Transcriptome-wide interrogation of RNA secondary structure in living cells with icSHAPE

Conventional RNA structure probing techniques

RNA structure is frequently measured by chemical modification or RNase digestion. Whereas RNase S1 recognizes and cuts single-stranded regions in RNA, RNase V1 will cleave at double-stranded residues⁷. Small-molecule chemical reagents can instead be used to increase the resolution of structure probing experiments. For example, dimethylsulfate (DMS) can alkylate the Watson-Crick face of adenosine and cytosine, as well as the N7 position of guanosine when not base-paired⁸. Kethoxal is used to modify the N1- and C2-exocyclic amines to reveal single-stranded residues in RNA⁹. Hydroxyl radical probing is widely used to measure RNA compaction and folding; the mechanism used cleaves RNA by hydrogen abstraction at the 5′ position¹⁰. The recent development of Selective Hydroxyl Acylation analyzed by Primer Extension, or SHAPE, has greatly extended chemical probing experiments because of its ability to interrogate every position in the RNA through 2′-OH reactivity¹¹. Overall, the field is well equipped to interrogate the structure of single RNAs.

The next step in the evolution of RNA structure probing is to transition from single-gene analysis to whole-transcriptome interrogation. A corollary can be made to our understanding of gene expression mechanisms and chromatin structure, which has greatly matured since the advent of deep sequencing¹². Knowledge of the structural content of RNA across the transcriptome will reveal unique facets of how diverse classes of RNA control the biology of the cell.

In vitro transcriptome-wide structure analysis

The first demonstration of the power of transcriptome-wide analysis of RNA structure was the development of Parallel Analysis of RNA Structure (PARS). PARS uses both single (nuclease S1)- and double-stranded (nuclease V1) nucleases to digest motifs in RNA and to generate a structure score¹³. The PARS scores from both experiments can be used to constrain the RNA structure predictions of thousands of transcripts in parallel. In a similar manner, Frag-seq uses nuclease P1 for the cleavage and identification of single-stranded residues¹⁴. A complementary method, using RNase One and RNase V1, has been used to describe the secondary structures of multiple metazoans¹⁵. The first chemical-based method to transform structural information into sequencing-based results for understanding the structures of thousands of RNAs in parallel was SHAPE-seq. In SHAPE-seq¹⁶, a pool of in vitro–transcribed RNAs are subjected to SHAPE chemical probing and deep sequencing. This approach has revealed structural changes that occur because of primary sequence alterations¹⁶. Although these methods have been very useful for determining the structure of refolded, in vitro transcripts, they are not suitable for describing the structures of RNA inside living cells.

In vivo transcriptome-wide structure analysis

DMS was the first chemical that was successfully applied to transcriptome-wide structure probing in living cells, in similar approaches called DMS-seq¹⁷, Structure-seq¹⁸ and Mod-seq¹⁹. These studies demonstrated the utility of chemical probes for understanding the structural elements in mRNAs and noncoding RNAs on a large scale. Rouskin et al.¹⁷ demonstrated that mRNA structures seem to sample many different conformations; however, highly important RNA structures, such as those involved in localization, have similar profiles within and outside the cell. In addition, they demonstrated that ATP-dependent RNA remodeling enzymes (probably helicases) were responsible for the structure probing variation. Ding et al.¹⁸ demonstrated with in vivo DMS profiling that mRNAs associated with stress responses tend to have more single-strandedness, longer maximal loop length and higher free energy per nucleotide. Such features may allow these RNAs to undergo conformational changes in response to environmental conditions. These key studies highlight important examples of information that in vivo RNA structural measurements garner about basic RNA biology. One important limitation of DMS is that it only interrogates two out of four bases in RNA, adenosine (A) and cytosine (C); hence, the structural information is incomplete. To address this limitation, we developed and previously described in vivo click-selective 2′-hydroxyl acylation and profiling experiment or icSHAPE²⁰; here we provide a detailed protocol for icSHAPE. Distinguishing features of our method include the ability to interrogate every position in an RNA, independently of its primary sequence content, and to enrich for modified RNAs to markedly increase the signal-to-noise ratio in transcriptome-wide measurements.

Development and overview of icSHAPE

icSHAPE was developed to provide both the experimental and computational tools necessary to study RNA structure in a transcriptome-wide manner. Existing high-throughput, chemical-based structure determination methods suffer from high background because of sequencing unmodified RNA^16–18. To address this issue, we took inspiration from methods such as RNA-binding protein immunoprecipitation (RIP)²¹ and individual-nucleotide resolution cross-linking and immunoprecipitation (iCLIP)²², in which the signal for RNA binding can be highly enriched by selective purification. We transformed an existing in vivo SHAPE probe 2-methylnicotinic acid imidazolide (NAI)²³ into a dual-function SHAPE probe (RNA structure probe and enrichment), by the addition of an azide to the nicotinic acid ring (at position 2), and we named this molecule NAI-N₃. Incorporation of the azido-moiety was crucial, as it allowed the ‘click’ of biotin to the SHAPE reagent using copper-free ‘click’ chemistry, which proceeds under conditions that are not detrimental to the SHAPE-modified RNA (Fig. 1a). By the addition of a biotin moiety to each RNA that was successfully SHAPE modified, we are able to subsequently enrich specifically for RNA fragments that have NAI-N₃ covalently attached. We construct icSHAPE cDNA libraries on enriched material, thereby only sequencing structurally informative regions of RNA, thus increasing our signal to noise.

An overview of the icSHAPE procedure is provided in Figure 1b. In brief, the bifunctional NAI-N₃ SHAPE reagent is used to label all RNAs inside living cells (or in vitro refolded RNA) to generate covalent modifications on targeted RNAs at single-nucleotide resolution. RNA is isolated, biotinylated using click chemistry and randomly fragmented to a size of 50–100 nt. To generate cDNA sequencing libraries, the RNA fragments are subjected to 3′-end repair, adapter ligation, size selection, biotin enrichment, reverse transcription (RT) and PCR. The end result is selective amplification of fragments corresponding to modified bases in the NAI-N₃–treated samples or background expression values from mock-treated samples. Sequencing reads generated from the NAI-N₃– and mock (DMSO)-treated samples are mapped to a transcriptome annotation, followed by computational identification of RT stops corresponding to SHAPE modifications—and hence reactive nucleotides—across the entire transcriptome (Fig. 1b). Traditionally, SHAPE reactivity directly reports on the structural flexibility of an individual nucleotide; however, in vivo icSHAPE reactivities can be influenced by protein binding²⁰, among other things.

We developed a highly modular and customizable computational pipeline to process raw icSHAPE sequencing data and to provide structural output files directly ready for interpretation. The pipeline was designed to take instructions from a centralized configuration file, and users can modify the parameters for each of the processing steps. To accomodate the numerous different strategies that laboratories use to analyze deep-sequencing data, each step taken by the icSHAPE pipeline can be achieved using alternative programs with similar functionality to those described here. The icSHAPE pipeline is capable of running in either ‘automatic’ or ‘manual’ modes, thus giving the user speed (automatic) or flexibility (manual) when implementing the analysis package. It should be noted that icSHAPE differs from regular RNA-seq experiments in that (i) icSHAPE profiles a transcript at base resolution, whereas RNA-seq experiments usually quantify transcriptomes and abundances of whole transcripts; (ii) the icSHAPE protocol generates relatively short cDNA inserts of ~20–50 nt; and (iii) in order to obtain base-resolution RNA structure profiles, the sequencing of icSHAPE should be very deep so as to generate sufficient coverage of individual bases. These features require analysis tools that can handle multiply mapping sequencing reads and that can efficiently process billions of reads with potentially limited computational resources and time.

Applications, advantages and limitations of icSHAPE

The use of NAI-N₃ provides an experimental framework for measuring the base flexibility of all four nucleotides in a transcriptome-wide experiment. This principally comes by leveraging SHAPE chemistry, whereas previous in vivo chemical methods have been restricted to interrogating only two of the four bases (As and Cs). Although integrating structural data from multiple probing strategies is ideal, icSHAPE provides a single experimental strategy to capture a global landscape of RNA structure, which is not biased for nucleotide content. An additional advantage of icSHAPE is the specific enrichment of NAI-N₃–modified RNAs; however, it is formally possible that reverse transcriptase does not always fully extend to the nucleotides modified with NAI-N₃, thereby possibly generating a false-positive signal. A recent study reported a method for mitigating these effects by digesting non-fully-extended RNA:cDNA hybrids with RNase One²⁴. Although this feature is not a part of this protocol, future designs could easily implement this concept.

The principle that ‘read counting’ allows the user to accurately and quantitatively assess the structural state of a given nucleobase is common to all sequencing-based structure determination methods. Variable amounts of starting material and differential complexity between samples are examples of parameters that contribute to variation across experiments. Generating icSHAPE (or DMS-seq) samples for sequencing requires PCR, and the above-mentioned experimental variables could contribute to differential amplification and/or overamplification of library samples. To combat this, we have incorporated a molecular barcode into the RT primer, which allows for accurate computational removal of PCR duplicates. The molecular barcode therefore allows reads with exactly the same sequence, but generated from unique RNA molecules, to be accurately counted as unique. Our barcode design allows for the precise resolution of >110,000 molecules with the exact same sequence.

Beyond comparing the reactivity profiles of specific transcripts, it is possible to use icSHAPE reactivity values to guide RNA secondary structure prediction. Software, such as RNAstructure²⁵, can incorporate icSHAPE and other chemical probing data to constrain the predicted secondary structures, thus improving the accuracy of the prediction. Future work applying novel²⁶ and integrated computational methods (reviewed in refs. 27–29) will allow for a global view of RNA secondary structure as it exists in vivo and in vitro.

Finally, icSHAPE can be applied to a wide variety of biological systems and questions. Implementation of icSHAPE in vitro and in vivo recapitulated known global structural features of mammalian transcriptomes; however, we also uncovered the ability to read out RNA-protein interactions²⁰. We demonstrated the ability to accurately predict protein-binding sites with a high true positive rate. This was accomplished for the proteins RBFOX and HuR, which are proteins that are important for regulating RNA splicing and stability, respectively²⁰. In addition, we found that using only icSHAPE reactivities accurately predicted 85% of known m⁶A chemical modifications across the mouse embryonic stem cell (ESC) transcriptome²⁰. An independent study further corroborated our finding, using focused biochemistry and traditional RNA structure footprinting³⁰. Concurrently, another group developed a biotin-conjugated SHAPE probe that enriches for RNA structural signal, although whether this probe works in living cells was not reported²⁴. Overall, these studies demonstrate the power and importance of chemical enrichment strategies, such as icSHAPE, for studying RNA structure in vivo.

Experimental design

REAGENTS

• Cells. icSHAPE has been implemented on bacterial, yeast, insect and mammalian cells²³ ! CAUTION It is important to regularly check cell cultures to ensure that they are authentic and are not infected with Mycoplasma.

• PBS (Life Technologies, cat. no. 10010049)

• Trypsin-EDTA, 0.25% (wt/vol) (Life Technologies, cat. no. 25200-114)

• Trypan blue solution, 0.4% (wt/vol) (Life Technologies, cat. no. 15250-061)

• UltraPure DNase/RNase-free dH₂O (Life Technologies, cat. no. 10977-023)

  ▲ CRITICAL All experimental procedures must be performed under DNase- and RNase-free conditions. All reagents should be made or purchased so as to be free of DNases and RNases.

• DMSO (Sigma-Aldrich, cat. no. D2650-5×10ML)

• NAI-N₃, 2 M (4-step organic synthesis is described²⁰)

• Trizol LS reagent (Life Technologies, cat. no. 10296-028)

• Chloroform (Ricca Chemical, cat. no. RSOC0020-500C) ! CAUTION Trizol and chloroform are hazardous, and caution should be taken when handling them; work should be performed in a chemical fume hood.

• Ethanol (200 proof; Gold Shield, cat. no. 412804)

• Buffer RWT (Qiagen, cat. no. 1067933)

• HEPES, 1 M (Life Technologies, cat. no. 15630-080)

• Magnesium chloride, 1 M (MgCl₂; Ambion, cat. no. AM9530G)

• Sodium chloride, 5 M (NaCl; Ambion, cat. no. AM9759)

• Click-IT biotin DIBO alkyne (Life Technologies, cat. no. C-10412)

• RiboLock RNase inhibitor (40 U/µl; Thermo Scientific, cat. no. EO0384)

• RNA fragmentation reagents (Ambion, cat. no. AM8740)

• T4 polynucleotide kinase (PNK; New England BioLabs, cat. no. M0201L)

• Tris, pH 7.0, 1 M (Ambion, cat. no. AM9850G)

• FastAP thermosensitive alkaline phosphatase (1 U/µl; Thermo Scientific, cat. no. EF0651)

• T4 RNA ligase 1 (ssRNA ligase), high concentration (New England BioLabs, cat. no. M0437M)

• Preadenylylated and 3′-biotin blocked RNA linker (DMSO samples, IDT, PAGE purified): /5rApp/AGATCGGAAGAGCGGTTCAG/3Biotin/

  ▲ CRITICAL 5′preadenylylated and 3′ blocked RNA linkers are important for efficient 3′end RNA ligations.

• Preadenylylated and 3′-ddC blocked RNA linker (NAI-N₃ samples, IDT, PAGE purified): /5rApp/AGATCGGAAGAGCGGTTCAG/3ddC/

• UltraPure DTT (Invitrogen, cat. no. 15508-013)

• SequaGel UreaGel System, 2.2 liter kit (National Diagnostics, cat. no. EC-833-2.2LTR)

• Ultrapure TEMED (Invitrogen, cat. no. 15524-010)

• Ammonium persulfate (APS; Thomas Scientific, cat. no. 0149N94)

• Ultrapure TBE buffer, 10× (Life Technologies, cat. no. 15581-044)

• EDTA (Ambion, cat. no. AM9260G)

• UltraPure 10% (wt/vol) SDS solution (Life Technologies, cat. no. 15553-027)

• SuperScript III reverse transcriptase (Invitrogen, cat. no. 18080-044)

• Deoxynucleotide solution mix, 10 mM (dNTPs; New England BioLabs, cat. no. N0447L)

• RT-primer-1 (IDT, standard desalting purification, underlined portion is the experimental barcode³¹) /5phos/DDDNNAACCNNNNAGATCGGAAGAGCGTCGTGGA/iSp18/GGATCC/iSp18/TACTGAACCGC; D = A/G/T and N = A/T/G/C are used to discriminate PCR duplicates, ‘AACC’ is the specific experimental barcode) ▲ CRITICAL The quality of the oligonucleotide synthesis should be verified by PAGE analysis; however, bulk PAGE purification before use in RT reactions is not necessary.

• Dynabeads MyOne streptavidin C1 (Life Technologies, cat. no. 65002)

• Tween 20 (Sigma-Aldrich, cat. no. P1379-500ML)

• UltraPure 1 M Tris-HCI buffer, pH 7.5 (Invitrogen, cat. no. 15567-027)

• Nonidet P40 (NP-40) (or substitute, IGEPAL CA-630; Roche, cat. no. 11332473001)

• N-Lauroylsarcosine sodium salt solution (20%, for molecular biology; Sigma-Aldrich, cat. no. L7414-10ML)

• Deoxycholic acid sodium salt (Fisher Scientific, cat. no. BP349-100)

• D-Biotin (Molecular Probes, cat. no. B20656)

• RNase cocktail enzyme mix (Ambion, cat. no. AM2286)

• RNaseH (Enzymatics, cat. no. Y9220L)

• Isopropanol (Molecular Biology Grade; Fisher Scientific, cat. no. BP2618-1)

• DNA ladder, 25 bp (Invitrogen, cat. no. 10597-011)

• Gel loading buffer II (Denaturing PAGE; Ambion, cat. no. AM8547)

• SYBR Gold nucleic acid gel stain, 10,000× (Life Technologies, cat. no. S-11494)

• BlueJuice gel loading buffer, 10× (Invitrogen, cat. no. 10816-015)

• 40% (wt/vol) acrylamide/Bis solution, 29:1 (Bio-Rad, cat. no. 161-0146)

• CircLigase II ssDNA ligase (Epicentre, cat. no. CL9025K)

• Phusion high-fidelity (HF) PCR master mix with HF buffer (New England BioLabs, cat. no. M0531L)

• SYBR Green I nucleic acid gel stain, 10,000× (Life Technologies, cat. no. S-7563)

• P5-Solexa PCR primer (IDT, PAGE purified):

• 5′-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT-3′

• P3-Solexa PCR primer (IDT, PAGE purified): 5′-CAAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCCTG CTGAACCGCTC TTCCGATCT-3′

EQUIPMENT

• Tissue culture plates (Denville, cat. no. T1115)

• Pipette tips (Denville, cat. nos. P1733, P1734, P1735, P1737)

• Micropipettors (Rainin, cat. nos. 17014382, 17014384, 17014388)

• Automatic pipettor (Drummond Pipet-aid, cat. no. 53498-001)

• Conical tubes, 15 and 50 ml (Fisher, cat. nos. 05-538-59A, 07-203-510)

• Filter, 500 ml, 0.22 µm (Millipore, cat. no. SCGVU05RE)

• Inverted light microscope (Leica DMi1)

• Hemocytometer (Reichert Bright-line)

• Cell culture hood (Thermo Forma, Class IIA/B3 Biosafety Cabinet)

• Tabletop centrifuge (Beckman Coulter Allegra 6R Centrifuge, cat. no. 366816)

• Benchtop centrifuge (Eppendorf Centrifuge 5424R)

• Cell scrapers (GeneMate, cat. no. T-2443-4)

• Razor blades (VWR, cat. no. 55411-050)

• Hybridization incubator (SciGene Model 400, cat. no. 1040-60-1)

• Vortexer (Scientific Industries Vortex Genie 2, cat. no. SI-0236)

• RNeasy mini and midi kits (Qiagen, cat. nos. 74106 and 75144)

• Siliconized microcentrifuge tubes, 1.5 ml (MidSci, cat. no. AVSC1510)

• Freeze dryer/lyophilizer (Labconco, cat. no. 7522700)

• NanoDrop spectrophotometer (Thermo Scientific)

• Poly(A)Purist MAG kit (Ambion, cat. no. AM1922)

• Thermomixer (Eppendorf)

• Rotator (Thermo LabQuake, with clip bar, cat. no. 400110Q)

• PCR machine (Eppendorf Mastercycler Pro, cat. no. 6321 000.515)

• Amicon Ultra-0.5 Centrifugal Filter Unit (Millipore, cat. no. UFC5010BK)

• High-voltage electrophoresis power supply (Bio-Rad, cat. no. 164–5056)

• PAGE apparatus (CSB Scientific, cat. no MGV-102-U)

• Glass plates, 8 × 10 cm (R. Shadel, custom order)

• Combs, 0.75-mm-thick (R. Shadel, custom order)

• Gel spacers, 0.75-mm-thick (R. Shadel, custom order)

• Steel heating plate, 8 × 10 cm (R. Shadel, custom order)

• Binder clips, 2 inch (OfficeMax, cat. no. 09015772)

• Milli-Q filtered water (EMD Millipore Milli-Q, for water purification)

• Scalpels (Fisher, cat. no. 14-823-45)

• Blue light transilluminator (Clare Chemical Research Dark Reader, cat. no. DR22A)

• Tubes, 0.5 ml (BioExpress GeneMate, cat. no. C-3267-1)

• Tubes, 2 ml (BioExpress GeneMate, cat. no. C-3261-1)

• Corning Costar Spin-X centrifuge tube filters, 0.45 µm (Sigma-Aldrich, cat. no. CLS8162-96)

• PCR tubes, 0.2 ml (Eppendorf, cat. no. 951010022)

• DNA Clean & Concentrator-5 columns (Zymo Research, cat. no. D4013)

• RNA Clean & Concentrator-5 (Zymo Research, cat. no. R1016)

• Plastic wrap (E&K Scientific, cat. no. EK-1344)

• Gel imager (Bio-Rad Gel Doc XR+, cat. no. 1708195)

• qPCR machine (Stratagene Mx3005p, capable of mid reaction removal of individual samples)

• Optically Pure 0.2-ml PCR tubes (Agilent Life Sciences, cat. no. 401428)

• Magnetic stand, for 1.5-ml tubes (Invitrogen DynaMAG-2, cat. no. 12321D)

• Magnetic stand, for 0.2-ml tubes (ThermoFisher Magnetic Stand-96, cat. no. AM10027)

REAGENT SETUP

EQUIPMENT SETUP

Cell culture ● TIMING ~10 min

• 1|

  Grow adherent or suspension cells of choice under standard conditions to ~70–80% confluency. Count the number of cells per plate.

• 2|

  Collect adherent cells as described in Option A. Collect suspension cells as described in Option B.

  ▲ CRITICAL STEP It is important not to trypsinize the plates that will subsequently be used for icSHAPE experiments, as we have found that this results in partially degraded total RNA.

  1. Collection of adherent cells

     1. Remove the culture medium and rinse the cells once with 1× PBS.
     2. Remove PBS and add 1.5 ml or 4 ml of 1× PBS to a 10-cm or 15-cm plate, respectively.
     3. Scrape the cells off the plate and prepare aliquots of 20 × 10⁶ cells per 15-ml conical tube.
     4. Centrifuge each 15-ml tube (1,000g for 2 min at 25 °C) and move directly to Step 3.
  2. Collection of suspension cells

     1. Homogenize the cell suspension by gentle swirling of the culture.
     2. Prepare aliquots of 20 × 10⁶ cells per 15-ml conical tube.
     3. Centrifuge each 15-ml tube (1,000g for 2 min at 25 °C) and move directly to Step 3.

icSHAPE modification and quality control ● TIMING ~2 h

• 3|

  Aspirate PBS from the pelleted cells (from Step 2), resuspend the cell pellet in 500 µl of 100 mM NAI-N₃ solution or 500 µl of DMSO solution with a cut-tip p1000 and incubate at 37 °C with rotation for 5 min.

  ▲ CRITICAL STEP All NAI-N₃ experiments must be paired with a DMSO experiment. DMSO will not modify the RNA, and thus it provides a background signal for determining the coverage level of any given nucleotide in the sequencing experiment.

  ▲ CRITICAL STEP If you are intending to compare in vivo and in vitro icSHAPE reactivity profiles, which we have shown can reveal biological events such as protein imprints²⁰, an additional DMSO sample should be included for subsequent in vitro modification at Step 17 (Box 2).

  ▲ CRITICAL STEP At least two independent biological replicates should be performed for each condition.

• 4|

  Stop the reaction by centrifuging the samples (2,500g for 1 min at 4 °C) and removing the supernatant.

• 5|

  Resuspend the cells in 500 µl of 1× PBS, and then transfer the cell solution to 6 ml of Trizol LS in a 15-ml tube.

• 6|

  Vortex the samples for ~15 s to homogenize, add 1.2 ml of chloroform (0.2× volumes), vortex again for ~15 s, and then let the samples incubate at 25 °C for 5 min.

• 7|

  Centrifuge the samples (2,500g for 15 min at 4 °C), move the aqueous phase to a new 15-ml tube, add 2× volumes of 96–100% ethanol, and vortex to mix.

• 8|

  Purify RNA using an RNeasy midi kit column according to the manufacturer’s protocol (starting at step 4 of the ‘RNA Cleanup’ section).

• 9|

  Elute RNA from the RNeasy midi column twice with 250 µl of water (final of 500 µl volume).

• 10|

  Quantify RNA yield, make 2-µg aliquots for quality control, and then freeze the RNA in 250-µg aliquots at −80 °C for storage.

• 11|

  Assess RNA quality by agarose gel (1%, wt/vol) analysis. Clear 18S and 28S rRNA bands should be present.

  ? TROUBLESHOOTING

  ■ PAUSE POINT Purified RNA can be stored at −80 °C for >6 months.

Isolation of poly(A)⁺ RNA fraction ● TIMING ~5 h to overnight

• 12|

  For each biological replicate, thaw one 250-µg aliquot of DMSO-treated RNA and in vivo NAI-N₃–modified RNA (from Step 11) on ice. If you are intending to include in vitro NAI-N₃-modified RNA in the analysis, thaw an additional DMSO-treated sample, which will be NAI-N₃–modified at Step 17 after polyA⁺ selection. For a standard experiment aimed at fully characterizing the RNA structure landscape, at least two biological replicates are needed, giving a total of six samples (2× DMSO, 2× in vivo NAI-N₃ and 2× in vitro NAI-N₃); however, additional biological replicates can be added for higher confidence.

• 13|

  Isolate poly(A)⁺ RNA using the Poly(A)Purist MAG kit or an equivalent enrichment procedure. Bring each sample to a final elution volume of 400 µl. When using the Poly(A)Purist MAG kit, this will be 400 µl of RNA storage solution (part of the kit).

• 14|

  Wash the poly-dT beads once with 500 µl of wash solution 1 and reuse the beads to perform a second poly(A)⁺ enrichment, without purifying the first-step enriched poly(A)⁺ RNA out of the RNA storage solution.

  ▲ CRITICAL STEP We perform two rounds of poly(A)⁺ RNA enrichment. This ensures the lowest amount of rRNA contamination while still yielding high amounts of poly(A)⁺ RNA for subsequent biochemical reactions.

• 15|

  After the second poly(A)⁺ enrichment, freeze each RNA sample (~400 µl each) for >10 min on dry ice and dry it on a lyophilizer (overnight for convenience).

• 16|

  Resuspend the enriched poly(A)⁺ RNA in 100 µl of water, add 350 µl of buffer RLT, 900 µl of 100% ethanol, and then clean up using an RNeasy mini column. Elute the RNA twice with 50 µl of water (100-µl final volume).

• 17|

  Quantify poly(A)⁺ RNA and make 500-ng aliquots into individual 1.5-ml tubes. For each icSHAPE library reaction, 400–500 ng of RNA is needed for successful library generation.

  ▲ CRITICAL STEP At this point, two DMSO-treated samples from Step 17 can be used for in vitro NAI-N₃ modification, as described in Box 2.

  ■ PAUSE POINT Poly(A)⁺ RNA can be stored at −80 °C for >6 months.

Biotin click reaction of NAI-N₃ ● TIMING ~3 h

• 18|

  Thaw purified DMSO- and NAI-N₃–modified poly(A)⁺ RNA (from Step 17 and Box 2) on ice, and then add 2 µl of 1.85 mM DIBO-biotin and 1 µl of RiboLock.

• 19|

  Incubate the samples at 37 °C in a Thermomixer for 2 h at 1,000 r.p.m. Under these conditions, the reaction will proceed to completion with all of the SHAPE adducts finally obtaining a biotin handle²⁰.

• 20|

  Stop the click reaction by purifying the RNA with a Zymo RNA Clean & Concentrator-5 column according to the manufacturer’s protocol. Elute RNA twice in 10 µl of water, and freeze-dry it on a lyophilizer.

  ■ PAUSE POINT Biotinylated RNA and DMSO samples can be stored at −80 °C for >6 months.

RNA fragmentation and RNA end repair ● TIMING ~1.5 h

• 21|

  Resuspend the RNA samples from Step 20 in 9 µl of water, preheat a PCR block to 95 °C and move RNA to a PCR tube.

• 22|

  Heat the RNA samples to 95 °C for 60 s, add 1 µl of 10× fragmentation reagent and incubate for 30–60 s.

  ▲ CRITICAL STEP RNA fragmentation time should be determined empirically by urea PAGE gel analysis. Once it is optimized, fragmenting conditions are highly reproducible. To facilitate accurate timing, only two samples should be processed in parallel for RNA fragmentation.

• 23|

  Add 1 µl of 10× stop solution to each sample and immediately place the samples on ice.

• 24|

  Purify fragmented RNA with a Zymo RNA Clean & Concentrator-5 column; freeze-dry as in Step 20.

  ■ PAUSE POINT Freeze-dried RNA samples can be stored at −80 °C for >6 months.

• 25|
  Resuspend RNA in 10 µl of RNA end-repair mix (tabulated below) and incubate it at 37 °C for 1 h.

  Component	Amount (µl)	Final concentration
  T4 PNK buffer, 5×	2	1×
  RiboLock, 40 U/µl	1	4 U/µl
  FastAP 1 U/µl	1	0.1 U/µl
  T4 PNK 10 U/µl	2	2 U/µl
  RNase-free water	4
  Total	10 (per reaction)

  Open in a new tab

  ■ PAUSE POINT End-repaired RNA samples can be stored at −80 °C for ~1 month.

RNA 3′-end ligation and size selection ● TIMING ~5 h

• 26|
  Add 10 µl of 3′ RNA ligation mix (tabulated below) to each sample from Step 25, mix by flicking, briefly centrifuge, and incubate at 25 °C for 3 h.

  Component	Amount (µl)	Final concentration
  RNA ligase buffer, 10×	1	1×
  DTT, 100 mM	1	5 mM
  Preadenylylated and 3′-blocked RNA linker, 10 µM	1	0.5 µM
  T4 RNA ligase 1, high concentration	1	0.66 U/µl
  PEG8000, 50% (vol/vol)	6	15%
  Completed end-repair reaction	10
  Total	20 (per reaction)

  Open in a new tab

  ▲ CRITICAL STEP DMSO-treated RNA samples require a 3′-biotin RNA linker, whereas NAI-N₃–treated RNA samples require a 3′-ddC RNA linker. This ensures that every RNA molecule from the DMSO sample is recovered, whereas only those RNAs that were icSHAPE-modified are recovered from the NAI-N₃ sample.

• 27|

  Add 30 µl of water to each sample (total volume of 50 µl) and purify RNA with a Zymo RNA Clean & Concentrator-5 column; freeze-dry the sample as in Step 20.

• 28|

  Resuspend RNA in 7 µl of gel loading buffer II, and flick the tube to mix the samples.

• 29|

  Prerun 6% (wt/vol) UreaGel denaturing PAGE, first at 25 W for 3 min and then at 50 W for 5 min. Use a metal (e.g., steel or aluminum) heating plate to evenly disperse the heat across the gel.

• 30|

  Load RNA samples onto the gel, and run them at 50 W for 4–10 min.

  ▲ CRITICAL STEP These pre-run and run conditions apply to home-poured mini-gels (~8 × 10 cm); longer times will be required to sufficiently heat up larger gels. The duration of the gel run should be sufficient to separate the 3′ RNA linker (~20 nt) from ligated RNA products (more than ~40 nt). However, excessive running times will require large gel slices to be cut at Step 32 and could lower the efficiency of RNA elution.

• 31|

  Stain the gel in SYBR Gold mix for 3 min at 25 °C and image for 0.3 s on a UV transilluminator for record keeping.

• 32|

  Move the gel to a blue light transilluminator, cut the samples above ~45 nt (Fig. 2a) and move the gel slices into 0.65-ml tubes nested in 2-ml tubes. Each 0.65-ml tube should be prepared with a needle (18–22 G) hole at the bottom of the tube.

• 33|

  Centrifuge the samples (16,000g for 5 min at 25 °C) and discard the 0.65-ml tube.

• 34|

  Add 250 µl of water to each sample and heat it to 70 °C for 10 min, with shaking at 1,000 r.p.m. in a Thermomixer.

• 35|

  Pulse-centrifuge each sample, transfer the gel slurry into a Spin-X column with a cut-tip p1000 and centrifuge (6,500g for 1 min at 25 °C) the sample.

• 36|

  Resuspend the gel in the top chamber of the Spin-X column with 250 µl of water, and repeat Steps 34 and 35 to obtain a final 500-µl elution volume. Reuse the Spin-X columns for the second elution.

• 37|

  Transfer the 500-µl eluate to a 0.5-ml 10K Amicon column and centrifuge the sample (13,500g for 5 min at 25 °C). This should concentrate the sample to ~100 µl.

• 38|

  Elute the samples as per the manufacturer’s instructions: flip the Amicon column over into a new 2-ml elution tube and centrifuge the samples (2,000g for 2 min at 25 °C).

• 39|

  Purify RNA with a Zymo RNA Clean & Concentrator-5 column; freeze-dry it as in Step 20.

  ■ PAUSE POINT Freeze-dried RNA samples can be stored at −80 °C for >6 months.

cDNA synthesis, isolation of NAI-N₃–modified molecules, and cDNA purification ● TIMING ~4.5 h

• 40|

  Resuspend RNA samples from Step 39 in 11 µl of RNase-free water and 1 µl of 1 µM RT primer; flick the tubes to mix, and move the samples into PCR tubes.

• 41|

  Heat the samples to 70 °C for 5 min in a PCR block, cool the samples to 25 °C by stepping down 1 °C every 3 s (45 steps); hold at 25 °C.

• 42|
  Add 8 µl of RT enzyme mix (tabulated below) and heat the samples to 25 °C for 3 min, 42 °C for 5 min, 52 °C for 30 min; hold at 4 °C.

  Component	Amount (µl)	Final concentration
  First strand buffer, 5×	4	1×
  RiboLock, 40 U/µl	0.75	1.5 U/µl
  DTT, 100 mM	1	5 mM
  dNTPs, 10 mM (10 mM each)	1	0.5 mM
  SuperScript III	1.25	12.5 U/µl
  Total	20 (per reaction)

  Open in a new tab

  ▲ CRITICA STEP Do not heat samples above 37 °C from this point until cDNA elution at Step 50. Premature heating could cause unwanted elution of samples.

  ■ PAUSE POINT Extended cDNA samples can be stored at −80 °C for ~1 month.

• 43|

  Add 10 µl of streptavidin bead solution, mix by pipetting up and down until the sample is homogeneous, and then incubate the sample at 25 °C for 45 min with rotation.

• 44|

  Apply the samples to DynaMag-96 side magnet for 1 min, and remove 25 µl of supernatant.

• 45|

  Add 100 µl of 4 M SHAPE wash buffer to each sample and transfer the samples to 1.5-ml siliconized tubes. Add an additional 400 µl (500 µl final) of 4 M SHAPE wash buffer and invert the tubes four times to mix.

• 46|

  Apply the samples to a DynaMag2 magnet for 1 min, and then remove the supernatant.

• 47|

  Add 500 µl of 4 M SHAPE wash buffer to each sample and repeat Steps 46 and 47 four more times (a total of 5 washes).

• 48|

  Add 500 µl of 1× PBS to each sample and repeat once more (a total of two washes).

• 49|
  Resuspend the beads in 50 µl of RNaseA/T1/H elution mix (tabulated below) and incubate them at 37 °C for 30 min at 1,000 r.p.m. in a Thermomixer.

  Component	Amount (µl)	Final concentration
  RNaseH buffer, 10×	5	1×
  D-biotin, 50 mM	12.5	12.5 mM
  P3-Solexa-PCR primer, 1 µM	1	20 mM
  RNase cocktail enzyme mix	1
  RnaseH	1	0.1 U/µl
  RNase-free water	30.5
  Total	50 (per reaction)

  Open in a new tab
• 50|

  Add 1 µl of 100% DMSO to each sample and heat to 95 °C for 4 min. Pulse-centrifuge the samples, apply them to DynaMag2 magnet for 1 min, and transfer the eluted cDNA to new 1.5-ml siliconized tubes.

• 51|

  Purify cDNA with a Zymo DNA Clean & Concentrator-5 column. Add a 7× volume (350 µl) of DNA binding buffer, mix by vortexing, and then add 350 µl (1 volume) of 96–100% ethanol and mix by vortexing. Wash the columns according to the manufacturer’s protocol. Elute the cDNA twice with 10 µl of water (final of 20 µl) and freeze-dry as in Step 20.

  ▲ CRITICAL STEP Use of the Zymo DNA Clean & Concentrator-5 columns under the conditions detailed in Step 51 will recover single-stranded DNA (cDNA) as small as 20 nt. Standard binding conditions recover DNA of 50 bp or larger, which would result in a loss of short single-stranded cDNA molecules.

  ■ PAUSE POINT Biotin-enriched cDNA samples can be stored at −80 °C for >6 months.

• 52|

  Resuspend cDNA in 7 µl of Gel loading buffer II, and flick to mix the samples.

• 53|

  Prerun 6% (wt/vol) denaturing PAGE, first at 25 W for 3 min, and then at 50 W for 5 min. During the pre-run, boil samples for 2 min at 95 °C, and then place them on ice for 1 min.

• 54|

  Load cDNA samples onto the gel and run them at 50 W for 4–10 min.

• 55|

  Stain, image and cut the gel as above for the RNA size selection (Steps 31 and 32). See Figure 2b for representative results.

  ▲ CRITICAL STEP It is important to separate the cDNA products from any excess unextended RT primer, which will run at ~50 nt. Isolation of cDNA >70 nt (~20 nt of cDNA and ~50 nt of RT primer) will provide sufficient insert sequence length for high-confidence mapping to mammalian genomes.

• 56|

  Crush the gel slices, elute cDNA, filter over a Spin-X and concentrate with 10K Amicon columns, as in Steps 37 and 38.

• 57|

  Purify the concentrated cDNA samples with a Zymo DNA Clean & Concentrator-5 column, as in Step 51. Elute cDNA twice in 8.5 µl of water (final of ~16 µl). Transfer the cDNA to PCR tubes.

  ■ PAUSE POINT PAGE-purified cDNA samples can be stored at −80 °C for >6 months.

cDNA circularization, library amplification and size selection ● TIMING ~5 h

• 58|
  Add 4 µl of circularization reaction mix (tabulated below) to each sample from Step 57, and incubate them at 60 °C for 2 h.

  Component	Amount (µl)	Final concentration
  CircLigase II buffer, 10×	2	1×
  MnCl2, 50 mM	1	2.5 mM
  CircLigase II	1	5 U/µl
  Total	20 (per reaction)

  Open in a new tab
• 59|

  Purify cDNA with a Zymo DNA Clean & Concentrator-5 column as in Step 51 and elute twice in 10 µl of water (final of ~20 µl). Transfer cDNA to optical PCR tubes.

  ■ PAUSE POINT Purified circularized cDNA samples can be stored at −80 °C for >6 months.

• 60|
  Add 21.4 µl of qPCR reaction mix (tabulated below) to each circularized cDNA sample.

  Component	Amount (µl)	Final concentration
  Phusion HF PCR master mix, 2×	20	1×
  SYBR Green I, 25×	0.4	0.25×
  Solexa PCR primer, 10 µM (10 µM forward and 10 µM reverse)	1.0	0.5 µM
  Total	21.4 (per reaction)

  Open in a new tab
• 61|
  Perform quantitative PCR (qPCR) on the samples from Step 60. Track the amplification of each library in real time on the qPCR machine and stop each sample as is necessary.

  Cycle number	Denature	Anneal	Extend
  1	98 °C, 45 s	—	—
  2–n	98 °C, 15 s	65 °C, 20 s	72 °C, 60 s

  Open in a new tab

  ▲ CRITICAL STEP To avoid overamplification, real-time qPCR is used to monitor the progress of each library individually. Samples become sufficiently amplified after reaching a fluorescence unit value of ~8,000–10,000 on the Agilent Mx3000P real-time PCR system. Amplification values for machines need to be determined empirically; however, for DMSO- and NAI-N₃–treated samples, n is expected to be cycles 5–8 and 8–11, respectively.

  ? TROUBLESHOOTING

• 62|

  Purify PCR DNA with a Zymo DNA Clean & Concentrator-5 column by adding 5× DNA binding buffer, according to the manufacturer’s protocol, eluting twice in 10 µl (final of ~18 µl) and freeze-drying the samples.

• 63|

  Load the samples onto a 6% (wt/vol) native PAGE gel and run them at 160 V for ~25 min.

• 64|

  Stain the gel in SYBR Gold mix for 3 min at 25 °C and image for 0.3 s on a UV transilluminator for record keeping.

• 65|

  Move the gel to a blue light transilluminator and cut the samples above ~150 bp. See Figure 2c for representative results.

  ▲ CRITICAL STEP Library PCR may generate other products, such as primer-dimer species. These will run at ~120–130 bp and should be avoided; contamination of library material with primer-dimer will lower the clustering density of the final icSHAPE library on deep-sequencing platforms.

  ? TROUBLESHOOTING

• 66|

  Crush the gel slices, elute the DNA twice and Amicon-concentrate the PCR DNA as in Steps 37 and 38.

• 67|

  Clean up the PCR samples with a Zymo DNA Clean & Concentrator-5 column as in Step 51, and elute twice in 13 µl of water (final of ~24 µl).

• 68|

  Quantify 1 µl of each library using a high-sensitivity Bioanalyzer chip, pool the desired samples for multiplexing if necessary, and sequence icSHAPE libraries on an Illumina-based machine using standard conditions. icSHAPE libraries are paired-end compatible; however, 75-bp single-end sequencing is sufficient to identify RT stops and to confidently map to mammalian transcriptomes. As the random and experimental barcodes are ‘in-line’, the standard Illumina 6- or 8-nt barcode read should be omitted.

icSHAPE data analysis ● TIMING ~1–2 d

• 69|
  Obtain the raw sequencing reads from Step 68 and unzip if necessary. Sequencing reads usually will be in the FASTQ file format. Sometimes they are zipped to save space (as in our example .fastq file, provided at https://github.com/qczhang/icSHAPE/blob/master/data/test/hiseq.nanog.fastq). If this is the case, they need to be unzipped by using the ‘unzip’ command (or ‘gunzip’ if the file is ‘gzipped’, which is usually implied by the suffix ‘.gz’):

  $ unzip example.fastq.zip
  

• 70|
  If the sequencing data contain multiple libraries, they should be demultiplexed according to their barcodes. To split the sequencing data, use the following command:

  $ splitFastq.pl -U your_fastq_file1:your_fastq_file2:…:your_fastq_fileN −l
  
  GGTT:LIB_DMSO1::TTGT:LIB_DMSO2::ACCT:LIB_NAI1.invivo::CAAT:LIB_NAI2.invivo::TGGC:LIB_N
  
  AI1.invitro::GGTC:LIB_NAI2.invitro::others:unmatched -b 5:4 -d library_split -s
  
  hiseq_barcode.stat
  

  ▲ CRITICAL STEP In this example, single-end sequencing reads are contained in N files, and they actually belong to two DMSO-treated samples and four NAI-N₃–modified samples (six total samples). The two DMSO libraries contain the barcodes GGTT and TTGT, and the four NAI-N₃ libraries contain the barcodes ACCT, CAAT, TGGC and GGTC at positions 5–8 (-b 5:4). The split files will be stored in the directory library_split, and the statistics on the number of reads of different barcodes are stored in the file hiseq_barcode.stat.

• 71|
  Check the quality of the reads from the FASTQ files using a sequencing quality control program such as FastQC. This tool can be used to produce summary graphs and tables that inform the user of potential problems in the data. For example, the software can readily identify low-quality sequence, and it usually suggests that the user trim the read to exclude this region. Use the following command to run FastQC:

  $ fastqc -o out_directory libX.fastq
  

• 72|
  Run the icSHAPE pipeline either automatically, as described in Step 72A, or manually, as described in Step 72B.

  1. Running the icSHAPE pipeline ‘automatically’

     1. Run the pipeline by typing the following example command:

        $ icSHAPE_pipeline.pl -i LIB_DMSO1.fastq:LIB_DMSO2.fastq -t
        
        LIB_NAI1.invivo.fastq:LIB_NAI2.invivo.fastq -o test -c configure
        

        Here LIB_DMSO1.fastq and LIB_DMSO2.fastq are two replicates of the DMSO (input) background libraries, and LIB_NAI1.invivo.fastq and LIB_NAI2.invivo.fastq are two replicates of the in vivo NAI-N₃–modified libraries. The configure file is a tab-delimited text file, and it contains information for the input data, and on where to look for specific programs and their associated parameters. For example, the key ‘FASTQCODING’ defines the coding of input FASTQ reads, and the key ‘COLLAPSEBIN’ specifies which program to use to remove PCR duplicates. Although many important keys will be discussed in the following sections, please see the associated ‘readme’ file (https://github.com/qczhang/icSHAPE/blob/master/README.md) for a complete description of the icSHAPE pipeline.
        The pipeline will create a directory called ‘test’ and put all intermediate files and final output into the directory. Please see Figure 1b for the workflow of the pipeline and Table 1 for the intermediate files generated in each step.
        The final output, ‘icSHAPE.out’, contains the filtered valid structural profile for each transcript. All other files are intermediate results, and in principle they could be deleted to save space after the job is accomplished. We suggest keeping some important intermediate files (denoted with an asterisk (*) in Table 1) that will allow rapid rerunning of the icSHAPE pipeline under different parameter settings. In this case, you do not need to run the whole pipeline from the beginning to the end and can choose specific steps to run.
        ▲ CRITICAL STEP Normally, if the pipeline is correctly configured, the ‘automatic’ running mode will run smoothly and generate the final results. However, incorrect configuration or error input data could result in failure of the pipeline. Each step in the pipeline is important for completion of the analysis and generation of structure profiles. The ‘manual’ implementation (Step 72B) can be used to perform each step individually if errors occur during the ‘automatic’ running mode.
  2. Running the icSHAPE pipeline ‘manually’

     1. Remove PCR duplicates for each library with the following command:

        $ readCollapse.pl –U libX.fastq –o libX.rmdup.fastq –f libX.fa
        

        Here the input file ‘libX.fastq’ is a single-end sequencing data set and the output file ‘libX.rmdup.fastq’ contains collapsed reads after PCR duplicate removal. The frequency of each sequence in the input file is recorded in the ‘libX.fa’ file. Please check ‘readCollapse.pl’ for more options.
     2. Trim adapter and barcode sequences from reads with the following command:

        $ trimming.pl –U libX.rmdup.fastq –o libX.trimmed.fastq –l 13 –t 0 –c phred33 –
        
        a adapter.fa –m 25
        

        Here the input file ‘libX.rmdup.fastq’ is the output-collapsed sequence file from the Step 72B(i), and the output file ‘libX.trimmed.fastq’ contains reads after trimming. This command will trim the leading 13-nt sequence (barcodes used in the current icSHAPE protocol) and the adapter sequence in the file ‘adapter.fa’. If the remaining read is <25 nt, it will be discarded. The default adapter trimmer used in icSHAPE is ‘Trimmomatic’, but it can be easily replaced by other tools such as ‘cutadapter’ (https://cutadapt.readthedocs.org/en/stable/).
     3. Map the reads to the transcriptome with the following command:

        $ bowtie2 –U libX.trimmed.fastq –S libX.sam –x mouse.transcriptome --non-
        
        deterministic --time
        

        Here the input file ‘libX.trimmed.fastq’ is the output-trimmed sequence file from Step 72B(ii), ‘libX.sam’ is the output alignment file and ‘mouse.transcriptome’ should point to the location of a Bowtie2 index.
        ? TROUBLESHOOTING
     4. Estimate transcript abundance with the following command:

        $ estimateRPKM.pl –i libX.sam –o libX.rpkm
        

        Here ‘libX.sam’ is the alignment file from the Step 72B(iii), and ‘libX.rpkm’ is the output transcript abundance file in the RPKM metric³².
     5. Calculate the RT stop in each transcript with the following command:

        $ calcRT.pl –i libX.sam –o libX.rt –r libX.rpkm –c 1
        

        Here ‘libX.sam’ and ‘libX.rpkm’ are the alignment file and the transcript abundance file from the Step 72B(iii) and 72B(iv), respectively, and ‘libX.rt’ is the output RT-stop file. By setting ‘-c 1’, transcripts of RPKM <1 will be skipped in the calculation.
     6. Merge background (DMSO) and target (NAI-N₃) libraries into two separate files with the following command:

        $ combineRTreplicates.pl –i libX.rt:libY.rt:libZ.rt –o background.rt
        

        Here ‘libX.rt’, ‘libY.rt’ and ‘libZ.rt’ are the input RT-stop files of the background libraries, and ‘background.rt’ is the output combined RT-stop file from Step 72B(v). Repeat this also to combine and obtain ‘target.rt’, the target-modified RT-stop file.
        ▲ CRITICAL STEP It is important to make sure that the replicates agree with each other well before combining them. Use the script ‘correlation.pl’ to calculate the correlation of RT-stop profiles of different replicates:

        $ correlationRT.pl −1 libX.rt −2 libY.rt 1> libX.Y.corr.txt
        

        For each common transcript in libX and libY, the script will calculate their correlation of RT-stop profiles. Investigate and/or plot the distribution with statistical programs such as R (http://www.r-project.org/) to check the overall correlation between the two replicates. The script has other parameters that allow calculation of correlation of specific regions of a transcript. Explore them by checking the help from the command line.
        ? TROUBLESHOOTING
     7. Normalize the combined background and target RT-stop files with the following command:

        $ normalizeRTfile.pl -i background.rt -o background.normalized.rt -m
        
        mean:vigintile2 -d 32 −l 32
        

        Here ‘background.rt’ is the input-combined background RT-stop file from the Step 72B(vi), and ‘background.normalized.rt’ is the output-normalized background RT-stop file. The other parameters normalize the mean of the RT stops of the 90–95% most reactive bases to 1 and scale all other RT stops proportionally. The leading 32 nt (‘-d 32’) and tailing 32 nt (‘-l 32’) will be excluded in the calculation of normalizing benchmarks.
     8. Calculate enrichment reactivity scores for all transcripts with the following command:

        $ calcEnrich.pl -f target.normalized.rt -b background.normalized.rt -o
        
        icSHAPE.tmp.out -w factor5:scaling1
        

        Here ‘target.normalized.rt’ and ‘background.normalized.rt’ are the input-normalized target and background RT-stop files from the Step 72B(vii), and ‘icSHAPE.tmp.out’ is the output file that contains enrichment scores. Setting ‘-w factor5:scaling1’, the enrichment reactivity score is scaled into the range of [0, 1], after removing the outliers by 90% Winsorization (the top 5th percentile is set to 1 and the bottom 5th percentile is set to 0).
     9. Filter valid enrichment reactivity scores for regions passing a certain threshold with the following command:

        $ filterEnrich.pl –i icSHAPE.tmp.out –o icSHAPE.out –T 2 –t 200 –s 5 –e 30
        

        Here ‘icSHAPE.tmp.out’ is the input raw icSHAPE enrichment score file from Step 72B(viii), and ‘icSHAPE.out’ is the output file that contains valid icSHAPE enrichment scores. The other parameters are defined to select the valid scores for those transcripts that have an average hit coverage higher than 2 (‘-T 2’) in the target library, and within regions that have a background base density higher than 200 (‘-t 200’), and to skip the leading 5 (‘-s 5’) and the ending 32 nt (‘-e 30’).
        ? TROUBLESHOOTING
• 73|
  Generate bigwig files for visualization with the University of California, Santa Cruz (UCSC) genome browser by using the following commands:

  $ enrich2Bedgraph.pl icSHAPE.out > icSHAPE.bedgraph
  
  $ sort -k1,1 -k2,3n icSHAPE.bedgraph -o icSHAPE.sorted.bedgraph
  
  $ uniqueTrack.pl icSHAPE.sorted.bedgraph icSHAPE.sorted.uniq.bedgraph
  
  $ cut -f1–4 icSHAPE.sorted.uniq.bedgraph | grep -v NULL > icSHAPE.sim.bedgraph
  
  $ bedGraphToBigWig icSHAPE.sim.bedgraph genome.chr.size icSHAPE.sim.bw
  

  Here we first convert the reactivity score for each base in the transcript coordinates into genome coordinates (file icSHAPE.out from the Step 72B(ix)) and then merge genome positions that are covered by multiple transcripts. Finally, convert the .bedgraph file to a .bw file by using the UCSC utility ‘bedGraphToBigWig’. The file ‘genome.chr.size’ contains the lengths of all chromosomes of the investigated genome.

? TROUBLESHOOTING

Troubleshooting advice can be found in Table 2.

● TIMING

Steps 1 and 2, cell culture: ~10 min

Steps 3–11, icSHAPE modification and quality control: ~2 h

Steps 12–17, isolation of poly(A)⁺ RNA fraction: ~5 h (to overnight)

Steps 18–20, biotin click reaction of NAI-N₃: ~3 h

Steps 21–25, RNA fragmentation and RNA end repair: ~1.5 h

Steps 26–39, RNA 3′-end ligation and size selection: ~5 h

Steps 40–57, cDNA synthesis, isolation of NAI-N₃-modified molecules and cDNA purification: ~4.5 h

Steps 58–68, cDNA circularization, library amplification and size selection: ~5 h

Steps 69–73, icSHAPE data analysis: ~1–2 d

Box 1, icSHAPE software installation and configuration: ~4 h

Box 2, in vitro icSHAPE modification: ~1 h