Quality control implementation for universal characterization of DNA and RNA viruses in clinical respiratory samples using single metagenomic next-generation sequencing workflow

A. Bal; M. Pichon; C. Picard; JS. Casalegno; M. Valette; I. Schuffenecker; L. Billard; S. Vallet; G. Vilchez; V. Cheynet; G. Oriol; S. Assant; Y. Gillet; B. Lina; K. Brengel-Pesce; F. Morfin; L. Josset

Determination of optimal internal quality control spiking 177 MS2 bacteriophage (MS2), a single-stranded RNA virus (ssRNA), was used as the IQC to 178 validate the whole metagenomic process for each sample. In order to optimize IQC spiking 179 level, the sensitivity of the metagenomic analysis workflow for MS2 detection was first 180 evaluated with a ten-fold serial dilutions of MS2 (from 10 -2 to 10 -5 ) in a nasopharyngeal swab 181 tested negative using FA RP (bioMérieux). MS2 was detected in internal QCT1 (IQCT1) for 182 all levels of MS2 spiking (Ct ranged from 17.5 at the 10 -2 dilution to 26.4 Ct at the 10 -5 183 dilution). Full to partial MS2 genome coverage was obtained for all MS2 spiking levels in 184 internal QCT2 (IQCT2; coverage ranged from 98% at the 10 -2 dilution to 69% at the 10 -5 185 dilution). For the highest spiking level, 66.0% of the total number of viral reads was mapped 186 to MS2; for the lowest spiking level, 0.9% were so (Fig. 2) . To limit the number of NGS reads 187 consumed for IQC detection, the optimal spiking condition was determined to be the 10 -188 5 dilution and was used for the rest of the study. Normalized read counts were significantly lower for linear dsDNA viruses than for other viral 237 genome types (Fig. 4b) . 238 compared to the other viral genome types were noticed. As previously described for EBV and 302 CMV, the necessary use of DNase to reduce host contamination may affect these fragile large 303 dsDNA viruses [9,10]. As the detection limit of mNGS analysis is mainly dependent on viral 304 load and total number of reads per sample, this effect could be overcome by increasing 305 sequencing depth; however, we chose to limit the costs of the workflow. 306