Quality control materials for pharmacogenomic testing in the clinic

Cell lines/ DNA samples

In 2010, the GeT-RM program characterized 107 genomic DNA reference materials (RMs) for the five most commonly tested genes (CYP2D6, CYP2C19, CYP2C9, VKORC1, and UGT1A1) [5]. Since more genes are now included in pharmacogenetic tests, the GeT-RM program has established an additional 137 genomic DNA samples for 28 pharmacogenes: CYP1A1, CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4, CYP3A5, CYP4F2, DPYD, GSTM1, GSTP1, GSTT1, NAT1, NAT2, SLC15A2, SLC22A2, SLCO1B1, SLCO2B1, TPMT, UGT1A1, UGT2B7, UGT2B15, UGT2B17, and VKORC1 [6]. The DNA samples were prepared from selected Coriell cell lines and genotyped in multiple laboratories by using a variety of commercially available platforms or laboratory-developed tests. These pharmacogenetic RMs and corresponding cell lines are publically available from Coriell Cell Repositories. Some of the cell lines and/or DNA samples extracted from these cells were successfully used in several PT programs for pharmacogenetics [7], [8], [9], [10]. Fifteen control samples with CYP2D6 copy number variants (CNVs) were applied in a test development [11]. The stably transformed cell lines with well-characterized variant alleles have provided a sustainable source of QC materials for pharmacogenetic tests. Since new mutations may possibly occur during cell division, ideally, the QC materials derived from each generation of cell lines should be validated before use. Nevertheless, from our experience in preparing EQA controls [8], [9], [10] from Coriell pharmacogenetic cell lines (mostly Epstein-Barr virus-immortalized lymphocytes), the cell lines were found to be stable. These cell lines were passaged up to 15 times, and yet no unwanted mutations were observed.

Because of the complexity of the gene and lack of well-characterized RMs, accurate genotyping of CYP2D6 is challenging. Fang et al. [12] developed 48 CYP2D6 DNA reference samples by multiple genotyping methodologies.

Plasmid controls

Plasmid controls can be created by cloning targeted sequences of the pharmacogenes into a plasmid. van der Straaten et al. [13] established plasmid controls for commonly tested pharmacogenetic alleles: TPMT∗2, ∗3B/C; CYP2D6∗3, ∗4, ∗6, ∗9, ∗41; CYP2C9∗2, ∗3; CYP2C19∗2, and ∗3 [13]. To create controls that can be used for different methodologies, they cloned a gene region of about 500 nucleotides up and downstream of the specific single nucleotide polymorphism (SNP). A heterozygous genotype can be produced by mixing two homozygous plasmids. Since Taq polymerases may incorporate base errors in the final polymerase chain reaction (PCR) product, plasmid controls should be validated by Sanger sequencing. Notably, for the preparation of a specific control, plasmid needs to be sequenced only once.

Residual patient specimens

Previously tested patient specimens are most suitable for QC; however, they have several significant shortcomings, which will be discussed in the subsequent section.

Commercial control products

Decisive diagnostics has 24 cell lines with known polymorphisms in the genes: CYP2C9, CYP2C19, CYP2D6, MTHFR, NAT2, and UGT1A1. Besides, synthetic DNA controls for testing CYP2C19, CYP2C9, and VKORC1 genes are available from Maine Molecular Quality Controls. Although these QC materials are synthetic constructs, they can still be used to “monitor the analytical performance of the extraction step,” as the vendor claimed.

Strengths and weaknesses of different formats

Genomic samples, such as human cell lines and residual patient specimens, have the advantage of being able to monitor the entire genotyping process including DNA extraction; therefore, they are more favorable for QC purposes. The main problem of using residual patient samples as controls is that the source is limited, and the desired samples with low allele frequency in certain ethnic populations are often not easily obtained. The pharmacogenetic RMs developed by the GeT-RM program have been comprehensively genotyped and are ready to use.

Plasmid controls do not resemble the real patient samples, and they cannot be used to monitor the DNA isolation step of genetic testing process. However, plasmid controls are relatively easy to be prepared in large quantities and are cost effective for use in multiplex-based genotyping techniques. Therefore, plasmid-derived controls are suitable for test development and validation.

QC materials for mutation detection: low-throughput assays

Low-throughput assays involve a wide range of PCR-based assays that amplify short DNA fragments for the analysis of a single or a few genes. For single gene-based mutation assays, QC materials can be prepared from residual patient samples or bought from commercial sources. The materials derived from surplus patient tumors are ideal for QC and widely used in EQA schemes for cancer therapy-associated gene mutational analysis [16], [17], [18], [19], [20], [21], [22], [23], [24]. However, there are several apparent shortcomings of this material: (1) the amount of tissue that can be obtained is limited; (2) samples with rare mutations are difficult to get; (3) the heterogeneity within a single tumor; (4) the precise amount of wild-type alleles versus mutated alleles is hard to determine. To address the above issues, artificial formalin-fixed paraffin-embedded (FFPE) controls were established.

FFPE cell line samples were first developed as QC materials for immunohistochemical assay of erb-b2 receptor tyrosine kinase (ERBB2) expression in breast cancer [25], [26], [27]. Later, the artificial FFPE cell line samples were applied in EQA schemes of KRAS [28], [29] and EGFR mutation testing [30]. The artificial control can be made by homogenously mixing mutant versus wild-type cell lines at defined allelic ratios [28], which closely mimics the FFPE tissue block processed in pathology laboratory, and it is highly reproducible. We have modified the original protocols to facilitate the preparation of FFPE cell line controls [29]. Generally, at least 100 tumor cells contain mutations should be present in each section. The precise allelic ratios of mutations in each sample can be quantified using droplet digital PCR (ddPCR) [30], [31]. For convenience, we listed 36 cancer cell lines harboring clinically significant gene mutations (see Supplemental data, Table 1). Using the above-mentioned method, it is possible for a genetic testing laboratory with minimum cell culture facilities to provide an almost limitless supply of appropriate QC material. Moreover, the artificial FFPE cell samples make it possible to create controls for rare variants and variants with an allelic frequency <20%. In addition, mutations may possibly occur during cell division. The solution for this is to produce large homogenized growths of cells for one batch, and the variants of interest in each batch of FFPE cell line samples should be validated before use [32].

Artificial QC materials for histological assessment of therapeutic response in cancer

Histological assessments of genomic aberrations, such as ERBB2 overexpression in breast cancer and anaplastic lymphoma kinase (ALK) rearrangement in NSCLC, are also applied in guiding targeted therapy in tumors. We prepared FFPE cell line samples as previously described [26] and used artificial controls in a pilot EQA study for ERBB2 testing in China. However, based on the feedback from the participating laboratories, the performance of the controls was not ideal. Owing to the matrix effect, the morphologic characteristics of the cell line-based controls were distinct from the tissue slides when fluorescent in situ hybridization (FISH) was used. To overcome this problem, we innovatively created xenograft tumor samples as QC materials [33]. In our EQA study for ALK rearrangement testing, the artificial tissue controls not only resembled real patient tumors in histomorphology, but also performed well with different methods including FISH, immunohistochemistry, reverse transcription PCR (RT-PCR), and NGS [33].

Approach of creating defined mutations in cell line

Human cell lines with target mutations can be created by modifying wild-type cell lines using genome editing technologies such as recombinant adeno-associated virus (rAAV), zinc finger nucleases (ZFNs), and clustered regularly interspaced short palindromic repeat (CRISPR)/associated protein (Cas) 9 system. The CRISPR/Cas9 system is a robust, extremely easy technology that can be manipulated in almost any laboratory. In addition, CRISPR/Cas9 has been used to introduce a range of variants including single-point mutations [34], multi-point mutations [35], insertion-deletion mutations (indels) [36], and translocation [37]. As PT samples, two artificial FFPE sections derived from genetically engineered cell lines from Horizon Diagnostics were used in an EQA scheme of RAS testing [22]. However, several laboratories failed to detect certain expected mutations in the synthetic samples. This was because of an amplification failure due to the engineering scars in the edited genome of the cell line. Nonetheless, if gene-modified cell lines were appropriately validated to assure that only wanted mutations were present, they can be used as proper resources for QC materials in single gene-based testing. Off-target effect is an innate defect of genome editing technologies, especially for CRISPR/Cas9 [38]. Therefore, the cell lines engineered by CRISPR/Cas9 system are currently not suitable for developing QC materials for NGS assays. The major advantage of using genetically modified cell lines is that QCs can still be produced, even when natural samples harboring rare mutations are difficult or impossible to obtain.

QC materials for mutation detection: high-throughput assays

NGS has been rapidly adopted by clinical laboratories for sequencing solid tumors owing to its ability to (1) detect all types of mutations, including single nucleotide variants (SNVs), indels, and structural variants (SVs); (2) survey an entire gene, a panel of genes, or a genome rather than a single or a few variants; (3) improve the limit of detection for somatic mutations and overcome the issue of tumor heterogeneity. An appropriate QC material is essential for assuring high-quality NGS testing in clinical laboratory settings.

Owing to the complexity of NGS technology, professional societies and government agencies including CDC, American College of Medical Genetics and Genomics (ACMG), and the College of American Pathologists (CAP) have published guidelines to assure the quality of NGS in the clinic [39], [40], [41]. The CDC has described approaches to develop characterized RMs for NGS [39]. QC materials containing mutations identical to those for which the NGS assay was designed to detect is recommended. However, it is impractical to develop a comprehensive set of QC materials that encompasses all possible disease-associated sequence variations. For instance, a targeted NGS assay was designed for assessing 380 actionable mutations [42]; however, it is difficult to create a single QC harboring all the variants. In such situations, the performance attributed to target-region sequencing assay could be based on several surrogate variants and not necessarily on the full complement of the disease-associated sequence variations within that region [43].

Commercial sources of QC materials for NGS assay

The Genome in a Bottle (GIAB) Consortium hosted by the National Institute of Standards and Technology (NIST) is developing RMs for human genome sequencing [44]. NIST RM 8398, the genomic DNA derived from the well-characterized cell line GM12878, is the world’s first RM used for whole-genome variant assessment [45]. In addition, the reference data collected for the physical genomic DNA can be used to assess the performance of bioinformatics pipelines [46]. Cell lines, DNA, and the data from the GIAB project (http://jimb.stanford.edu/) are publicly available.

Commercial QC materials for NGS, including the AcroMetrix Oncology Hotspot Control from Thermo Fisher Scientific, Seraseq Solid Tumor Mutation Mix-II and Seraseq Solid Tumor Mutation Ladder-II from SeraCare Life Sciences, use another cell line (GM 24385) from GIAB project as the background “wild-type” material. Several QC products from Horizon Diagnostics are available in different formats, such as genomic DNA, FFPE, and formalin-compromised DNA.

Development of QC materials for NGS mutation assay

The QCs for NGS assays can be produced from a variety of potential sources, each with its own strengths and weaknesses (Table 1). For NGS-based assay intended for guiding cancer therapy, patient samples that are comprehensively sequenced with high-quality consensus variants are preferred as QC materials. However, the residue sample from a single patient has limited supply. As standard FFPE tumor samples are more regularly used as starting materials for NGS, FFPE controls, which have the ability to assess the entire process including sample DNA purification and quantitation, are perhaps the most suitable type of controls. Dumur et al. [47] developed artificial FFPE QC materials for AmpliSeq Cancer Hotspot Panel v2 (CHP2) assay by mixing four characterized cancer cell lines containing therapeutically actionable mutations at different allelic frequencies. However, this preparation method may encounter the hassle of sourcing, blending, and characterizing the cell line mixtures. A more convenient approach is to prepare DNA controls from NCI-60 cell lines (http://www.ingenuity.com/nci60_wes), whose whole exome sequence data are publicly available [48]. Although synthetic DNA constructs might not perfectly resemble the human genome, they could be designed to assess particular genetic variants, such as SVs, CNVs, and repetitive sequences. Moreover, they can be either spiked into samples for routine QC of each run or used as controls for the analytical validation of NGS assays [42]. Electronic data files from real patient samples or created (simulated sequence data), as well as reference data from GIAB, can be used for QC of informatics pipeline.

By taking advantage of synthetic constructs and the NIST reference genomic DNA, a robust, cost-effective, and consistent source of QC material can be prepared from a mixture of synthetic DNA constructs and NIST whole-genome reference material. To meet the current mainstream NGS technologies, the SNVs and indels in synthetic controls may be designed to have flanking sequences of 300–500 base pairs (bp), and the SVs may have from 700 to 1100 bp on either side, spanning the breakpoint. The genomic DNA extracted from GM12878 cell line might serve as “wild-type” background. A DNA mixture of mutations can be produced in the range of minor allele frequencies to emulate real specimens. To filter out the irrelevant mutations, pure GM12878 genomic DNA can be used as normal control besides the mixture controls. We used this approach to produce EQA samples including variants of SNVs and indels, which were successfully applied in a national EQA for the detection of somatic mutations employing NGS in 2015 [43].

Sequins (http://www.sequin.xyz/), a set of novel synthetic DNA standards that simulate the features of a real human chromosome, may serve as qualitative and quantitative spike-in controls for NGS [49]. This control system is based on an artificial in silico reference chromosome, which constitutes sequins and “background” derived from reference genome sequences. Sequins are short DNA molecules (<10 kb), with genetic variants of interest flanked by inverted reference sequences. Sequencing reads from sequins align only to the in silico chromosome, which allows them to be separated for parallel analysis as internal controls. The subtle design of sequins make them promising controls for use in almost all the stages of an NGS assay, including informatics analyses.

QC materials for oncology circulating tumor DNA (ctDNA) assays

Cell-free tumor DNA is a promising liquid biopsy with numerous advantages over invasive biopsies [50]. As a non-invasive biomarker, ctDNA provides information about mutations that are highly consistent with those of the paired tumors, which can be used to monitor tumor dynamics and track the development of acquired resistance. Moreover, it overcomes the issue of tumor heterogeneity and the failure of biopsy procedure in certain tumors. To produce QC materials for the analysis of ctDNA, an approach similar to NGS controls can be adopted. First, blend the synthetic DNA constructs containing mutations with NIST reference genomic DNA at a pre-defined ratio. To mimic natural ctDNA fragments (180–200 bp), the DNA mixture was fragmented by sonication, stabilized, and purified. Human proteins may be added to the mixture to simulate real plasma samples. Minor allele frequencies <5% could be achieved by titration and precisely quantitated though ddPCR. SeraCare Life Sciences and Horizon Diagnostics provided the commercial controls for ctDNA assays.