Bacterial diversity in two Neonatal Intensive Care Units (NICUs)

Abstract

Infants in Neonatal Intensive Care Units (NICUs) are particularly susceptible to opportunistic infection. Infected infants have high mortality rates, and survivors often suffer life-long neurological disorders. The causes of many NICU infections go undiagnosed, and there is debate as to the importance of inanimate hospital environments (IHEs) in the spread of infections. We used culture-independent next-generation sequencing to survey bacterial diversity in two San Diego NICUs and to track the sources of microbes in these environments. Thirty IHE samples were collected from two Level-Three NICU facilities. We extracted DNA from these samples and amplified the bacterial small subunit (16S) ribosomal RNA gene sequence using 'universal' barcoded primers. The purified PCR products were pooled into a single reaction for pyrosequencing, and the data were analyzed using QIIME. On average, we detected 93+/-39 (mean +/- standard deviation) bacterial genera per sample in NICU IHEs. Many of the bacterial genera included known opportunistic pathogens, and many were skin-associated (e.g., Propionibacterium). In one NICU, we also detected fecal coliform bacteria (Enterobacteriales) in a high proportion of the surface samples. Comparison of these NICU-derived sequences to previously published high-throughput 16S rRNA amplicon studies of other indoor environments (offices, restrooms and healthcare facilities), as well as human- and soil-associated environments, found the majority of the NICU samples to be similar to typical building surface and air samples, with the notable exception of the IHEs which were dominated by Enterobacteriaceae. Our findings provide evidence that NICU IHEs harbor a high diversity of human-associated bacteria and demonstrate the potential utility of molecular methods for identifying and tracking bacterial diversity in NICUs.

Figure 1. PCoA analysis of NICU samples and previously published indoor studies.

PCoA of pair-wise weighted UniFrac distances (see Methods) both with biplots that include taxonomy (A), and without biplots (B). The different colored points indicate the various indoor sampling environments Most of the NICU samples cluster with other indoor surface samples, except for nine NICU1 samples in the top left which cluster with a single office surface sample. Order-level taxonomy illustrates that the presence of Enterobacteriales contributes to the distinct clustering of these samples.

Figure 2. Likely sources of microbes in the two NICUs.

SitePainter images display the results from SourceTracker. The NICU sites are colored on a heatmap scale, where blue indicates that low similarity between a sink and a source and red indicates high similarity between a sink and a source Many surfaces have microbial compositions that are not similar to any of the sources (represented by Unknown), while the handles of the drawers, door and faucet, and the keyboard of the incubator, resemble the communities of human skin.