Guiding RNAi Design Through Characterization of Endogenous Small RNA Pathways

Abstract

RNAi is a common eukaryotic gene regulation process driven by small RNA effectors. Mechanisms that govern small RNAs behavior have only been extensively described in a handful of organisms, which suggests that the most effective RNAi approach in many organisms, such as insect pests, remains to be determined. Taking advantage of advances in high throughput sequencing, characterization of small RNA molecules can be achieved through bioinformatic approaches without the need for genetic experiments. This chapter describes pipelines for the characterization of the three main classes of small RNAs (microRNAs, small-interfering RNAs, and piwi-associated RNAs) using computationally determined small RNA biogenesis signatures. Understanding the abundance of different small RNA classes will lead to better-informed RNAi strategies.

1. Introduction

Post-transcriptional gene silencing through RNA interference (RNAi) is mediated by small RNAs. These non-coding RNA act through complementary base pairing with target RNAs, triggering mRNA destruction or translational inhibition as well as epigenetic regulation.[1, 2] Diverse biogenesis pathways and activities of small RNAs have been identified over the last decade, mostly due to advances in sequencing technologies. Unlike approaches for mRNAs that involve fragmentation prior to sequencing, small RNA libraries are cloned through adapter ligation, which preserves their endogenous ends. These mean that small RNA sequencing not only captures expression information but also processing signatures. This in turn, makes it feasible to use computational approaches as a means to distinguish small RNA classes. Three major classes of small RNAs are recognized: microRNAs (miRNAs), small-interfering RNAs (siRNAs), and piwi-associated RNAs (piRNAs). Each has been used in successful RNAi strategies[3]. Unfortunately, the function of different small RNAs varies by animal. This makes characterization of small RNA biology in a particular species valuable for identifying an RNAi strategy. Wet lab and computational approaches are described here for dissecting small RNA biogenesis without using genetic manipulation methods, which are often untested in insect pests.

miRNAs Biogenesis

miRNAs are non-coding RNAs (~22 nt) derived from conserved short hairpins. They are cleaved from heterogenous transcripts by the microprocessor complex, which contains the RNase III enzyme Drosha. Several sequence elements such as a mismatch (bulge) near the base of the hairpin recruit the microprocesses to produce a pre-miRNA hairpin that exhibits a 3’ 2nt overhang characteristic of RNase III processing [4]. In the cytoplasm, another RNase III enzyme, Dicer cleaves the loop from the pre-miRNA [5] The resulting ~22nt asymmetric duplex RNA typically exhibits a central mismatch. The guide strand with the most unstable 5’ end is loaded into a miRNA-specific Argonaute protein (AGO1) to form a silencing complex. The passenger strand is degraded, leading to highly asymmetric accumulation of guide strand relative to passenger strand.[6] Detecting miRNAs can be achieved through identifying precursor miRNAs hairpins (pre-miRNAs) using a combination of hairpin fold prediction and small RNA expression. Software such as miRDeep2, miRanalyzer and many others can be used for this purpose (Figure 2a, b) [7] miRNAs are routinely used as RNAi reagents in a host of animals. For instance, an artificial miRNA (amiRNA) was successfully developed to target the chitinase gene of one of the most damaging polyphagous crop pest, Helicoverpa armigera. When the amiRNA was inserted into a vector and used as a transgene in the host plant, H. armigera larva failed to molt and eventually died upon feeding on the host plant.[8]

Figure 2. WCR miRNA NW_021040030.1_748375:

2a shows a portion of miRDeep2 output with 2a showing the hairpin structure possessing GA at the 3’ end suggesting Drosha cleavage. 2b is a graphical representation of read frequency (y-axis) against read length (x-axis). In 2a and 2b, the red portion (mature) is the 5’ arm, purple portion (star) is the 3’ arm and the yellow portion is the hairpin.

siRNA Biogenesis

siRNAs are ~21nt long and produced from endogenous or exogenous long double-stranded RNA [9]. Endogenous-siRNAs (endo-siRNAs) derive from cis-Nats RNAs and long hairpin RNAs, while exogenous-siRNAs arise from sources like viral dsRNA. Insect siRNAs are produced by dedicated Dicer (Dcr2), which like miRNA-dicer leaves 2nt 3’ overhangs on resulting small RNA duplexes (Figure. 1b). Insect siRNAs are loaded into a dedicated AGO (AGO2), after which they are 2’-O-methylated by Hen1. siRNA-loaded AGO2 “slices” the target transcript following near-perfect base pairing of the target. Signatures of dicer cleavage (2nt 3’ overhang) and origin can together be used to distinguish siRNAs in sequencing libraries. Consistent with a robust siRNA pathway, some insects are sensitive to exogenous dsRNA and mount a potent RNAi response. Hence, dsRNA is already used as novel insecticidal technologies[10, 11]. The commercial corn product Stax Pro elicits a host induced RNAi approach to control the Western Corn Rootworm (WCR) Dabrotica virgifera virgifera. The transgenic corn crop expresses a hairpin dsRNA that targets the essential snf7 gene of WCR such that upon feeding leads to mortality through silencing of the target gene.[12, 13]

Figure 1. Major small RNA classes and biogenesis:

1a shows a typical miRNA precursor with downward-facing arrow showing region of Drosha cleavage and upward-facing arrow showing region of dicer cleavage. N represent n number of nucleotide and * represents mismatched regions. 1b shows a typical siRNA duplex with dicer activity on 21 nt region and a signature staggered cleavage pattern (arrow) with 2nt 3’ overhang. (c) Shows pingpong generated piRNA pair with a characteristic 10nt overlap. PIWI slicing positions are represented with arrows (Antoniewski, 2014). 1d is an example fastqc result showing over-represented reads that contains the adapter used for library preparation.

piRNA Biogenesis

Insect piRNAs are longer (25–30 nt) than miRNAs and siRNAs.[14] piRNAs were originally thought to be gonad restricted where they suppress transposable elements. Recently, it was discovered that most insects have somatic piRNAs that appear to regulate gene expression in addition to suppress transposable elements [15]. piRNAs are Dicer independent, 2’-Omethylated, and associate with PIWI proteins. They exhibit a bias for 5’ uracil. Two biogenesis pathways have been identified. Both involve recruitment of RNA fragments created by PIWI-mediated slicing. AGO/PIWI protein slicing occurs at the target RNA base complementary to the 10^th base of the small RNA. The first pathway is the ping-pong cycle where two PIWI proteins (AUB and AGO3) collaborate to generate piRNAs through alternating slicing (Figure 1c).[16] When a piRNA bound to AUB cleaves a transcript a fragment will load into AGO3. If a transcript is cleaved by AGO3 the resulting fragment is in-turn recruited by AUB. Thus, reads generated in this manner can be identified by sense/antisense read pairs that overlap by 10 nt. The second pathway, the phasing pathway is initiated when PIWI proteins slice transcripts which then become substrates for consecutive endonucleolytic cleavages by the zucchini/mito-PLD nuclease to produce piRNAs end to end.[17] Like miRNA and siRNA induced RNAi, piRNA-triggered RNAi have been achieved in the white fly (Bemisia tabeci), another invasive crop damaging pest that transmits plant viruses throughout the world. By using bioinformatics approaches to identify piRNA producing loci, specific piRNAs were designed and fused to aquaporin1 and alpha glucosidase1 to serve as silencing triggers against the respective targets following feeding.[18] Although, hundreds of small RNAs have been identified in animals, high-throughput sequencing and bioinformatics tools repeatedly reveal the existence of new miRNAs, siRNAs and piRNAs. Insects are one of the most diverse and successful groups, so it is not surprising that they also have diverse and elaborate RNAi pathways. This suggests in depth knowledge of a particular insect’s RNAi biology would be valuable to identifying an effective RNAi strategy. Moreover, these insights could lead to novel strategies in pest management. This chapter will provide methods and computational pipelines that can be used to identify small RNA biogenesis modes using small RNA sequencing data alone. To demonstrate tool usage, small RNA sequencing datasets from WCR will be analyzed in the protocol that follows.

2. Materials

2.1. Required Installations

• 2.1.1Numpy, python (versions 2 and 3) and pysam are required for the pipeline. The following software are also required.
• 2.1.2fastqc: https://anaconda.org/bioconda/fastqc
• 2.1.3fastx_toolkit: https://anaconda.org/bioconda/fastx_toolkit
• 2.1.4bowtie (version 1): https://anaconda.org/bioconda/bowtie
• 2.1.5samtools: https://anaconda.org/bioconda/samtools
• 2.1.6mirdeep2: https://www.mdc-berlin.de/content/mirdeep2-documentation
• 2.1.7sra-tools: https://anaconda.org/bioconda/sra-tools
• 2.1.8R-programming language: https://anaconda.org/r/r

2.2. Requirements for cloning and dsRNA production

• 2.2.1APE: Plasmid Editor https://jorgensen.biology.utah.edu/wayned/ape/
• 2.2.2Thermocycler
• 2.2.3Taq Polymerase Enzyme
• 2.2.4Phire Hot Start II DNA polymerase
• 2.2.5T7 promoter containing primers for amplification of template DNA
• 2.2.6Agarose and gel electrophoresis system
• 2.2.7GeneJET Gel Extraction Kit (ThermoFisher)
• 2.2.8pGEM®-T Easy Vector Systems (Promega)
• 2.2.9MiniPrep Kit (QIAGEN)
• 2.2.10E. coli DH5-α
• 2.2.11LB medium/agar plate with Ampicillin and LB medium liquid without Ampicillin
• 2.2.12Isopropyl β-D-thiogalactopyranoside (IPTG) solution.
• 2.2.135-bromo-3-Indolyl β-D-galactopyranoside (Blue Gal)
• 2.2.14MEGAScript Kit (ThermoFisher)
• 2.2.15Trizol or Tri-reagent

3. Methods

3.1. Preliminary Small RNA Quality assessment

• 3.1.1Initial Quality determination: After total RNA extraction with trizol, determine purity and amount with Nano drop, measure quality with bioanalyzer, determine intactness of RNA with gel electrophoresis.

3.2. Library preparation and sequencing

• 3.2.1small RNA prep kit: NEBNext small RNA prep kit, NextFlex, SMARTer, smRNA kit, CATS and TruSeq small RNA library prep kit are all commercially available.
• 3.2.2Follow kit-specific protocol for library preparation
• 3.2.3Size selection: Use bead-based selection or PAGE purification to select desired small RNA reads.
• 3.2.4Sequence reads on an Illumina sequencing platform or any other appropriate platform.

3.3. Obtaining and processing the sequencing data

• 3.3.1in this Chapter, a dataset from the Western Corn Rootworm (WCR), SRA database accession number: SRR10430839 is used to demonstrate the pipeline presented in this protocol.

  Following download, the .sra file is decompressed with the fastq-dump tool from the sra-toolkit.

3.4. Preparing the data

• 3.4.1Adapter identification: use the fastqc tool to identify over-represented sequences that will contain the adapter used for the library prep (Figure 1d). Small RNA sequencing libraries with adapter-only contamination greater than 60% should not be used for analysis.
• 3.4.2Clip adapters from the 3’ end of reads using fastx-toolkit fastx_clipper. Only retain reads for subsequent analysis that are clipped and are longer than 15nt.

3.5. Computational Identification of miRNA in RNA-seq data

• 3.6.1miRNA identification: Many tools have been developed for miRNA identification that incorporates sequencing data. This includes miRDeep2[7] and miRanalyzer [19].

  These tools are freely available and are provided with instructions for installation and use.

• 3.6.2miRDeep2: https://www.mdc-berlin.de/content/mirdeep2-documentation
• 3.6.3miRanalyzer (sRNAbench): https://arn.ugr.es/srnatoolbox/srnabench/

  To demonstrate miRNA discovery with these tools, miRdeep2 was used to identify miRNAs in Western Corn Rootworm small RNA data. Following a successful run of miRdeep2, the generated files include a html link (Figure 2a and 2b). The example novel miRNA in Figure 2b comprise three major segments, the mature (red = 5’ arm), the Star (purple = 3’ arm) and the loop region (yellow). The characteristic loop and bulges facilitate Drosha cleavage and essentially differentiating miRNA precursors from siRNA precursors. Hence, one can successfully identify miRNAs from seq-data leaving behind small RNAs (such as siRNAs) that did not result from short hairpin structures.

3.7. Computational Identification of siRNA and ping-pong piRNA expressing loci

• 3.7.1siRNAs can be identified by inspecting small RNA dataset for dual strand reads with dicer signatures. piRNA identification in sequencing datasets rely on ping-pong or phasing read identification as explained earlier.
• 3.7.2Executing the provided scripts in this order will yield siRNA and piRNA producing loci from sequencing datasets.
• 3.7.3The first process is to map a small RNA sequencing data to the relevant genome sequence. This can be done with the “mapping_smallRNAs.sh”

  Usage: ./mapping_smallRNAs.sh genome.fasta RNAseq_data.fastq

• 3.7.4The mapping script takes two inputs: a genome sequence file in multi-fasta format and sequencing data file in fastq format. The provided python script (overlapping_reads.py) is required to be in the current directory [20]. This script is needed to identify reads that exhibit 2nt 3’ overhangs for Dicer processed reads or the 10nt overlap for the pingpong piRNA signature.
• 3.7.5To identify loci that produce siRNAs and piRNA a script is provided (smallRNA1.sh). It computes files created by the mapping script, thus usage is: ./smallRNA1.sh. The script’s operations are as followed:
• 3.7.6The mapped reads are used to identify regions of higher dicer and pingpong activity by searching for loci were read depth exceed a threshold.
• 3.7.7The results are then sorted in a defined order and the best candidate loci are selected. Usually, these top loci would be most efficient for designing RNAi reagents due to the high siRNA/piRNA expression levels.
• 3.7.8The selected top loci (100 loci in this example) are individually queried for the presence of dual strand reads that meets either dicer or pingpong criteria. These are then counted and collected as read depth.
• 3.7.8The depth of the reads for all loci are then plotted as a bar graph using ggplot2 (Figure 3a, 3b)
• 3.7.8siRNA producing loci comprise regions of high dual strand activity where reads overlap perfectly leaving 2nt overhang on either 3’ ends. These loci are listed in the script output named: siRNALoci.txt (top 100 only)
• 3.7.9ping-pong piRNA producing loci represent regions of high ping pong activity where piRNA sized reads overlap each other by 10nt. This sort of overlap signature is often observed in the ping-pong activity seen when Aub and AGO3 produce piRNAs in flies. These loci are listed in the script output named: piRNALoci.txt (only top 100)

Figure 3. small RNA Biogenesis:

3a, and 3b respectively represents the top 100 high-expressed siRNA and pingpong piRNA producing loci with y-axis showing read count (nt) and x-axis showing each locus. 3c and 3d show distribution of each nt in 2 of 100 locus. 3c is an example of a non-phasing locus and 3d is a phasing piRNA locus. In 3d, arrow points to the trailing U (nt) of phasing piRNA biogenesis (usually in position 1–4 in the 5’ to 3’ direction). Red line = U, y-axis is the z-score and x-axis is the distance toward the trailing 1U.

3.8. Computational Identification of phasing piRNA in RNA-seq data

• 3.8.1Executing the following script will generate phasing piRNAs. This script will use some of the files generated earlier by the first script.
• 3.8.2Usage: ./smallRNA2phasing.sh genome.fasta
• 3.8.3Using the longer reads (20 to 30nt) that mapped to the reference genome, the reads are converted into a bed format that is then intersected with all regions where read depth exceeds a certain threshold.
• 3.8.4The result of this intersection is then sorted in descending order so that the top 100 loci are separated and queried for presence of phasing piRNAs.
• 3.8.5Each line in the resulting file (top 100) representing each locus is then passed into a python script (this should be present in the directory at the time of script execution) piPipes_nuc_percentage.py [21].
• 3.8.6The python script searches the reads in each location (locus) and count each nucleotide in each position across all reads, calculates and assign z-scores to each nucleotide position.
• 3.8.7This generates 100 files (for each locus) containing z-scores for “A”, “C”, “G” and “T”.
• 3.8.8The script then uses ggplot2 to create 100 line graphs using Z-score information of only U and all nt positions (Figure 3c, 3d).
• 3.8.9The line graphs are then scanned for the presence of a bias usually around 1– 4 position (absent in Figure 3a but present in Figure 3b) trailing the first U at the 5’ end.
• 3.8.10This is evidence of phasing biogenesis and should be considered a potential phasing piRNA locus.

3.9. Cloning and dsRNA production

Based on results from the pipeline, a choice can be made regarding the best small RNA class to exploit for RNAi technology. This decision can be based on the abundance or read depth of a class. For example, in the WCR small RNA analysis, more piRNA loci are present relative to siRNA loci (3b and 3a respectively). However, the siRNA loci appear to have greater read abundance. Widespread piRNA loci in this example could mean that piRNAs could be more active in WCR, and that RNAi reagents based on piRNAs might produce more significant gene silencing.

• 3.9.1Design primers using APE plasmid editor for SOEing PCR with sequences from the best loci obtained from small RNA analysis as well as target sequence. (see Note 4.1)
• 3.9.2Perform SOEing PCR to add the target sequence region to the small RNA using Taq polymerase Enzyme (see Note 4.2)
• 3.9.3Run on agarose gel to confirm the fragment size.
• 3.9.4Cut the band and purify DNA using the protocol described in the GeneJET Gel Extraction Kit (see Note 4.3)
• 3.9.5Perform ligation reaction using the purified DNA fragment in 3.9.4 by following the protocol in pGEM®-T Easy Vector Systems kit and use https://nebiocalculator.neb.com/#!/ligation for ligation calculation. (see Note 4.4)
• 3.9.6
  Mix the ligation reaction with competent E. coli cells. Keep on ice for 20 minutes.

  • 3.9.6.1
    Heat-shock the mixture at 42°C for 50–70 sec and immediately transfer on ice for 4 min.
  • 3.9.6.2
    Add 1mL of LB medium to the Microfuge tube and incubate on rotor at 37°C for 45 min.
  • 3.9.6.3
    Mix IPTG (100 mM:40 μL) and Blue Gal (20 mg/mL: 40 μL) then add the mixture to Ampicillin plates. Incubate plates at 37°C for about 15 minutes before to use.
  • 3.9.6.4
    Spin down the Microfuge tube at 5000 g for 5 min and discard almost all the supernatant. (see Note 4.5)
  • 3.9.6.5
    Spread transformed competent cells with sterile spreader or glass beds in LB/agar plate and incubate overnight at 37°C
  • 3.9.6.6
    Pick up a very small portion of 10 blue colonies and mix with the PCR reaction.
• 3.9.7Do PCR using Phire Hot Start II DNA polymerase and M13 forward and SP6 reversed primers.
• 3.9.8Run an agarose gel to verify the fragment size and growth the colony with the right insert in LB medium with Ampicillin.
• 3.9.9Purify the vector pGEMT easy with the insert by MiniPrep Kit and reed in Nanodrop the concentration.
• 3.9.10Use the vector purified to do a PCR with Phire Hot Start II DNA polymerase and M13 forward and SP6 reversed primers.
• 3.9.11Use the PCR product as a template to do in vitro transcription with MEGAScript Kit as well as clean up the final reaction by following the protocol.
• 3.9.12Determine concentration with nanodrop and run an agarose gel to check the integrity.

  After those steps, the dsRNA containing small RNA sequence and target sequence are ready to apply to insects through oral, injection or soaking delivery methods.

• 3.9.13Time required for drRNA processing is dependent on the insect and target of interest being investigated
• 3.9.14Total RNA extraction should be done using trizol or tri-reagent followed by gene expression analysis

4. Notes

• 4.7It is very important to use Taq Polymerase to enhance TA cloning.
• 4.8Minimal exposure of DNA to UV should be ensured to prevent DNA cross-linking.
• 4.9For better results, keep the reaction overnight.