Skin microbiome sampling in the preterm neonate

Abstract

Despite advances in our understanding of the human microbiome, there exist significant knowledge gaps in our understanding of the skin microbiome of the preterm neonate. Herein we describe skin microbiome sampling of 6 preterm neonates at multiple timepoints, and compare the skin microbiome samples to environmental (crib/isolette swabs) and negative controls. Samples of the same type (skin, crib, control) were more similar than when compared by week or by patient.

Introduction

Little is known about sampling the skin microbiome in the preterm neonate,^1-3 due in part to presumed low biomass of the skin microbiome in this population. To explore whether skin microbiome samples differed from control and environmental swabs, an interim analysis was performed on a subset of samples from the larger ongoing Pediatric Dermatology Research Alliance funded study "Unraveling determinants of the early cutaneous microbiome." Data from 6 patients (27 swabs, including controls), which were interrupted due to cessation of research during the COVID-19 pandemic, are presented.

Methods

For each low-birthweight preterm infant, a total of 4 skin microbiome samples were collected shortly after birth, and then weekly for the first month after birth. At each sample collection time point, a flocked polystyrene swab was used to sample the skin microbiome from the volar forearm, and samples were additionally collected from the inside of the infant's isolette ("crib" swabs) and a control swab was opened to air during sample collection, but left on the sterile tray without touching any surface ("air" swabs). Demographic and clinical data for the 6 subjects is provided in supplemental Table 1.

Skin microbiome sampling method:

After testing several methods of microbiome sampling, the study protocol utilized polystyrene flocked swabs with perforated breakpoints, moistened with 2 drops of sterile saline, to collect microbiome samples from the skin of the inner forearm. The swab was rolled in alternating directions within a 2x2 cm square of skin (parallel, perpendicular, and diagonal) for a total of 20 seconds. The swab was then inserted into an Eppendorf tube, and broken at the breakpoint with the tube lid. This method was well-tolerated without adverse reactions in all subjects.

Sample extraction, sequencing, and analysis:

Bacterial DNA from each swab was extracted using DNeasy PowerLyzer PowerSoil kit from Qiagen. DNA was quantified and PCR amplified using V1-V3 primers (27F-534R) with adaptors and indexes for MiSeq sequencing. Demultiplexed reads were processed using DADA2 pipeline⁴ for primer sequence removal, quality filtering, correction of Illumina amplicon sequencing errors and dereplication, followed by amplicon sequence variants (ASVs) generation and chimera removal. ASVs were checked for possible contamination using decontam R package⁵ using the prevalence method and threshold set to 0.1, and no contaminants were detected. Taxonomic classification was performed using DADA2 assignTaxonom and addSpecies functions. Principal Coordinate Analysis (PCoA) was generated using the phyloseq package⁶ from Bray-Curtis dissimilarity matrix after count normalization and log₁₀ transformation.^7,8 Taxa bar plots were generated using QIIME 1.9.1⁹ summarize_taxa_through_plots.py function. Difference in beta diversity was tested using PERMANOVA using the vegan R package command adonis (version 2.5-5) with 1000 permutations and stratified on infant ID to account for multiple measures from the same subject. 16S sequencing reads have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under bioproject ID: PRJNA851790.

Results

A total of 27 microbiome samples were included in the analysis (10 skin microbiome samples from 6 infants, 9 environmental samples from the crib/isolette, and 8 negative control samples.) In total, 306 bacterial genera were identified via 16S sequencing. Bacterial taxa from air, skin, and crib swabs appeared to cluster with similar sample types on PCoA (Figure 1). Samples of the same type (skin, crib, air) were more similar (P = 0.002) than when compared by day after birth (P = 0.995) or by skin sampling site (right versus left arm, P = 0.330). Staphylococcus (57.1%) was the most abundant genus identified on skin microbiome samples, followed by Raoultella (20.4%), Klebsiella, (7.3%), Serratia (3.1%), Streptococcus (2.8%), Enterococcus (2.4%), and Enterobacter (1.2%) (Figure 2). Staphylococcus epidermidis was the most prevalent of Staphylococci species (50.7%), followed by S. haemolyticus (3.0%). S. aureus only comprised 0.1% of skin samples.

Figure 1.

Beta-diversity, Principal Coordinate Analysis of all samples

Figure 2.

Relative abundance of all skin microbiome samples. Each patient has 1 to 2 skin swab (denoted as “s”) as demonstrated on the x-axis; first number is the subject number and the second number is week after birth of sample collection.

Discussion

This study supports the demonstrated methods for skin sampling, despite presumed low microbial biomass in preterm infants, as the skin microbiome swabs clustered separately from crib and control swabs. Generalizability is limited, however, due to the small sample size (n = 6 infants, 27 total samples, including 10 skin microbiome samples and 17 control samples). Demonstration of the feasibility of the methods described here was essential to determine that skin microbiome sampling in the preterm infant is both well-tolerated and effective. Results are comparable to previous reports of the preterm infant skin microbiome, with the exception of higher than expected relative abundance of Raoultella.^1,3,10 Further results will be reported when the study is completed (141 infants enrolled to date), including analysis of clinical and demographic factors which are significantly associated with the skin microbiome.

Supplementary Material

supinfo

Supplemental Table 1. Selected subject demographic and clinical data