Background: Cultivation of targeted pathogens has been long recognized as a gold standard for healthcare surveillance. However, there is an emergent need to characterize all viable microorganisms in healthcare facilities to understand the role that both clinical and nonclinical microorganisms play in healthcare-associated infections. Metagenomic sequencing allows detection of entire microbial communities, in contrast to targeted identification by cultivation. Widespread application of metagenomic sequencing has been impeded in part because the sensitivity and specificity are unknown, which inhibits our ability to interpret results for risk assessment. To assess the impact of sample preparation methods on sensitivity and specificity, we compared several pretreatment steps followed by metagenomic sequencing, and we performed culture-based analyses. Methods: We collected 120 surface swabs from the medical intensive care unit at Rush University Medical Center, which we aggregated to create a representative microbiome sample. We then subjected aliquots to different processing methods (DNA extraction methods, internal standard addition, propidium monoazide (PMA) treatment, and whole-cell serial filtration). We evaluated the effects of these methods based on DNA yields and metagenomic sequencing outcomes. We also compared the metagenomic results to the microbial identifications obtained by cultivation using environmental microbiology methods and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). Results: Our results demonstrate that bead-beating and heat lysis followed by liquid-liquid extraction is the optimal method for the identification of low-biomass surface-associated microbes, as opposed to widely used column-based and magnetic bead-based methods. For low-biomass surface-associated samples, ~590,000 reads per sample are sufficient for ≍90% coverage in metagenomic sequencing (Fig. 1). The ZymoBIOMICS microbial community standard is not appropriate for methods assessing membrane integrity. For the identification of putatively viable microorganisms, PMA treatment is promising, although elimination of signals from nonviable organisms will reduce the overall detectable signal. Combining PMA-treated metagenomic sequencing with cultivation yields the most comprehensive results, particularly for low-abundance taxa, despite high sequencing coverage (Fig. 2). To distribute more detection resources to bacteria, our target domain, we tried whole-cell filtration prior to extraction, attempting to isolate bacterial cells from eukaryotic cells and other particles. For low-biomass surface-associated samples, the sample loss and the difficulties in performing filtration outweigh the slight increase of bacterial signal. Conclusions: Despite optimization, we observed certain blind spots in both cultivation and metagenomic sequencing. This information is essential for informed risk assessment. Further research is needed to identify additional limitations to ensure that results from metagenomic sequencing can be interpreted in the context of healthcare-acquired infection prevention.

Funding: This work was supported by the Centers for Disease Control and Prevention (BAA FY2018-OADS-01 Contract 02915).

Disclosures: None