidermal growth factor receptor, other receptor tyrosine kinases, and cyclin and cyclin-dependent kinases. Conclusion The mutational profiles of GCs and GEJs varied according to TP53 mutation status. These mutational differences can be used when designing future studies assessing the predictive ability of TP53 mutation status when targeting differentially affected molecular pathways.


Introduction
Gastric and esophageal cancer are the third and sixth leading cause of cancer deaths worldwide. 1 In the US alone, 26 250 new cases of gastric cancer and 19 260 new cases of esophageal cancer were diagnosed in 2021, with males representing 61% and 79% of these new diagnoses. 2 The western population, in particular, has seen the incidence of gastric cancer (GC) and steady increase of gastroesophageal junction adenocarcinomas (GEJs) due to the increasing prevalence of obesity and gastro-esophageal reflux disease. [3][4][5] GCs and GEJs, collectively known as gastroesophageal adenocarcinomas (GEAs), are molecularly diverse. Attempts have been made to molecularly categorize these tumors to organize future studies and create tailored treatment strategies. In 2014, the Cancer Genome Atlas Research Network (TCGA) analyzed 295 GC samples using array-based somatic copy number analysis, whole-exome sequencing, arraybased DNA methylation profiling, messenger RNA (mRNA) sequencing, microRNA (miRNA) sequencing, reverse-phase protein array, and microsatellite instability (MSI) testing to propose separating GC into 4 general molecular subtypes. 6 The first 2 molecular subtypes were generated on the basis of CpG island hypermethylation (CIMP): Epstein-Barr virus (EBV)+ CIMP tumors (9%) with distinct hypermethylation patterns not associated with the epigenetic silencing of MLH1, and microsatellite instability (MSI) enriched CIMP EBV-tumors (22%) with hypermethylation patterns consistent with MLH1 epigenetic silencing. The remainder of the GCs were then stratified into one of 2 groups based on their relative burden of somatic copy number aberrations (SCNAs): genomically stable (GS) tumors (20%) had less SCNAs as compared to those deemed to have chromosomal instability (CIN, 50%). 6 Commonly occurring genetic mutations in GEAs include ARID1A, PIK3CA, KRAS, and CDH1. 7 However, TP53 is the most frequently altered gene with approximately 50% of GEAs and 70% of CIN GCs harboring mutations. 6,8 TP53 encodes the tumor suppressor p53 protein that functions as a key member of the G1/S checkpoint to help maintain genetic integrity through the cell cycle. 9 In times of cellular stress TP53 expression increases to mediate G1 phase cell cycle arrest to facilitate repair before cellular division or promote apoptosis in cells with an overabundance of molecular derangements that cannot be overcome. When there is a loss of TP53 function, the G1/S checkpoint is effectively lost, and genetically damaged cells have significantly less barriers to unchecked proliferation with negative physiologic consequences. Therefore, it is not surprising that inactivating TP53 mutations have been implicated in the carcinogenesis of a multitude of malignancies, with GEAs being no exception. 10 In fact, the Asian Cancer Research Group (ACRG) included TP53 expression as part of their proposed molecular classification of GC. 11 Using gene expression panels derived from 300 GCs, the ACRG first separated GCs molecularly into 2 large categories: MSI and microsatellite stable (MSS). MSS tumors were then stratified by the presence of, or lack thereof, an epithelialto-mesenchymal gene signature (MSS/EMT). The remaining MSS/EMTtumors were divided into the final 2 categories, MSS/TP53 + and MSS/TP53 − , based on levels of TP53 expression with noted differences in patient characteristics between these 2 groups. 11 Next-generation sequencing (NGS) is now widely used for most advanced cancers as an aid to elucidate the genomic characteristics of tumors. While therapeutic interventions for GEAs taking TP53 mutation status into account have yet to be developed, its high prevalence makes it a subject of interest to clinical investigators as a potential predictive marker. 4,10,[12][13][14] An understanding of commonly co-occurring genetic alterations in TP53 wild-type (TP53 WT ) and TP53 mutated (TP53 MUT ) GEAs may help guide clinically practical investigations in the future assessing this potential. This study identified the unique mutational profiles of TP53 WT and TP53 MUT GEAs by analyzing the DNA sequencing results of 3741 GEA tumor samples.

Patients and Methods
De-identified data were acquired from Foundation Medicine for all available patients with GEA for whom the FoundationOne CDx (F1CDx) assay was performed within the US. F1CDx is a commercially available NGS diagnostic test developed by Foundation Medicine that uses targeted high throughput hybridization-based capture technology for the detection of substitutions, insertion and deletion alterations, and copy number alterations in 324 genes and select gene rearrangements using DNA isolated from formalin-fixed paraffin-embedded tumor tissue specimens. 15 The data provided included age at the time of the assay results, gender, tumor mutational burden (TMB), and the distinct genomic alterations noted on DNA sequencing. Information on MSI and PD-L1 status was not available in the dataset. The dataset was sorted by TP53 mutation status. Differences in mutation frequency were detected using Fisher's exact test of independence with a P-value of <.01 designated as the cutoff value for statistical significance. Statistical significance for continuous data was tested using a 2 samples t-test for mean values and the Mann-Whitney U test for distribution/median values.
Pathway analysis was completed by first isolating the genes with statistically different mutation frequencies between TP53 WT and TP53 MUT tumors. After studying these results, 12 predetermined functional gene groupings were created to help elucidate molecular pathways suggested to be differentially affected in TP53 WT and TP53 MUT tumors. The number of tumors containing a mutation in at least one of the prespecified genes in the molecular pathways of interest was counted. Then, the association of the frequency of tumors containing a gene mutation in the molecular pathways of interest and TP53 mutation status was assessed using Fisher's exact test of independence with a P-value of <.01 deemed statistically significant.

Demographics
The dataset consisted of 3741 patients with GEA: 2946 GCs and 795 GEJs. The entire study population was 65.2% male and 34.8% female. GC patients were 60.6% male and 39.4% female, whereas patients with GEJ were 82.4% male and 17.6% female (P < .001). The median patient age was 63, which was similar between TP53 WT and TP53 MUT patients.

Pathway Analysis
The pathway analysis results are represented in Fig. 3. TP53 WT GEAs were more likely to contain a mutation in a gene related to DNA mismatch repair, homologous recombination repair, DNA and histone methylation, Wnt/B-catenin, PI3K/Akt/mTOR, and chromatin remodeling complexes.
Furthermore, TP53 WT tumors were more likely to contain any mutation in the gene panel as compared to TP53 MUT tumors. TP53 MUT GEAs were more likely to contain a mutation in a gene related to fibroblast growth factor, epidermal growth factor receptor, other receptor tyrosine kinases, and cyclin and cyclin-dependent kinases. When broken down by tumor type, GCs had pathway analyses results that closely mirrored those of the entire dataset with the only difference being that the frequency of mutations in DNA and histone methylation genes was no longer statistically different between

Discussion
The most common predictive biomarkers currently used in GEAs are microsatellite instability or mismatch repair deficiency status and expression of human epidermal growth factor receptor 2 (HER2) and programmed death-ligand 1 (PD-L1). 16,17 There are other biomarkers that will soon have therapeutic implications, such as Claudin 18.2 and FGFR2b. 18,19 The use of these predictive biomarkers has helped us tailor treatments for patients with GEA, yet there is still an unmet need to identify additional targets for novel treatment strategies. In our study, we were able to demonstrate that the mutational profiles of GEAs differed when stratified according to TP53 mutation status. TP53 mutations were identified at higher rates in our sample of GEAs as compared to those described by TCGA and other manuscripts reporting such data. 4,8,20,21 Genes directly involved in the DNA damage response and epigenetic regulation of DNA were more frequently mutated in TP53 WT tumors. Meanwhile, growth factor and receptor tyrosine kinase genes were more frequently mutated in TP53 MUT tumors.
GEJ tumors were more likely to be TP53 MUT than their GC counterparts (81.4% vs. 61.6%, P-value = <.001) in this dataset, which is consistent with DNA sequencing results previously described. 22,23 Given this variability, there was interest in comparing GEJ and GC mutation profiles against one another to assess any further genetic differences stratified by TP53 mutational status that may distinguish the 2 GEA malignancies. Homologous recombination repair genes were more frequently mutated in both TP53 WT GEJs and GCs. Additionally, the ratio of mutational involvement of DNA mismatch repair, DNA and histone methylation, epidermal growth factor, and receptor tyrosine kinase genes appeared to be consistent in GEJs and GCs when broken down by TP53 mutation status. The remaining functional gene groupings, however, did not seem to have as much mutational variation between TP53 WT and TP53 MUT tumors in GEJs as compared to GCs. This can be most readily seen when looking at the involvement of cyclin and cyclin-dependent kinase genes, which play an integral role in the maintenance of the cell cycle in concert with TP53. Twenty-seven percent vs. 44.2% of TP53 WT and TP53 MUT GCs had cyclin and cyclin-dependent kinase gene mutations. Meanwhile, 54.7% vs. 56.9% of TP53 WT and TP53 MUT GEJs had cyclin and cyclin-dependent kinase gene mutations. On a gene-specific level, ZNF217 (15.5%) and AURKA (8.1%) amplification mutations were disproportionately seen in TP53 WT GEJs as compared to the rest of the dataset (4.4% and 2.1%, respectively). As both ZNF217 and AURKA genes are located on chromosome 20q13, it stands to reason that 20q13 amplification may play a prominent role in the carcinogenesis of TP53 WT GEJs. 20,24 These findings suggest that despite their anatomic proximity the genetic makeup of GEJs and GCs are distinct from one another.
Studies assessing the prognostic impact of TP53 mutations on the outcomes of GEAs have demonstrated conflicting  results. [25][26][27][28][29] The molecular characterization of GEAs in relation to TCGA subtype, co-occurring mutations, and the type of TP53 mutation itself all seem to influence TP53's prognostic significance. 11,12,30 Equally, the ability of TP53 mutation status to function as a predictive marker of treatment response is also unknown. A meta-analysis of 13 studies including 564 cases concluded that GCs with high levels of p53 expression on immunohistochemistry (IHC) have a decreased response to chemotherapy. 31 Unfortunately, the correlation of p53 expression on IHC with TP53 mutational status is not consistent, making it difficult to extrapolate the findings of this analysis toward tumor DNA sequencing results. 32,33 More recent studies have attempted to delineate the predictive potential of TP53 mutations in the era of molecularly targeted therapies and immunotherapies. One manuscript supporting its predictive ability in targeting the vascular endothelial growth factor (VEGF) pathway was recently published. 34 In 48 patients with GC who received second-line ramicurimab/paclitaxel combination therapy the median OS was 9.5 months for carriers of TP53 Inactive mutations, 8.6 months for carriers of other TP53 mutations, 6.0 months for carriers of TP53 Active missense mutations, and 4.5 months for TP53 WT patients (P = .01). VEGFA was accordingly more mutated in our group of TP53 MUT GEAs as compared to TP53 WT GEAs (4.8% vs. 1.6%, −log 10 (P-value) = 6.3). In another review of 356 GCs from the TCGA database, it was found that TP53 MUT tumors had lower levels of anti-tumor immunity and a decreased response to immune checkpoint inhibitor therapy as compared to TP53 WT tumors. 35 On the other hand, sub-group analysis of the phase III GOLD trial investigating olaparib and paclitaxel combination therapy in recurrent and metastatic GC found no difference in therapeutic response based on TP53 mutation status. 36 More studies are needed to assess the predictive ability of TP53 mutation in isolation and in conjunction with other molecular alterations. The limitations of this study include missing data regarding clinical outcomes, microsatellite instability status, expression of other protein biomarkers, and additional patient characteristics necessary to organize these tumors into one of the 4 molecular subtypes as proposed by the Cancer Genome Atlas Research Network. In addition, while tissuebased assays offer a comprehensive genomic profile, there can be intra-and intertumoral (primary vs. metastatic site) heterogeneity which may not be fully encompassed on tissue NGS. 37 This potentially limits the clinical utility of tissuebased mutation analysis as the full mutational landscape of the patient's tumors may not be elucidated due to the sampling limitations inherent with this method of genetic profiling. This notion, along with the invasiveness of obtaining tumor samples for molecular analysis, has led to the increased use of liquid biopsies assessing circulating tumor DNA (ctDNA) NGS results in conjunction with tissue NGS to guide clinical decision-making. Nonetheless, as tissuebased DNA/RNA sequencing results yet remain standard in clinical practice, the observations of this study could serve as a foundation for practical future preclinical and clinical investigations that can help elucidate the clinical correlations of our findings. For example, genes in the PI3K/AKT/ mTOR pathway were significantly more mutated in TP53 WT tumors, which was largely driven by the greater presence of PIK3CA mutations in TP53 WT as compared to TP53 MUT tumors (16.2% vs. 7.9%). Trials are currently underway assessing the role of PI3K inhibitors for PIK3CA mutated tumors in GEAs. It would be useful to perform a subgroup analysis of the results of these trials stratified by TP53 mutation status to determine if these findings signal a pathophysiologic difference that carries predictive relevance in the real-world setting. Similar analyses could be completed for additional studies investigating therapeutic interventions targeting genes and cellular products related to the DNA damage response, epigenetic regulation of DNA, growth factors, receptor tyrosine kinases, cyclins and cyclin-dependent kinases, and the PI3K/AKT/mTOR pathway.
Despite the inherent limitations of this qualitative study, compelling trends within a large database of tumor DNA sequencing results derived from clinically accessible information were identified. A clinical trial assessing the efficacy of berzosertib and irinotecan in TP53 mutant GEA cancers is currently ongoing (NCT03641313) which is the first clinical trial selected for patients based on TP53 status. The mutational differences of GEAs stratified by TP53 mutation status as described in this manuscript can be used as a tool when designing and analyzing future pre-clinical and clinical trials to assess the predictive ability of TP53 mutational status when targeting differentially affected molecular pathways.

Funding
This work was supported in part by the Biostatistics and Bioinformatics Shared Resource at the H. Lee Moffitt Cancer Center & Research Institute, an NCI designated Comprehensive Cancer Center (P30-CA076292).