This protocol demonstrates how to purify extracellular microRNAs from cell culture media for small RNA library construction and next generation sequencing. Various quality control checkpoints are described to allow readers to understand what to expect when working with low input samples like exRNAs.
Extracellular and circulating RNAs (exRNA) are produced by many cell types of the body and exist in numerous bodily fluids such as saliva, plasma, serum, milk and urine. One subset of these RNAs are the posttranscriptional regulators – microRNAs (miRNAs). To delineate the miRNAs produced by specific cell types, in vitro culture systems can be used to harvest and profile exRNAs derived from one subset of cells. The secreted factors of mesenchymal stem cells are implicated in alleviating numerous diseases and is used as the in vitro model system here. This paper describes the process of collection, purification of small RNA and library generation to sequence extracellular miRNAs. ExRNAs from culture media differ from cellular RNA by being low RNA input samples, which calls for optimized procedures. This protocol provides a comprehensive guide to small exRNA sequencing from culture media, showing quality control checkpoints at each step during exRNA purification and sequencing.
Extracellular and circulating RNAs (exRNAs) are present in various bodily fluids and are resistant towards RNases1,2. Their high abundance, stability and ease of accessibility are attractive for clinical assessment as diagnostic and prognostic markers3. The mode of transport for exRNAs include extracellular vesicles (EVs), association with lipoproteins (such as high-density lipoprotein; HDL) and ribonucleoprotein complexes (such as with Argonaute2 complexes)4.
A subset of exRNAs are microRNAs (miRNAs), which are small non-coding RNAs of about 22 nt that regulate posttranscriptional gene expression. Ex-miRNAs have been implicated in cell-cell communication and regulation of cell homeostasis5. For example, HDL delivers ex-miR-223 to endothelial cells to repress intercellular adhesion molecule 1 (ICAM-1) and inflammation6. Interestingly, miR-223 is also seen transported by extracellular vesicles from leukocytes to lung cancer cells, programming them to take on a more invasive phenotype7. Thus, the transcriptome of ex-miRNAs from various bodily fluids and cell culture medium will greatly improve our understanding of ex-miRNA signaling.
Small RNA sequencing (small RNA seq) is a powerful tool that can be used to understand the transcriptomics of small RNAs. Not only can different samples be compared amongst differentially expressed known RNAs, but novel small RNAs can also be detected and characterized. Consequently, it is also a robust method to identify differentially expressed miRNAs under different conditions. However, one of the hurdles of small RNA seq is the difficulty in generating small RNA seq libraries from low exRNA input fluids like cerebrospinal fluids, saliva, urine, milk, serum and culture media. The TruSeq Small RNA Library Prep protocol from Illumina requires approximately 1 µg of high quality total RNA and the NEBNext Small RNA Library Prep Set protocol from New England Biolabs requires 100 ng-1 µg of RNA8,9. Oftentimes, total RNA from these samples are below detection limit for conventional UV-vis spectrophotometers.
Ex-miRNAs derived from bodily fluids are potentially good prognostic and diagnostic markers. However, in order to study the functional effects or to determine the origin of specific ex-miRNAs, cell culture systems are often used instead. Mesenchymal stem cells (MSCs) have been studied extensively because their EVs have been implicated in alleviating many diseases including myocardial infarction, Alzheimer's disease and graft versus host disease10. Here, we demonstrate the purification of ex-miRNAs from bone marrow-derived MSCs (BMSCs) and the specific steps used to optimize small RNA library construction, sequencing and data analysis (Figure 1).
NOTE: Mesenchymal stem cell growth medium (MSC media) is prepared beforehand as indicated in the Table of Materials.
1. Cell culture
NOTE: Human mesenchymal stem cells can be obtained from the bone marrow, adipose tissue or other sources11. Alternatively, hMSCs can be bought through a supplier. The BMSCs used in this protocol were derived from the bone marrow of patients and bought from a company.
2. EVs and RNA-associated biomolecules collection
NOTE: EV collection media is prepared beforehand (Dulbecco's Modified Eagle's Medium [DMEM] with 10% fetal bovine serum [FBS] and 1% penicillin/streptomycin [P/S]). EV collection media is normal MSC media, but prepared with commercial EV-depleted FBS (Table of Materials). This is to avoid bovine exRNA contamination from FBS, which normally contains exRNAs associated with EVs, ribonucleoproteins, and lipoproteins. For small RNA library preparation, exRNAs derived from 5 confluent flasks of MSCs are required to enable library construction.
3. EVs and RNA-associated biomolecules collection of differentiated cells
NOTE: EVs and RNA associated biomolecules can also be collected from the cell culture media while the cells undergo differentiation. The example depicted in the protocol describes osteoblastic differentiation and exRNA collection at day 0 and 7 of differentiation. If no differentiation is required, then skip Section 3 and go to Section 4.
4. RNA extraction and quality control
5. Library construction and quality control
NOTE: Small RNA libraries are constructed using commercial kits (Table of Materials) with adjustments due to the low RNA input. Library construction is performed on the chilled block.
6. Bioinformatics pipeline
NOTE: This is an in-house pipeline and the programs used here are listed in the Table of Materials.
The method described in this protocol is optimized to collect exRNA from MSC culture for next generation sequencing. The overall scheme of the workflow is in Figure 1 on the left and the respective quality control checkpoints are on the right.
The morphology of the cells on the day of collection for undifferentiated (Figure 3A) and differentiated (Figure 3B) cells are shown. Furthermore, representative normalized levels of ALP activity of cells induced for 7 days are also shown (Figure 3C). Here, ALP activity is approximately 2.5 times that of the uninduced cells.
Three T175 flasks of confluent MSCs yield 2-5 x 1010 particles/mL (Figure 2). Mean particle size is approximately 160-165 nm while most of the particles are 130-140 nm. There were a few smaller peaks between 200 and 500 nm, which would indicate large complexes or large aggregates.
Since RNA can be degraded quickly, adhere strictly to the conditions of RNA extraction as per the kit or protocol used. Use RNase-free conditions, in particular RNase-free water to resuspend the RNA and keep RNA on ice when possible. After extraction, the RNA can be quantified using a chip-based capillary electrophoresis machine. Representative electropherograms of the RNA after being analyzed are shown in Figure 4. The electropherogram of exRNAs includes a broad range of RNAs (Figure 4A). In contrast, cellular RNA has very distinct peaks at around 70-120, 1800, and 3800 nt, which correspond to tRNA/5S rRNA/5.8S rRNA, 18S rRNA, and 28S rRNA, respectively (Figure 4B). The RNA integrity number (RIN) represents the quality of the RNA. A good RIN of cellular RNA is >8 (indicating almost no RNA degradation). The algorithm on which RIN is based is the ratio between 28S and 18S. This means that RIN is not a good indicator of RNA quality of exRNAs because full-length rRNAs are usually not detectable in exRNAs.
Following RNA extraction, the miRNA libraries were constructed and the cDNA libraries were separated by gel electrophoresis. The adaptor-ligated cDNA of the miRNAs is typically between 140 and 160 nt. The quality of the library was then visualized on a chip-based capillary electrophoresis machine (Figure 5). Figure 5A and Figure 5B are the libraries from the cellular RNA at Day 0 and Day 7, respectively. Figure 5C and Figure 5D are the exRNA libraries at Day 0 and Day 7 respectively. All four samples had peaks at around 140 nt, which indicate successful miRNA library construction. The amount of cDNA in the libraries was quantified by qPCR using a library quantification kit. The quantification cycle (Cq) of the standards was plotted against the log of the concentration to obtain a standard curve (Figure 6). The equation of the standard curve was then used to calculate the concentration of the libraries (Table 1). In our sample, the concentration of the libraries from exRNAs was 5-8 nM, which was significantly lower than libraries from cellular RNAs (8-30 nM). The libraries were then pooled with equal amount and sent for sequencing.
After the libraries were sequenced, the low-quality reads were trimmed away with FASTX-Toolkit and the resulting quality of the library exRNAs was high and comparable to that of the libraries from the cells (Figure 7A,B). The sequence length of the libraries from cells had a single peak at around 22 nt (Figure 7C), which correlates with miRNAs. In contrast, the libraries from exRNAs were more heterogeneous and contained 3 major peaks: 22 nt, 30 nt, and 33 nt (Figure 7D). Similar to their cellular counterparts, the peak at 22 nt is the miRNAs. Closer inspection revealed that the peaks at 30 and 33 nt are tRNAs halves/fragments or piRNAs. The annotations of the mapped reads were then plotted. Most of the reads (65.1%) from cellular RNA mapped to human miRNAs and each of the other small RNAs accounted for a small portion of the mapped reads (Figure 8A). Contrastingly, only 8% of exRNAs mapped to human miRNAs (Figure 8B). The most abundant small RNA to which the reads mapped were tRNAs. Most of the reads were "unmatched reads" (43%). Further examination of the unmatched reads from exRNAs to global databases led to the surprising finding that most of them correspond to bacterial sequences (74% and 85% for D0 and D7 exRNA, respectively; Table 2). In contrast, only 0.9% of the cellular RNA was unmatched and, of those, only 10% mapped to bacteria. A comparison to the bovine genome showed less than 1% match suggesting that FBS was not a source of contamination (Table 2). Hence, exRNAs exhibit an atypical profile compared to normal cellular RNA.
Figure 1: Workflow of exRNA sequencing and analysis. The entire workflow is depicted on the left in the grey boxes and quality control associated with each step is in the red boxes on the right. Please click here to view a larger version of this figure.
Figure 2: Size distribution of particles secreted by the cells at day 0 and day 7 of differentiation. Representative NTA results of particles collected from 3 x T175 flasks at (A) Day 0 and (B) Day 7 of differentiation. The respective mean and mode sizes as well as the concentration of the particles are tabulated beneath the graphs. SD: standard deviation. Please click here to view a larger version of this figure.
Figure 3: Cell morphology and osteogenic differentiation prior to extracellular RNA collection. (A) Micrograph of undifferentiated BMSCs cultured with collection media on the day of collection. (B) Micrograph of BMSCs differentiated for 7 days with collection media on the day of collection. Scale bars = 100 µm. (C) ALP Activities of the cells were normalized to their respective cell viabilities. The values plotted were that of three independent experiments. Please click here to view a larger version of this figure.
Figure 4: Representative electropherogram of RNA integrity after RNA extraction. RNA analyses of extracellular RNA (A) and cellular RNA (B). The x-axis is the length of the RNA (nt) and the y-axis is the fluorescence. The first peak is the ladder at 25 nt. Ribosomal RNAs are shown as 18S (1,800 nt) and 28S (3,800 nt). RNA integrity number (RIN) is an indicator of RNA quality based on the 18S and 28S ribosomal RNA integrity. Higher RIN numbers equal better quality. Please click here to view a larger version of this figure.
Figure 5: Representative electropherogram of the cDNA libraries after library construction. DNA analyses of the libraries from (A) cells at Day 0, (B) cells at Day 7, (C) extracellular RNA from Day 0, and (D) extracellular RNA from Day 7. Green and purple numbers are the ladders at 35 nt and 10,380 nt, respectively. Please click here to view a larger version of this figure.
Figure 6: cDNA library quantification and calculation. Libraries were diluted between 500-1,000 times for those from exRNA and between 1,000 and 4,000 times for cell RNA. The libraries were run alongside the standards from the library quantification kit. The average Cq values of the DNA standards were plotted against the log10 concentration (pM) to generate a standard curve, an equation for the standard curve and an R2 value. Please click here to view a larger version of this figure.
Figure 7: Quality scores and length distribution of sequences. Low quality reads and adaptor sequences were trimmed with FASTX-Toolkit for (A) cells and (B) exRNAs. The length distribution of clean reads for (C) cells and (D) exRNAs. Please click here to view a larger version of this figure.
Figure 8: Annotation of small RNAs after sequencing. Annotations represent percentages of clean reads for (A) cells and (B) exRNAs. Please click here to view a larger version of this figure.
Sample | Dilution factor | Cq value | log(pM) | Concentration (pM) | Concentration mean (pM) | Concentration mean * (452/143) | Multiply by dillution factor (pM) |
BMSC D0 exRNA | 500 | 7.72 | 0.702 | 5.036 | 5.276 | 16.678 | 8338.993 |
BMSC D0 exRNA | 500 | 7.57 | 0.754 | 5.677 | |||
BMSC D0 exRNA | 500 | 7.7 | 0.709 | 5.117 | |||
BMSC D0 exRNA | 1000 | 8.64 | 0.383 | 2.416 | 2.365 | 7.477 | 7476.846 |
BMSC D0 exRNA | 1000 | 8.68 | 0.369 | 2.340 | |||
BMSC D0 exRNA | 1000 | 8.68 | 0.369 | 2.340 | |||
BMSC D7 exRNA | 500 | 8.52 | 0.425 | 2.659 | 3.221 | 10.182 | 5090.913 |
BMSC D7 exRNA | 500 | 8.42 | 0.459 | 2.880 | |||
BMSC D7 exRNA | 500 | 7.97 | 0.615 | 4.125 | |||
BMSC D7 exRNA | 1000 | 8.87 | 0.303 | 2.011 | 1.933 | 6.110 | 6110.391 |
BMSC D7 exRNA | 1000 | 8.92 | 0.286 | 1.932 | |||
BMSC D7 exRNA | 1000 | 8.97 | 0.269 | 1.857 | |||
BMSC D0 cell | 1000 | 6.86 | 1.000 | 10.005 | 9.810 | 31.007 | 31006.974 |
BMSC D0 cell | 1000 | 6.91 | 0.983 | 9.614 | |||
BMSC D0 cell | 4000 | 8.68 | 0.369 | 2.340 | 2.398 | 7.581 | 30322.041 |
BMSC D0 cell | 4000 | 8.59 | 0.400 | 2.514 | |||
BMSC D0 cell | 4000 | 8.68 | 0.369 | 2.340 | |||
BMSC D7 cell | 1000 | 8.44 | 0.452 | 2.834 | 2.880 | 9.104 | 9104.439 |
BMSC D7 cell | 1000 | 8.43 | 0.456 | 2.857 | |||
BMSC D7 cell | 1000 | 8.39 | 0.470 | 2.950 | |||
BMSC D7 cell | 4000 | 10.36 | -0.213 | 0.612 | 0.702 | 2.218 | 8872.689 |
BMSC D7 cell | 4000 | 10.27 | -0.182 | 0.658 | |||
BMSC D7 cell | 4000 | 9.97 | -0.078 | 0.836 | |||
Calculations: | |||||||
log(pM) = -0.3467 x (Sample Cq value) + 3.3786; use the standard curve and formula | |||||||
concentration (pM) = 10^log(pM) | |||||||
Size adjustment calculation = Size of DNA standard (452 bp) divided by the average fragment length (143 bp) |
Table 1: miRNA library quantification.
Sample: | # reads not mapping human | # reads mapping to bacteria | % bacterial reads | # reads mapping to cow | % bovine reads |
BMSC D0 Cell | 131007 | 9858 | 8% | 99 | 0.08% |
BMSC D7 Cell | 169730 | 7935 | 5% | 188 | 0.11% |
BMSC D0 exRNA | 6122833 | 4551477 | 74% | 891 | 0.01% |
BMSC D7 exRNA | 7046691 | 5970086 | 85% | 771 | 0.01% |
Table 2: The percentage of unmatched reads that map to bacteria and bovine reads.
Here, we describe a protocol for next generation sequencing of exRNAs that enables differential expression analyses from low input samples. Adhering to a specific protocol for EV and exRNA isolation is important because even small alterations (i.e., the ultracentrifugation step or a change in rotor type) can influence the transcriptome and miRNA levels13,14. Thus, regardless of how the exRNA is isolated, it is important to apply the same experimental and bioinformatic procedure to all the samples in an experiment to be able to compare the results.
RNA integrity is usually assessed on the bioanalyzer. However, since exRNAs are devoid of full length rRNA, their RIN value is very low, but this does not necessarily reflect low quality RNA samples. Additionally, the RNA concentration of the exRNAs varied greatly and is often inaccurate. Hence, instead of assessing RNA quality and quantifying the RNA on the bioanalyzer, use the same volume of media each time for further processing. Also, resuspend the extracted RNA in the same volume of resuspension buffer and use the same volume for library construction.
Small RNA library construction was optimized by reducing the amount of adaptors to a tenth of the standard protocol and using one-half of the other reagents. Reducing the adaptors not only prevents adaptor dimers, it also prevents intramolecular RNA circularization15. The reduction in reagents is optimized for the Illumina TruSeq Small RNA Sample Prep Kit. Should other kits be used, it is recommended to also lower adaptor and reagents accordingly, unless the kit specifies that it is tailored to low input samples. Finally, the PCR cycle for library construction was increased from 12 to 15 cycles in this protocol to account for the low concentration of exRNAs. We have found this to be the optimal adjustment without seeing amplification biases15.
Various quality control metrics can be assessed including base quality scores and read length profile. When doing bioinformatics analyses, the base quality of the reads in cellular RNA and exRNAs were similar; however, the length distribution and the annotated origin of sequences were completely different.Most of the reads from cellular RNA mapped to human miRNAs, however, a large proportion of the exRNA reads were unmatched reads. Upon closer inspection, the unmatched reads turned out to be bacterial reads (Table 2). This was surprising because antibiotics were routinely used in our culture. Our result aligns well with other reports that also showed large proportion of unmatched reads when sequencing cell culture derived exRNA16. One of the reasons for these contaminants is that bacteria overlap in size with extracellular complexes or EVs, and hence co-purify during the ultracentrifugation step17. Previous publications have identified widespread contamination of certain bacteria in culture media which remain undetected under normal cell lab procedures18. To date, this problem has largely been ignored but should be taken into account when purifying and analyzing exRNAs.
This protocol details a complete guide to harvesting exRNAs from culture media, optimizing small RNA library preparation and processing the raw library data. This protocol specifically highlights the various quality control checkpoints throughout the process to demonstrate how low input samples (like exRNAs) can deviate from normal sample preparations so that others working with low input samples may know what to expect.
The authors have nothing to disclose.
We are grateful to Mr. Claus Bus and Ms. Rita Rosendahl at iNANO for their technical assistance. Special thanks to Dr. Daniel Otzen for allowing our frequent use of his ultracentrifuge. This study was supported by the Innovation Fund Denmark (MUSTER project).
Bone Marrow-Derived Mesenchymal Stem Cells | ATCC | PCS-500-012 | Cells used in this protocol was bought from ATCC |
MSCGM BulletKit | Lonza | PT-3001 | Termed as Mesenchymal Stem Cell Growth Medium (MSC media) |
Exosome-depleted FBS | Gibco | A2720801 | |
ExRNA collecting media: MSCGM but with the FBS replaced by exosome-depleted FBS | |||
Trypsin-EDTA | Gibco | 25200056 | |
T175 Flask | Sarstedt | 833,912 | |
Penicillin-Streptomycin | Gibco | 15140122 | |
Phosphate Buffered Saline | Sigma | 806552 | |
Ultracentrifuge | Beckman Coulter | ||
Polycarbonate Bottle with Cap Assembly | Beckman Coulter | 355618 | |
Beckman Coulter Type 60 Ti | Rotor used here | ||
NucleoCounter | Chemometec | NC-3000 | Cell Counter |
β-glycerophosphate | Calbiochem | 35675 | Components of the osteogenic differentiation media |
Dexamethasone | Sigma | D4902-25MG | Components of the osteogenic differentiation media |
2-Phospho-L-ascorbic acid trisodium salt | Sigma | 49752-10G | Components of the osteogenic differentiation media |
1α,25-Dihydroxyvitamin D3 | Sigma | D1530-1MG | Components of the osteogenic differentiation media |
miRNeasy Mini Kit | Qiagen | 217004 | miRNA and total RNA purification kit for step 4.8 |
Agilent RNA 6000 Pico Kit | Agilent Technologies | 5067-1514 | Chip-based capillary electrophoresis machine and chips for RNA and DNA analysis |
Agilent 2100 Bioanalyzer | Agilent Technologies | G2939BA | Chip-based capillary electrophoresis machine and chips for RNA and DNA analysis |
Agilent High Sensitivity DNA Kit | Agilent Technologies | 5067-4626 | Chip-based capillary electrophoresis machine and chips for RNA and DNA analysis |
KAPA Library Quantification Kits | Roche | KK4824 | Library quantification kit used here |
TruSeq Small RNA Library Prep Kit -Set A (24 rxns) (Set A: indexes 1-12) | Illumina | RS-200-0012 | Small RNA library prepartion kit used in this protocol – used in step 5 |
Pippin Prep | Sage Science | Automated DNA gel extractor used in this protocol; manual extraction can be done too | |
MinElute PCR Purification Kit | Qiagen | 28004 | PCR purification in step 5.17 |
FASTX_Toolkit | Cold Spring Harbor Lab | Trimming low-quality reads in step 6 | |
cutadapt | Adaptor removal in step 6 | ||
Bowtie | Mapping of clean reads in step 6 | ||
Samtools | To make the expression profile in step 6 | ||
Bedtools | To make the expression profile in step 6 |