3′ End Sequencing Library Preparation with A-seq2

1. Cell Growth and mRNA Isolation

1. Grow cells according to your experimental design in 6-well plates to ~1 x 10⁶ cells per well at 80% confluence.
2. Remove the growth medium and wash the cells once with phosphate buffered saline. Directly lyse the cells on the plate by adding 1 mL of lysis buffer from the mRNA-isolation kit. Transfer the viscous lysate into a 15 mL plastic tube with a 1 mL pipette tip. Use a rubber spatula to completely detach the cell material from the plate surface.
3. Shear the lysate containing viscous DNA with a 1 mL syringe attached to a 23 G hypodermic needle by several vigorous up and down movements of the plunger until the lysate is no longer viscous. Point the syringe needle to the center of the bottom to avoid ejecting the lysate out of the tube.
4. Transfer the lysate into a 1.5 mL tube using the syringe. Spin 5 min at 20,000 x g and 4 °C to remove the debris. Use DNA low bind 1.5 mL vials throughout the protocol.
5. While the centrifuge is running, wash 300 µL of resuspended oligo (dT)₂₅ magnetic beads on a magnetic rack with 500 µL of lysis buffer. Mix the tubes 2-3 times on the rack. Remove the buffer after the solution is clear. Collect the clear supernatant from step 1.4 and add to the beads. Resuspend and place tubes on a rotating wheel for 10 min.
6. Place the tubes on a magnetic rack. Remove the clear liquid after 2 min. Add 0.8 mL buffer A from the mRNA-isolation kit. Turn the tube by 180° degrees on the rack, 2-3 times. Repeat this washing step once more with buffer A.
7. Wash the beads 2 times with 0.8 mL of buffer B as described in step 1.6.
8. To elute the bound mRNA from the beads, add 33 µL H₂O and resuspend the beads. Heat to 75 °C for 5 min on a heated block. Immediately spin the tubes for 1 s and place them on the magnetic rack. Transfer the supernatant to a new tube. Samples can be stored at -80 °C until further use.
9. Add 66 µL alkaline hydrolysis buffer to the 33 µL mRNA (step 1.8), mix and heat for exactly 5 min at 95 °C on a heating block. Immediately chill tubes on ice.
10. Isolate RNA with an RNA cleanup kit.
    NOTE: Confirm the volume; it should be 100 µL.
    1. Add 350 µL RLT buffer from the kit and 250 µL ethanol. Load onto the column and spin for 30 s at 8,000 x g at room temperature (RT). Wash with 500 µL RPE buffer from the kit. Wash with 500 µL 80% ethanol. Spin for 5 min at 20,000 x g to dry the column. Add 36 µL H₂O to column and spin for 1 min at 20,000 x g. Discard the column and save the eluate.

2. 5' end Phosphorylation and DNase Treatment

1. Add 5 µL polynucleotide kinase buffer, 5 µL 10 mM ATP, 1 µL ribonuclease inhibitor, 1 µL DNase and 2 µL polynucleotide kinase to samples and incubate at 37 °C for 30 min. Optionally prepare master reaction mixes throughout the protocol by mixing 1.1 volumes x n (n = number of samples) of each component.
2. Change buffer and remove ATP on a spin-column to prevent poly(A) addition in the next step.
   1. Prespin spin-columns at 735 x g for 1 min. Transfer the columns to new 1.5 mL vials and load the kinase reactions onto the columns. Spin the columns 2 min at 735 x g. Discard the columns and place the tubes with collected reactions on ice or store at -80 °C.

3. Blocking 3' Ends with Cordycepin Triphosphate

NOTE: It is essential to block the 3' ends of RNA fragments to avoid their concatemerization in the subsequent ligation reactions. 3' ends that are not already blocked by a (cyclic) phosphate after hydrolysis are treated by addition of a 3' dATP (cordycepin triphosphate) chain terminator nucleotide with the aid of poly(A) polymerase. Here, yeast poly(A) polymerase (yPAP), that was expressed and purified as described in ⁸ was used at a concentration of 0.5 mg/mL. Yeast or E. coli PAP both have almost the same activity for addition of 3'dATP and can be purchased commercially (see the Table of Materials).

1. Add 13.5 µL 5x concentrated poly(A) polymerase reaction buffer, 2 µL of 10 mM 3' dATP, 1 µL RNase inhibitor and 1 µL poly(A) polymerase to the reaction from step 2.2.1. Mix and spin for 1 s. Incubate at 37 °C for 30 min. Add 32.5 µL H₂O to each reaction. Purify the RNA as in step 1.10.1. Elute the RNA with 14 µL H₂O.

4. Ligation of Reverse 3' Adapters to the 5' End of RNA Fragments

1. Place the reactions in a vacuum concentrator for 10 min to reduce the volume to 6 µL. Add 3 µL 10x T4 RNA ligation buffer, 3 µL 10 mM ATP, 15 µL PEG­-8000, 1 µL RNase inhibitor, 1 µL of 0.1 mM reverse complement 3' adapter "revRA3" (see the Table of Materials) and 1 µL high concentration RNA ligase 1, mix.
2. Incubate the reactions at 24 °C for 16 h on a heated mixer with intermittent mixing at 1,000 rpm. Add 70 µL H₂O to each reaction and mix. Purify the RNA as in step 1.10.1. Elute the RNA with 14 µL H₂O. Samples can be stored at -80 °C at this point.

5. Reverse Transcription (RT)

1. Place the eluates in a vacuum concentrator for 3 min to reduce the volume to 11 µL. Transfer reactions to 200 µL PCR tubes. Add 1 µL 0.05 mM RT primer "Bio-dU-dT25". Heat for 5 min at 70 °C in a PCR cycler and leave at RT for 5 min.
2. Add 1 µL 10 mM dNTPs, 4 µL 5x reverse transcriptase buffer, 1 µL 0.1 M DTT, 1 µL RNase inhibitor, and 1 µL reverse transcriptase. Mix and heat the reactions for 10 min to 55 °C and 10 min to 80 °C in a PCR cycler. Keep on ice or at -80° C for longer storage.

6. Digestion with Uracil DNA Glycosylase Enzyme Mix

1. Pipet 100 µL Streptavidin-beads into a 1.5 mL vial, resuspend in 800 µL biotin binding buffer and place on a magnetic rack. Invert tubes 2-3 times. Remove the buffer when clear. Repeat the washing step. Resuspend the beads in 200 µL biotin binding buffer.
2. Add reverse transcription reaction to the beads solution and incubate 20 min at 4 °C on a rotating wheel. Wash the beads 2x with biotin binding buffer as in step 6.1 and 2x with TEN buffer on a magnetic rack. Resuspend the beads in 50 µL TEN buffer, add 2 µL Uracil DNA glycosylase enzyme mix, and incubate 1 h at 37 °C in a mixer with intermittent mixing.
3. Add 50 µL H₂O, 11 µL of RNase H buffer and 1 µL RNase H to the reactions. Incubate at 37 °C for 20 min. Place tubes on a magnetic rack and transfer the liquid containing the cleaved cDNA to a new tube
4. Purify the cleaved cDNA.
   1. Add 550 µL of buffer PB from the PCR purification kit to the cleavage reactions. Add 10 µL of 3 M sodium acetate, pH 5.2 to lower the pH. Load the reactions on minimal elution spin columns and spin at 17,000 x g for 1 min.
   2. Add 750 µL buffer PE to columns and spin at 17,000 x g for 1 min. Discard the flow-through. Spin the columns at 17,000 x g for 1 min to dry. Transfer the columns to a 1.5 mL vial, add 16 µL H₂O and spin at 17,000 x g for 1 min. Place the reactions in a vacuum concentrator for 8 min to concentrate to a volume of 7 µL.

7. Ligation of 5' Adapters to 5' Ends of cDNA

1. To the isolated cDNA, add 3 µL 10x T4 RNA ligase 1 buffer, 3 µL 10 mM ATP, 15 µL PEG­-8000, 1 µL 50 µM "revDA5" oligo, and 1 µL high concentration T4 RNA ligase 1. Incubate at 24 °C for 20 h. Add 70 µL H₂O to each reaction. Samples can be stored at -20 °C at this point.

8. Pilot PCR, Amplification of Libraries and Size Selection

1. In a pilot reaction, determine the optimal number of PCR cycles to reach library amplification within the exponential phase.
   1. Pipet 25 µL DNA polymerase mix, 20 µL ligation reaction, 2 µL H₂O, 1.5 µL 10 µM forward PCR primer (RP1) and 1.5 µL 10 µM reverse PCR index primer into 200 µL PCR tube.
   2. Run the cycler with the following program: 3 min 95 °C, followed by 20 cycles of 20 s 98 °C, 20 s 67 °C and 30 s 72 °C. Collect 7 µL aliquots after 6, 8, 10, 12, 14, 16 and 18 cycles directly from the cycler. Add 1 µL 10x loading buffer (50% glycerol, 0.05% xylene cyanol).Note: Please follow the recommendations of the supplier if using multiplexing when combining barcodes.
   3. Separate products in small slots on a 2% agarose gel in 1x TBE buffer containing a 1:10,00 dilution of fluorescent green dye.
      1. Load aliquots on a 2% agarose gel and run the gel at 100 volts for 15 min. Visualize migration of PCR products on a gel documentation system.
2. Use the number of cycles at the beginning of exponential amplification in the pilot reaction for a large-scale PCR reaction with twice the volumes as used for the pilot reaction (Figure 2).
   1. For large-scale PCR reactions, concentrate and desalt the reactions first with a PCR purification kit and separate the products on wide slots on 2% agarose gels in 1x TBE buffer.
3. Cut out gel slices containing 200-350 nt DNA products.Melt the gel in the chaotropic buffer at RT for up to 30 min.Extract DNA from the gel slices with a gel extraction kit. Do not heat to 50 °C to prevent bias in binding of A-rich DNA ⁹.
4. Submit for sequencing.
   NOTE: Typically, 50 cycles single-read (SR50) are sufficient (see, for e.g., https://www.illumina.com/technology/next-generation-sequencing.html).

9. Data Processing

NOTE: The resulting sequencing data (in fastq format) are processed with software available in the gitlab repository (https://git.scicore.unibas.ch/zavolan_public/A-seq2-processing). The analysis includes four main steps: (1) downloading the git repository, (2) installation of a virtual environment, (3) setting specific parameters in the configuration file and (4) launching the analysis through ‘snakemake’ ¹⁰. The entire analysis done in step 4 requires only one command. A detailed step-by-step description of the analysis can be found in the README file in the gitlab repository and a short description is available below. All individual processing steps are accomplished by the execution of publicly available tools, either from external sources or prepared in-house. The computational pipeline depends on an anaconda-based ¹¹ python 3 virtual environment with the snakemake package available ¹⁰. It runs on machines with Unix-like operating system and was tested in a Linux environment with the CentOS 6.5 operating system installed and 40 GB RAM available.Software dependencies are automatically controlled within the virtual environment. The following publicly available software tools are required and thereby installed together with the environment: snakemake (v3.9.1) ¹⁰,fastx toolkit (v0.0.14) ¹², STAR (v2.5.2a) ¹³, cutadapt (v1.12) ¹⁴, samtools (v1.3.1) ^14,15, bedtools (v2.26.0) ^16,17.

1. Data pre-processing from reads to cDNAs
   NOTE: The sequencing depth may vary between runs and, depending on the instrument, data from one sample can be split over multiple sequence files. If this is the case, concatenate the files that correspond to one sample into a single input file that is used in the following steps.
   1. Convert the file from fastq to fasta format.
   2. Extract reads with a correct structure (3 thymidines at positions 5, 6 and 7 of the read).
      NOTE: A read that is correctly prepared according to the experimental protocol described above should have the structure (from the 5' end): 4-nucleotide barcode – 3 thymidines – reverse complement of transcript 3' end.
   3. Store the information about the starting tetramer in the description line of the sequence.
      NOTE: The tetramer serves as a unique molecular identifier (UMI) that facilitates the correction of amplification artifacts later in the analysis.
   4. Remove the first seven nucleotides from the read's 5' end.
   5. Correct for amplification artifacts by keeping only one copy of the reads with the same insert sequence and UMI.
   6. Remove the part of the 3' end that matches the adapter sequence and then reverse complement the sequence. Only proceed with reads that have a minimum length (default: 15 nt).
      NOTE: Depending on the length of the original mRNA fragment and the number of sequencing cycles, the 3' end of the read may contain part of the 3' adapter, which is removed in this step.
2. Extract all reads that fulfill the following criteria: maximum 2 unknown nucleotides ('N'), maximum 80 % As, and last nucleotide of the read not A. These reads are considered to be of sufficient quality to be used in the analysis.
3. Map the reads to the genome with a tool that handles spliced reads and generates an output file in BAM format.
   1. If STAR is used, create a file with the index of the genome to which the reads should be mapped. For the human genome, this step requires 35 GB of memory (RAM).
   2. Map the reads to the genome.
      NOTE: (STAR-specific notes) Soft-clipping is disabled in order to force the mapping of the 3' end of each read as this is the nucleotide immediately upstream of the cleavage site.
4. Convert the BAM to a BED-file. If a read maps to multiple locations, keep only those with the lowest edit distance.
   NOTE: The copy number of the read mapped at a specific location is used as score. Reads that map to multiple locations are counted fractionally at each location with a weight equal to 1/number of locations to which a read maps.
5. Collapse reads that vary by a likely sequencing error. If two distinct reads map to the same location (start and end position of the mappings are identical) and they share the same UMI, consider them as PCR duplicates and keep only one.
6. Infer all individual pre-mRNA 3' end processing sites.
   NOTE: An individual read provides evidence for a 3' end when its last four nucleotides are mapped to the genome without error. The position to which the 3' end of the read maps is stored as cleavage site.
7. Detect 3' end sites that could have originated from internal priming. Define the site as internal priming artifact when the 10 nt downstream of the cleavage site in the genome satisfy one of the following criteria: contains more than six As, contains six consecutive As, or starts with one of the following tetramers: AAAA, AGAA, AAGA, AAAG.
8. Generate a table of individual 3' end processing sites in BED format.
9. Identify independently regulated poly(A) site clusters.
   NOTE: The steps described here follow the procedure that was introduced in a prior publication ⁵.
   1. Start by collecting individual 3' end processing sites that were obtained in all samples of the study.
   2. Annotate known poly(A) signals ⁷ in the region of -60 to +10 nucleotides around each individual 3' end processing site.
   3. Identify poly(A) sites expressed above the background in each sample as follows.
      1. Sort the sites by their raw expression within the current sample. Traverse the list of sites from top to bottom, associating lower ranked sites with a higher ranked site if they are located within a predefined distance in the genome (default: 25 nt up- or downstream) from the high-ranking site.
         NOTE: All low-ranking sites associated with a high-ranking site define a cluster whose expression is the number of reads documenting all of these sites.
      2. Sort these clusters by expression and traverse the list of clusters from highest to lowest expression, determining the expression threshold c at which the percentage of clusters with an annotated poly(A) signal drop below a predefined threshold (default: 90%).
      3. Discard sites from any cluster below the cutoff.
   4. Cluster closely spaced 3' end sites obtained across samples.
      NOTE: Sort 3' end processing sites first by the number of supporting samples and then by the sum of the normalized read count (reads per million (RPM)) across samples. Traversing the list from top to bottom, associating lower-ranked sites with higher-ranked sites when their distance to the higher-rank site is not larger than a predefined limit (default: 12 nt). Whenever any of the constituting 3' end site overlaps with an annotated poly(A) signal or has a poly(A) signal directly downstream, the corresponding cluster is marked for further inspection to detect internal priming.
   5. Merge poly(A) site clusters.
      NOTE: When a cluster is marked as a putative internal priming candidate, it is either merged into a downstream cluster if the two clusters share their poly(A) signals or retained if the most downstream site in the cluster has a poly(A) signal located at a minimum distance upstream (default: 15 nt). Finally, closely spaced clusters are merged if: (i) they share the same poly(A) signal(s), or (ii) the span of the resulting cluster does not exceed a maximum (default: 25 nt).
   6. Store clusters in BED-file format with the total normalized read count from all 3' end sites in each cluster as score.