Limited Correlation of Shotgun Metagenomics Following Host Depletion and Routine Diagnostics for Viruses and Bacteria in Low Concentrated Surrogate and Clinical Samples

Abstract

The etiologic cause of encephalitis, meningitis or meningo-encephalitis is unknown in up to 70% of cases. Clinical shotgun metagenomics combined with host depletion is a promising technique to identify infectious etiologies of central nervous system (CNS) infections. We developed a straightforward eukaryotic host nucleic acid depletion method that preserves intact viruses and bacteria for subsequent shotgun metagenomics screening of clinical samples, focusing on cerebrospinal fluid (CSF). A surrogate CSF sample for a CNS infection paradigm was used to evaluate the proposed depletion method consisting of selective host cell lysis, followed by enzymatic degradation of the liberated genomic DNA for final depletion with paramagnetic beads. Extractives were subjected to reverse transcription, followed by whole genome amplification and next generation sequencing. The effectiveness of the host depletion method was demonstrated in surrogate CSF samples spiked with three 1:100 dilutions of Influenza A H3N2 virus (qPCR Ct-values 20.7, 28.8, >42/negative). Compared to the native samples, host depletion increased the amount of the virus subtype reads by factor 7127 and 132, respectively, while in the qPCR negative sample zero vs. 31 (1.4E-4 %) virus subtype reads were detected (native vs. depleted). The workflow was applied to thirteen CSF samples of patients with meningo-/encephalitis (two bacterial, eleven viral etiologies), a serum of an Andes virus infection and a nose swab of a common cold patient. Unlike surrogate samples, host depletion of the thirteen human CSF samples and the nose swab did not result in more reads indicating presence of damaged pathogens due to, e.g., host immune response. Nevertheless, previously diagnosed pathogens in the human CSF samples (six viruses, two bacteria), the serum, and the nose swab (Human rhinovirus A31) were detected in the depleted and/or the native samples. Unbiased evaluation of the taxonomic profiles supported the diagnosed pathogen in two native CSF samples and the native and depleted serum and nose swab, while detecting various contaminations that interfered with pathogen identification at low concentration levels. In summary, damaged pathogens and contaminations complicated analysis and interpretation of clinical shotgun metagenomics data. Still, proper consideration of these issues may enable future application of metagenomics for clinical diagnostics.

Keywords: CSF; NGS; bacteria; central nervous system infection; diagnostics; host depletion; shotgun metagenomics; viruses.

Figure 1

Evaluation of the host NA depletion method. Primer-probe targets listed on x-axis: the actin B gene host cell genomic DNA (gDNA), the mitochondrial NADH-ubiquinone oxidoreductase chain 1 DNA (mtDNA), the eukaryotic 18S ribosomal RNA (host rRNA), the matrix proteins of Influenza A virus (virus), the putative siderophore biosynthesis protein of Y. pseudotuberculosis (bacterium). Y-axis shows the net amount of depleted NA after host cell lysis as fold reduction in decimal logarithm scale comparing depleted to native samples of which the naturally free-floating NA were depleted previous to NA extraction using AMPure® XP beads reagent. Test was performed in 5 biological replicates and in 4 for bacteria. The error bars depict the standard deviation of the mean fold reduction of the replicates.

Figure 2

Whole transcriptome amplification kit (WTA2) for the metagenomic fragment library preparation for RNA virus spiked surrogate CSF samples with three 100-fold dilutions (v1to1, v1to100, v1to10,000) of Influenza A virus H3N2 virus (v; Inf A). (A) qPCR and (B–D) sequencing results of native and host NA depleted (depl) samples WTA2 amplified and NA extracted by automated MagNA Pure 96 system with the DNA and Viral NA Large Volume Kit and the Viral NA Universal LV 1000 3.0.1 run protocol. (A) Primer-probe targets listed on x-axis: the actin B gene host cell genomic DNA (host gDNA), the mitochondrial NADH-ubiquinone oxidoreductase chain 1 DNA (mtDNA), the eukaryotic 18S ribosomal RNA (host rRNA), the matrix proteins of Inf A H3N2 (virus). The y-axis shows the cycle threshold value (Ct-value) in steps of 3.33, representing an approximated 10-fold reduction of copy number. (B) Taxonomical classification of reads to Inf A H3N2 by exact k-mer mapping (Kraken) to a custom database of human, viruses and bacteria assemblies from RefSeq or Genbank complemented with local alignment to the NCBI's nt (BLAST®) of unclassified reads by Kraken. (C) % Reference sequence coverage and (D) reads aligned with maximum depth of coverage (Max depth) results of reference sequence alignments (Bowtie2) to Inf A H3N2. (B–D) Y-axis shows the number reads in % respective to the total reads that were taxonomically classified or assigned as unclassified by Kraken in decimal logarithm scale.

Figure 3

Comparing whole transcriptome amplification (WTA) to a reverse transcription and whole genome amplification (RT-WGA) approach for shotgun metagenomics analysis of a host NA depleted surrogate CSF sample spiked with RNA virus Influenza A virus H3N2 (Inf A) and bacteria Y. pseudotuberculosis (Y. pseud.). (A) qPCR and (B–D) sequencing results WTA2 compared to SuperScript™ III First-Strand Synthesis SuperMix followed by PicoPLEX® WGA Kit amplification systems for sequencing library preparation of NA extracted by MagNA Pure 96 with the DNA and Viral NA Large Volume Kit and Viral NA Universal LV 1000 3.0.1 run protocol. (A) Primer-probe targets listed on x-axis: the actin B gene host cell genomic DNA (host gDNA), the mitochondrial NADH-ubiquinone oxidoreductase chain 1 DNA (mtDNA), the eukaryotic 18S ribosomal RNA (host rRNA), the matrix proteins of Inf A (virus), the putative siderophore biosynthesis protein of (bacterium). The y-axis shows the cycle threshold value (Ct-value) in steps of 3.33, representing an approximated 10-fold reduction of copy number. (B) Taxonomical classification of reads to Inf A and Y. pseud., respectively, by exact k-mer mapping (Kraken) to a custom database of human, viruses, and bacteria assemblies from RefSeq or Genbank complemented with local alignment to the NCBI's nt (BLAST®) of unclassified reads by Kraken. (C) % Reference sequence coverage and (D) reads aligned with maximum depth of coverage (Max depth) results of reference sequence alignments (Bowtie2) to respective species. (B,D) Y-axis shows the number reads in % respective to the total reads that were taxonomically classified or assigned as unclassified by Kraken in decimal logarithm scale.

Figure 4

Nose swab sequencing results of previously defined Human rhinovirus (HRV) infection by qPCR. Taxonomical classification of reads to HRV-A, HRV-A31, and HRV other than HRV-A by exact k-mer mapping (Kraken) to a custom database of human, viruses and bacteria assemblies from RefSeq or Genbank complemented with local alignment to the NCBI's nt (BLAST®) of unclassified reads by Kraken. X-axis indicates the native and the host NA depleted (depl) preparation. Y-axis shows the number reads in % respective to the total reads that were taxonomically classified or assigned as unclassified by Kraken in decimal logarithm scale.

Figure 5

Abundance overview of most prominent virus species in shotgun metagenomics profiles in at least one across a set of different samples comprising native (nat) and host NA depleted (depl) surrogate CSF samples spiked with Influenza A virus (Inf A) and amplified using whole transcriptome amplification (WTA2) and samples of patients and negative controls (nc) amplified using sequential reverse transcription and whole genome amplification (SuperScript™ III First-Strand Synthesis SuperMix, PicoPLEX® WGA Kit) including CSF samples of patients the pathogen was detected in high (VZV) and low (HSV-1) amounts, diagnostically negative (neg.), and of unknown disease etiology accompanied by PCR grade water nc (H2O), furthermore a nose swab (nose HRV) accompanied by PBS nc, and a serum (serum ANDV). Majority of samples were sequenced on an Ion Torrent™ Ion S5™ using Ion 540™ Chips, few were sequenced on an Illumina® HiSeq 2500 V4 paired-end (^*) on ¹/₂ lane (nat and depl EnteroV (samples 10, 11 of Table 1), nat and depl neg.), respectively on ¹/₆ lane (nat and depl v1to1 Inf A). HRV, Human rhinovirus; EnteroV, Enterovirus genus; ANDV, Andes virus; HSV-1,-2, Herpes simplex virus 1 and 2; VZV, Varicella zoster virus; EBV, Epstein-Barr virus; M.s., Mollivirus sibericum; GemycV, Gemycircularvirus genus; ZEBOV, Zaire ebolavirus; WNV, West Nile virus; TBEV, Tick-borne encephalitis virus; Bl.sh.V, Blueberry shoestring virus; RSV, Respiratory syncytial virus spp.; PBV, Picobirnavirus; M.c.c.a.V, Montastraea cavernosa colony-associated virus; SV-40, Macaca mulatta polyomavirus 1; LassaV, Lassa virus; HPV, Human papillomavirus spp.; HIV-1, Human immunodeficiency virus type 1; Guanar.V, Guanarito virus; Gr.a.t.V, Grapevine associated totivirus-1; DengueV, Dengue virus spp.; C.end.p.V, Citrus endogenous pararetrovirus; CaudoV, Caudovirales order (only Kraken results).