Cyndi Goh, Clinical Research Fellow, J Knight Group (Wellcome Trust Centre for Human Genetics, Oxford)
Sepsis, the dysregulated host immune response to infection (Singer et al, 2016), is a major public health burden, accounting for 5% of deaths in England (McPherson et al, 2013) and 27% of intensive care admissions (Harrison et al, 2006). Despite this high disease incidence, no specific therapeutic options for sepsis are currently in widespread clinical use with over a hundred failed phase 2 and 3 clinical trials for sepsis drugs (Marshall et al, 2014).
Progress in sepsis research has been severely limited by a heterogeneous disease phenotype. Application of omics-based methodologies is advancing understanding of the dysregulated host immune response to infection in sepsis.
However, the frequently elusive nature of the infecting organism in sepsis has limited efforts to understand the effect of disease heterogeneity involving the pathogen (Goh and Knight, 2017). A 2016 epidemiological study (Gupta et al, 2016) of nearly 7 million patients with sepsis admitted to hospital between 2001 and 2010 in the USA estimated the incidence of culture-negative sepsis at 47%, with culture negativity identified as an independent predictor of mortality. Obtaining a microbiological diagnosis is a key clinical and research priority for sepsis.
This pilot study involves patients recruited through the UK Genomic Advances in Sepsis (GAinS) study in collaboration with Professor Charles Hinds (Barts and The London NHS Trust) and the GAinS investigators. This UK multi-centre prospective study of genomic modulators of sepsis has recruited over 1200 intensive care patients with sepsis from community-acquired pneumonia (CAP) or faecal peritonitis (the two most common causes of adult sepsis in this country) (Rautanen et al, 2015; Davenport et al, 2016).
Within the CAP cohort, 48% of patients currently have no microbiological diagnosis based on clinical testing. For this pilot study, we analysed plasma from 10 individuals including CAP patients with blood-culture positive S. pneumoniae infection (Samples 1-4; n=4), CAP patients with influenza infection diagnosed by respiratory secretion PCR (Samples 5-7; n=3), and controls with no evidence of infection or sepsis (Samples 8-10; n=3).
Total nucleic acid extraction was performed using the NucliSENS easyMag platform (Biomerieux). For each patient, each of two 500ul aliquots of plasma was eluted in 25ul of buffer. For the Affymetrix Axiom Microbiome Array, 21ul extracted sample input was used.
Data analysis was performed using the Axiom Microbial Detection Analysis Software (MiDAS). Results were compared with metagenomic sequencing and droplet digital PCR (ddPCR) (Bio-Rad QX-100).
The Axiom Microbiome Array was not utilised for the detection of RNA-based pathogens (e.g. influenza) due to limited sample availability (this would have included performing a second assay involving cDNA synthesis for each sample).
Two out of the four samples from the S. pneumoniae cohort (Samples 2 and 3) tested positive for S. pneumoniae using the Axiom Microbiome Array. None of the remaining 6 samples tested positive for S. pneumoniae.
Escherichia coli was detected in all the samples and likely reflects sample contamination or cross-reactivity with nucleic acid extraction reagents. The DNA-based torque teno virus was the only positive viral hit, detected in two out of ten samples.
Metagenomic sequencing and ddPCR detected S. pneumoniae in all four S. pneumoniae samples.
Of the three methods, the Axiom Microbiome Array was least sensitive for the detection of S. pneumoniae with failure to detect the pathogen in half the samples. These two false-negative results occurred in the two samples with lower levels of S. pneumoniae reads on sequencing with correspondingly lower concentrations on ddPCR. The manufacturers recommend a quantity of 50-100ng of DNA per sample, which we were unable to achieve, limiting the usefulness of this technique for this sample type.
This technique may be better suited to body sites with richer and more diverse microbiomes than human plasma (e.g. gut, oropharnyx, skin). However, advantages to the Axiom Microbiome Array include ease of sample processing and data analysis when compared to sequencing, and the ability to detect multiple pathogens without prior expectation of a particular organism a single technique when compared with digital droplet PCR. The Axiom Microbiome Array was as specific as ddPCR in calling the non-S. pneumoniae samples as negative.
The Axiom Microbiome Array is better suited to sample types with a richer and more diverse microbiome than plasma and cases where sample quantity to achieve the recommended DNA input quantity is less of a consideration.