Interaction between HER2 and ATM predicts poor survival in bladder cancer patients

Abstract Human Epidermal Growth Factor Receptor 2 (HER2) overexpression is considered one of the interesting prognostic biomarkers in bladder cancer. However, the mechanism of bladder cancer development in relation to HER2 status remains to be elucidated. In this study, we investigated HER2‐Ataxia telangiectasia mutated (ATM) kinase interaction and their impact on patient survival and cancer aggressiveness. Using the Cancer Genome Atlas (TCGA) cohorts, we demonstrated that ATM expression (protein/mRNA) is increased in HER2 deficient compared with proficient HER2 patients. This finding was then validated using the Gene Expression Omnibus database (GEO). Correlation analysis (using low expression vs high expression as a discriminator) revealed a significant association of ATM low and HER2 high status with several clinicopathological variables such as high tumour grade, late disease stage and tumour shape. Kaplan–Meier survival analysis indicated that ATM low and HER2 high is a powerful prognosticator of both overall survival (OS) and disease‐free survival (DFS). Furthermore, using bioinformatics and protein/protein interaction analyses, we identified 66 putative overlapping proteins with direct link between HER2 and ATM most of which are functionally involved in transcription regulation, apoptotic process and cell proliferation. Interestingly, the results showed that these proteins are strongly linked with PI3K‐Akt pathway, p53 pathway and microRNAs in cancer. Altogether, our data pinpoint an important biological role of the interconnection between HER2 and ATM. The latter appear to be an independent prognostic biomarker and may serve as targets to develop novel combination therapies to improve the outcome of patients with bladder cancer.


| INTRODUC TI ON
Bladder cancer is the 10th most diagnosed type of cancer globally.
In men, it is the 4th most common cancer and the 8th leading cause of cancer death. 1,2 Based on the latest Globocan data, approximately 580,000 new cases and 220,000 deaths due to bladder cancer occurred in 2020 and are expected to double in the upcoming years. 2 Urothelial bladder carcinoma is the most common type of the disease as it originates mainly from the inner layer of the bladder (urothelium). Approximately 75% of urothelial cancer cases are classified as non-muscle invasive bladder cancer (NMIBC) with the remaining 25% being muscle invasive bladder cancer (MIBC) which is likely to metastasize to lymph nodes or other organs. 3 Despite advances in cancer diagnosis and therapy, bladder cancer remains a source of challenge to clinicians and healthcare providers due to high recurrence rates and the aggressive phenotype. 4 Hence, prognostic biomarkers are desperately needed to predict outcome and optimize the treatment protocol for bladder cancer patients. Recent advances in sequencing and genomics have yielded a wealth of information that could be used in personalized medicine and targeted cancer therapy. For instance, Human Epidermal Growth Factor Receptor 2 (HER2), Fibroblast Growth Factor Receptor (FGFR), Mammalian target of rapamycin (mTOR) and immune checkpoint inhibitors; currently used in clinical practice. [5][6][7][8][9][10] Several studies provide evidence on the prognosis benefit of HER2 amplification or overexpression levels in bladder cancer, suggesting the potential benefits of HER2 targeted therapies on patients' survival. [11][12][13][14] HER2 is a transmembrane glycoprotein receptor tyrosine kinase of the EGFR (growth factor receptor family). Increased activity of HER2 has been evaluated and was associated with poor prognostic in breast, gastric and bladder cancers. 15 Recently, 2411 bladder tumours were sequenced and six distinct molecular subtypes were identified; HER2-like is one of them. 16,17 Today, HER2 is considered one of the important prognostic biomarkers in bladder cancer. 18,19 Early data revealed that HER2-targeted therapy is beneficial for metastatic or advanced carcinoma patients with HER2 overexpression. 20,21 However, a phase II trial for patients with advanced or metastatic urothelial cancer overexpressing HER2 treated with trastuzumab combined with chemotherapy showed similar results achieved with chemo alone. 22,23 A more recent study indicated that patients with recurrent urothelial bladder cancer and amplified HER2 gene benefited from trastuzumab and chemotherapy. 23 The aforementioned studies highlight the clinical relevance of HER2 and the utility of anti-HER2 targeted therapy as an alternative treatment in bladder cancer. However, the exact molecular mechanisms underlying the effect of HER2, including the crosstalk between HER2 and other signalling pathways such as Ataxia telangiectasia mutated (ATM) kinase, remain to be elucidated. 24 ATM is a tumour suppressor gene that works as a genomic stability guardian due to its essential role in DNA damage response and repair. 25 In bladder cancer, ATM/RB1 mutations predicted poorer survival. 26,27 In breast cancer, ATM activity reduced recurrence time in patients with invasive HER2-positive; moreover ATM was found to be involved in HER2 tumour progression. 28,29 In gastric cancer, ATM low protein expression subtype was exclusive with HER2 high protein expression. 30 The aim of the current study was to describe an integrative analysis of HER2 and ATM interaction using the Cancer Genome Atlas (TCGA) bladder cancer cohorts to highlight the importance of both genes as potential prognosticators for bladder cancer patients. We sought to determine the expression of HER2 and ATM at the protein and mRNA levels in bladder cancer cohorts to understand their relationship and investigate their impact on patient survival and cancer aggressiveness.
Also, we attempted to identify the overlapping proteins between HER2 and ATM pathways to provide a deeper insight into the molecular interactions and functional mechanisms between these two biomarkers.

| Study cohorts
The current study is a retrospective study using four cohorts. Cohort one; ERBB2 and ATM mRNA expression in a panel of different cancer types (bladder, breast, colon, kidney renal clear cell, kidney renal papillary cell, kidney chromophobe, uterine corpus endometrial, thyroid, liver and stomach) extracted from TCGA datasets along with normal match. Excluding cancer types with less than 19 samples and cancer types with no significant different between the mRNA levels in tumours and the respective normal tissues. Data were examined in UALCAN a publicly available interactive online portal (http://ualcan.path.uab.edu/ index.html). 31 Cohort two; TCGA MIBC dataset (n = 413) was used to evaluate HER2 and ATM mRNA, protein expressions and clinicopathological information provided by cBioPortal. 32,33 In this cohort, mRNA expression z-scores (RNA Seq V2 RSEM) measured by Agilent microarray and protein expression z-scores measured by Reverse Phase Protein Array (RPPA). All data extracted from cBioPortal (https://www.cbiop ortal.org/) originally from Bladder Cancer (TCGA, Cell 2017) 34 and can be found in Table S1. We defined mRNA and protein under-or overexpression if the value is greater/less than the median cut-point of HER2 factor. Cohort three; From the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/gds/), the GSE13507 (Platform GPL6102) dataset was obtained. 35,36 In this cohort, 10 normal bladder mucosae samples, 58 normal looking bladder mucosae surrounding cancer and 165 primary bladder cancer samples were profiled for ERBB2 and ATM mRNA expression using Illumina human-6 v2.0 expression beadchip. Cohort four; GSE32548 (Platform GPL6947) also from GEO database. This cohort is 131 primary bladder cancer tumour samples analysed with Illumina HumanHT-12 V3.0 expression beadchip. 37

| Data processing, Protein-Protein Interaction (PPI) network construction and co-expressed proteins identification
In order to obtain a PPI network between HER2 and ATM, we first used Reactome (https://react ome.org/) a pathway database, 38 to find all related pathways/functions where both of our protein targets are involved. The median score of each target from Cohort two RPPA data was used. To find both functional and physical networks between our targets, we proceed with STRING (version 11.0), multiple proteins database, which will evaluate the interactive relationships (https://strin g-db.org/). 39 Using experiments, co-expression and co-occurrence as active interaction sources at high confidence (0.700). All nodes with direct interaction with both HER2 and ATM were obtained, then all co-expressed proteins were visualized by Cytoscape (https://cytos cape.org/). 40 To analyse the network, NetworkAnalyzer, 41 a plugin in Cytoscape, was applied to calculate the topology parameters. Then a centrality calculation was performed by Cytohubba, 42 a plugin in Cytoscape, using degree as a topological method in order to explore the important nodes in our sub-network.

| Functional and pathway enrichment analysis
The Database for Annotation, Visualization and Integrated Discovery tool (DAVID; version 6.8: https://david.ncifc rf.gov/home.jsp) was used to provide Gene Ontology analysis including biological process, molecular function, cellular component and also Kyoto Encyclopedia of Gene and Genomes pathway analysis (KEGG). 43 Pathway enrichment analysis was performed with the threshold of p < 0.05.

| Statistical analyses
Data analysis was performed using JMP Pro 15 (SAS Institute Inc., USA). In the univariate analysis, Chi-square test (χ 2 ) for more than five subjects per cell and Fisher exact test for less than five subjects per cell were used to evaluate the relationship between HER2 and ATM factors expression and clinicopathological variables. For the prognostic significance survival curves, Kaplan-Meier analysis was used with log-rank comparison test. In multivariate analysis, to emphasize on HER2-ATM interaction, Cox proportional hazard model was used for the multivariate survival analysis including all potential confounder factors. The proportional hazards assumption was checked, the relationship between log cumulative hazard and a covariate was linear.
Where appropriate, two-tailed Student's t-test was performed using GraphPad Prism (version 8.4.3, USA). All differences were considered statistically significant at p < 0.05, p values were two-sided; all confidence intervals were at 95%.

| Analysis of HER2 and ATM expression levels in human cancers
We initially profiled the expression pattern of ERBB2 and ATM mRNA levels in a panel of different normal and tumour tissues with bioinformatics analyses using the TCGA database (Cohort one). The data revealed an inverse relationship between the mRNA levels in tumours and the respective normal tissues in all organs.
Interestingly, we noticed that when the ERBB2 level is high in cancer compared with the matching normal tissue, the ATM level in cancer shows deregulation compared with ATM levels in normal matched tissue. This is more obvious in bladder, breast, uterine corpus endometrial and thyroid. However, when the ERBB2 expression in cancer is impaired compared with the matching normal tissue, the ATM level increases compared with the matching normal tissue, as illustrated in colon, kidney renal clear cell and kidney renal papillary cell cancers. ERBB2 and ATM levels both increases in liver cancer, stomach cancer and both decreases in kidney chromophobe cancer ( Figure 1A). It is also important to emphasize that the thresholds of ERBB2 expression are higher than the thresholds of ATM expression in all tissues.

| Relationship between HER2 and ATM in bladder cancer
We investigated the balancing mechanism between HER2 and ATM at protein and mRNA levels using bladder cancer TCGA cohort from the cBioPortal database. The current bladder cancer TCGA cohort (Cohort two) included 413 patients diagnosed with MIBC. The mean age at diagnosis was 68 years, ranging from 34 to 90 years old with a median age of 69; the median follow-up time is 17.61 months (ranging from 0 to 165.9 months). The distribution of the clinicopathological characteristics of the patients is presented in Table 1. In this cohort, first, we found that HER2 protein expression, measured by RPPA, was significantly higher than ATM protein expression as expected; p = 0.0002 ( Figure 1B). Interestingly, when we sub grouped ATM-positive and ATM-negative patients according to HER2 status, we found that ATM expression levels increased significantly in HER2 deficient patients compared with proficient HER2 patients (p = 0.008).
Similarly, at mRNA levels ( Figure 1C), total ERBB2 was significantly higher than ATM levels (p < 0.0001), and high ATM expression was observed when ERBB2 was low compared with high ERBB2 (p = 0.0009). This data was validated in an independent cohort using GEO GSE13507 dataset accessible from the National Center for Biotechnology Information (NCBI). 35,36 In this cohort (Cohort three), 10 normal bladder mucosae samples, 58 normal looking bladder mucosae surrounding cancer and 165 primary bladder cancer samples were profiled for ERBB2 and ATM mRNA expression using Illumina human-6 v2.0 expression beadchip. Figure 1D, shows total ERBB2 expression (mean = 7.8) significantly low than ATM (mean = 9) in both normal sets (p < 0.0001).
However, in primary cancer samples, ERBB2 mean increase to 8.1 with a decrease of ATM to 8.4 (p < 0.0001). Interestingly, the ATM levels increased significantly in ERBB2 deficient patients compared with ERBB2 proficient patients (p < 0.0001). We validated this finding with second dataset from GEO (Cohort four) GSE32548 (Platform GPL6947). This cohort consist of 131 primary bladder cancer tumour samples analysed with Illumina HumanHT-12 V3.0 expression beadchip. As expected ( Figure 1E), total ERBB2 was significantly higher than ATM levels (p < 0.0001) and ATM levels increased significantly in ERBB2− compared to ERBB2+ patients (p = 0.0187).

| Relationships between ATM, HER2 factors and clinical outcome
Considering the inverse relationship between the HER2 level and the ATM level, we sought to investigate whether HER2 and ATM ex- Then, we stratified patients into two subgroups according to the HER2 status. Our data indicated that patients with low ATM and high HER2 expression strongly predicts poor OS and DFS; p = 0.008 and p = 0.018, respectively (Figure 2A,B). In contrast, ATM levels did not show any difference on patients' survival when HER2 was low ( Figure 2C,D). Also, no significance was seen for ATM mRNA levels with different ERBB2 status ( Figure S4A-D). The combination of low HER2 and high ATM expression was significantly associated with late disease stage (p = 0.026), no association was observed with any other clinicopathological variables (Table S2).

| HER2-ATM Co-expression in bladder cancer
Interestingly, when stratified the whole population based on both HER2/ATM protein expression status, patients with low ATM/high

| Relationship between ATM/HER2 and clinicopathological features
To evaluate the relationship of ATM/HER2 status and the clinicopathological parameters, we used univariate analysis. Our data indicate that mRNA expression of ATM/ERBB2 was significantly associated with tumour grade (p = 0.011), disease stage (p = 0.009) and tumour shape (p = 0.001). Similarly, it was identified in the ATM/HER2 protein level with tumour grade (p < 0.001), disease stage (p = 0.008), tumour shape (p < 0.001), in addition to lymph node (p = 0.046) and metastasis (p = 0.001) ( Table 2).
Multivariate analysis was conducted to investigate whether HER2/ATM expression is an independent prognostic factor. As shown in Table 3 Table 4 is a PPI ranked by degree method illustrating the degree, betweenness centrality and closeness centrality. Also, 11 topological analysis methods were applied (Table S3). Then, we assessed the significant differentness in the overlapping proteins in the stratified patients cohort (Cohort two) according to HER2-ATM status. Data identified that in the low HER2 cohort and different ATM levels (high/low); ABL1, SMAD4, RB1 and PARP1 were significantly upregulated and AKT1, AKT2, TSC2, RPTOR and mTOR were significantly downregulated in bladder cancer (Table S4). The data demonstrated significant alteration in ATM expression according to HER2 status. Our data were in agreement with another study that attempted to explore the association of immune markers in gastric cancer patients. The authors showed that a subgroup of the analysed cohort is enriched in ATM low protein expression and HER2 high protein expression. 30 The finding that ATM mRNA expression increased significantly in ERBB2 low patients compared with ERBB2 high patients was further validated in independent GEO cohorts. This transcriptomic and proteomic-based analysis provides evidence of a strong correlation between inhibition of HER2 expression and increased ATM expression in bladder cancer.

| Functional and pathway enrichment analyses
We observed that HER2 overexpression or/and amplified ERBB2 of bladder tumours strongly associated with clinicopathological variables characteristic of poor prognosis, including high tumour grade, tumour stage, late disease stage, tumour shape, lymph node and metastasis in bladder cancer, our data are consistent with previous reports. 18,49 ATM expression alone did not show a strong association with any clinical features including patients' survival. However, previous studies revealed that ATM mutation was an indicator for poor overall survival in bladder cancer compared with the wild-type. 26,50 Recent emerging data indicated that bladder cancers harbouring

| CON CLUS ION
To our knowledge, this is the first time where the inverse relationship between ATM and HER2 was highlighted in bladder cancer.
The impact of such a relationship on the prognostic outcome of patients with bladder cancer was also described. Even though our data demonstrated potential overlapping proteins between ATM-HER2 pathways, which could contribute to bladder cancer pathogenesis, the exact molecular mechanism and biological significance of the crosstalk between HER2 and ATM still require further investigations to improve prognosis and treatment efficacy in bladder cancer. The main limitation of our study is that it was a retrospective observational study and further analyses with larger sample size are needed to investigate the relation between HER2, ATM and bladder cancer prognosis. Also, the results of the present study were based on bioinformatical analysis and must be validated further by experimental test.

ACK N OWLED G EM ENTS
This work was supported by a grant from King Abdullah International Medical Research Center (Protocol No.: RJ17/072/J).

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data generated or analysed during this study are included in this published article and its supplementary information files. Also, both TCGA and GEO are publicly available as indicated in the materials and methods/study cohorts section.