Discordance of PIK3CA and TP53 mutations between breast cancer brain metastases and matched primary tumors

There is limited knowledge of the biology of breast cancer (BC) brain metastasis (BM). We primarily aimed to determine the mutations in BCBM and to compare the mutational pattern with the matched primary breast cancer (BC). Secondary aims were to determine mutations in each subgroup (Luminal A-/B-like, HER2+ and TNBC) of BCBM, and to determine survival according to specific mutations. We investigated 57 BCBMs, including 46 cases with matched primary tumors (PT) by targeted Next Generation Sequencing (NGS) using the Cancer Hotspot Panel v2 (ThermoFisher Scientific) covering 207 targeted regions in 50 cancer related genes. Subtype according to immunohistochemistry was re-evaluated. NGS results fulfilling sequencing quality criteria were obtained from 52 BM and 41 PT, out of which 37 were matched pairs. Pathogenic mutations were detected in 66% of PTs (27/41), and 62% of BMs (32/52). TP53 mutations were most frequent; 49% (20/41) of PTs and 48% (25/52) in BMs, followed by PIK3CA mutations; 22% (9/42) in PTs and 25% (13/52) in BMs. Mutations in CDH1, EGFR, HRAS, RB1 CDKN2A and PTEN were detected in single pairs or single samples. Mutational pattern was discordant in 24% of matched pairs. We show a discordance of PIK3CA and TP53 mutations of roughly 25% indicating the need to develop methods to assess mutational status in brain metastasis where analysis of cell-free DNA from cerebrospinal fluid (CSF) has shown promising results.


Scientific Reports
| (2021) 11:23548 | https://doi.org/10.1038/s41598-021-02903-x www.nature.com/scientificreports/ CDK4/6, PARP and PIK3CA is underway. Hence, the need for analysis of BC tissue in the metastatic setting is apparent. BM have become increasingly common in BC 6,11 , consequently, there is need for adequate information regarding the evolution of mutations, and histology in the metastatic process for correct therapy decisions [12][13][14] . There are studies investigating gene expression in metastatic lesions 15 . However, due to difficulties to sample tissue, BMs represent very few of investigated metastatic lesions.
We primarily aimed to determine the mutational status of key driver genes in BCBMs and to compare the results with mutations in matched primary BC using next generation sequencing (NGS). Secondly, we aimed to investigate the mutations in each BC subgroup (Luminal A-/B-like, HER2+ and TNBC) and to determine survival according to specific mutations.

Materials and methods
Patients. From hospital records of all diagnostic codes, patients with BMs from BC between 1994 and 2014 were identified. Patients received treatment of the primary BC four hospitals in the western region of Sweden. The diagnoses were confirmed in the patient's charts and patients with available FFPE tissue from BMs were selected. Clinical characteristics, type of metastasis, progression, and survival were extracted from patient charts. PTs and BM tissue were evaluated, as specified below, when sufficient material was available. Data from the original report was utilized if deemed appropriated by the responsible breast pathologist if there was insufficient material available for re-evaluation by IHC. The study was conducted in accordance with the Declaration of Helsinki and the Sahlgrenska University Hospital Ethical Review Board; Gothenburg, Sweden approved the study (460-09, T592-14). Approval for the chart review and bio bank extractions was granted from each head of the participating departments.
Immunohistochemistry. The available material was re-evaluated for histological type, nuclear grade, and receptor status. ER/PR, Ki67 and HER2 (HercepTest) IHC status was found as per standard procedures using the Dako Autostainer Link and the EnVision™ FLEX detection systems according to the manufacturer's instructions. HercepTest was followed by SISH when the IHC was judged as 2+ or 3+. IHC was used to construct the subgroups. Luminal A-like was defined as ER positive, PR present and/ or low Ki67 (< 14%). Luminal B-like was defined as ER positive, PR absent and/or high Ki67 (> 15%). HER2+/ER+ was defined as amplified HER2 and ER positive. HER2+/ER− was defined as HER2 amplified and ER negative. TNBC was defined as lacking HER2 amplification, ER positivity and PR positivity.

Next generation sequencing. Preparation of sample library and next generation sequencing (NGS). DNA
isolation from FFPE sections containing a minimum of 25% neoplastic cells as assessed by breast pathologists was performed using the QIAamp DNA FFPE tissue kit (Qiagen Gmbh, Hilden, Germany). DNA concentration was determined using NanoDrop™ 3300 with Quant-IT Picogreen dsDNA assay kit (ThermoFisher Scientific, Waltham, MA, USA). Ten ng of DNA was used to prepare barcoded libraries with the Ion AmpliSeq™ Library kit 2.0 (ThermoFisher Scientific, Waltham, MA, USA). The Cancer Hotspot Panel v2 (ThermoFisher Scientific, Waltham, MA, USA) covering 207 targeted regions in 50 cancer related genes was used (https:// tools. lifet echno logies. com/ conte nt/ sfs/ broch ures/ Ion-Ampli Seq-Cancer-Hotsp ot-Panel-Flyer. pdf). Template preparation and enrichment was performed with the IonChef™ Instrument (ThermoFisher Scientific, Waltham, MA, USA). Eight barcoded samples were pooled per Ion 318™ v2 BC chip and sequenced on the Ion PGM™ System (ThermoFisher Scientific, Waltham, MA, USA). All steps were performed according to the manufacturer's instruction. Matched germline DNA from the patients was not available in this retrospective study.
Data processing. After alignment to the hg19 human reference genome, variant calling by the Torrent Suite Software™ v.4.2.1.0 (ThermoFisher Scientific, Waltham, MA, USA) with subsequent manual filtering of known SNPs. Variants were visually inspected with the Integrative Genomics Viewer (IGV; Broad Institute, Cambridge, MA, USA). Mutations were manually curated as pathogenic, likely pathogenic, variants of unknown significance or as benign. For ten samples, NGS results could not be obtained due to insufficient quality of DNA. For individual variants the minimum accepted read depth was 500 and if the read depth was < 1000, an allele frequency of > 5% was required for positive variant calling. Genetic profile of BM and PT. Twenty-seven PTs (66%) and 32 BMs (62%) exhibited at least one pathogenic mutation (Fig. 3). A pathogenic JAK2 p.V617F mutation in a patient with previously diagnosed myelopro-

Mutational findings in matched pairs of BM and PT.
In the matched pairs, similarly, the most prevalent mutation was in TP53 in 43% (16/37). Mutated TP53 was found in three Luminal A-and B-like PTs and four paired BMs (27%/36%), eight TNBC PTs and ten paired BMs (47%/59%), and six HER2 positive PTs and five paired BMs (67%/55%). Mutations in PIK3CA was the second most prevalent mutation in this material. In matched PT/BM pairs, PIK3CA mutations were found in two PTs and four BMs of Luminal B-like (18%/36%); three PTs and BMs in TNBC (18%/18%), and in one matched case of HER2 positive tumors (11%) (Fig. 3, Table 3). The numerical differences of TP53 and PIK3CA mutations in BC subgroups did not reach statistical significance (p = 0.228) (p = 0.552) respectively.
Mutational concordance in matched material. The mutational profile of the genes present in the targeted panel was discordant between the PTs and BMs in nine cases out of 37 (24%). The mutational differences displayed no discernible pattern. There were three cases of mutations present only in the PT, four with only mutations found in the BM, and one case of mutation in the same gene, but at another position BM. Of note, in two PT/BM pairs, a pathogenic PIK3CA mutation and five TP53 mutations were detected only in the BM. (Table 4) with a median of 28 months (6-64) and 34 months (7-62) respectively, whilst patients with HER2+/ER− and TNBC had BMSS of 11 months (0-46) and eight months (0-46) respectively (p = 0.0851). We found equal BMSS when patients with mutations in PIK3CA were compared to patients without PIK3CA mutation (p = 0.467). There was no difference in BMSS when patients with a TP53 mutation were compared with patients without documented TP53 mutation (p = 0.456).

Discussion
Concordant with previous published data, we confirm that TP53 and PIK3CA are the most common mutations in BCBM, as well as, the matched PTs, in to our very best knowledge so far one of the largest cohorts of genetically characterized BCBM [16][17][18][19][20][21] . Difficulties in accessing intra-cranial tissue samples have hampered development of systemic treatments that could have an effect in the treatment of patients with BMs. Our cohort, consists of 52 analyzed cases, and 37 cases with matched PT, all from one region in Sweden. The difficulty in obtaining tissue from the brain results in few tumors which limits the study when comparing BC subgroups or in survival analyses. Although the tissue material is partly old, with the first samples collected in 1994, sequencing results by NGS were successfully obtained in 90% of the BMs. At least one mutation, among 50 cancer driver genes, was present in 62% of the analyzed sample, similar to previous publications [16][17][18][19][20][21] . Concordant with these previous studies, we found TP53 and PIK3CA as the most commonly mutated driver genes, both in the BMs and PTs. Of these two mutations, PIK3CA are considered targetable, whilst TP53 mutations can be indirectly attacked through restoration of the transcriptional activity resulting in a functional wild-type TP53 protein 22,23 . The drug alpelisib, is a PI3Kα-specific inhibitor available for patients with recurrent PIK3CA mutated HER2 negative Luminal A-/B-like BC 23 . Previous published data show that approximately 40% of ER positive PT harbor a PIK3CA mutation 24 . Two studies that compared PIK3CA mutations in PTs and BC metastases reported mutations in 33% of PTs and 30% of metastases and 45% of PTs and 53% in metastases respectively 18,25 . There were few or no patients with BMs in the above-mentioned studies. Our results in BCBM had a lower fraction of PIK3CA mutations, with 20% in PTs and 23% in the BMs. A recently published systematic review compared 164 BMs with its matched PT in 126 patients extracted from 13 studies and found PIK3CA mutations in 22% of the patients with BMs which in line with our results 26 . It might be hypothesized that PIK3CA mutations is lower in BMs compared to other metastatic sites, however, additional studies are required to adequately adress this question. As expected, we found the highest proportion of PIK3CA mutations in Luminal A-and B-like BC. Of note, in our material two out of eight pathogenic PIK3CA mutations and five TP53 mutations were found in the BMs only, and not in the PT underlining the need for re-evaluation of metastatic tissue or possibly an added value by analysis of cell free DNA (cfDNA) 27 .
We found PIK3CA mutations as the second most common mutation after TP53 in matched pairs of TNBC with 18% PIK3CA mutations all concordant in PTs and BMs. This is in line with previous published data in which PIK3CA mutations was the second most common mutation after TP53, especially in basal-like and androgen receptor expressing subtypes of TNBC [28][29][30] . The combined treatment of alpelisib and nab-paclitaxel is currently under investigation in pre-treated TNBC with either loss of PTEN expression or a PIK3CA mutation (NCT04251533).
The prognostic role of PIK3CA mutations in HER2+ BC has been extensively investigated. Mutations in PIK3CA have been associated with diminished effect of HER2 blocking therapy, both in the neo-adjuvant setting, and for recurrent BC in terms of lower pCR rates, and shorter survival respectively 31,32 . We found PIK3CA mutations in only 10% of the HER2+ cases, which is lower than the previously reported frequency of 20-40% 24 . This discrepancy may be due to the limited number of patients with HER2+ BC. The effect of alpelisib in HER2+ breast cancer is under investigation in one ongoing and one completed clinical trial (NCT02038010; NCT04208178). Table 3. TP53 and PIK3CA mutations in primary breast cancer tumors and brain metastases with NGS results fulfilling sequencing quality criteria. The bottom row shows the concordance in mutation status in the 37 matched brain metastasis and primary breast tumors.

All (n = 57) Luminal B-like (n = 14) TNBC (n = 26) HER2 + (n = 17)
Mean (SD) brain metastasis specific survival (months) 17   www.nature.com/scientificreports/ Almost 50% of the PTs, out of which many exhibited TNBC subtype, in the present study population harbored a TP53 mutation, in contrast to the 20-35% mutation prevalence reported in unselected primary BC 33 . The Cancer Genome Atlas reveals an enrichment of TP53 mutations in basal-like and HER2 enriched BC 24 . Interestingly, previous data reveal that a high proportion of patients with a TP53 mutation in the primary BC developed BM 34,35 . We lack data on the molecular subtypes in our material, but find an increased number of TP53 mutations in TNBC and HER2/ER-BC. Enrichment of TP53 mutations in BMs was not seen in the HER2 amplified subgroup, a finding that must be interpreted with caution due to the very small sample size of the HER2 group. Other detected mutations in our series, CDH1, EGFR, HRAS, RB1, CDKN2A and PTEN were rare, in general found in single or both samples from one patient.
Roughly 25% of the matched pairs changed IHC based BC subtype in the BMs. The most common change was from Luminal A-like in PT to other subtypes, in most cases Luminal B-like in BMs. This is in accordance with previous results in which 219 patients showed a 36% overall discordance with the most common change in form of loss of PgR 36 . In a review pooling a total of 3384 matched pairs of BC and metastases from all organs, BM showed a discordant median rate of 22% compared to 45% in liver metastases and 16% in lymph node metastasis 10 . TNBC was the most stable group with less than 10% showing a gain in ER (one case) and HER2 (one case).
A limitation of our study is that genetic profiling was performed using a relatively small NGS panel of 50 genes. Although the panel provides broad coverage of important cancer driver genes, more comprehensive sequencing might provide a more granular view of mutation discordancy between PTs and BMs. Still, we believe that the relatively large number of matched PT and BM add to the knowledge about the biology of BMs, and changes in the mutational landscape of actionable genes during the metastatic process.
In conclusion, we confirm mutations in TP53 and PIK3CA to be common in both primary breast tumors and BMs but the proportion varied depending on the subgroup. Mutation pattern, as well as IHC based subtypes were discordant in approximately 25% of the patients underlining that re-biopsy at disease progression may be warranted. In this context, analysis of cfDNA may be a fruitful avenue in patients with BMs, placing hope in promising results with liquid biopsy from CSF 27,37,38 .

Data availability
All pathological and uncertain mutations are added as a Supplemental Table. Clinical data is not included as this would allow for possibly identifying individuals.