Coexistent ARID1A-PIK3CA mutations are associated with immune-related pathways in luminal breast cancer

Up to 40% of luminal breast cancer patients carry activating mutations in the PIK3CA gene. PIK3CA mutations commonly co-occur with other mutations, but the implication of this co-occurrence may vary according to the specific genes involved. Here, we characterized a subgroup of luminal breast cancer expressing co-mutations in ARID1A and PIK3CA genes and identified their effect on important signaling pathways. Our study included 2609 primary breast cancer samples from the TCGA and METABRIC datasets that were classified based on tumor subtype and the existence of mutations in PIK3CA and ARID1A genes. Differential expression and WGCNA analyses were performed to detect molecular modules affected by the existence of the mutations. Our results reveal various evidence for the involvement of immune-related pathways in luminal tumors harboring ARID1A and PIK3CA mutations, as well as a unique Tumor-infiltrated immune cells composition. We also identified seven key hub genes in the ARID1A-PIK3CA mutated tumors associated with immune-related pathways: CTLA4, PRF1, LCK, CD3E, CD247, ZAP70, and LCP2. Collectively, these results indicate an immune system function that may contribute to tumor survival. Our data induced a hypothesis that ARID1A and PIK3CA mutations' co-occurrence might predict responses to immunotherapy in luminal BC and, if validated, could guide immunotherapy development.


Data resource and study design
Normalized gene expression profiles, probe annotation and clinical information of 3054 samples were downloaded from the TCGA 17,18 (1073 samples) and METABRIC [19][20][21] (1,981 samples) databases using the MetaGx-Breast R package 22 (version 1.12.0).In addition to the clinical information obtained from MetaGxBreast, PAM50 classification (2,695 samples), information regarding somatic non-synonymous mutations (2,914 samples), tumor mutational burden (TMB) (2,606 samples), and protein expression measured by reverse-phase protein array (RPPA) z-score for TCGA samples (731 samples) were downloaded from cBioPortal for Cancer Genomics 23,24 .Samples without mutational or PAM50 information were excluded from the analysis, leaving 2,609 BC primary tumor samples.Throughout the analysis, we grouped luminal A and luminal B samples together ("luminal samples").The classification was done according to PAM50.
Data processing was performed using R studio software (version 4.1.1) 25.The study design and the comparisons performed are presented in Fig. 1a,b.The control group in this study consisted of tumor samples with no mutations in either ARID1A or the PIK3CA genes.

Identification of differentially expressed genes (DEGs)
The limma 26 (version 3.48.3)and edgeR 27,28 (version 3.34.1)R packages were used for batch effect correction and detection of DEGs among the different groups.Unless specified otherwise, a false discovery rate (FDR) of < 0.05 and fold change > 1.5 were used as the statistical cutoff for differential expression.

Weighted correlation network analysis (WGCNA)
The WGCNA 29,30 R package (version.1.70.3) was used to perform signed network analysis from all gene expression profiles for the different cohorts.For network construction, the selected power (β) was 12, and scale-free topology was set to a minimum of 0.8.Cluster dendrogram visualized the modules represented in different colors, and dynamic modules with high similarities were combined into one at a cutline of 0.

Pathways enrichment analysis
Pathways enrichment analysis was performed against the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways database through the WebGestalt 2019 31 tool.DEGs found in each comparison and genes included in WGCNA selected modules were used as input for over-representation analysis (ORA).Ranked gene lists with their corresponding logFC from each comparison were used as input for gene set enrichment analysis (GSEA).Only ontologies and pathways with FDR < 0.05 are shown.

Identifying Hub genes
Hub genes are genes in candidate modules highly correlated to a specific trait in the WGCNA analysis.Unless otherwise mentioned, the top 10% of genes in a candidate module with MM > 0.8 and GS > 0.2 were considered hub genes 32 .These genes were used as input to the Search Tool for Retrieval of Interacting Genes/Proteins (STRING) 33 database to perform protein-protein interaction (PPI) analysis.Visualization was performed using the Cytoscape softwere 34,35 .Genes with the highest degrees of connectivity in the PPI network play a critical role in the module 32 and were therefore defined as key hub genes.

Cell-type identification by estimating relative subsets of RNA transcripts (CIBERSORT)
Tumor-infiltrated immune cells were estimated using CIBERSORTx 36,37 .The gene expression profile was used as input.The proportion of twenty-two immune cell subtypes was measured in the absolute mode for each sample using an LM22 signature with 100 permutations for p value calculation.In WGCNA analysis, the abundance of each immune cell was treated as a continuous variable.

Tumor infiltrated immune cells and gene expression correlation
The Pearson correlation between gene expression and tumor infiltrated immune cells absolute score detected by CIBERSORTx and the correlation significance were calculated using the corrplot 38 R package (Version 0.92).

Statistical analysis
This study's statistical analyses were conducted using R 4.1.1statistical framework.Differences in categorical variables between groups were analyzed using Fisher's exact test.The nonparametric Kruskal-Wallis test followed by the Dunn post hoc test were performed to compare continuous parameters between different groups.Survival curves were estimated using the Kaplan-Meier method and visualized by Survival 39,40 (Version 3.5-3) and Survminer (Version 0.4.9)R packages, and p values were calculated using the log-rank test.Effects were considered statistically significant at a two-sided p < 0.05.All statistical analyses were performed using the R statistical framework (v.4.4.1).A ggplot2 41 R package (Version 3.3.5)was used for generating plots.Venn diagrams were prepared using the VennDiagram 42 (Version 1.7.3)R package.

Coexistent ARID1A-PIK3CA mutations in luminal breast cancer
To assess the prevalence of ARID1A and PIK3CA mutations co-occurring in BC, we examined their frequency in various BC subtypes using the TCGA 17,18 and METABRIC [19][20][21] datasets, which included 2,609 patients.Of these patients 1,780 were classified as Luminal (including 1,134 Luminal A and 646 Luminal B patients), 291 were Her2-enriched, 371 were Basal and 167 Normal subtype.Most patients were over 50 years old, had tumor sizes ranging from 2.1 to 5 mm and exhibited tumor grades of 2-3.Among the patients with luminal tumors, 51 (2.86%) had mutations in ARID1A gene, 780 (43.82%) had mutations in PIK3CA gene, 50 (2.8%)patients had mutations in both the ARID1A and PIK3CA genes, and 899 (50.5%) women served as a control group showing no mutations in any of these genes (Table 1 and Fig. 1a,b).Our analysis revealed that PIK3CA mutations, as well as ARID1A-PIK3CA mutational co-occurrence, were significantly more common (p ≤ 0.05) in luminal tumors than in other BC subtypes (Fig. 2a,b).The distribution of ARID1A and PIK3CA mutation types in breast tumors is illustrated in Fig. 2c.
The PIK3CA gene exhibited a predominance of missense mutations, whereas ARID1A displayed a wide variety of mutation types.Nevertheless, approximately 40% of ARID1A mutations were identified as nonsense or frame-shift mutations, which contribute to the decay of mRNA and the loss of ARID1A protein function.In fact, reduced expression levels of PIK3CA and ARID1A proteins were observed in mutated luminal tumors of PIK3CA and ARID1A, respectively (Suppl Fig. 1).As our investigation revealed significant enrichment in ARID1A-PIK3CA mutational co-occurrence exclusively in the luminal subtype, our research aimed to analyze this specific BC subtype.
Subsequently, we investigated the potential correlation between coexisting ARID1A-PIK3CA mutations and various clinicopathological parameters in luminal tumors, such as disease grade, tumor size, patient age, and overall survival.However, our analysis indicated that none of the aforementioned clinicopathological features exhibited a significant association with coexisting ARID1A-PIK3CA mutations.PIK3CA mutations alone were enriched within grade 3 tumors (Table 1, Suppl Fig. 2).

Immunity-related signaling pathways are associated with coexistent ARID1A-PIK3CA mutations in luminal tumors
To obtain an overview of the main gene expression variations in luminal BC patients with coexisting ARID1A-PIK3CA mutations, we conducted several gene expression comparisons (DEGs detected in each comparison are listed in Suppl file 1).Specifically, we identified differentially expressed genes (DEGs) between patients with coexisting ARID1A-PIK3CA mutations and control samples (with no mutations in either ARID1A or in the PIK3CA genes), patients with either ARID1A or PIK3CA mutations alone, and control samples.The Venn diagram in Fig. 3a displays the number of DEGs identified exclusively in each comparison group, as well as the overlapping DEGs.Upon comparing gene expression in ARID1A-PIK3CA mutational co-occurrence versus control samples, we identified a total of 135 DEGs, with 120 genes being unique to this pairwise comparison and not overlapping with other comparison groups.To gain insight into the biological functions of these unique DEGs, we performed an overrepresentation enrichment analysis (ORA) on these genes.Our analysis revealed 13 pathways which were significantly enriched (FDR < 0.05), most of which were immune-related, such as T-cell receptor, B-cell receptor, Th1 and Th2 cell differentiation, Natural killer cell-mediated cytotoxicity, etc. (Fig. 3b).Additionally, to identify potential biological functions associated with ARID1A-PIK3CA mutational co-occurrence tumors, we conducted gene set enrichment analysis (GSEA) and found 49 enrichment pathways, of which 35 were upregulated and 15 downregulated.The majority was related to immune responses (Fig. 3c).
Comparing PIK3CA mutated tumors to the control identified 90 unique DEGs (Fig. 3a).GSEA analysis revealed upregulation of 28 pathways and downregulation of 22 of which only nine were immune-related (Suppl Fig. 3).Notably, comparing ARID1A mutated tumors to the control revealed only 12 unique DEGs (FDR < 0.05 with no fold change cutoff).To better comprehend the significance of the coexisting ARID1A-PIK3CA mutations to immune system participation in luminal malignancies, we compared gene expression between patients with coexisting ARID1A-PIK3CA mutations and patients with PIK3CA mutations alone and detected 33 DEGs (FDR < 0.05 with no fold change cutoff) of which 18 are immune related.GSEA analysis revealed upregulation of 33 pathways, 20 of them are immune related (Fig. 3d).
To confirm our findings with an additional approach, we identified the co-expression network most strongly associated with ARID1A-PIK3CA mutational co-occurrence tumors using Weighted Gene Co-expression Network Analysis (WGCNA).This approach allowed us to identify genes that are highly correlated with specific traits through their co-expression across samples.All luminal tumors were included in the WGCNA analysis.Figure 4a displays the correlations of the module eigengenes with the traits.As expected, most of the gene modules are correlated with pathological features such as the grade.When looking for modules correlated with molecular features, the brown module was found to be the most significantly related to the ARID1A-PIK3CA mutational co-occurrence trait, despite the low correlation (correlation = 0.097, p value = 4E-05), followed by the Greenyellow (correlation = 0.091, p value = 1E-04) and Orange modules (correlation = 0.078, p value = 9E-04).Over-representation analysis (ORA) of the brown module genes showed that most of them were mainly associated with immune pathways, such as primary immunodeficiency, natural killer cell-mediated cytotoxicity, and Th1 and Th2 cell differentiation (Fig. 4b).ORA of the Greenyellow and Orange module genes revealed mainly pathways related to autoimmune diseases and drug metabolism respectively (Suppl.Figure 5).
The tan module was the most correlated to the PIK3CA mutation-only trait (correlation = 0.24, p value = 3E-24) (Fig. 4a).Genes included in this module were involved among others in ECM-receptor interaction and protein digestion absorption (Fig. 4c).The immune-related brown module was also significantly associated with PIK3CA mutation-only trait albeit with lower correlation (correlation = 0.087, p value = 2E-04) than that of the co-mutation trait.Notably, no module was significantly related to the ARID1A mutation-only trait.To ensure that the presence of an ARID1A mutation in the co-mutated trait is indeed contributing to the association with the immune system, we performed a secondary round of WGCNA analysis.This time, we focused solely on luminal patients who had mutations in the PIK3CA gene.In this analysis, the examined trait is the addition of ARID1A mutation as all patients included here are PIK3CA-mutated.

Identification of Hub genes associated with ARID1A-PIK3CA mutational co-occurrence luminal tumors
PIK3CA is frequently mutated in 40-50% of estrogen receptor-positive (ER +) tumors 43 .Luminal tissues harboring PIK3CA mutations were selected for additional analysis using WGCNA.The correlation between relevant modules and clinical characteristics is demonstrated through a heatmap depicted in Fig. 5a.The Yellow module exhibited the strongest correlation with ARID1A-PIK3CA mutational co-occurrence trait (correlations = 0.14, p value = 6E-05).ORA analysis revealed a significant association with immune signaling pathways (as depicted in Fig. 5b).Subsequently, 87 hub genes representing the top 5% of genes with trait gene significance ≥ 0.1 were identified within the yellow module.The enriched pathways associated with these hub genes were mainly related to the immune system, as shown in Fig. 5c.Using the Cytoscape software and STRING database, a protein-protein interaction (PPI) network was constructed for the hub genes, revealing seven key hub gene with a high degree of connectivity (CTLA4, PRF1, LCK, CD3E, CD247, ZAP70, and LCP2).These genes were found to be significantly associated with immune-related pathways, such as the T-cell receptor signaling pathway and natural killer cellmediated cytotoxicity, as illustrated in Fig. 5d,e.Notably, CTLA4, one of the hub genes, is known to play a crucial role in immune escape 2 .The expression of CTLA4 was found to be upregulated in ARID1A-PIK3CA mutational co-occurrence tumors compared to the control (as shown in Fig. 5f).Lastly, a Venn-diagram in Fig. 5g revealed an overlap between 135 unique DEGs genes of ARID1A-PIK3CA mutational co-occurrence tumors, 80 hub genes, and the seven key hub genes.These findings collectively suggest that immune-related pathways may contribute to tumorigenesis in patients with ARID1A-PIK3CA mutational co-occurrence tumors.

The hub genes exhibit strong positive correlations with immune cell infiltration in ARID1A-PIK3CA mutational co-occurrence luminal breast cancer
Due to the link between immune pathways and ARID1A-PIK3CA mutational co-occurrence in luminal tumors, we proceeded to investigate the infiltration of immune cells in the tumor microenvironment by utilizing the CIBERSORTx algorithm.This algorithm employs gene expression profiles to evaluate the abundance of 22 immune cell subsets in the samples 44 .Our analysis revealed a higher proportion of B cell memory, T cells follicular helper, T cells regulatory (Tregs), and Macrophages M1 in ARID1A-PIK3CA mutational co-occurrence luminal tumors compared to the PIK3CA mutation-only group.Conversely, mast cells and eosinophils infiltration were significantly higher in the latter group (Fig. 6a).
We also investigated the immune cells infiltration in ARID1A-PIK3CA mutational co-occurrence luminal tumors compared to both ARID1A mutation-only and control groups (no mutations in these genes at all), and found similar results (Suppl Fig. 6).Additionally, we performed correlation analysis to corroborate the relationship between tumor infiltrated immune cells and the expression of the seven key hub genes (CTLA4, PRF1, LCK, CD3E, CD247, ZAP70, and LCP2) in ARID1A-PIK3CA mutational co-occurrence tumors.Our results indicate a positive correlation between immune cells and these key hub genes, whereas ARID1A and PIK3CA expression exhibited a negative correlation with immune cells (Fig. 6b).A weaker but similar correlation was observed in the PIK3CA mutation-only group (Fig. 6c).Furthermore, we identified gene sets related to immune infiltration cells in ARID1A-PIK3CA mutational co-occurrence tumors using WGCNA analysis (Fig. 6d).The turquoise module displayed a highly positive correlation with B cells memory, T cells CD4 memory activated, Tregs, Macrophages M1, and resting Dendritic cells.By leveraging the Cytoscape software and STRING database, we identified nine key hub genes in the turquoise module associated with T and B cells' signaling pathways and natural killer-cell-mediated cytotoxicity (Fig. 6e).

High Tumor Mutation Burden (TMB) in ARID1A-PIK3CA mutational co-occurrence tumors
Tumor mutational burden (TMB) is a major immune biomarker demonstrated as a useful marker of checkpoint inhibitors' effectiveness and is predictive for immune therapies in advanced BC 45 .Considering the involvement of the immune system in ARID1A-PIK3CA mutational co-occurrence tumors, we evaluated the TMB distribution of the tumors.The median TMB in ARID1A-PIK3CA mutational co-occurrence tumors was higher than in the PIK3CA mutation-only group (Fig. 7a).Estimating the prevalence of patients with TMB above 10 nonsynonymous mutations per MB (high TMB) in ARID1A-PIK3CA mutational co-occurrence tumors, we found that high TMB were significantly enriched in ARID1A-PIK3CA mutational co-occurrence tumors compared with PIK3CA mutation-only group (p = 0.009, OR = 2.28) (Fig. 7b).Moreover, as shown in Fig. 6d, the Sienna3 module was highly positively correlated with TMB in ARID1A-PIK3CA mutational co-occurrence tumors.Key hub genes of this module are related to the immune system, i.e., defense response to viruses and interferon alpha/   beta signaling (Fig. 7c).These results support the potential suitability of luminal ARID1A-PIK3CA mutational co-occurrence tumors for immunotherapy.

Discussion
Luminal BC represents a highly heterogeneous subgroup in terms of somatic mutations and copy number alterations, which can be characterized by different risks of relapse 4,15,17 .Considering the great genetic heterogeneity of luminal tumors, it seems that classification of a new sub-group in this type of tumor based on genetic alterations could be fundamental, both in predicting the clinical outcome and in therapeutics improvement.
Since ARID1A and PIK3CA mutations are known to cooperate in initiating ovarian and gastric cancer 12,16 , we focused here on luminal tumors with a PIK3CA gene mutation and an additional mutation in ARID1A gene in an attempt to identify a new sub-group of luminal tumors.We analyzed the frequencies and possible involvement in pathogenesis of ARID1A and PIK3CA mutations co-occurrence in luminal tumors.We found that PIK3CA mutations were significantly enriched in luminal tumors, in agreement with other studies 5 .We also characterized, an enriched sub-group of luminal tumors with ARID1A and PIK3CA mutations co-occurrence.
In our analysis of differential gene expression, we observed a relatively high number of DEGs when comparing samples with co-occurring mutations in ARID1A and PIK3CA to the control group.Similarly, we found a substantial number of DEGs when comparing samples with PIK3CA mutations to the control group.However, there were significantly fewer DEGs detected when comparing samples with ARID1A mutations to the control group.We assume, that the presence of various mutation types in the ARID1A gene contributes to increased variability in gene expression within the mutated samples, resulting in a relatively low number of differentially expressed genes.Expanding the sample size would enable a more detailed analysis based on mutation types, potentially reducing sample to sample variability and revealing a greater number of mutation-specific DEGs.
Importantly, we found a strong link between ARID1A and PIK3CA mutation co-occurrence and immunerelated pathways, in particular B-cell and T-cell signaling.We also identified seven key hub genes in ARID1A-PIK3CA mutational co-occurrence tumors: CTLA4, PRF1, LCK, CD3E, CD247, ZAP70, and LCP2.As expected, these genes were associated with immune-related pathways.The presence of infiltrated immune cells in these tumors, some of which inhibit or promote disease progression, also supports the idea that immune pathways are involved in this subgroup.
A link between coexistent ARID1A-PIK3CA mutations and immune pathways was previously demonstrated in ovarian cancer 12 .Interestingly, the PIK3CA mutation combines with another gene mutation in immune-related breast tumors.As evidence, PIK3CA and CDH1 mutations act together to induce immune-related invasive lobular carcinoma of the breast 46 .
Our data revealed an association between ARID1A-PIK3CA mutational co-occurrence tumors and immune pathways.A link between the immune response and luminal BC cells where the PI3K pathway is disrupted has already been published, providing evidence that tumor infiltration of CD8-positive lymphocytes is associated with PIK3CA mutations in luminal tumors 47 .The next question would be whether oncogenic ARID1A-only mutations also contribute to immune signaling regulation in this subgroup.We could not answer this question using our data because comparing ARID1A-mutated tumors to the control revealed a very small number of DEGs.However, a recent report shows that ARID1A mutation can enhance the immunogenicity of tumor cells in BC 48 and were found to be associated with immune activity in gastrointestinal and microsatellite-stable colorectal cancers 49,50 .While we cannot determine with certainty that the ARID1A mutation, in our case, also directly activates or suppresses the immune pathways, we suggest that it contributes to and encourages the general activation of the immune system and increases the activity of the PIK3CA mutation.
Notably, we observed a significant negative correlation between the expression of ARID1A and PIK3CA genes and the presence of infiltrated immune cells.This intriguing correlation remains consistent even though we did not identify any significant alterations in these genes expression when comparing mutated and non-mutated samples.However, both genes exhibited significant decrease of their protein levels.The fact that we don't see difference in mRNA levels can be explained due to heterogenous mutation types which collectively lead to an overall absence of significant change.Furthermore, the observed negative correlation between gene expression and immune cell infiltration, may result from even small changes in mRNA levels, underscoring the sensitivity of this relationship.
Among luminal tumors, a high proportion of immune cells predicted a worse prognosis, suggesting that an efficient immune escape could be an important factor influencing recurrence in luminal BC 2,51 .CTLA4 has an important role in immune escape, as its expression on Tregs could suppress the anti-tumor immune response by inhibiting the proliferation of effector T lymphocytes and was associated with worse disease-free survival and overall-survival 2 .Notably, our work revealed that the key hub gene CTLA4 was associated with ARID1A-PIK3CA mutational co-occurrence luminal tumors, and its expression was higher in these tumors than in the control.Although the exact underlying mechanism remains unclear, considering these cancer-immune interactions in luminal tumors might provide useful information for further development of immunomodulatory combination treatments for luminal tumors.
Nowadays, the efficacy and safety of immunotherapy in luminal tumors have been tested in various ongoing clinical trials 52,53 .For example, a Phase II Trial of Ipilimumab (Anti-CTLA4) and Nivolumab (Anti-PD-1) conducted in unresectable or metastatic metaplastic BC, part of the patients in this trial have luminal tumors 2, 54 .It is very important to focus on identifying luminal patients who may benefit from immunotherapy and offer them the appropriate immunotherapy treatment.We hypothesize that the new sub-group of luminal tumors with ARID1A and PIK3CA mutation co-occurrence we identified could benefit from immunotherapy treatment due to the strong association between these tumors and immune-related pathways.Specifically, it is worth testing the efficacy of ICBs, such as CTLA4 inhibitors in this new sub-group, as CTL4A was defined as a key hub gene in these tumors.Moreover, high TMB, a useful marker of checkpoint inhibitors' effectiveness in advanced BC 45 , was significantly enriched in the new sub-group of luminal tumors.
In conclusion, we identified an enriched sub-group of luminal tumors with ARID1A and PIK3CA mutation co-occurrence.Our results reveal different evidence for the involvement of immune-related pathways and tumorinfiltrating immune cells in these tumors.Particularly, our data raise the hypothesis that ARID1A and PIK3CA mutations' co-occurrence may be predictive for responses to immunotherapy in luminal BC and, if validated, could guide immunotherapy development in this context.

Figure 1 .
Figure 1.Study design and datasets.(a) Study design and (b) patients' cohorts used for analysis.

Figure 4 .
Figure 4. Identification of key modules associated with ARID1A-PIK3CA mutational co-occurrence luminal tumors utilizing weighted gene co-expression network analysis (WGCNA).(a) Heatmap of the correlation between consensus module eigengenes and clinical traits.Each column corresponds to a clinical trait, and each row corresponds to a module.Numbers and colors in the table represent correlation coefficients; red marks positive correlation, while green represents negative correlation.The p values are shown in parentheses under the correlations.The RNA-seq expression dataset obtained from patients with coexistent ARID1A-PIK3CA mutations (n = 50), PIK3CA only mutations (n = 780) and ARID1A only mutations (n = 51).(b-c) Over Representation Analysis (ORA) for (b) 'brown' and (c) 'tan' genes modules.

Figure
Figure Identification of hub genes associated with immune pathways in ARID1A-PIK3CA mutational co-occurrence luminal tumors.WGCNA analysis of luminal tissues exhibiting PIK3CA mutations (n = 780).(a) A module-trait heatmap.(b) ORA for the 'yellow' module.(c) ORA for the top-ranked 5% (with trait gene significance ≥ 0.1 and module membership ≥ 0.8) hub genes in the 'yellow' module.(d) The protein-protein interaction (PPI) network of hub genes in the 'yellow' module is visualized by the Cytoscape software.Colored circles represent the connectivity degree of the hub gene.Darker colors correspond to higher degrees of connectivity.(e) The top seven key hub genes with the highest connectivity degrees in a STRING PPI network of the 'yellow' module.(f) mRNA expression levels of CTL4A were investigated in luminal tumors according to their ARID1A and PIK3CA mutation status.The RNA-seq expression dataset obtained from patients with coexistent ARID1A-PIK3CA mutations (n = 50) and controls (n = 897).The box plots display the expression distribution of CTL4A across the different groups.The line in the middle represents the median value (**p ≤ 0.01, Wilcoxon test).(g) Venn diagram showing the overlap between 135 unique DEGs genes of ARID1A-PIK3CA mutational co-occurrence tumors (n = 50) and the seven key hub genes.

Figure 7 .
Figure 7. Distribution of Tumor Mutation Burden (TMB) across ARID1A-PIK3CA mutational co-occurrence luminal tumors.(a) The box plots display the distribution of the TMB value across ARID1A-PIK3CA mutational co-occurrence tumors (n = 50) compared to PIK3CA mutation only tumors (n = 779).The line in the middle represents the median value (*p ≤ 0.05, Wilcoxon test).(b) Bar plot representing the enrichment of tumors with TMB higher than 10 (**p ≤ 0.01, Fisher's exact test).(c) The 'Sienna3' module is highly positively correlated with TMB in ARID1A-PIK3CA mutational co-occurrence tumors (n = 50), as demonstrated in Fig. 5d.The hub genes (top-ranked 10%) in the 'Sienna3' module are visualized by the STRING PPI network.

Table 1 .
) Clinical features of the study samples.*p value represents enrichment in luminal tumors versus the rest as calculated using Fisher exact test.