Total Extracellular Small RNA Profiles from Plasma, Saliva, and Urine of Healthy Subjects

Abstract

Interest in circulating RNAs for monitoring and diagnosing human health has grown significantly. There are few datasets describing baseline expression levels for total cell-free circulating RNA from healthy control subjects. In this study, total extracellular RNA (exRNA) was isolated and sequenced from 183 plasma samples, 204 urine samples and 46 saliva samples from 55 male college athletes ages 18-25 years. Many participants provided more than one sample, allowing us to investigate variability in an individual's exRNA expression levels over time. Here we provide a systematic analysis of small exRNAs present in each biofluid, as well as an analysis of exogenous RNAs. The small RNA profile of each biofluid is distinct. We find that a large number of RNA fragments in plasma (63%) and urine (54%) have sequences that are assigned to YRNA and tRNA fragments respectively. Surprisingly, while many miRNAs can be detected, there are few miRNAs that are consistently detected in all samples from a single biofluid, and profiles of miRNA are different for each biofluid. Not unexpectedly, saliva samples have high levels of exogenous sequence that can be traced to bacteria. These data significantly contribute to the current number of sequenced exRNA samples from normal healthy individuals.

Figure 1. Distribution of total input reads and reads mapped to the genome.

(A) Displays the alignment of the input reads for each biofluid to the human genome, human rRNA, reads that were too short (<15 nts), and unaligned to the human genome. (B) Displays the distribution of the reads mapped to the human genome to RNA biotypes: miRNA, tRNA, piRNA, protein-coding fragments, miRNA hairpins, Mt_tRNA (mitochondrial tRNA), oncRNA (other non coding RNA), snoRNA, snRNA, Vault RNA, YRNA, tRNA flanking regions, 3′ and 5′ (50 bps flanking the mature tRNA sequence), more than one RNA biotype, and reads that are unassigned (intergenic, intronic, and overlapping with >40 regions to the genome).

Figure 2. Reads aligning to tRNA and piRNAs.

(A) Read overlap between piRNA and tRNA. A large number of sequences detected by sequencing simultaneously align to both piRNA and tRNA. We assessed the distribution of reads per million mapped to the human genome and the numbers that were uniquely classified as piRNA, uniquely classified as tRNA, or overlapped between the two RNAs. In urine and saliva samples, there were few reads that exclusively mapped to piRNA. This does not rule out the presence of piRNA, but the origin of these sequences would have to be further investigated. (B) The upper and lower panels display the number of tRNA and YRNA fragments displays the number of tRNA fragments normalized as reads per million mapped to the human genome (RPM) found in each biofluid. Urine has very high levels of tRNA fragments compared to plasma and saliva, normalized as reads per million mapped to the human genome (RPM) found in each biofluid. Urine has very high levels of tRNA fragments compared to plasma and saliva and the lower panel demonstrates that there are a large number of YRNA fragments found in plasma compared to urine and saliva. (C and D) These two panels display the lengths for the tRNA and YRNA fragments respectively, identified in each biofluid and their abundance.

Figure 3. Distribution of miRNAs in biofluids.

Panel A is a principal components analysis of the miRNAs detected in each biofluid. Each biofluid has a distinct miRNA pattern. Panel B displays the miRNAs detected in each biofluid with >10 or >50 counts in at least 80% of the samples. There are only a handful of miRNAs uniquely detected in urine and saliva at this level of expression. Most miRNAs can be detected in plasma. (C) shows the number of detected miRNAs at 1 count, 10 counts or 50 counts, as a function of input reads. (D) shows the number of reads mapped to the human genome as a function of the input reads. Saliva samples require larger numbers of input reads to achieve the same numbers of reads aligned to the genome as plasma and urine. Urine samples behave similarly to plasma samples with respect to input reads that map to the genome (D), but have fewer miRNAs detected (C).

Figure 4. Coefficient of variation for multiple samples taken from the same individual compared with the coefficient of variation from all subjects.

(A) displays the distribution of CVs calculated for each miRNA detected at >50 counts in at least 80% of the plasma and urine samples. Some of the individual subjects that provided more than 5 samples for sequencing over the course of ~70 weeks, show a closer distribution of CVs when compared to samples from all subjects (asterisk). And some of the subjects with >5 samples display a higher CV than when examining all subjects at once (star). Distribution of miRNA counts for the 15 miRNAs with the lowest CV (B) and the highest CV (C) in each biofluid. At this time, we are unable to determine if the CVs are related to biological variability or to technical variability.