Genetic mutation profile of Chinese HER2-positive breast cancers and genetic predictors of responses to Neoadjuvant anti-HER2 therapy

Purpose Despite the therapeutic success of existing HER2-targeted therapies, tumors respond quite differently to them. This study aimed at figuring out genetic mutation profile of Chinese HER2-positive patients and investigating predictive factors of neoadjuvant anti-HER2 responses. Methods We employed two cohorts. The first cohort was comprised of 181 HER2-positive patients treated at Guangdong Provincial People’s Hospital from 2012 to 2018. The second cohort included 40 patients from the first cohort who underwent HER2-targeted neoadjuvant chemotherapy. Genetic mutations were characterized using next-generation sequencing. We employed the most commonly used definition of pathological complete response (pCR)-eradication of tumor from both breast and lymph nodes (ypT0/is ypN0). Results In Chinese HER2-positive breast cancer patients, TP53 (74.6%), CDK12 (64.6%) and PIK3CA (46.4%) have the highest mutation frequencies. In cohort 2, significant differences were found between pCR and non-pCR groups in terms of the initial Ki67 status, TP53 missense mutations, TP53 LOF mutations, PIK3CA mutations and ROS1 mutations (p = 0.028, 0.019, 0.005, 0.013, 0.049, respectively). Furthermore, TP53 LOF mutations and initial Ki67 status (OR 7.086, 95% CI 1.366–36.749, p = 0.020 and OR 6.007, 95% CI 1.120–32.210, p = 0.036, respectively) were found to be predictive of pCR status. Conclusion TP53 LOF mutations and initial Ki67 status in HER2-positive breast cancer are predictive of pCR status after HER2-targeted NACT.


Introduction
HER2 positivity accounts for about 15-20% of breast cancers and the development of HER2-targeted therapies has profoundly changed the course of these patients [1]. More and more HER2-targeted drugs, such as trastuzumab, pertuzumab, T-DM1 and neratinib, have become available for treatment of HER2-positive breast cancer. Despite this progress, however, many patients still die of HER2-positive breast cancer, calling for the identification and investigation of genetic profiles of HER2-positive breast cancer and predictors of responses to HER2-targeted therapies.
Neoadjuvant chemotherapy (NACT) is used commonly to downstage locally advanced cancer to allow breast-conserving surgery and to predict responses to systemic therapy based on pathological assessment. Based on the recent progress, NACT has gained momentum as an ideal setting in which to investigate predictive biomarkers of treatment Kai Li, Ning Liao, Bo Chen and Guochun Zhang have contributed equally to this work. 1 3 responses. The Collaborative Trials in Neoadjuvant Breast Cancer (CTNeoBC) pooled analysis has confirmed NACT to be a good way to discriminate patients who have different clinical outcomes by their responses to the therapies [2]. Patients who attain complete eradication of tumor after NACT in both breast and lymph nodes have improved survival, which is defined as pathological complete response (pCR) [2,3]. The NeoSphere and NeoALTTO trials both tried to seek higher pCR rates by different anti-HER2 combinations [4,5] and new treatments in the past decade have significantly improved the prognosis of HER2-positive breast cancer with a pCR rate as high as 75% [6]. Despite these achievements, however, HER2-positive breast cancer patients still have a high death rate [6]. According to the latest interim analysis of the landmark KATHERINE trial, patients with HER2-positive early breast cancer who had residual invasive disease after NACT have improved survival when they receive adjuvant T-DM1 therapy after surgery [7]. As these great trials have provided effective regimens for patients with different responses, it is important to investigate the response predictors.
Large-scale, next-generation sequencing studies have provided large amounts of genetic information and produced valuable insights into the genomic landscape of primary breast cancers [8][9][10][11][12]. These studies have highlighted that TP53 and PIK3CA were the two most prevalent mutated genes in HER2-positive breast cancers [8] and enriched in residual tissues after HER2-targeted therapies. PIK3CA mutation rate in HER2-positive breast cancers is about 23%, and patients with PIK3CA mutations have a lower pCR rate after HER2-targeted NACT [13].
In this study, we investigated the mutation profiles of HER2-positive breast cancer patients in China and analyzed the mutation differences between primary HER2-positive breast cancers with pCR and non-pCR after HER2-targeted NACT. We hypothesized that there would be mutations predictive of the anti-HER2 therapy responses and could be used as biomarkers for guiding treatment decisions.

Study cohorts
This study was comprised of two cohorts. The first cohort included 181 HER2-positive patients treated at GPPH from 2012 to 2018. The second cohort included 40 patients from the first cohort who underwent HER2-targeted neoadjuvant therapy (NACT). Twenty-three patients received NACT with docetaxel 75 mg/m 2 , carboplatin (6 mg/min/ml carboplatin AUC area under curve) and trastuzumab (8 mg per kilogram intravenously as a loading dose, followed by 6 mg per kilogram intravenously every 3 weeks), while oral lapatinib was added in other 17 patients. Lapatinib was given daily at a dose of 750 mg (250 mg tablets) for the first week, followed by 1000 mg daily for a year. All 40 patients completed the scheduled 6 NACT cycles.
Detailed information of our study cohorts is listed in Table 1.
Pathological examination of tumor specimens was performed in the Department of Pathology at GPPH. ER, PR, ROS1 and HER2 status were reconfirmed by two experienced pathologists based on IHC and fluorescence in situ hybridization (FISH) [14]. The cutoff for ER-negative and PR-negative IHC status was less than 1% staining in the nuclei. HER2 status was considered negative when an IHC score was 0 or 1 or when HER2 amplification was absent (ratio < 2.2) by FISH analysis. If any disagreements arose during the evaluation of the IHC and FISH results, a third pathologist was consulted.

NGS library preparation
DNA fragmentation was performed using Covaris M220, followed by end repair, phosphorylation and adaptor ligation. Fragments of size 200-400 bp were selected by bead (Agencourt AMPure XP Kit, Beckman Coulter, California, USA) followed by hybridization with capture probes baits, hybrid selection with magnetic beads and PCR amplification. Subsequently, a high-sensitivity DNA assay was performed to assess the quality and size of the fragments. Indexed samples were sequenced on Nextseq500 sequencer (Illumina, Inc., California, USA) with pair-end reads.

Capture-based targeted DNA sequencing
Genomic profiling was performed using a panel covering 520 cancer-related genes (Burning Rock Biotech Ltd.). Among them, whole exons of 312 genes and critical exons, introns and promoter regions of the remaining 208 genes were captured.

Sequence data analysis
Sequence data were mapped to the human genome (hg19) using BWA aligner 0.7.10. Local alignment optimization, variant calling and annotation were performed using GATK 3.2, MuTect and VarScan. Variants were filtered using the VarScan filter pipeline, with loci with depth less than 100 filtered out. At least 5 supporting reads were needed for INDELs, while 8 supporting reads were needed for SNVs to be called. According to the ExAC, 1000 Genomes, dbSNP and ESP6500SI-V2 database, variants with population frequency over 0.1% were grouped as SNP and excluded from further analysis. Remaining variants were annotated with ANNOVAR and SnpEff v3.6. DNA translocation analysis was performed using both Tophat2 and Factera 1.4.3.

Statistical analyses
Statistical analyses were performed using GraphPad Prism version 7.00 for Mac (GraphPad Software, La Jolla California, USA). Pearson's Chi-square test and Yate's continuitycorrected Chi-square test were employed for significance of differences between groups. A two-sided p value less than 0.05 was considered significant unless otherwise stated. To determine which covariates affected pCR, we used pathological and mutational variables by univariate and multivariate regression. We included variables which are < 0.05 for p value and those supposed to affect pCR status.

Clinicopathologic features and genetic mutations of Chinese HER2-positive breast cancer patients
The patients' clinicopathological parameters are listed in Table 1. In cohort 1, the median age was 48, and 56.9% of patients were pre-menopausal. The population was comprised of stage I, II and III patients, which account for 20.4%, 59.7% and 19.9%, respectively. The cohort 2 was derived from cohort 1, in which the median age was 53, more II, III stage patients were included and there was no difference in hormone receptor expression. No difference of pCR status was found between the two treatment groups. However, the two groups of Ki67 < 40% versus ≥ 40% had different pCR rates of 63.2% and 28.6% (p < 0.028) ( Table 1). We next analyzed the mutation profiles in Chinese HER2-positive breast cancers ( Fig. 1 and Table 2). The most commonly mutated genes in Chinese HER2-positive breast cancer patients were TP53 (74.6%), CDK12 (64.6%) and PIK3CA (46.4%). Most of the TP53 mutations were missense mutations (40.35%) and LOF mutations (21.5%). Interestingly, the ROS1 mutations were only detected in HR + HER2 + patients. Mutation sites of TP53, PIK3CA and ROS1 are listed in Fig. 1. No TP53 mutation hotspots were found in these Chinese HER2 + patients, while p.H1047R was investigated to be hotspot in PIK3CA mutations ( Fig. 1 and Table 3). In cohort 2 ( Table 2 and Fig. 2), TP53 (90.0%), CDK12 (77.5%) and PIK3CA (55%) were still the most mutated genes. And 5 patients got ROS1 mutations, all of whom belong to HR + HER2 + subgroup. The mutation sites of TP53, PIK3CA and ROS1 in cohort 2 are also listed in
Then we detected the differences between the pCR and non-pCR groups according to pathological and mutational variables and found significant differences in terms of the initial Ki67 status, TP53 missense mutations, TP53 LOF mutations, PIK3CA mutations and ROS1 mutations One PIK3CA mutation hotspot was examined to be p.H1047R which belongs to be a missense mutation. Three mutation sites (p.V650M, p.E545G and p.E542K) were detected to be located in the PI3Ka subunit which has been found to be crucial for PIK3CA function.  There was a significant difference of PIK3CA mutation frequency between pCR and non-pCR group (33.3% vs 72.7%; p = 0.013, Table 2 and Fig. 3).
ROS1 amplification was only investigated in 5 hormone receptor-positive patients who all got pCR. The ROS1 amplification breast cancers were found to have higher pCR rate (p = 0.049) ( Table 2 and Fig. 2). However, when we go further to investigate whether ROS1 had amplification in protein level by immunohistochemistry, the result was negative (Figs. 3, 4).

Predictors for pCR or non-pCR
As shown above, some mutations and pathological factors seemed to affect pCR status and, to determine which of those are the predictive factors, we did univariate and multivariate regression. For pCR status, three covariates were significant in the univariate analyses (Tumors size, TP53 missense mutations, TP53 LOF mutations), but only one was retained and a new one appeared after forward selection in the multivariate analysis-TP53 LOF mutations and initial Ki67 status (OR 7.086, 95% CI 1.366-36.749, p = 0.020 and OR 6.007, 95% CI 1.120-32.210, p = 0.036, respectively, in Table 4). Some factors selected in the univariate model seemed to be highly correlated and for this reason were rejected or appeared in the multivariate model.

Discussion
Our findings indicate significant associations (p = 0.020 and p = 0.036) of TP53 LOF mutations and lower initial Ki67 status (< 40%) with a high probability of pCR in HER2positive breast cancer patients receiving NACT.
TP53 has been reported to have heterogenous types of mutations which include attenuation of function, separation of function or neomorphic function [15][16][17]. There are gainof-function mutations mostly TP53 missense mutations and LOF mutations which have a common characteristic of losing functions of wild-type TP53 and composed of splice site, frame shift and nonsense mutations [16,18].
HER2-positive breast cancer has a high frequency of TP53 mutations (up to 72%) [8], and in our previous study, the mutation frequency of TP53 in the triple-positive HR + /HER2 + group is 66.1% and up to 89.3% in the HR-/HER2 + group in Chinese breast cancer patients [19].
Interestingly, the TP53 mutation rate was similar in those of HER2-positive patients who achieved a pCR or not (88.9% v 81.8%), and there was no statistical difference of TP53 mutation rate between two groups (p = 0.859). However, as far as specific types of TP53 mutations were considered, such as missense mutations, LOF mutations and so on, quite significant mutational differences were manifested. TP53 missense mutation rates in pCR and non-pCR groups were quite different (22.2% vs 59.1%, p = 0.019, Table 2 and Fig. 3). Similarly, patients in pCR group were detected  And why is that? How TP53 LOF mutation mediates a good response to NACT is complicated. TP53 has been reported to be activated in response to mitotic stress caused by agents such as taxanes that disrupt microtubules [20]. Evidences from breast cancer models and tumors have shown p53 directs cells to undergo cell cycle arrest and senescence [21,22]. One mechanism that contributes to senescent cell survival and persistence in the residual disease after chemotherapy treatment is the engulfing and cannibalizing of neighboring cells [23]. We believe that TP53 LOF mutations lose capacity to activate cell cycle arrest and senescence to escape apoptosis induced by therapy, and as a result mediate a good response to NACT.
Some TP53 missense mutations are associated with enhanced characteristics of invasion and metastasis when they acquire a gain-of-function effect [24]. More than 80% of TP53 alterations are missense mutations that will produce a stable but transcriptionally deficient protein. These mutant-TP53-expressing tumors are aggressive and associated to poor prognosis [25,26]. Compelling evidences have proved that TP53 missense mutations promote cell migration and metastasis and dramatically influence tumor progression [27][28][29]. In our study, TP53 missense mutations showed significant difference between pCR and non-pCR groups, and also a predictor for non-pCR by univariate regression. We speculated that TP53 missense may affect therapy response depending on other variables. TP53 mutations were not predictive of neoadjuvant chemotherapy response in the EORTC 10,994/BIG 1-00 trial [30]. However, in that study, patients with HER2 + breast cancers were randomly assigned to different NACT groups without HER2-targeted therapy, and the yeast assay was used to assess TP53 mutations which does not distinguish between pure loss-of-function mutations compared to mutations with simultaneous gain and loss-of-function [31,32]. Several molecular alterations are thought to contribute to trastuzumab resistance, including TP53 mutation [33,34] and PIK3CA alteration [13,35,36], but results evaluating these biomarkers as response predictors have been inconsistent.
Two retrospective studies [37,38] reported TP53 mutations were significantly predictive of HER2 + patient treatment response (pCR) to neoadjuvant chemotherapies. Soley et al. report that for patient samples with concordant Blue-Print/MammaPrint and PAM50 data, the pCR plus non-pCR rate among patients whose tumors were TP53 mutant was 17/39 (44%), whereas in patients whose tumors were TP53 wild type, it was 5/31 (16%), p = 0.020 [37]. And Stefan et al. report that the response rate among TP53mutated patients was 30%, significantly higher than TP53 wild-type patients (10%; p = 0.0032) [38]. However, both studies used the AmpliChip TP53 assay (Roche Molecular Systems, Pleasanton, CA), a DNA microarray-based resequencing assay designed to detect single-base substitutions and single-base deletions in all coding regions of the TP53 gene, which needs a reference sequence and is unable to detect all the possible mutations like a NGS assay [39,40]. Two other studies found no associations of TP53 mutations with HER2 + NACT treatment response [41,42]. However, all these studies only examined associations of TP53 mutations as a whole with treatment response, despite the fact that TP53 missense mutations and loss-of-function mutations have quite different functions during breast cancer progression [16,24,26]. In our study, we assessed the associations of TP53 missense mutations and LOF mutations separately with treatment response and found quite different predictive characteristics.
From a tumor biological point of view, Ki67 should be viewed as a continuous variable, as it reflects the percentage of proliferating cells in the tumor, which can reach any value between 0 and 100%. The fact of defining our cut points should not be seen as a limitation of the marker but point to a strength of Ki67, as studies have shown a wide range of cut points was significant for various endpoints and subgroups [43]. Therefore in our study, Ki67 is still a predictive biomarker for HER2-positive breast cancer subgroup receiving chemotherapy and HER2-targeted therapy.
Tumor size was analyzed to be a predictor of pCR status by univariate regression and not by multivariate regression. It is easy to understand that tumor size is a factor susceptible to other pathological and mutational factors.
In conclusion, our study reports TP53 LOF mutations and initial Ki67 status predict pCR status for HER2-positive breast cancer patients receiving NACT. As this study is an exploratory retrospective study of small size, further prospective clinical research with large sample is still needed.

Compliance with ethical standards
Conflict of interest The authors have no conflicts of interest to declare.
Ethical approval Primary tumor biopsies were obtained using an Institutional Review Board approved protocol, and this study had been approved by the Ethics Committee of Guangdong Provincial People's Hospital (no. GDREC2014122H). All patients provided written informed consent for translational research.

Informed consent
We thank all the patients and their families for participation. Informed consent was obtained from all individual participants included in the study.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.