Identification of causative agents of infective endocarditis by metagenomic next-generation sequencing of resected valves

Abstract

Background: Infective endocarditis (IE) is a rare and life-threatening condition with considerable mortality rates. Diagnosis is often complicated by negative blood culture results, limiting the accurate identification of causative pathogens. This study aimed to evaluate the effectiveness of metagenomic next-generation sequencing (mNGS) of cardiac valve specimens compared to conventional clinical laboratory methods for identifying pathogens in IE.

Methods: Nineteen patients with suspected IE who were scheduled for surgical valve removal were prospectively enrolled. The metagenomic workflow included bacterial DNA enrichment from resected valves using the Molzym Ultra-Deep Microbiome Prep, sequencing of metagenomic libraries using the Illumina MiSeq platform, and Kraken 2 taxonomic assignments based on read data.

Results: Valve mNGS achieved a sensitivity of 82.4% and a specificity of 100% relative to the final adjudicated pathogen diagnosis. Blood culture, considered the reference standard, exhibited slightly higher sensitivity (88.2%) with comparable specificity (100%). In comparison, valve culture (sensitivity: 29.4%, specificity: 50.0%) and microscopy (sensitivity: 35.3%, specificity: 100%) showed lower diagnostic performance. Delays between blood culture negativization and valve resection impacted mNGS sensitivity, likely due to pathogen clearance. Notably, valves resected within 12 days from blood culture negativization achieved 100% diagnostic accuracy, emphasizing the importance of timing for optimal mNGS results.

Conclusion: This study underscores mNGS as a valuable diagnostic tool for detecting IE pathogens, complementing traditional diagnostic methods. The detection of antibiotic resistance genes and multi-locus sequence typing profiles in some samples further demonstrated its utility.

Keywords: cardiac surgery; clinical metagenomics; heart valve; microbiome; next-generation sequencing.

Figure 1

Major contaminant and pathogen species identified in valve samples and negative extraction controls (NEC). Relative abundances determined by Kraken 2 are plotted according to the colour gradient scale given at the bottom. (a) Contaminant species. Among 86 putative contaminant species, those with the relative abundances >2.5% in at least one NEC are reported. (b) IE-causing pathogens. Relative abundances are reported for the most abundant species of each genus given that they were, in both extracts, >1%, and higher than the summed up relative abundances of contaminant species.

Figure 2

Putative reagent contaminants in the host-depleted mNGS dataset. (a) Percentage of reads assigned to H. histolytica. The statistical significance of variations related to the extraction method (MOLZ vs MOLZ-CH) for valve samples or NECs was evaluated using the Wilcoxon signed-rank test. Differences between valve samples and NECs were compared using the Wilcoxon rank-sum test. Only changes that were statistically significant are reported. (b) Percentage of reads assigned to putative contaminant species. The reads assigned to putative contaminant species were summed up and expressed relative to the total number of assigned non-human reads. Wilcoxon rank-sum test was used to assess the differences between mNGS positive valve samples, mNGS-negative valve samples and negative extraction controls (NECs), processed with the same extraction method. Only significant changes are reported.

Figure 3

Correlation between qPCR and mNGS-based estimates of the relative bacterial load. The bacterial fraction is expressed relative to the sum of human and bacterial components in MOLZ (a) or MOLZ-CH (b) samples.

Figure 4

Differences in the bacterial load between samples from mNGS-positive valves, mNGS-negative valves and negative extraction controls. (a) qPCR-based assessment of bacterial DNA in the extracts. (b) The relative abundance of bacterial DNA, expressed as a percentage of the total DNA content (i.e., the sum of human and bacterial DNA abundances as estimated by qPCR). (c) mNGS-based estimate of the relative bacterial DNA abundance, expressed in relation to the combined counts of human and bacterial reads. Statistical significance between sample groups processed with the same extraction method was evaluated using the Wilcoxon rank-sum test, while differences between extraction methods were assessed with the Wilcoxon signed-rank test. Only statistically significant differences are presented. MOLZ and MOLZ-CH denote the two extraction methods used for valve samples, and ‘NEC’ represents negative extraction controls.

Figure 5

Assignments of reads to streptococcal species. The abundance of the dominant Streptococcus species in a given patient is set to 100% for both (MOLZ and MOLZ-CH) valve extracts. The abundance of other streptococcal species (with at least 1% relative abundance in both MOLZ and MOLZ-CH datasets) is expressed as a percentage relative to the dominant species according to the color gradient scale given at the bottom (right). The scale bar (bottom left) shows genetic distance (100–ANI). S. epidermidis was used as an outgroup to root the dendrogram.