A Multiplex PCR-Based Next Generation Sequencing-Panel to Identify Mutations for Targeted Therapy in Breast Cancer Circulating Tumor Cells

: Targeted therapy has become the preferred approach to treat most cancers, including metastatic breast cancer. Using liquid biopsies, which can act as a dynamic diagnostic tool, is an appealing concept to identify e ﬀ ective therapies. identify mutations from circulating we have developed a multiplex PCR-based next generation sequencing-panel. The CTCs were enriched using the CellSearch system and isolated by micromanipulation followed by whole genome ampliﬁcation of their DNA. Afterwards, mutation hotspot regions in the PIK3CA , the ESR1 , the AKT1 , and the ERBB2 genes were ampliﬁed and barcoded. Sequencing was performed on a MiSeq system. The assay was validated with cells from various cell lines displaying the expected mutations. Mutations that provide the basis for potential targeted therapies were detected in 10 out of 13 patients in all analyzed genes. In four patients, mutations in more than one gene were observed—either in the same cell or in di ﬀ erent cells, suggesting the presence of di ﬀ erent tumor cell clones, which might be targeted with combination therapies. This assay is a time and cost e ﬀ ective tool to investigate the most relevant genomic positions indicative for targeted therapies in metastatic breast cancer. It can support therapy decision to improve the treatment of cancer patients.


Introduction
In the last decades, targeted therapy has become the preferred approach to treat most cancers. However, obtaining information to choose a targeted therapy is challenging: biopsies of recurrent or metastatic lesions are invasive and cannot be performed when clinical conditions have worsened or when a tumor is inaccessible [1]. Furthermore, the genomic profile of biopsy tissues provides a picture limited to a single point in space and time, and may, thus, under-represent intratumoral heterogeneity [2]. The predictive utility of tissue biopsies is limited by such factors, worsened still by the continuous evolution of the tumor in response to endogenous and exogenous selective pressures [3]. Deriving information about the primary tumor or metastatic lesions from liquid biopsies, which may act as a dynamic diagnostic tool, is an appealing concept to overcome this challenge [4,5].

Whole-Genome Amplification
DNA of single isolated cells was amplified by WGA with the Ampli1 WGA Kit (Menarini) according to the manufacturer's protocol. Afterwards DNA integrity was determined with the Ampli1 QC Kit (Menarini).

Multiplex PCR-Based Next Generation Sequencing
DNA fragments were amplified and barcoded in two steps using KAPA2G Fast Multiplex Mix (Merck, Darmstadt, Germany). In the first step, fragments coding for mutation hotspot regions were amplified with gene specific primers ( Table 2) containing primer binding sites for universal Multiplicom MID Dx primers (Agilent technologies, Santa Clara, CA, USA) in a multiplex PCR. 1 µL WGA product and in total 0.5 µM gene specific primers were applied (concentrations of all primers are shown in Table S1). The concentration of each primer is shown in Table S1. In the second step, Multiplicom MID Dx primers were used to attach unique barcodes for sequencing to the PCR products. 0.5 µL of Multiplicom MID Dx primers and 0.5 µL of first round PCR product were used. The applied PCR protocols are listed in Table S2.
Final PCR products were purified using Agencourt AMPure XP beads (Beckmann Coulter, Brea, CA, USA) and quantified with the Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA) using the Broad Range Quant-iT dsDNA Assay Kit (Thermo Fisher Scientific).
DNA libraries of 0.0143 pM per sample were sequenced on a MiSeq system (Illumina, San Diego, CA, USA) with 151 bp paired end reads.
As controls, DNA from the reference cell lines mentioned below as well as a negative control was included in every run.

Validation Experiments with Spiked Cells
For validation of the workflow, 200 cells of the cell lines SK-BR-3, T47-D, and MCF7 (ATCC, Manassas, United States; catalog numbers: SK-BR-3: HTB-30, T-47D: HTB133, and MCF7: HTB-22) as well as long term cultured CTCs were spiked into 7.5 mL healthy donor blood. For detachment from the culture flask, an enzyme-free cell dissociation buffer (Thermo Fisher Scientific) was used. The tumor cells were enriched using the CellSearch, isolated using the CellCelector and the genome was amplified with the Ampli1 WGA Kit similarly to CTCs.

Statistical Analysis
Statistical analyses were performed using GraphPad Prism (Graphpad Software, San Diego, CA, USA). p-values < 0.05 were considered statistically significant.

Development of the Assay
We aimed to develop a multiplex PCR-based NGS panel to identify mutations in CTCs relevant for targeted therapies in metastatic breast cancer, considering that the genomic DNA of single CTCs was amplified using an MseI restriction-based method. We included hotspot regions in the PIK3CA, ESR1, AKT1, and ERBB2 genes frequently mutated in breast cancers. In the PIK3CA gene, hotspot regions within the exons 5, 10, and 21 were covered. In the ESR1 gene, the hotspot regions in exons 5, 7, and 8 were analyzed. In the AKT1 gene, we investigated the hotspot regions in exon 3 and in the ERBB2 gene our assay covers the hotspot regions in exons 8, 18, 19, and 20. For each amplicon, a panel of primers was tested and the most specific primer combination was chosen for the multiplex PCR (Table 2, Figure S1). Primer concentrations in the first PCR were titrated resulting in an equal ratio for each fragment of 7.0% ± 1.4% to 10.2% ± 0.8% after sequencing ( Figure S2). In total, we achieved an on target rate of 95.3% ± 0.9% of all reads.
Next, a group of CTCs of different quality determined by Ampli1 quality control was sequenced and the obtained coverage was compared. For cells with highest amplified DNA quality showing 4 bands in the quality control PCR a mean coverage of 10.2 ± 1.3 fragments per cell was observed ( Figure 1). The coverage correlated with the quality determined with the quality control PCR. Cells displaying 3 or 4 bands in the quality control PCR had a significantly increased coverage compared to those with less than three bands (p < 0.0001 determined by two-tailed t-test). Thus, we focused on cells displaying at least 3 bands in the quality control PCR for further analysis to yield reliable results.

Statistical Analysis
Statistical analyses were performed using GraphPad Prism (Graphpad Software, San Diego, CA, USA). p-values < 0.05 were considered statistically significant.

Development of the Assay
We aimed to develop a multiplex PCR-based NGS panel to identify mutations in CTCs relevant for targeted therapies in metastatic breast cancer, considering that the genomic DNA of single CTCs was amplified using an MseI restriction-based method. We included hotspot regions in the PIK3CA, ESR1, AKT1, and ERBB2 genes frequently mutated in breast cancers. In the PIK3CA gene, hotspot regions within the exons 5, 10, and 21 were covered. In the ESR1 gene, the hotspot regions in exons 5, 7, and 8 were analyzed. In the AKT1 gene, we investigated the hotspot regions in exon 3 and in the ERBB2 gene our assay covers the hotspot regions in exons 8, 18, 19, and 20. For each amplicon, a panel of primers was tested and the most specific primer combination was chosen for the multiplex PCR (Table 2, Figure S1). Primer concentrations in the first PCR were titrated resulting in an equal ratio for each fragment of 7.0% ± 1.4% to 10.2% ± 0.8% after sequencing ( Figure S2). In total, we achieved an on target rate of 95.3% ± 0.9% of all reads.
Next, a group of CTCs of different quality determined by Ampli1 quality control was sequenced and the obtained coverage was compared. For cells with highest amplified DNA quality showing 4 bands in the quality control PCR a mean coverage of 10.2 ± 1.3 fragments per cell was observed ( Figure 1). The coverage correlated with the quality determined with the quality control PCR. Cells displaying 3 or 4 bands in the quality control PCR had a significantly increased coverage compared to those with less than three bands (p < 0.0001 determined by two-tailed t-test). Thus, we focused on cells displaying at least 3 bands in the quality control PCR for further analysis to yield reliable results.

Validation of the Assay
To validate the assay, we performed spike in experiments using cells from four cell lines exhibiting different mutations. Those cells were processed applying the same workflow used later for clinical samples. As a negative control, cells from the Her2-enriched breast cancer cell line Sk-Br-3 harboring no mutations in the analyzed regions were analyzed. Cells from the luminal breast cancer cell lines T47-D and MCF7 were expected to show the PIK3CA mutations H1047R and E545K, respectively. Further, long term cultured CTCs derived from the DLA product of a metastatic luminal breast cancer patient harboring the ESR1 mutation E380Q and the AKT1 mutation E17K were analyzed. All investigated cells displayed the expected mutations demonstrating the reliability of our assay (Figure 2). 3 harboring no mutations in the analyzed regions were analyzed. Cells from the luminal breast cancer cell lines T47-D and MCF7 were expected to show the PIK3CA mutations H1047R and E545K, respectively. Further, long term cultured CTCs derived from the DLA product of a metastatic luminal breast cancer patient harboring the ESR1 mutation E380Q and the AKT1 mutation E17K were analyzed. All investigated cells displayed the expected mutations demonstrating the reliability of our assay ( Figure 2).

Analysis of CTCs from Metastatic Breast Cancer Patients to Identify Targeted Therapies
After validation with cell line cells we analyzed CTCs from 13 metastatic breast cancer patients. 11 of those patients had a primary tumor of luminal subtype and two patients were diagnosed with triple negative breast cancer. CellSearch analyses determined 6 to 15,340 CTCs per 7.5 mL blood. The median CTC count was 94 per 7.5 mL blood. For our study, 1 to 15 CTCs per patient were analyzed.
In seven patients, mutations in the PIK3CA gene were identified. Five of those mutations are well-described mutations located in the nucleotide triplets coding for the amino acid N345, E545, and H1047, respectively. ESR1 mutations were detected in CTCs from two patients. One of those in the coding triplet for L536 is located in a well-known hotspot region. In CTCs from three patients, the AKT1 mutation E17K was found, and in CTCs from five patients we discovered mutations in the ERBB2 gene, including hotspot mutations in the nucleotide triplets coding for S310, L755, and V777 ( Figure 3).

Analysis of CTCs from Metastatic Breast Cancer Patients to Identify Targeted Therapies
After validation with cell line cells we analyzed CTCs from 13 metastatic breast cancer patients. 11 of those patients had a primary tumor of luminal subtype and two patients were diagnosed with triple negative breast cancer. CellSearch analyses determined 6 to 15,340 CTCs per 7.5 mL blood. The median CTC count was 94 per 7.5 mL blood. For our study, 1 to 15 CTCs per patient were analyzed.
In seven patients, mutations in the PIK3CA gene were identified. Five of those mutations are well-described mutations located in the nucleotide triplets coding for the amino acid N345, E545, and H1047, respectively. ESR1 mutations were detected in CTCs from two patients. One of those in the coding triplet for L536 is located in a well-known hotspot region. In CTCs from three patients, the AKT1 mutation E17K was found, and in CTCs from five patients we discovered mutations in the ERBB2 gene, including hotspot mutations in the nucleotide triplets coding for S310, L755, and V777 ( Figure 3).

Discussion
Mutations detected in cancer can help to choose an appropriate targeted therapy. In the advanced disease, analysis of a liquid biopsy can facilitate to identify such druggable targets. Various studies reported suitable tools for the analysis of single CTCs' WGA-products, either by investigation of single genomic positions by e.g., Sanger sequencing [22,23] or by sequencing multiple genomic positions by, e.g., NGS panels [24,25]. However, most available assays are either cost intensive, not adapted to single cell whole genome amplified DNA, or cover only single or not all relevant genomic positions for therapy selection in breast cancer. Here, we developed and provided the protocol for a multiplex PCR-based, Ampli1 WGA-adapted NGS panel that covers mutation hotspots in the PIK3CA, the ESR1, the AKT1, and the ERBB2 genes that are most relevant for a liquid biopsy-based targeted therapy in breast cancer.
The assay was validated with spiked cell line cells and finally applied to CTCs from metastatic breast cancer patients. Mutations that provide the basis for potential targeted therapy were detected in 10 out of 13 patients. Although one CTC from patient 4 was found to harbor the ERBB2 A775T mutation, it was not considered as clinically targetable since this mutation has not been characterized in breast cancer, yet. For the treatment of hormone receptor-positive, HER2-negative advanced breast cancers harboring PIK3CA variants such as N345K, E545K, H1047L, or H1047R, alpelisib plus fulvestrant is indicated [26]. Furthermore, HER2-positive tumors of patients with this variant exhibited resistance to trastuzumab [27]. The detected ESR1 L536P mutation leads to a conformational change of the protein and to a ligand-independent ERα activation, thereby conferring resistance to aromatase inhibition [13]. However, treatment with second-generation selective ERα degraders such as fulvestrant is recommended [28]. The AKT1 E17K mutation hyperactivates the mTOR pathway that could be targeted by mTOR inhibitors, such as everolimus, or AKT-inhibitors, such as ipatasertib and capivasertib [29,30]. Mutations such as S310F, L755S, or V777L in the ERBB2 gene offer to be treated with the tyrosine kinase inhibitor neratinib [16].
The frequency of PIK3CA and AKT1 mutations in our patient cohort exceeded the frequencies published in literature. AKT1 E17K mutation frequencies between 1.4% and 8.2% have been described in breast cancer tissue [31] and mutations of the PIK3CA gene are found in 25%-40% of all breast cancers [32,33]. One reason for that is that such mutations can be acquired during the metastasis process leading to a discrepancy of the mutation status of the primary tumor and a liquid biopsy [34]. Furthermore, mutations in the PIK3CA and the AKT1 gene lead to an increased proliferation of the tumor combined with reduced apoptosis [10,15] and might thereby cause a higher probability of CTCs detected in general and of CTCs with a high genome integrity in particular.
Most of the mutations we detected were located in well-described hotspot regions. Some mutations, however, are novel and have not been reported yet, such as the PIK3CA M318T, ERBB2 L313P, or the ERBB2 A775T mutations. These mutations were only detected in single WGA products and due to prior WGA, a false positive result cannot be totally excluded. To further investigate these mutations, tools to predict the effect on the protein structure have been applied [35][36][37]. Although they are not fully consistent, the results indicate that these mutations might affect the function of the protein (Table S3).
Heterogeneity was observed within CTCs from almost all patients. Such heterogeneity indicates that different subclones exist within the tumor and that the patient might therefore benefit from the combination of treatments targeting different proteins. However, it cannot be excluded that the observed assumed heterogeneity may at least in part be caused by false negativity due to allelic dropout during WGA. Thus, the absence of mutations does not necessarily mean that these mutations were not present in the initial cells. One example for allelic dropout is the T47-D reference cell 3: in this cell, only the mutated allele was amplified during WGA. This effect of allelic dropout is limiting all approaches for the analysis of DNA from single cells. The probability of allelic dropout increases with reduced DNA integrity. To minimize it, we focused on the analysis of WGA products that showed at least three bands in the quality control PCR.
Most studies on the clinical utility on mutations have been performed on ctDNA. However, the clinical utility of ctDNA and CTCs has not been compared so far and the analysis of CTCs and ctDNA might be complementary [34].
Altogether, we have developed a time and cost effective multiplex PCR-based assay for the analysis on a MiSeq system. Our assay covers the most relevant positions indicative for targeted therapies in breast cancer and can thereby improve therapy decision and help to improve the treatment of cancer patients eventually.

Conflicts of Interest:
The authors declare no conflict of interest.