|Home | About | Journals | Submit | Contact Us | Français|
The diversity of small non-coding RNAs (sRNA) is rapidly expanding and their roles in biological processes, including gene regulation, are emerging. Most interestingly, sRNAs are also found outside of cells and are stably present in all biological fluids. As such, extracellular sRNAs represent a novel class of disease biomarkers and are likely involved in cell signaling and intercellular communication networks. To assess their potential as biomarkers, sRNAs can be quantified in plasma, urine, and other fluids. Nevertheless, to fully understand the impact of extracellular sRNAs as endocrine signals, it is important to determine which carriers are transporting and protecting them in biological fluids (e.g., plasma), which cells and tissues contribute to extracellular sRNA pools, and cells and tissues capable of accepting and utilizing extracellular sRNA. To accomplish these goals, it is critical to isolate highly pure populations of extracellular carriers for sRNA profiling and quantification. We have previously demonstrated that lipoproteins, particularly high-density lipoproteins (HDL), transport functional microRNAs (miRNA) between cells and HDL-miRNAs are significantly altered in disease. Here, we detail a new protocol that utilizes tandem HDL isolation with density-gradient ultracentrifugation (DGUC) and fast-protein-liquid chromatography (FPLC) to obtain highly pure HDL for downstream profiling and quantification of all sRNAs, including miRNAs, using both high-throughput sequencing and real-time PCR approaches. This protocol will be a valuable resource for the investigation of sRNAs on HDL.
Extracellular non-coding small RNAs (sRNAs) represent a new class of disease biomarkers and potential therapeutic targets and likely facilitate cell-to-cell communication1. The most widely studied type of sRNA are microRNAs (miRNA) which are approximately 22 nts in length and are processed from longer precursor forms and primary transcripts2. miRNAs have been demonstrated to post-transcriptionally regulate gene expression through suppression of protein translation and induction of mRNA degradation2. Nevertheless, miRNAs are just one of many types of sRNAs; as sRNAs can be cleaved from parent tRNAs (tRNA-derived sRNAs, tDR), small nuclear RNAs (sRNA-derived sRNAs, sndRNA), small nucleolar RNAs (snoRNA-derived sRNAs, snRNA), ribosomal RNAs (rRNA-derived sRNAs, rDR), Y RNAs (yDR), and other miscellaneous RNAs1. A few examples of these novel sRNAs have been reported to function similar to miRNAs; however, the biological functions of many of these sRNAs remains to be determined, although roles in gene regulation are likely3-6. Most interestingly, miRNAs and other sRNAs are stably present in extracellular fluids, including saliva, plasma, urine, and bile. Extracellular sRNAs are likely protected from RNases through their association with extracellular vesicles (EV), lipoproteins, and/or extracellular ribonucleoprotein complexes.
Previously, we reported that lipoproteins, namely high-density lipoproteins (HDL), transport miRNAs in plasma7. In this study, HDL were isolated using a sequential method of density-gradient ultracentrifugation (DGUC), fast-protein liquid chromatography (size-exclusion chromatography gel filtration, FPLC), and affinity chromatography (anti-apolipoprotein A-I (apoA-I) immunoprecipitation)7. Using both real-time PCR-based low-density arrays and individual miRNA assays, miRNA levels were quantified on HDL isolated from healthy and hypercholesterolemic subjects7. Using this approach, we were able to profile miRNAs and quantify specific miRNAs in highly pure HDL preparations. Since 2011, we have determined that although affinity chromatography enhances HDL purity, antibody saturation greatly limits yield, and can be cost-prohibitive. Currently, our protocol recommends a two-step sequential tandem method of DGUC followed by FPLC, which also produces high quality HDL samples for down-stream RNA isolation and sRNA quantification. Due to recent advances in high-throughput sequencing of sRNAs (sRNAseq), e.g., miRNAs, and the increased awareness of other non-miRNA sRNA classes, sRNAseq is the current state-of-the-art in miRNA and sRNA profiling. As such, our protocol recommends quantifying miRNAs and other sRNAs on HDL samples using sRNAseq. Nonetheless, total RNA isolated from HDL can also be used to quantify individual miRNAs and other sRNAs or validate sRNAseq results using real-time PCR approaches. Here we describe in detail a protocol for the collection, purification, quantification, data analysis, and validation of highly pure HDL-sRNAs.
The overall goal of this paper is to demonstrate the feasibility and process of sRNA quantification in highly pure HDL isolated from human plasma.
This protocol is a series of established methods linked together to allow for the quantification of sRNAs on highly pure HDL by high-throughput sequencing or real-time PCR (Figure 1). To demonstrate the feasibility and impact of this protocol, HDL was purified from human plasma by the tandem DGUC and FPLC method. Collected FPLC fractions corresponding to HDL (by cholesterol distribution) were concentrated and total RNA was isolated from 1 mg of HDL (total protein). sRNA libraries were generated with HDL collected from n = 4 healthy human subjects. All human blood/plasma samples were collected under active Institutional Review Board protocol (# 070416), and human subjects were enrolled with informed consent as approved by Vanderbilt University School of Medicine. Prepared sRNA libraries were sequenced at Vanderbilt Technologies for Advanced Genomics (VANTAGE) DNA sequencing center and data analyses were performed using in-house pipelines. Data visualizations were created by graphing software.
The critical need to perform a second method (FPLC) for HDL isolation after DGUC is due to possible co-purification of EVs. EVs are a generalized class of membrane-derived vesicles that originate from multivesicular bodies (exosomes), plasma membrane budding (microvesicles), and other processes, including apoptotic bodies10. EVs, namely exosomes, have been reported to have a similar density as small HDL (Figure 2A), and thus, could be present in the HDL density fraction (1.063 - 1.21 g/mL) after DGUC11,12. To remove EVs and other lipoproteins (e.g., LDL), from DGUC-HDL, FPLC was used to separate HDL from other vesicles and particles by size; HDL (7 - 12 nm in diameter) and EVs (40 - 220 nm in diameter) are easily fractionated due to their size differences (Figure 2A,B). HDL fractions were readily apparent due to the distribution of cholesterol, and DGUC-HDL FPLC fractions were pooled and concentrated using 10 kDa cut-off centrifugation filters. Concentrated HDL were collected and used for downstream RNA analysis (Figure 2B). To illustrate this point, plasma pre-DGUC was fractionated using FPLC and the distribution of phospholipids (phosphatidylcholine, PC); total cholesterol (TC), and protein levels were quantified and plotted across fractions. PC, TC, and proteins levels were quantified using three colorimetric kits. All three parameters clearly defined the HDL fractions (19 - 24) and VLDL/LDL fractions (14 - 18). In FPLC-fractionated plasma, protein and lipid signals were minimally detected or not detected at all in fractions corresponding to vesicle sizes greater than VLDL (left of orange box) (Figure 3A). To demonstrate DGUC separation of HDL, DGUC-HDL was also fractionated by FPLC and PC, TC, and proteins levels were quantified (Figure 3B). Plotting these values illustrates the co-fractionation of a small amount of LDL using DGUC and reinforces the need to use the tandem DGUC-FPLC approach to isolate highly pure HDL. Although EVs are predicted to co-fractionate with HDL during DGUC, there was little to none found in the lipid and protein in fractions corresponding to vesicle sizes greater than LDL (Figure 3B). The EV lipid and protein signature in the predicted size range is also minimally detectable or absent from DGUC-VLDL and DGUC-LDL when fractionated by FPLC (data not shown). For RNA analysis, total RNA was isolated from 1 mg of tandem purified HDL for sRNA library generation. Briefly, adapters were ligated to the 3’ and then 5’ terminal ends of sRNAs, including miRNAs and tDRs (Figure 4A). Ligated products were reverse transcribed (RT) and PCR amplified using PCR primers harboring specific indexes designed for multiplexing samples during downstream sequencing reactions (Figure 4A). Each adapter is 61 nts in length; therefore, a ligated miRNA (approximately 22 nts in length) would produce a product of 144 nts in length. Many non-miRNA sRNAs are slightly longer than miRNAs (e.g., 30 - 35 nts in length). As such, our targeted length of product was 156 bps in length. Amplified products were size-selected using an automatic DNA size selector and all cDNA products 135 - 177 bps in length were collected (Figure 4A). The qualities of size-selected libraries were assessed by high-sensitivity DNA chips using the bioanalyzer (Figure 4B). sRNA cDNA libraries of this size appeared slightly larger in size due to possible “bubbling” effects during the analysis. For multiplexed sequencing, equimolar concentrations of cDNA libraries were pooled, cleaned, concentrated, and reanalyzed using the bioanalyzer (Figure 4B). The multiplexed sample was then sequenced using a single read 50 bp protocol. In-house data analysis pipelines were used to demultiplex samples based on indexes, trim adapters, run quality control metrics, align to the human genome and count annotated sRNAs. The distribution of normalized reads (Reads Per Million, RPM), > 16 nts in length, illustrate the enrichment of sRNAs 20 - 22 nts in length and 34 - 35 nts in length on HDL (Figure 4C).
Alignment and counting analysis of the four HDL-sRNA libraries found that 38% of reads aligning to annotated regions of the human genome were miRNAs (Figure 5A). Equally abundant (37%) were sRNAs-derived from tRNAs (tDRs). sRNAs-derived from miscellaneous RNAs (e.g., Y RNAs), rRNAs, smRNAs, snoRNAs, and long non-coding RNAs (lincRNAs) were also present on HDL (Figure 5A). After normalization to the total number of reads for each class, the top 50 most abundant miRNAs (Reads Per Mapped miRNA, RPMM, Figure 5B), tDRs (Reads Per Mapped tDRs, RPMT, Figure 5C), and other non-miRNA non-tDR sRNAs (Reads Per Mapped sRNA, RPMS, Figure 5D) were organized by heatmaps. These data illustrate the similarities (e.g., HDL-miRNAs) and discrepancies (e.g., tDRs) of the most abundant HDL-sRNAs across healthy subjects (Figure 5B-D). These results demonstrate the power of this strategy and protocol to provide remarkable insights into the transport of sRNAs by HDL. Nevertheless, sRNA sequencing should not be relied on for absolute quantitation. sRNAs of interest should be validated by real-time PCR. Likewise, many investigators may only require the quantification of individual miRNAs on HDL. As such, total HDL RNA was isolated from each subject for relative quantification of HDL-miRNAs using real-time PCR. Five of the more abundant miRNAs detected on HDL by sRNAseq were used for real-time PCR assays to determine their levels on each HDL sample (Figure 5E). The relative quantitative values were calculated using an arbitrary housekeeping Ct of 32 (RQV= 2-ΔCtArb) and normalized to 1,000 μg of HDL protein input into the RNA isolation13. Results from the real-time PCR study demonstrate that this protocol is also valuable in detecting HDL-miRNAs and quantifying differences across individuals. Collectively, the presented schematics define the critical details of the methods, and the results from both high-throughput sequencing and real-time PCR approaches represent the deliverables of this protocol.
This protocol is designed to quantify miRNAs and other sRNAs by high-throughput sequencing or real-time PCR on highly pure HDL. As with any approach, special considerations should be given to each step in the process of purifying HDL and RNA and then quantifying sRNAs. This protocol is designed for projects starting with ≥2 mL of plasma. Nevertheless, high quality RNA analyses can successfully be completed with HDL purified from as little as 80 μL of human or mouse plasma using affinity chromatography; however, larger starting plasma volumes produce better data using the methods presented here. Sample processing and extraction methods can drastically alter miRNA/sRNA quality and yield. The use of anticoagulants, while essential for plasma collection, likely affects downstream extraction. For example, heparin is not easily removed during RNA extraction14,15, and can inhibit downstream enzymes utilized in PCR16-18. In addition, there are reports that sodium citrate may also interfere with qPCR measurements where miRNAs from EDTA samples are shown to have improved expression profile compared to sodium citrate samples16,19,20. These studies demonstrate the need for careful consideration when choosing an anticoagulant. Furthermore, due to increased potential for cell lysis during coagulation, serum samples are not recommended as lysed cellular miRNAs may artificially associate with lipoproteins. Likewise, plasma is ideally isolated immediately from whole blood to avoid contamination from other peripheral blood cells, including lysed leukocytes. As cold temperature is widely known to affect platelet activation, whole blood samples are spun away from erythrocytes, leukocytes and platelets at 3,000 × g for 15 min at room temperature. Hemolysis is also ill advised as ruptured erythrocytes have been reported to also influence miRNA outcome21,22, as they contain miRNAs that may be shed during hemolysis23,24. Plasma can then be stored at 4 °C for 2 - 4 weeks or alternatively plasma may be frozen at -80 °C where miRNA content is stable and viable in both short term and long term storage (max 4 years)25. To our current knowledge, freezing plasma samples is not predicted to harm HDL-miRNAs and frozen plasma samples are suitable to analyze. As lipoproteins and other whole blood components can be affected by food intake, fasting blood collection is preferred.
As mentioned above, plasma HDL can be isolated through a number of standard methods, including DGUC, FPLC, and affinity chromatography against apoA-I. Isolation of plasma lipoproteins by DGUC using KBr, as described by Redgraves et al.26, separates VLDL (d = 0.94 - 1.006 g/ mL), LDL (d = 1.006 - 1.063 g/mL) and HDL (d = 1.063-1.21 g/mL) and is an ideal method for large volumes of plasma (2 - 60 mL per sample). The limitation to using this approach alone is that exosomes and other miRNA containing EVs are reported to have a similar density to HDL and may be present in DGUC HDL (Figure 2A). Other disadvantages include possible structural damage to proteins from high salt exposure in buffers and strong ultracentrifugation forces as well as differential stabilities of HDL sub-classes27,28. Despite this, DGUC is commonly used as it results in a high yield of HDL protein, producing close to 1 mg of HDL protein per 1 mL of plasma29. The FPLC approach, which involves size-exclusion gel filtration, requires less time, depending on the flow rate (approximately 6 h), and has greater precision separating HDL from apoB-containing lipoproteins (by size) and their sub-classes, including EVs that are much larger than HDL but have similar densities. It should be noted that using the methods detailed in this protocol, plasma vesicles between approximately 220 - 75 nm in diameter do not elicit a detectable lipid or protein signature when separated by FPLC. Depending on the set-up, FPLC requires relatively small input, 0.3 - 1 mL, which is useful for rodent plasma and small frozen aliquots. Quick colorimetric cholesterol assays are also important following FPLC to confirm the distribution of HDL or other lipoprotein fractions. Fractionation dilutes samples and lipoproteins; however, samples can then be concentrated using standard centrifugal size-filters (e.g., 10 kDa m.w. cut-off filter). HDL isolation by apoA-I IP using micro-spin columns are the quickest of the three approaches, but are limited by low input (0.08 - 1 μL) volume and low yield (approximately 0.1 mg). While apoA-I IP can be laborious, this method is ideal for low volume cell culture supernatants. While each of these methods has strengths and weaknesses, these can be reduced if used in tandem (e.g., DGUC followed FPLC). Using two approaches in tandem, results in greater purity of HDL for miRNA and sRNA analysis.
Here, we detailed the steps required to isolate highly pure HDL using the tandem method of DGUC followed by FPLC and outlined the steps used to quantify HDL-miRNAs and sRNAs using either high-throughput sequencing or real-time PCR. Although this protocol was designed for HDL, it has great applicability in quantifying sRNAs on other lipoproteins, including LDL and VLDL, based on their densities and sizes.
This work was supported by awards from the National Institutes of Health, National Heart, Lung and Blood Institute to K.C.V. HL128996, HL113039, and HL116263. This work was also supported by awards from the American Heart Association to K.C.V. CSA2066001, D.L.M POST26630003, and R.M.A. POST25710170.
The authors have nothing to disclose.
The video component of this article can be found at http://www.jove.com/video/54488/