The Number of Target Molecules of the Amplification Step Limits Accuracy and Sensitivity in Ultradeep-Sequencing Viral Population Studies

Abstract

The invention of next-generation sequencing (NGS) techniques marked the coming of a new era in the detection of the genetic diversity of intrahost viral populations. A good understanding of the genetic structure of these populations requires, first, the ability to identify the different isolates or variants and, second, the ability to accurately quantify them. However, the initial amplification step of NGS studies can impose potential quantitative biases, modifying the variant relative frequencies. In particular, the number of target molecules (NTM) used during the amplification step is vastly overlooked although of primary importance, as it sets the limit of the accuracy and sensitivity of the sequencing procedure. In the present article, we investigated quantitative biases in an NGS study of populations of a multipartite single-stranded DNA (ssDNA) virus at different steps of the procedure. We studied 20 independent populations of the ssDNA virus faba bean necrotic stunt virus (FBNSV) in two host plants, Vicia faba and Medicago truncatula FBNSV is a multipartite virus composed of eight genomic segments, whose specific and host-dependent relative frequencies are defined as the "genome formula." Our results show a significant distortion of the FBNSV genome formula after the amplification and sequencing steps. We also quantified the genetic bottleneck occurring at the amplification step by documenting the NTM of two genomic segments of FBNSV. We argue that the NTM must be documented and carefully considered when determining the sensitivity and accuracy of data from NGS studies.IMPORTANCE The advent of next-generation sequencing (NGS) techniques now enables study of the genetic diversity of viral populations. A good understanding of the genetic structure of these populations first requires the ability to identify the different isolates or variants and second requires the ability to accurately quantify them. Prior to sequencing, viral genomes need to be amplified, a step that potentially imposes quantitative biases and modifies the viral population structure. In particular, the number of target molecules (NTM) used during the amplification step is of primary importance, as it sets the limit of the accuracy and sensitivity of the sequencing procedure. In this work, we used 20 replicated populations of the multipartite faba bean necrotic stunt virus (FBNSV) to estimate the various limitations of ultradeep-sequencing studies performed on intrahost viral populations. We report quantitative biases during rolling-circle amplification and the NTM of two genomic segments of FBNSV.

Keywords: DNA sequencing; FBNSV; Medicago truncatula; faba bean; next-generation sequencing; number of target molecules; rolling-circle amplification; sequencing accuracy; sequencing sensitivity.

FIG 1

Overview of the protocol for experiment 1. (A) Faba bean plants (V. faba) were agroinoculated with a mix of A. tumefaciens cultures carrying plasmids containing the FBNSV segments. Ten A. pisum aphids were used as vectors to transmit FBNSV from symptomatic faba bean plants to Medicago truncatula plants. Total DNA was collected from faba bean and Medicago plants. This procedure was performed on 20 independent lines. Vf1, Vicia faba; Mt, Medicago truncatula. (B) Rolling-circle amplifications were performed on each of the 40 DNA samples collected from faba bean and Medicago plants (from the 20 lines described above for panel A). DNA purification was performed after RCA in order to remove the RCA buffer from the samples. Cleaned DNA samples were sent for sequencing. Genome formulae were measured by qPCR on pre- and post-RCA samples and estimated by counting the number of reads per segment after NGS analysis.

FIG 2

FBNSV genome formulas after DNA extraction (white), after RCA (light gray), and after NGS (dark gray) in faba bean (A) and Medicago (B) plants. Each box plot represents the distribution of the relative frequencies of one segment in 20 replicated plants.

FIG 3

Rolling-circle amplification rates for the different segments of FBNSV with DNA samples from faba bean (A) or Medicago (B) plants. Box plots show the distribution of RCA rates observed for 20 independent plants. Letters on top of the box plots determine groups of values not statistically different after a Tukey HSD test was performed on these data.

FIG 4

Correlation between the relative proportions of segments (genome formula) and rolling-circle amplification rates in faba bean (A) and Medicago (B) samples. The regression formulae are y = 0.0001x + 0.087 (R² = 0.056) for panel A and y = −5.10⁵x + 0.165 (R² = 0.042) for panel B. Pearson's product-moment correlations were not significant for faba bean (t = 0.60, df = 6, P value = 0.57, and correlation coefficient = 0.24) or Medicago (t = −0.51, df = 6, P value = 0.63, and correlation coefficient = −0.21).