Clinicopathological Features, Tumor Mutational Burden, and Tumour-Infiltrating Lymphocyte Interplay in ERBB2-Mutated Breast Cancer: In Silico Analysis

Recent evidence suggests that somatic mutations in ERBB2 activate ERBB2 signaling. These mutations occur at a frequency of approximately 3% in breast cancer (BC). ERBB2 mutations indicate poor prognosis as they are associated with recurrence and metastasis. This study aimed to evaluate the clinicopathological features, immune infiltration levels, tumor mutational burden (TMB), and tumor-infiltrating lymphocytes (TILs) in ERBB2-mutated breast cancer (ERBB2-mutated BC) using a bioinformatic approach and publicly available datasets (i.e., TCGA-BRCA and TIMER2.0). ERBB2-mutated BCs were associated with a high histological grade. ERBB2-mutated BCs comprised invasive breast carcinoma of no special type (21/35, 60%), classic invasive lobular carcinoma (12/35, 34.3%), and pleomorphic invasive lobular carcinoma (2/35, 5.7%). A Kaplan-Meier survival curve demonstrated that ERBB2-mutated BC was associated with a significantly worse prognosis compared to ERBB2 non-mutated BC (p < 0.01). Furthermore, 40% (14/35) of the patients with ERBB2-mutated BC harbored CDH1 mutations. Mutations at L755 and V777 accounted for 30.5% of these mutations in ERBB2-mutated BC, suggesting that these sites are mutational hot spots in BC, particularly in invasive lobular carcinoma. Of the ERBB2-mutated BCs, 8.6% were classified as TIL-high, whereas 77.1% were TILs-low; TMB significantly correlated with TILs (p < 0.05). CD8+ T cell infiltration levels were significantly higher in ERBB2 non-mutated BC. Among ERBB2-mutated BCs, 22.9% were classified as TMB-high, which was significantly higher than the rate in the ERBB2 non-mutated BC (p < 0.01). These findings provide evidence for a link between ERBB2 mutations and high TMB in BC.


INTRODUCTION
It has been recently reported that in rare cases, somatic mutations in ERBB2 can activate ERBB2 signaling [1]. In vitro analyses have demonstrated that some ERBB2 mutations are oncogenic and promote cancer cell growth, invasion, and survival [1,2]. A systematic review revealed that the frequency of ERBB2 mutations in breast cancer (BC) is relatively low (2.7%) [3]. ERBB2 mutations are more common in invasive lobular carcinoma (ILC) than in invasive breast carcinoma of no special type (IBC-NST) [4]. ILC is classified into two subtypes, classic ILC (c-ILC) and pleomorphic ILC (p-ILC). CDH1 is located on chromosome 16q22.1 and encodes E-cadherin [5], a transmembrane glycoprotein expressed in epithelial tissues and contributes to calcium-dependent cell-to-cell adhesion [6]. Patients with relapsed CDH1-mutated ILC exhibit a higher frequency of ERBB2 somatic mutations than those with non-CDH1-mutated BC [7]. ERBB2 mutations indicate a significantly poor prognosis, regardless of the histological type [8,9] as they are associated with relapse and bone metastasis [7,10]. Previous studies have demonstrated that the ERBB2 mutation is frequently associated with ILC [7,[10][11][12][13][14]. The CDH1 mutation is a major gene mutation in ILC, but only a few studies have focused on histopathological images and mutations of ERBB2 and CDH1 [12,14]. Moreover, the number of ERBB2-mutated BC was relatively fewer, 5 cases and 18 cases, respectively. Therefore, the relationship between the histological type and genetic alteration of ERBB2 and CDH1 has not been completely elucidated.
In this study, we investigated the clinicopathological characteristics and the frequency of cooccurrence of the CDH1 mutation and tumour-infiltrating lymphocytes (TILs) in 35 ERBB2-mutated BCs via a public database (TCGA-BRCA). Furthermore, we compared the tumor mutational burden (TMB) as well as CD8 + T cell, CD4 + T cell, and Treg immune infiltration levels between ERBB2-mutated BC and ERBB2 non-mutated BC via in silico analysis.

Data Collection
Data on ERBB2 and CDH1 somatic mutations were obtained from UCSC Xena (http://xena.ucsc.edu/). Genomic Data Commons (GDC) TCGA Breast Cancer (BRCA), comprising 986 BC samples, was used to obtain mutation data. Clinicopathological information (age of onset, ethnicity, sex, histology, and subtype) and genomic information from TCGA-BRCA were obtained using the GDC Data Portal (https://portal.gdc.cancer.gov) (Supplementary Table S1). The ERBB2 mutation type, amino acid change, and mutation site were identified. The ERBB2 mutation site was classified as follows: Receptor-L domain, Furin-like cysteine-rich domain, kinase domain, interdomain region, or C-terminal region. The pathological significance of each ERBB2 or CDH1 mutation was assessed using COSMIC FATHMM, Ensembl Variant Effect Predictor (VEP), SHIFT, and PolyPhen (Supplementary

Clinicopathological Features
Clinicopathological features, including age, sex, ethnicity, and intrinsic subtype data were obtained from the GDC Data Portal. Intrinsic subtypes, as determined using multi-gene assay PAM 50, were classified into the following five types: Luminal A (LumA), Luminal B (LumB), Her-2, Basal, and normal. However, there were cases in which subtypes were not assigned and others with two subtypes; these were excluded. Finally, there were 32 cases in ERBB2-mutated BC and 859 cases in ERBB2 non-mutated BC. The prognosis of ERBB2 mutated BC was compared with that of ERBB2 non-mutated BC and analyzed on the Kaplan-Meier survival curve using UCSC Xena (http://xena.ucsc.edu/).

Pathological Review
Samples with ERBB2 mutations (n 35) were reassessed. Whole slide images of diagnostic sections in ERBB2-mutated cases (n 35) were downloaded from the GDC Data Portal and analyzed using Aperio ImageScope (Sausalito, CA, United States). The histological types were classified according to the 5th edition of the World Health Organization classification system [15]. The histological grade was evaluated in accordance with the Elston and Ellis modification of the Nottingham grading system [16]. All samples were reviewed by two pathologists (S.U. and T.S.). Interobserver differences were resolved through re-evaluation and discussion to reach consensus. Samples were defined as c-ILC when all tumor cells were small and uniform with round nuclei and inconspicuous nucleoli ( Figure 1B). p-ILC was defined based on previously reported cytological characteristics (i.e., a greater degree of cellular pleomorphism, nuclear membrane irregularities, more prominent nucleoli, increased hyperchromasia, and more frequent mitotic activity) ( Figure 1C) [17].

Assessment of TILs
Stromal TILs were quantified on each virtual slide. TILs were assessed in accordance with the guidelines proposed by the international TILs working group [18]. Stromal TILs counts were estimated as the percentage of immune cells in stromal tissues within the tumor. TILs were categorized into three: low (0-10%), intermediate (11-59%), and high (60-100%).

TMB Estimation
TMB is a measure of the total number of mutations per megabase of tumor tissue. It can also be interpreted as the mutation density in tumor genes, defined as the average number of mutations in the tumor genome, including the total number of coding sequence errors, base substitutions, insertion, or deletions [20]. TMB was estimated for TCGA-BRCA (n 951) and ERBB2-mutated BC (n 35) and was calculated as the total number of mutations per sample/38, with an estimated exome size of 38 Mb [21]. The samples were classified as TMB-high if they had ≥10 mutations per megabase (mut/Mb) as previously described [22].

Statistical Analysis
The subtype, CDH1 mutation, and TMB between ERBB2 mutated BC and ERBB2 non-mutated BC were analyzed by Fisher exact tests, with a significance threshold of p < 0.05. The correlations for each TMB and TILs were evaluated using the Spearman rank correlation coefficient. Results with p < 0.05 were considered statistically significant.
Comparison Between the Level of Immune Cell Infiltration in ERBB2-Mutated and Non-mutated BC CD8 + T cell infiltration ( Figure 4A) was significantly higher in ERBB2-mutated BC than in their non-mutant counterparts (p < 0.05). In contrast, no differences in CD4 + T cell and Treg infiltration were observed between ERBB2 mutated and nonmutated BC ( Figures 4B,C), as determined by TIMER and CIBERSORT analyses using TIMER2.0.

DISCUSSION
We characterized the clinicopathological features, immune infiltration levels, TMB (mut/Mb), and TILs (%) associated with ERBB2-mutated BC using bioinformatics. ERBB2mutated BC was identified in 3.5% of TCGA-BRCA samples, concurrent with a previous report [3]. In this study, among the 35 samples, 60% were IBC-NST and 40% were ILC (c-ILC, p-ILC). As ILC typically accounts for 5-15% of all BCs [15], ILC appears to be enriched in ERBB2 mutations. A similar trend has been reported previously [4,7]. In ERBB2-mutated BC, ILC is frequently related to a high frequency of CDH1 mutations. In this study, 40% (14/35) of the ERBB2 mutant cases harbored CDH1 mutation.   In our study, missense mutations in the kinase domain were the most common ERBB2 mutation type in BC. The mutation type and site differ among tumor types. For example, ERBB2 mutations in micropapillary urothelial carcinoma of the urinary bladder are predominantly present in the extracellular domain [23].
In this study, L755 mutations and V777L accounted for 34.4% of all ERBB2 mutations in BC. Remarkably, more than half of c-ILC and p-ILC samples revealed the presence of amino acid change at L755 and V777L. A correlation has been reported between p-ILC and amino acid change at L755 [14]; however, in this study, a similar correlation was also observed for c-ILC. Together with previous reports [14,24], our study shows that kinase domain (predominantly L755 site and V777L substation, especially in lobular carcinoma) is a hot spot for ERBB2 mutation in BC. A previous study integrated three cohorts (METABRIC, TCGA, MSK-IMPACT) and calculated the frequency of the ERBB2 mutation (2.2%, 34/ 1,580). The ERBB2 mutation was enriched in ILC with a prevalence of 5.7% (n 16) vs. 1.4% in IDC (n 18) [25]. Based on these findings, it is thought that an analysis of other data sets (METABRIC and MSK-IMPACT) will yield the same results.
Previous studies have reported that irreversible tyrosine kinase inhibitors (TKIs, such as neratinib and afatinib) are useful for treating ERBB2-mutated BC [1,26,27]. Ongoing clinical trials are investigating the effect of neratinib (SUMMIT trial; NCT01953926) and afatinib (NCI-MATCH; NCT02465060) in patients with ERBB2-mutant cancers. However, neratinibresistant BC has been previously reported [26]. Acquired tolerance to TKIs occurs via multiple mechanisms, including gatekeeper mutations and "bypass" resistance [28]. The acquisition of therapeutic resistance in cancer cells is a major challenge of molecular targeted therapy. Therefore, we evaluated TMB and TILs to investigate the possibility of immunotherapy as an alternative treatment, including the use of TKIs, such as neratinib and afatinib.
TMB is a predictive biomarker for the response to immune checkpoint inhibitor (ICI) therapy, and some clinical studies have reported a response to immunotherapy based on a high TMB [29,30]. In fact, in June 2020, the US-FDA approved pembrolizumab for treating solid tumors in adults and children with unresectable or metastatic high TMB (≥10 mut/Mb) solid tumors based on the results of the KEYNOTE-158 trial [31]. In the present study, we found that the TMB is significantly higher in ERBB2-mutated BC than in ERBB2 non-mutated BC samples without ERBB2 mutations (p < 0.01). In BC, the TMB is usually lower (2.6 mut/Mb) than in other carcinomas, such as lung cancer [21,22]. In a previous study based on 3,969 primary and metastatic BC samples, approximately 5% of the samples were classified as TMB-high [22]. However, in this study, 22.9% of the ERBB2-mutated BC samples were TMB-high, and ERBB2-mutated BC was regarded as a subset of TMB-high BC. TILs are a favorable prognostic factor and are associated with the response to PD-1/PD-L1 inhibitors in BC [32]. The samples used in our study exhibited the following    [33]. In another study, TILs were lacking in 16% of the samples [34]. In our study, ERBB2-mutated BC was characterized by low TILs count. TILs are reportedly higher in triple-negative and Her2-positive BC subtypes than in the luminal subtype. Moreover, ILCs are TILs-low compared with other histological types [33]. The TILs pattern in ERBB2mutated BC resembled the pattern observed in hormone receptor-positive/Her2-negative BC and ILC [34]. Although ERBB2-mutated BCs are classified as TMB-high and TILs-low, a positive correlation was observed between TMB and TILs in this study. TIMER2.0 analysis revealed that the CD8 + T infiltration level of ERBB2-mutated BC was significantly higher than that of ERBB2 non-mutated BC. In ERBB2mutated BC, the CD8 + T cell infiltration level was considered to be up-regulated, and subpopulations of TILs may contain much CD8 + T cells. However, further investigations are needed to confirm this.
This study demonstrated that 22.9% of the ERBB2-mutated BCs were TMB-high BC; in contrast, 77.1% of them were TILs-low. From this result, it is questionable whether ERBB2mutated cases would really benefit from ICI therapy. Further research is needed to determine whether it is an indication for ICI therapy. Since TMB estimation are highly variable [35], with slower turn-around times and high costs, further studies are required to validate the correlations with between other biomarkers, such as PD-L1 immunostaining and microsatellite instability (MSI), and TMB. A recent study reported that a machine learning algorithm, Image2TMB, can predict the TMB directly from images of H&E-stained histopathological sections [36]; however, this method is still not used in the clinical setting, although it may be a novel method of estimating the TMB easily.
This study has some limitations. This result was obtained from one dataset (TCGA-BRCA), and it is necessary to verify the obtained result by laboratory research in the future.
In conclusion, The TMB and CD8 + T cell infiltration level in ERBB2-mutated BC samples was significantly higher than that in ERBB2 non-mutated BC. Additionally, 22.9% of the ERBB2-mutated BC samples were identified as TMB-high, and a positive correlation was identified between TMB and TILs.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
SH and TS designed the experiments and performed the pathological reassessment. TK performed the statistical analyses, SU and TK analyzed the data. All authors were involved in writing the paper and approved of the submitted manuscript.