Improving Small RNA-seq: Less Bias and Better Detection of 2'-O-Methyl RNAs

Summary

We present a detailed small RNA library reparation protocol with less bias than standard methods and an increased sensitivity for 2'-O-methyl RNAs. This protocol can be followed using homemade reagents to save cost or using kits for convenience.

Abstract

The study of small RNAs (sRNAs) by next-generation sequencing (NGS) is challenged by bias issues during library preparation. Several types of sRNA such as plant microRNAs (miRNAs) carry a 2'-O-methyl (2'-OMe) modification at their 3' terminal nucleotide. This modification adds another difficulty as it inhibits 3' adapter ligation. We previously demonstrated that modified versions of the 'TruSeq (TS)' protocol have less bias and an improved detection of 2'-OMe RNAs. Here we describe in detail protocol 'TS5', which showed the best overall performance. TS5 can be followed either using homemade reagents or reagents from the TS kit, with equal performance.

Introduction

Small RNAs (sRNAs) are involved in the control of a diversity of biological processes¹. Eukaryotic regulatory sRNAs are typically between 20 and 30 nt in size; the three major types are microRNAs (miRNA), piwi-interacting RNAs (piRNA) and small interfering RNAs (siRNA). Aberrant miRNA expression levels have been implicated in a variety of diseases². This underscores the importance of miRNAs in health and disease and the requirement for accurate, quantitative research tools to detect sRNAs in general.

Next-generation sequencing (NGS) is a widely used method to study sRNAs. Main advantages of NGS as compared with other approaches, such as quantitative PCR or microarray techniques (qPCR), are that it does not need a priori knowledge of the sRNA sequences and can therefore be used to discover novel RNAs, and in addition it suffers less of background signal and saturation effects. Further, it can detect single nucleotide differences and has a higher throughput than microarrays. However, NGS also has some drawbacks; the cost of a sequencing run remains relatively high and the multistep process required to convert a sample into a library for sequencing may introduce biases. In a typical sRNA library preparation process, a 3' adapter is first ligated to the sRNA (often gel-purified from total RNA) using a truncated version of RNA ligase 2 (RNL2) and a preadenylated 3' adapter (Figure 1) in the absence of ATP. This increases the efficiency of sRNA-adapter ligation and reduces the formation of side reactions such as sRNA circularization or concatemerization. Subsequently, a 5’ adapter is ligated by RNA ligase 1 (RNL1), followed by reverse transcription (RT) and PCR amplification. All these steps may introduce bias^3,4. Consequently, read numbers may not reflect actual sRNA expression levels leading to artificial, method-dependent expression patterns. Specific sRNAs may be either over- or underrepresented in a library, and strongly underrepresented sRNAs may escape detection. The situation is particularly complicated with plant miRNAs, siRNAs in insects and plants, and piRNAs in insects, nematodes and mammals, in which the 3' terminal nucleotide has a 2'-O-methyl (2'-OMe) modification¹. This modification strongly inhibits 3' adapter ligation⁵, making library preparation for these types of RNA a difficult task.

Previous work demonstrated that adapter ligation introduces serious bias, due to RNA sequence/structure effects^{6,7,8,9,10,11}. Steps downstream of adapter ligation such as reverse transcription and PCR do not significantly contribute to bias^6,11,12. Ligation bias is likely due to the fact that adapter molecules with a given sequence will interact with sRNA molecules in the reaction mixture to form co-folds, that may either lead to favorable or unfavorable configurations for ligation (Figure 2). Data from Sorefan et al⁷ suggest that RNL1 prefers a single stranded context, while RNL2 prefers a double strand for ligation. The fact that the adapter/sRNA co-fold structures are determined by the specific adapter and sRNA sequences explains why specific sRNA are over- or underrepresented with a given adapter set. It is also important to note that within a series of sRNA libraries to be compared, the same adapter sequences should be used. Indeed, it has previously been observed that changing adapters by the introduction of different barcode sequences alters miRNA profiles in sequencing libraries^9,13.

Randomization of adapter sequences near the ligation junction likely reduces these biases. Sorefan and colleagues⁷ used adapters with 4 random nucleotides at their extremities, designated "High Definition" (HD) adapters, and showed that the use of these adapters lead to libraries that better reflect true sRNA expression levels. More recent work confirmed these observations and revealed that the randomized region does not need to be adjacent to the ligation junction¹¹. This novel type of adapters was named "MidRand" adapters. Together, these results demonstrate that improved adapter design can reduce bias.

Instead of modifying the adapters, bias can be suppressed through the optimisation of reaction conditions. Polyethylene glycol (PEG), a macromolecular crowding agent known to increase ligation efficiency¹⁴, has been shown to significantly reduce bias^15,16. Based on these results, several "low bias" kits appeared on the market. These include kits that use PEG in the ligation reactions, either in combination with classical adapters or HD adapters. Other kits avoid ligation altogether, and use 3' polyadenylation and template switching for 3' and 5' adapter addition, respectively¹². In yet another strategy, 3' adapter ligation is followed by a circularization step, thus omitting 5' adapter ligation¹⁷.

In a previous study, we searched for a sRNA library preparation protocol with the lowest possible levels of bias and the best detection of 2'-OMe RNAs¹². We tested some of the above-mentioned 'low bias' kits, which had a better detection of 2'-OMe RNAs than the standard protocol (TS). Surprisingly however, upon modification (the use of randomized adapters, PEG in the ligation reactions and removal of excess 3' adapter by purification) the latter outperformed the other protocols for the detection of 2'-OMe RNAs. Here, we describe in detail a protocol based on the TS protocol, 'TS5', which had the best overall detection of 2'-OMe RNAs. The protocol can be followed using reagents from the TS kit and one reagent from the 'Nf' kit or, to save money, using homemade reagents, with equal performance. We also provide a detailed protocol for the purification of sRNA from total RNA and the preparation of preadenylated 3' adapter.  

Protocol

1. Isolation of small RNAs

1. Extract total RNA using phenol-based reagents or any other method. Verify if the RNA is of good quality.
2. Pre-run a 15% TBE-urea gel (see Table of Materials) for 15 min at 200 V.
3. While the gel is pre-running, mix 5-20 µg of total RNA in a 5-15 µL volume with an equal volume of formamide loading dye (see Table of Materials; 95% deionized formamide, 0.025% bromophenol blue, 0.025% xylene cyanol, 5 mM EDTA pH 8) in a 200 µL PCR tube. Likewise, mix 10 µL (200 ng) of small-RNA ladder (see Table of Materials) with an equal volume of formamide loading dye. Incubate for 5 min at 65 °C in a thermocycler with heated lid, then place the tubes immediately on ice.
4. Load the ladder and the sample on the same gel with at least one lane between them and run at 200 V until the bromophenol blue (dark blue) has migrated about two-third of the gel length (approximately 40 min).
5. Prepare a system to elute the RNA from gel as follows: puncture the bottom of a nuclease-free 0.5 mL micro centrifuge tube with a 21-gauge needle. Place the punctured 0.5 mL tube in a nuclease-free round-bottom 2 mL micro centrifuge tube.
6. Remove the gel, and incubate at room temperature with 3 μL nucleic acid gel stain (10,000 x concentrate; see Table of Materials) in 30 mL water for 10-15 min.
7. View the gel on a 'Dark Reader' trans illuminator (it is strongly recommended to avoid UV as this might damage the RNA) and cut out the sample RNA between the 17 nt and the 29 nt ladder bands. Transfer the gel piece to the 0.5 mL tube from step 1.5.
8. Centrifuge the 0.5 mL tube in the 2 mL tube in a micro centrifuge at maximum speed for 2 min. Remove the 0.5 mL tube, which should be empty now.
9. Add 300 µL of nuclease-free 0.3 M NaCl to the 2 mL tube containing the crushed gel and rotate for at least 2 h at room temperature or at 4 °C overnight (16 h).
10. Transfer the suspension of crushed gel pieces to a spin column (see Table of Materials) and centrifuge for 2 min at maximum speed in a micro centrifuge.
11. Add 1 µL (20 µg/µL) of glycogen (see Table of Materials) and 950 µL of room temperature 100% ethanol. Incubate for at least 30 min at -80 °C.
12. Centrifuge for 20 min at maximum speed in a micro centrifuge at 4 °C. Remove the supernatant, wash the pellet with 800 µL of cold 80 % ethanol. Centrifuge again for 5 min at 4 °C, and carefully remove all supernatant. Resuspend RNA pellet in 15 µL of nuclease free water. Typically, ~5-20 ng of small RNA should be recovered, depending on the amount of input total RNA (~1 ng of small RNA per 1 µg of input total RNA).
13. Recommended additional step: Check the quantity and quality of the recovered sRNA (e.g., by capillary gel electrophoresis using a small RNA kit; see Table of Materials).

2. Preparation of preadenylated 3' HD adapter

NOTE: Preadenylation of 3' HD adapter was done in a manner similar to the protocol described by Chen et al¹⁸. Note that preadenylated adapter can be ordered directly (/5rApp/ modification), but this is quite expensive.

1. Order 5' phosphorylated, 3' blocked 3' HD adapter oligonucleotide. See Table 1 for sequence and modifications. Note that '3AmMO' is a 3' amino modifier group, most suppliers can produce oligonucleotide with this modification. Dilute in nuclease free water to 100 µM.
2. Set up a 100 µL reaction containing the following reagents: 10 µL of oligo (100 µM), 10 µL of T4 RNA ligase buffer (10x), 10 µL of ATP (10 mM), 40 µL of 50% PEG8000, 5 µL of T4 RNA ligase 1 (50 units), 25 µL of nuclease-free water. Incubate overnight at 20 °C.
3. Perform a classical phenol-chloroform extraction of the preadenylated oligonucleotide followed by ethanol precipitation. Add 100 µL of acid (pH 4.5) phenol:chloroform and vortex. Spin for 5 min at room temperature, maximum speed. Carefully transfer 90 µL of the upper phase to a new tube; add 10 µL of 3 M sodium acetate pH 5.2, 1 µL of ultra-pure glycogen and 250 µL of cold 100% ethanol.
4. Keep at -20 °C for at least 30 min. Centrifuge for 30 min at 4 °C maximum speed. Remove the supernatant, wash the pellet once with 500 µL cold 80% ethanol. As an alternative to phenol-chloroform extraction and ethanol precipitation, a nucleotide removal kit can be used (see Table of Materials). Resuspend in 25 µL water.
5. Measure the concentration of the adapter using a kit for specific detection of single-stranded DNA (see Table of Materials). Dilute to 80 ng/µL (10 µM). 
6. Recommended additional step: Verify the efficacy of preadenylation by migrating 1 µL of 10 µM adapter on a 15% TBE-urea gel (Table of Materials) along with untreated oligonucleotide. The preadenylated adapter should migrate slightly slower than untreated oligonucleotide. If desired, the preadenylated adapter can be gel-purified; proceed as described above (steps 1.2-1.12).

3. Library preparation - Protocol TS5

NOTE: We present here the modified TS protocol 'TS5' that we described previously¹² and that can be performed either with reagents from the kit or with self-provided reagents. It should be noted that we obtained similar or even slightly better results with a different protocol, 'TS7'. However, with TS7 it is more difficult to eliminate adapter dimers. We have therefore preferred to describe TS5 in detail, but TS7 can be followed by simply replacing the adapters. For TS7 use the 'MidRand-Like (MRL)' adapter sequences (Table 1). Note that here the randomized regions are in the middle of the adapters. Primers for reverse transcription and PCR will hybridize to the sequences downstream of the randomized region in the 3' adapter and upstream of the randomized region in the 5' adapter. Sequencing will start from the first randomized nucleotide in the 5' adapter.

1. 3' adapter ligation.
   1. Combine 1 µL of preadenylated 3' HD adapter (10 µM) with 1 µL of purified small RNA (~0.1-1 µM) in a 0.2 mL micro centrifuge tube. Incubate for 2 min at 72 °C in a thermo cycler, then put directly on ice.
   2. Add 4 µL of 50 % PEG 8000 (viscous solution; pipet slowly), 1 µL of RNA ligase buffer (10x), 1 µL of H₂O, 1 µL of T4 RNA ligase2 truncated, and 1 µL of RNase inhibitor. Incubate overnight at 16 °C.
2. Elimination of unligated 3' adapter
   1. Add 10 µL of nuclease-free water and mix well. Add 6 µL of 3 M NaOAc pH 5.2 or 'Adapter Depletion Solution' from the Nf kit (Table of Materials) and mix well. Add 40 µL of magnetic purification beads (Table of Materials) and 60 µL of isopropanol and mix well. Incubate for 5 min at room temperature.
   2. Put the sample in a magnetic rack until the solution appears clear. Remove and discard the supernatant.
   3. Add 180 µL of freshly prepared 80 % ethanol. Incubate for ~30 s, then remove. Take care to use freshly prepared 80% ethanol and do not incubate with 80% ethanol for extended periods.
   4. Briefly spin the tube and remove residual liquid that may have collected at the bottom of the well. Let the beads dry for 2 min, then resuspend in 22 µL of 10 mM Tris pH 8 or resuspension buffer from the Nf kit. Incubate for 2 min, then magnetize the sample until the solution appears clear.
   5. Add 6 µL of 3 M NaOAc pH 5.2 or ‘Adapter Depletion Solution’ from the Nf kit (Table of Materials) to a new tube. Transfer 20 µL of the supernatant from the previous step to this new tube and mix by pipetting. Add 40 µL of magnetic beads and 60 µL of 100% isopropanol and mix well by pipetting. Incubate for 5 min.
   6. Magnetize the sample until the solution appears clear, then remove and discard the supernatant.
   7. Add 180 µL of freshly prepared 80 % ethanol. Incubate for ~30 s, then remove. Take care to use freshly prepared 80% ethanol and do not incubate for with 80% ethanol for extended periods.
   8. Briefly spin the tube and remove residual liquid that may have collected at the bottom of the well. Let the beads dry for 2 min, and resuspend in 10 µL of nuclease-free water. Incubate for 2 min, then magnetize the sample until the solution appears clear.
   9. Transfer 9 µL of supernatant to a new tube. Add 1 µL of T4 RNA ligase buffer (10x) and 1 µL of water. Alternatively, add 2 µL of ligase buffer from the TS kit.
3. Ligation of 5' adapter.
   1. Add 1 µL of 5' HD adapter (10 µM; Table 1) to a 200 µL PCR tube in a thermo cycler with heated lid. Incubate for 2 min at 70 °C, then put the tube directly on ice.
   2. Add 1 µL of 10 mM ATP and 1 µL of T4 RNA ligase 1. Mix well by gently pipetting. Add 3 µL of this mix to the 3' ligated RNA from step 3.2.9 and mix by pipetting. Incubate for 1 h at 28 °C.
4. Reverse transcription (RT)
   1. Transfer 6 µL of 3' and 5' adapter ligated RNA to a new 200 µL PCR tube (keep the remaining ~8 µL at -80 °C for later use if necessary). Add 1 µL of RT primer (10 µM; Table 1) and mix by pipetting. Incubate for 2 min at 70 °C, then put the tube directly on ice.
   2. Add the following reagents for RT: 2 µL of 5x first strand buffer, 0.5 µL of 12.5 mM dNTP mix, 1 µL of 100 mM DTT, 1 µL of RNase inhibitor, and 1 µL of reverse transcriptase (Table of Materials). Incubate for 1 h at 50 °C.
5. PCR amplification
   1. Add the following reagents to the 12.5 µL of RT reaction mixture: 10 µL of PCR polymerase buffer, 2 µL of universal P5 primer (10 µM; Table 1), 2 µL of P7-index primer (10 µM; Table 1), 1 µL of 12.5 mM dNTPs, 0.5 µL of DNA polymerase (Table of Materials), and 22 µL of water. Keep the reaction on ice until use.
   2. Run the following PCR program: 98 °C for 30 s, 11 cycles (98 °C for 10 s, 60 °C for 30 s and 72 °C for 15 s) and 72 °C for 10 min. Keep the reaction at 4 °C when finished.
      NOTE: Concerning the number of PCR cycles: we typically perform 11 cycles, but this should be optimized by the user. Try to perform the smallest possible number of PCR cycles.
6. Gel purification
   1. Run 5, 10 and 20 µL of PCR product on a native 6% TBE gel (see Table of Materials) along with a suitable ladder (see Table of Materials). Run the gel for about 1 h at 145 V (until the bromophenol blue reaches the bottom; this dye migrates at the 65 bp position).
   2. Remove the gel, and incubate with nucleic acid gel stain (see Table of Materials) in water for 10-15 min.
   3. View the gel on a "Dark Reader" trans illuminator (it is strongly recommended to avoid UV as this might damage the RNA) and cut out the library band at 150 bp. Prepare a system to elute the RNA from gel as described in step 1.5 and transfer the gel piece to the 0.5 mL tube.
   4. Centrifuge in a micro centrifuge at maximum speed for 2 min. Remove the 0.5 mL tube, which should be empty now.
   5. Add 300 µL of nuclease-free water to the 2 mL tube containing the crushed gel and rotate for at least 2 h at room temperature or at 4 °C overnight.
   6. Transfer the suspension of crushed gel pieces in water to a spin column and centrifuge for 2 min at maximum speed.
   7. Add 1 µL of (20 µg/µL) glycogen, 30 µL of 3 M NaOAc and 975 µL of ice cold 100% ethanol. Centrifuge for 20 min at max speed at 4 °C.
   8. Resuspend the pellet in 20 µL of 10 mM Tris pH8. Use 1 µL for concentration measurement and 1 µL for quality control.

4. Data analysis

NOTE: The data analysis procedure described below is based on the Linux operating system Ubuntu 16.04 LTS.

1. Treatment of raw sequence files
   1. Download the FASTQ sequence file(s) generated during the sequencing run. If required, perform demultiplexing with bcl2fastq2 (version V2.2.18.12; a manual can be downloaded from the following link: http://emea.support.illumina.com/sequencing/sequencing_software/bcl2fastq-conversion-software/documentation.html).
      Use the following command:
      nohup Pathway_of_bcl2fastq/bcl2fastq --runfolder-dir Pathway_of_Run --ignore-missing-bcl --output-dir Pathway_of_Output_Directory --barcode-mismatches 1 --aggregated-tiles AUTO -r 16 -d 16 -p 16 -w 16
   2. Remove adapter sequences using Cutadapt¹⁹ version 1.15. A manual can be downloaded here: https://cutadapt.readthedocs.io/en/stable/guide.html.
      Use the following command:
      Pathway_to_cutadapt/cutadapt -a TGGAATTCTCGGGTGCCAAGG  -n 5 -O 4 -m 10 -j 0 --nextseq-trim 10 -o Output_File_Read1_cutadapt.fastq.gz Input_File_Read1.fastq.gz
      Note that the sequence in the command corresponds to the 3’ HD adapter without the 4 random nucleotides; these will therefore not be removed during this step.
   3. Use seqtk (https://github.com/lh3/seqtk) for the removal of the terminal random nucleotides in the sequencing reads. Use the following command (variable names in bold):
      seqtk trimfq -b 4 -e 4 Output_File_Read1_cutadapt.fastq > Output_File_Read1_trimmed.fastq
   4. Use the following awk command in order to discard the sequences shorter than 10 nt:
      awk 'BEGIN {FS = "\t" ; OFS = "\n"} {header = $0 ; getline seq ; getline qheader ; getline qseq ; if (length(seq) >= 10) {print header, seq, qheader, qseq}}' Output_File_Read1_trimmed.fastq > Length_Filtered.fastq
2. Mapping of the trimmed sequences
   1. Download the database corresponding to the organism of study from miRBase as follows. Go to http://www.mirbase.org/ftp.shtml and download the ‘mature.fa’ file. Note that the sequences are indicated in RNA notation. Replace the U residues by T with the following command:
      sed -i '/^>/! s/U/T/g' mature.fa
      NOTE: This will yield a complete list of all miRNAs in miRBase, originating from a variety of organisms.
   2. Select the miRNA sequences of your organism of interest with the following command:
      awk 'name_of_the_organism{print; nr[NR+1]; next}; NR in nr' mature.fa > mature_name_of_the_organism_mirs.fa
   3. Map the reads to the above-created file using Bowtie2²⁰ (version 2.3.0) allowing no mismatches. First, build an index for your file with the following command:
      bowtie2-build mature_name_of_the_organism_mirs.fa mature_name_of_the_organism_mirs
   4. Align the sequencing reads to the database, requiring that a read maps entirely to a miRNA of the database, without any mismatches. To this end, use the following tool:
      bowtie2 -N 0 -L 10 --score-min C,0,0 --end-to-end --time -x mature_name_of_the_organism_mirs -U Length_Filtered.fastq -S Length_Filtered_ALIGNMENT.sam
      NOTE: The option: --score-min C,0,0 ensures that alignment is without any mismatches. For an explanation of the various parameters in the tool, please visit the following website: http://bowtie-bio.sourceforge.net/bowtie2/manual.shtml
   5. To discard the reads that did not align, use the following command:
      samtools view -F 4 Length_Filtered_ALIGNMENT.sam > Reads_aligned_to_Mirs.sam
      NOTE: As a result of these steps, you should now have obtained the aligned reads, corresponding to miRNAs.

Results

Critical steps are the isolation of the small RNA fraction of the starting total RNA material (Figure 3) and the desired final library product (Figure 4). Both steps involve polyacrylamide gel purification; small RNA is isolated from 15% TBE urea gels, while the final libraries are isolated from 6% native TBE gels. Small RNA isolated from gel can be analyzed on a small RNA capillary electrophoresis chip (Table of Materials; ...

Discussion

Small RNA library preparation remains challenging due to bias, mainly introduced during adapter ligation steps. RNAs with a 2'-OMe modification at their 3' end such as plant miRNAs, piRNA in insects, nematodes and mammals, and small interfering RNAs (siRNA) in insects and plants are particularly difficult to study because the 2'-OMe modification inhibits 3' adapter ligation. A number of solutions have been proposed in the literature to improve sRNA library preparation protocols, but most commerc...

Materials

Name	Company	Catalog Number	Comments
2100 Bioanalyzer Instrument	Agilent	G2939BA
Acid-Phenol:Chloroform, pH 4.5 (with IAA, 125:24:1)	ThermoFisher	AM9720
Adenosine 5'-Triphosphate (ATP)	Nex England Biolabs	P0756S
Agencourt AMPure XP beads	Beckman Coulter	A63880
Bioanalyzer High Sensitivity DNA Kit	Agilent	5067-4626
Bioanalyzer Small RNA Kit	Agilent	5067-1548
Corning Costar Spin-X centrifuge tube filters	Sigma Aldrich	CLS8162-96EA
Dark Reader transilluminator	various suppiers
HotStart PCR Kit, with dNTPs	Kapa Biosystems	KK2501
NEXTflex small RNA-seq V3 kit	BIOO Scientific	NOVA-5132-05	optional
Novex TBE gels 6%	ThermoFisher	EC6265BOX
Novex TBE Urea gels 15%	ThermoFisher	EC6885BOX
QIAquick Nucleotide Removal Kit	Qiagen	28304
Qubit 4 Quantitation Starter Kit	ThermoFisher	Q33227
Qubit ssDNA Assay Kit	ThermoFisher	Q10212
RNA Gel Loading Dye (2X)	ThermoFisher	R0641
RNA Gel Loading Dye (2X)	ThermoFisher	R0641
RNase Inhibitor, Murine	Nex England Biolabs	M0314S
SuperScript IV Reverse Transcriptase	ThermoFisher	18090200
SYBR Gold Nucleic Acid Gel Stain	ThermoFisher	S11494
T4 RNA Ligase 1 (ssRNA Ligase)	Nex England Biolabs	M0204S
T4 RNA Ligase 2, truncated	Nex England Biolabs	M0242S
TrackIt 50 bp DNA ladder	ThermoFisher	10488043
TruSeq Small RNA Library Prep Kit	Illumina	RS-200-0012/24/36/48	optional
UltraPure Glycogen	ThermoFisher	10814010
XCell SureLock Mini-Cell	ThermoFisher	EI0001
XCell SureLock Mini-Cell	ThermoFisher	EI0001
ZR small RNA ladder	Zymo Research	R1090
			the last two numbers correspond to the set of indexes