Characterization of the viral microbiome in patients with severe lower respiratory tract infections, using metagenomic sequencing

Abstract

The human respiratory tract is heavily exposed to microorganisms. Viral respiratory tract pathogens, like RSV, influenza and rhinoviruses cause major morbidity and mortality from respiratory tract disease. Furthermore, as viruses have limited means of transmission, viruses that cause pathogenicity in other tissues may be transmitted through the respiratory tract. It is therefore important to chart the human virome in this compartment. We have studied nasopharyngeal aspirate samples submitted to the Karolinska University Laboratory, Stockholm, Sweden from March 2004 to May 2005 for diagnosis of respiratory tract infections. We have used a metagenomic sequencing strategy to characterize viruses, as this provides the most unbiased view of the samples. Virus enrichment followed by 454 sequencing resulted in totally 703,790 reads and 110,931 of these were found to be of viral origin by using an automated classification pipeline. The snapshot of the respiratory tract virome of these 210 patients revealed 39 species and many more strains of viruses. Most of the viral sequences were classified into one of three major families; Paramyxoviridae, Picornaviridae or Orthomyxoviridae. The study also identified one novel type of Rhinovirus C, and identified a number of previously undescribed viral genetic fragments of unknown origin.

Figure 1. A flow-chart describing the classification pipeline.

A flow-chart describing the entire process from sample collection through the various data-analysis steps. Step 1 and 2 illustrate sample collection, preparation and sequencing while step 3 through 6 illustrate in silico efforts.

Figure 2. Major classification of sample content.

The contigs defined by relative position-wise homology (see Methods) and split into Viruses, Bacteria, Mammals, Others or none of the four where in the cases where it could not clearly be determined (Undefined). The numbers are the derived number of reads (panel A) and the number of sequences (panel B).

Figure 3. Class distribution of bacterial content.

The bacterial part of the sample where sequences were defined by closest homolog, and split into classes. The numbers are the derived number of reads.

Figure 4. An overview of the viral part of the sample.

The virus part of the sample, where sequences were defined by closest homolog, split into families (panel A), species found in the Paramyxoviridae family (panel B) and species found in the Picornaviridae family (panel C). The numbers are the derived number of reads and the family and species designations were manually curated, only alignments with an e-value<1e-5 considered.

Figure 5. Phylogenetic analysis of the HRV-C35 contigs.

Phylogram, based on the VP1 region, showing genetic relationships between the new HRV-C35 (bottom, highlighted) with other publically available HRV-C. Some representative HRV-A, HRV-B serotypes were included in the analysis and two HEV-A serotype sequences were used as outgroup. The neighbor-joining tree was evaluated by 1,000 bootstrap pseudoreplicates. Only bootstrap values over 70% are shown. The distinct clusters representing distinct types are indicated by vertical bars. The scale bar represents the genetic distance.