Tissue culture and compound administration (TS-543)
We seeded TS-543 neurospheres (passage #11) on a 96-well plate (Corning, #3799) at a density of 7500 cells per well (50,000 cells/mL) in 150µL NS-A complete medium (containing 10% v/v NeuroCult NS-A Proliferation Supplement, 20 ng/mL EGF, 10 ng/mL bFGF, and 2ug/mL heparin) (STEMCELL Technologies #05,751). We incubated the plate of cells for 36 h prior to the start of the experiment at 37 °C and 5% CO2 in a tissue culture incubator. We separately prepared stock solutions of PP242 (Tocris, #4257), MNK-i1 (Sigma, #534,352), NVP-BKM120 (Selleck, S2247), AZD8055 (Selleck, S1555), and 4EGi-1 (Tocris, #4800) in DMSO vehicle (Sigma, #472,301). After dilution with NS-A basal culture medium (without supplement, cytokines, or heparin), we administered the drugs or pure DMSO to the experimental and control wells, respectively, in 1µL doses. Final concentrations were 50 nM AZD8055, 625 nM PP242, 1 μM BKM120, 100 nM MNK-i1, and 50 μM 4EGi-1, including in pairwise combination-treated samples. Drug treatment proceeded for 6 h in the tissue culture incubator prior to lysis.
For the dose response analysis shown in Supplementary Fig. 1 (also see Supplementary Data S1 and Supplementary Data S2 for raw scans), we treated TS-543 neurospheres with the indicated concentrations of MNK-i1 as described above. Western blot analysis was performed using p-EIF4E anti-rabbit monoclonal antibody (Abcam ab76256, 1:5000 dilution) with AlexaFluor 488-labeled goat anti-rabbit secondary antibody (A-11034, ThermoFisher) and EIF4E anti-mouse monoclonal antibody (BD Biosciences 610,270, 1:2000 dilution) with AlexaFluor 647-labeled goat anti-mouse (A-21236, ThermoFisher).
Cell lysis (TS-543)
Following treatment, we centrifuged the plate of TS-543 for 7 min at 1800RPM on a Sorvall Legend XTR at room temperature and removed supernatants by aspiration. Placing the plate on ice, we resuspended the pelleted cells in each well in 30µL of polysome lysis buffer (20 mM Tris–HCl, pH = 7.4, 250 mM NaCl, 15 mM MgCl2),0.1 mg/mL cycloheximide, 0.5% Triton X-100, 1 mM DTT, 0.5U/µL SUPERase-In (ThermoFisher, AM2696), 0.024U/µL TURBO DNase (Life Technologies, AM2222), 1 × Protease Inhibitor (Sigma, P8340)), mixed 5 times by pipetting, and rested the plate on ice for 5 min. We then centrifuged the plate for 5 min at 1400RPM at 4 °C to remove bubbles before performing a quick freeze–thaw, placing the plate first in a -80 °C freezer and then resting at room temperature for 5 min each. Following an additional 10 min rest on ice, we viewed the plate under a microscope to check the extent of cell lysis. We then prepared a new 96-well plate containing 3.5µL 2 × TCL buffer (Qiagen, #1,070,498) per well, to which we transferred 3.5µL of lysate (approximately 10% total volume).
Automated pan-ribosome immunoprecipitation
To the remaining lysate, we added 1 µL of SUPERase-in (ThermoFisher, AM2696) and 1 µL of biotinylated y10b antibody (ThermoFisher, MA516060) to each well, then sealed the plate and allowed it to incubate while gently shaking for 4 h at 4 °C. During this incubation, we washed 500µL of Dynabeads MyOne Streptavidin C1 streptavidin-coated magnetic beads (ThermoFisher, #65,001) 3 times with polysome wash buffer (20 mM Tris–HCl (pH 7.4), 250 mM NaCl, 15 mM MgCl2, 1 mM DTT, 0.1 mg/mL cycloheximide, 0.05% v/v Triton X-100), using 1 mL per wash and resuspending in 500µL. We added 5µL of washed beads to each well, then incubated while gently shaking at 4 °C for an additional hour. After this short incubation, we placed the plate on a magnet, removed and reserved supernatants, and washed the wells 3 times with 200µL per well of polysome wash buffer supplemented with 1µL/mL SUPERase-in on the Biomek 4000 automated liquid handling system.
Following the final wash, we resuspended the beads in 15µL of ribosome release buffer (20 mM Tris–HCl (pH 7.4), 250 mM NaCl, 0.5% Triton X-100, 50 mM EDTA) per well. During a 15-min incubation at 4C on a Peltier module, with continuous pipet mixing on the Biomek 4000 in order to maximize elution, we distributed 15µL of 2 × TCL buffer to each well of a new 96-well plate. Finally, we replaced the eluted sample plate on the magnet and transferred eluants to the TCL-containing plate.
Tissue Culture and Cell Lysis for Non-Automated riboPLATE-seq in Wi-38
Separately, we seeded WI-38 human fibroblast cells on a 96-well plate at a density of 3,000 cells per well in 60µL cell culture media per well (DMEM (ThermoFisher #11,965,092) + 10% FBS (ThermoFisher #A3160501)), 36 h prior to cell lysis. After removing media by aspiration and gently washing wells once with cold PBS supplemented with 0.1 mg/mL cycloheximide, we added 30µL cold polysome lysis buffer to each well and mixed by pipetting up and down five times. Additionally, we added 1µL of 1:5000 ERCC spike-in mix 1 (ThermoFisher #4,456,740) to every other column of wells on the plate. We rested the plate at room temperature for 5 min, then ice for 10 min more, following which we centrifuged the plate at 1400RPM for 5 min to remove bubbles. We then reserved 10µL from each well (33% initial lysate volume) in a second plate for PLATE-seq, and added 10µL 2 × TCL buffer (Qiagen) to each well before freezing at -80C.
Manual Immunoprecipitation of Ribosome Bound RNA (Wi-38)
We first added 0.6 µL each of biotinylated antibody y10b and SUPERase-IN to each well of the remaining lysate, mixed well by pipetting, then sealed the plate and incubated for 4 h at 4C with gentle shaking. We then added 4µL of streptavidin-coated magnetic beads to each well, which we had washed 3 × and resuspended in polysome wash buffer, mixed carefully, and allowed the plate to incubate for one additional hour at 4C with gentle shaking. After placing samples on a 96-well plate magnet, we washed all wells 3 × with polysome wash buffer, and eluted after the final wash by 15 min of incubation in 15µL ribosome release buffer. Finally, we removed the supernatant using a 96-well plate magnet, and added 15µL of 2 × TCL buffer to each well before freezing the plate.
Ribosome profiling and RNA sequencing
We seeded TS-543 neurospheres in a 6-well plate at a starting density of 50,000 cells/mL in 2 ml of NS-A complete medium per well, and allowed the plate to rest for 36 h. After preparing PP242 solution in DMSO as above, we treated two wells each with 625 nM PP242 or DMSO vehicle for 6 h in the tissue culture incubator. In a separate experiment, we treated two wells each with the same concentrations of PP242 and DMSO for 30 min. Following treatment, we transferred samples to 15 mL conical vials for centrifugation at 640 RCF for 7 min, then removed supernatants and added 400 µL polysome lysis buffer (recipe above). After mixing by rapid pipetting, we transferred samples to 1.8 mL microcentrifuge tubes, rested them on ice for 5 min, and triturated by 5 passages through a 23-gauge needle. Following a clarifying spin of 11 K RCF for 10 min at 4C on a benchtop centrifuge, we transferred supernatants to a new set of microcentrifuge tubes and discarded pellets. We prepared ligation-free ribosome profiling and total RNA-seq libraries from the clarified polysome lysates treated for 6 h following the instructions provided with their respective kits (smarter-seq smRNA-seq kit, Takara-Clontech; NEBnext Ultra-Directional II) augmented with our previously-published ligation-free ribosome profiling protocol59. We additionally prepared conventional ribosome profiling and total RNA-sequencing libraries from the samples treated for 30 min, using previously-described55 modifications to the protocol by Ingolia et al60. We sequenced 6 ribosome profiling libraries or up to 12 RNA-seq libraries in one NextSeq 550 high-output 75-cycle kit. Library construction methods and experimental conditions for each sample are presented in Supplementary Table S2. Both PLATE-seq and ligation-free ribosome profiling library preparation protocols are available on our laboratory website (http://www.columbia.edu/~pas2182/index.php/technology.html).
PLATE-seq library preparation and sequencing
We submitted plates of ribosome-associated and previously reserved total lysate in TCL buffer to the Columbia Genome Center for processing by the previously-described PLATE-seq method of RNA-seq library preparation27, which involves poly-A selection of transcripts, incorporation of sequence barcodes in poly(T)-primed reverse transcription, and pooling for subsequent library preparation steps, generating a single 3’-end RNA-seq library from each 96-well plate. We pooled total and ribosome-associated PLATE-seq libraries, sequencing the pooled pairon the Illumina NextSeq 550 with a 75-cycle high-output kit. With paired-end sequencing, the first read corresponds to the 3’ end of a transcript, and the second read contains the barcode identifying the library in which the read was obtained.
Read alignment and data analysis
With a custom processing pipeline, we first trim reads of trailing polyA sequence and adapters with cutadapt (v 3.5)61, then align the whole set of multiplexed reads to the hg38 assembly of the human genome, plus additional sequences corresponding to ERCC spike-in transcripts added for depletion experiments, with STAR62 (v 2.7.9a). We then demultiplex the aligned fragments from Read 2 to their original riboPLATE- or PLATE-seq library indices according to their barcodes present in Read 1, as described in the original PLATE-seq paper27. We use a similar pipeline to process and align ribosome profiling and RNA sequencing libraries, first trimming polyA tails and adapters with cutadapt, then removing reads that align to the 45S pre-ribosomal RNA and 5S ribosomal RNA with bowtie263 (v 2.2.5) before aligning with STAR. We then use featureCounts64 (v 2.0.1) to count the number of fragments aligned to each gene in each library, counting all exon-aligned reads as valid. Barcode sequences for each PLATE- or riboPLATE-seq library generated are available as separate worksheets in Supplementary Table S3. Library quality control data are available in Supplementary Table S4.
Definition of gene sets of interest
As PLATE- and riboPLATE-seq depend on isolation of RNA by poly(T) pulldown, they can only be used to measure polyadenylated transcripts. We first combined two sets of poly(A)- predominant transcripts from HeLa and H9 cells determined in a screen of polyadenylation status across the transcriptome65, and removed these genes from consideration in our study to leave only consistently polyadenylated transcripts. We also obtained a set of known 5’ terminal oligopyrimidine motif-containing genes (TOP genes), as well as novel TOP candidates with and without known TOP-containing analogues in mice, from a comprehensive search of transcription start sites50.
Variance-stabilizing transformation and outlier removal
After subsetting the count matrices for all libraries to remove alignments to non-polyadenylated and spike-in transcripts, we constructed an overall count matrix of all 192 libraries for all 96 samples. We read this matrix into DESeq2 with corresponding column data describing the sample ID, library type (ribo or RNA PLATE-seq), and drug treatment for each library. We then used the variance-stabilizing transform in DESeq2 (version 1.3.4) with default parameters to obtain an approximately homoscedastic, log-scale transformation of raw counts for all libraries. We used this transformed count matrix to perform two-dimensional principal component analyses (PCA) in Python, utilizing the scikit-learn package for analysis and matplotlib for visualization. We limited these analyses to genes determined significant by DESeq2 for differential RA in any drug tested (Benjamini–Hochberg adjusted FDR < 0.05; 1813 genes total). For each sample, we computed RA for each gene as the ratio of normalized riboPLATE-seq to PLATE-seq counts; for log-scale transformations such as vst, this corresponds to their difference. With the average RA across all vehicle-treated controls as a reference for baseline RA, we computed the log-fold change from baseline for all genes in each drug-treated sample. We first performed PCA on RA and log-fold change in RA for the full set of samples from the plate, then removed 11/96 samples as PCA outliers (2 DMSO, 2 4EGi-1, 2 BKM120 + MNK-i1, 1 AZD8055, 1 PP242, 1 BKM120, 1 PP242 + MNK-i1, 1 PP242 + BKM120) for subsequent analyses of differential ribosome association with DESeq2. We used the remaining samples to generate final matrices of raw counts and vst-transformed counts for the RNA quality control and principal component analyses in Figs. 2 and 3.
Differential count analysis in DESeq2
We performed differential expression and differential ribosome association analyses using the DESeq2 package in R66. We first read the entire matrix of counts across all samples, plus its corresponding column data table describing sample ID, drug treatment, and library type for each sample, into DESeq2. For each condition, we subset the matrix of gene counts to samples corresponding only to that condition and DMSO controls, then analyzed that subset using a likelihood ratio test with the following parameters:
dds_sub <- DESeq(dds_sub, fitType=’local’, test=’LRT’, full=~condition+type+condition:type, reduced=~condition+type)
where condition and type in the design formulas refer to experimental condition/drug treatment and sequencing library type (riboPLATE or PLATE-seq), respectively. Following each subset DESeq2 analysis, we retrieved results for the interaction term condition:type, corresponding to changes in the ratio of riboPLATE- to PLATE-seq counts between conditions, i.e. differential RA:
res <—results(dds, name = ’condition < X > .typeRIBO’).
We analyzed ribosome profiling and RNA sequencing data in an identical fashion, comparing PP242 vs DMSO-treated samples at both 30 min and 6 h of treatment to generate two signatures of differential TE.
Comparison of sequencing library types with gene set enrichment analysis
We constructed ranked lists for gene set enrichment analysis (GSEA) using the per-gene differential translation efficiencies calculated with ribosome profiling and RNA sequencing data at 30 min or 6 h of PP242 treatment. For each drug tested via riboPLATE-seq, we identified its targets as genes exhibiting significant change in ribosome association (RA) by DESeq2 (Benjamini–Hochberg adjusted FDR < 0.05), split by up- or downregulation (lfcRA > 0 / lfcRA < 0). We then removed genes from each such that all sets were mutually exclusive, i.e., no two drugs share a common upregulated or downregulated gene. We used the preranked function in the GSEA desktop app to compare the ranked lists for differential TE, using each gene’s log fold change in TE as a ranking metric, against these differential RA-derived gene sets with default parameters and scoring method set to ‘classic’.
Network visualization
To create a basic network, we interpreted the genes exhibiting significant reductions in RA under treatment with kinase inhibitors (FDR < 0.05) as positive targets of the kinases inhibited. We loaded these gene sets into CytoScape67 (v2.9.0) as individual networks for each kinase, merged the three networks, and organized the resulting merged network with the yFiles68 Organic automatic layout. We then color-coded the sets of canonical and novel TOP motif-containing genes present in the network, based on lists obtained from Yamashita et al50.
Data visualization and code
All code was run in Python 3.9.5 and R 4.0.5. R packages were installed via Bioconductor v 3.12, including DESeq2 v1.34.0 for differential RA and TE analysis and normalization, and BiocParallel v 1.28.0 for use of multicore processors.
Python libraries were installed via Anaconda (v 4.10.3). We generated plots and diagrams using matplotlib (v3.4.3) and Jupyter Notebook (IPython 7.28.0, jupyter_core v4.8.1)69,70. Our analyses use NumPy71 (v1.21.2) for data manipulation, SciPy72 (v1.6.3) for statistical tests, and scikit-learn73 (v1.0) for PCA. We additionally generated strip plots and heatmaps using Seaborn74 (v0.11.2).
