Enhancing infectious intestinal disease diagnosis through metagenomic and metatranscriptomic sequencing of 1000 human diarrhoeal samples

Abstract

Background: Current surveillance of diarrhoeal disease is hindered by limitations of traditional diagnostic approaches, which often fail to identify the causative organism, particularly for novel or hard-to-culture bacterial pathogens. Sequencing nucleic acids directly from stool can overcome such constraints, but such approaches need to reliably detect pathogens identifiable by conventional methods.

Methods: As part of the INTEGRATE study, we analysed stool microbiomes from 1067 patients with gastroenteritis symptoms using direct sequencing, and compared findings with standard diagnostic techniques (culture, immunoassay, microscopy, and single-target PCR) and molecular assays (Luminex xTAG GPP) for detection of bacterial and viral pathogens in the UK.

Results: We found strong positive correlations between metatranscriptomic reads and traditional diagnostics for six out of 15 pathogens. The metatranscriptomic data were highly correlated with the Luminex assay for eight out of 14 pathogens. In contrast, metagenomic sequencing only showed a strong positive correlation with traditional diagnostics for three of 15 pathogens, and with Luminex for four of 14 pathogens. Compared with metagenomics, metatranscriptomics had increased sensitivity of detection for four pathogens, while metagenomics was more effective for detecting five pathogens. Metatranscriptomics gave near-complete transcriptome coverage for Human mastadenovirus F and detected Cryptosporidium via identification of Cryptosporidium parvum virus (CSpV1). A comprehensive transcriptomic profile of Salmonella enterica serovar Enteritidis was recovered from the stool of a patient with a laboratory-confirmed Salmonella infection. Furthermore, comparison of RNA/DNA ratios between pathogen-positive and pathogen-negative samples demonstrated that metatranscriptomics can distinguish pathogen-positive/negative samples and provide insights into pathogen biology. Higher RNA/DNA ratios were observed in samples that tested positive via gold-standard diagnostics.

Conclusions: This study highlights the power of directly sequencing nucleic acids from human samples to augment gastrointestinal pathogen surveillance and clinical diagnostics. Metatranscriptomics was most effective for identifying a wide range of pathogens and showed superior sensitivity. We propose that metatranscriptomics should be considered for future diagnosis and surveillance of gastrointestinal pathogens. We assembled a rich data resource of paired metagenomic and metatranscriptomic datasets, direct from patient stool samples, and have made these data publicly available to enhance the understanding of pathogens associated with infectious intestinal diseases.

Keywords: Culture-independent; Diagnostics; Genomics; Metagenome; Metatranscriptome; Microbiome; Pathogens.

Fig. 1

Visual overview and comparison of DNA (metagenomic) and RNA (metatranscriptomic) sequencing reads assigned to GI pathogens of relevance to the UK setting. For all graphs, the dashed (black) intercept line is provided to highlight the skew of sensitivity towards either DNA or RNA. This skew highlights the likelihood of identifying a pathogen in either DNA or RNA (e.g. Adenovirus can be detected more sensitively with DNA, whilst Cryptosporidium is detected more sensitively with RNA). Values shown are expressed as logarithmic units

Fig. 2

Statistically significant correlations were observed between sequencing data and laboratory tests for 10 out of 14 major GI community pathogens in the United Kingdom. Results where at least one statistically significant correlation was observed are shown. All correlations, whether significant or not, are displayed in Additional File 4: Fig S2. No statistically significant correlation was found between the sequencing and diagnostic test for Astrovirus, E. histolytica, Giardia or V. cholera. The darker the colour of a quadrant in a heatmap, the stronger the correlation (coefficient) between the detection of a pathogen in sequencing data (metagenomic or metatranscriptomic) and a laboratory result (Luminex or Traditional). Asterisks in quadrants indicate the statistical significance of correlations as follows: *: p < 0.25; **: p < 0.05; ***: p < 0.01; ****: p < 0.001

Fig. 3

Adenovirus can be detected through its genomic material and the expression of transcript, directly from stool. Coverage graphs display the breadth of coverage (%) for both DNA and RNA across nine samples, chosen on the basis of positive results through gold-standard laboratory methods. Coverage values were generated via mapping to (A) human adenovirus F40 (NC_001454.1) and (B) F41 (DQ315364.2). For RNA, Study 1680 and 6229 are omitted from this visualisation due to insufficient coverage when mapping to F40. The x-axis shows genomic coordinates, while the y-axis indicates sequencing depth at each position. Colours group sequencing data by patient, with sample labels shown on the left-hand side

Fig. 4

Higher RNA/DNA ratios were observed in samples that tested positive for Campylobacter and C. difficile by gold-standard diagnostics. Violin plots display the distribution of RNA/DNA ratios (logged for visualisation purposes) in samples classified as positive (0) or negative (1) based on gold-standard diagnostics. Panels correspond to the following pathogens and diagnostic tests: A Campylobacter Traditional, B Campylobacter Luminex, C C. difficile Traditional, D C. difficile Luminex. White lines indicate mean RNA/DNA ratios and p-values are the result of statistical comparisons between ratios in positive and negative samples, performed using the Wilcoxon rank-sum test