Exosomes circulating in the blood carry regulatory RNA molecules, thereby allowing for long distance cell-cell communication. Because diseased cells, including tumor cells, actively release exosomes into the blood stream, the circulating exosomes may provide a stable source of RNAs for disease diagnosis, prognosis and treatment management [2, 6, 39, 48]. In this study, we developed a protocol for isolating exosomal small RNA from a very low volume of plasma. We performed deep sequencing analysis of the exosomal RNAs, and generated expression profiles of the important extracellular RNAs. Our findings will not only help characterize the RNA content of exosomes but will also contribute to understanding exosome function and biology.
Exosomal RNA profiling analysis is not possible without high quality RNA. Compared to cellular RNAs, exosomal RNAs are more stable , and are reportedly resistant to physical degradation such as prolonged storage and freeze/thaw cycles . The circulating exosomal RNAs have been found to be resistant to biochemical degradation by ribonuclease in serum as well as by RNase A under an in vitro condition. This stability makes reproducible and consistent evaluation of blood-based non-coding RNA possible . Indeed, our study strongly supports the protective role of the microvesicles or other proteins in the stability of the circulating plasma RNA. Recently, Argonaute 2 was reported to bind and protect miRNAs from degradation in the circulation . It appears that Argonaute 2-protected miRNAs contribute to a significant proportion of the RNA circulating in the blood. Therefore, RNAs (at least miRNAs) in the blood stream are protected by multiple mechanisms and may be more stable than previously believed .
The dominant size of the exosomal RNA that was detected in this study was 18–28 nt. This size range is apparently smaller than that of the small RNAs derived from culture medium [34, 54, 55], where the sizes were centered at about 70 nt. Different isolation methods may account for the size discrepancies. Ultracentrifugation at 100,000 g seems to be less capable of discriminating exosomes from other microvesicles, especially when the exosomes are large. The mixed sizes of the isolated microvesicles may have caused more heterogeneity of RNA biotypes, which in turn impacted on the size and abundance of the RNAs in the libraries. In addition, the ExoQuick-based assay that we used to precipitate the exosomes may co-precipitate non-exosomal microparticles or RNA-binding proteins. Therefore, technically, the exosomal RNA may account for a fraction of all RNAs isolated by this assay. To obtain reproducible and reliable expression data, further study of the isolation methods is highly recommended.
The highly enriched exosomal miRNAs may have significant impacts on the target cells. For example, miR-99a-5p, the most abundant miRNA in the plasma exosomes, functions in a tissue-dependent manner. In prostate tumor tissue, miR-99a-5p was found to be down-regulated and its overexpression in a prostate cancer cell line was reported to inhibit the growth of the recipient cells and decreased the expression of the prostate-specific antigen . However, overexpression of the miR-99a was also reported to be responsible for increased proliferation, migration and fibronectin levels in a murine epithelial cell line NMUMG, possibly via modulating the TGF-β pathway . The functional role of miR-124 as a tumor suppressor has been established in glioblastoma, breast cancer, hepatocellular carcinoma, gastric cancer, and prostate cancer [58–62]. Another study demonstrated that miR-124 silencing in neuroblastoma cells led to cell differentiation, cell cycle arrest and apoptosis . In support of the important functions of the highly expressed exosomal miRNAs, our GO-based target prediction showed their potential roles in phosphorylation, RNA splicing, chromosomal abnormality, and angiogenesis; however, these predictions need further functional confirmation. Clearly, once released into target cells, the highly enriched miRNAs may participate directly in the regulation of mRNA translation and influence cell functions.
We also observed low level of “long” RNA fragments such as mRNA and lncRNA in the small RNA sequencing libraries. Our library preparation protocols were designed to capture small non-coding RNAs (~20–40 nt long). Therefore, the mRNAs and lncRNAs that were identified in this study should all be treated as fragmented RNAs. The procedures that were used for RNA extraction and library preparation may have caused partial RNA degradation, enabling the detection of fragments of the long RNAs in the small RNA libraries. Another possible explanation for the presence of long RNA fragments is that the exosomes also function as a “reservoir” to remove degraded mRNA and lncRNA derived from the cytosol. The exact mechanism underlying the presence of fragmented long RNAs in exosomes remains to be unraveled.
The current study demonstrated the reproducibility for each library preparation kit. Both Pearson correlation and hierarchical cluster analysis showed highly correlated RNA profiles between technical replicates, suggesting the consistency of these commercial kits. However, the study also showed significant biases between the library preparation methods. Each kit preferentially captured specific RNA sequences. For high abundant RNAs, this bias does not seem to be problematic because all three kits detected these RNAs. For low abundant RNAs, however, the bias could be an issue because these RNAs may be detected by one kit but not by another. Protocol-based bias may also create problems in data interpretation if different commercial kits are used. We suggest that separate validation using qPCR should be performed for all sequencing-based detections.
The ever growing number of novel sequences in the miRNA database implies that human miRNA annotation is far from complete . To identify novel miRNAs, next generation sequencing is the most powerful and the most popular approach. However, systematic bias during library preparation and the limited power of prediction algorithms means that some of the novel miRNAs may have been falsely predicted. We strongly recommended using other complementary methods such as Northern blot and qPCR for subsequent validation. Additionally, this study used only three plasma samples and, therefore, our findings may not fully represent all exosomal RNAs in human populations. To completely survey the exosomal transcriptome more samples from diverse populations and with different disease status are required.
The plasma exosomes are believed to be derived from a variety of cell populations. Their heterogeneous origin may limit the detection of disease-specific exosomes in peripheral blood samples. Vast numbers of exosomes shed from other cell types may dilute the exosome population derived from tumor cells, significantly reducing the proportion of tumor-derived miRNAs in the sequencing libraries. Because the less common tumor-derived miRNA may be a direct reflection of the disease status and critical for tumor development, the increased read depth of RNA sequencing is required. It is worth mentioning, that although the detection of rare RNA transcripts will increase as sequencing depth increases, the rare sequences still account for a tiny fraction of the exosomal RNAs. Whether or not the rare exosomal miRNAs are functional remains to be determined.