A duplex sequencing approach for high-sensitivity detection of genome-edited plants

Abstract

In this paper, we have evaluated a targeted high-throughput massive parallel sequencing approach for detecting single nucleotide mutations or small genomic changes generated by new genomic techniques (NGT). We used unique molecular identifiers (UMIs) for the quantification of the mutant alleles and duplex sequencing to confirm a mutation on both strands to avoid polymerase chain reaction (PCR) artefacts or sequencing miss-calls. We tested the approach in blinded analyses on a set of mixed NGT-modified tomato lines and identified each single nucleotide mutation or small insert/deletion (InDel) down to a 0.1 % level. To our knowledge, this is the first performance evaluation of a duplex sequencing approach for detecting and quantifying small NGT DNA changes without a priori knowledge of the mutation type and position in a target region. Our study advances the scientific discussion on detecting NGT-induced DNA modifications in plants and food products, evaluating the potential and current limitations of a cutting-edge NGS-approach.

Keywords: Detection; Duplex sequencing; Genetically modified organisms (GMOs); Mutation; New genomic techniques (NGT); Next generation sequencing (NGS); Solanum lycopersicum.

Fig. 1

Duplex sequencing workflow. A) DNA preparation: Genomic DNA is extracted, mechanically fragmented, size-selected (500–600 bp) and subjected to end-repair and A-tailing. B) Duplex adapter generation: duplex adapters are designed to include an Illumina adapter sequence, a sample-specific barcode (sample tag), a Unique Molecular Identifier (UMI) to distinguish individual DNA molecules (molecular tag), and TT/GG nucleotides to differentiate the top and bottom strands (strand tag). The adapter preparation involves the annealing of the TOP and BOTTOM oligonucleotides, followed by an extension reaction to generate the duplex form of the adapter. C) Ligation: the DNA fragments produced in the first step (A) are ligated to the duplex adapters. D) Target enrichment: the target regions in the Psy1 and CrtR-B2 genes are amplified using a universal primer (IL2), which binds to the Illumina adapter sequence, and four gene-specific primers (SPE), that recognise the target sequences in the respective loci (PCR1). In the first cycle, the original TOP and BOTTOM strands are amplified with the SPE and the IL2 primers respectively. In the second cycle, the universal (IL2) and SPE primers amplify respectively their corresponding complementary strands. E) Universal amplification: Illumina P5 and P7 index adapters are inserted in the second PCR amplification step (PCR2) to allow Illumina paired-end sequencing of the target regions. F) Final library: final DNA fragments generated in the experimental procedure. G) Sequencing: paired-end sequencing is carried out on an Illumina platform (NovaSeq 6000). H) Bioinformatics workflow: the pipeline includes controls to filter out reads of low quality, alignment to the tomato reference genome, UMI-based grouping of reads, the generation of single-strand and duplex consensus sequences and high-confidence variant identification. I) Variant identification: the strategy involves the grouping of sequence reads sharing the same UMI, which originates from the same DNA molecule and the merging of complementary strands information, which can be distinguished by their TT and GG strand-specific tags. Single-strand consensus sequences generated from multiple reads with the same UMI allow to discriminate sequencing errors. Duplex consensus sequences generated from both strands of the original DNA molecule allow to distinguish polymerase errors introduced during the PCR amplification steps. Only variants present at the same position on both complementary strands are retained for variant calling.

Fig. 2

Bubble plot showing the mutations detected in the sequencing libraries and their frequencies. Expected mutations are plotted in red colour, while unexpected mutations are plotted in grey. The size of each bubble is proportional to the mutation frequency; for better clarity, all bubbles with frequency greater than 20 % have the same size. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Frequency distribution of duplex molecules per unexpected or expected mutations. The mutations are grouped according to the number of duplex consensus reads supporting each mutation. For each group, the percentage of expected (orange) and unexpected (blue) mutations is reported. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Plot of A.) the coefficient of variation (%) and B.) the bias (%) versus the target level (copy number ratio). Values on the X-axis are reported on a logarithmic scale to better visualize the variation across two orders of magnitude. Psy1+1 (het) and Psy_Δ1 (het) are two alleles of the same heterozygous line Psy_±1.