Ligation-free ribosome profiling of cell type-specific translation in the brain
- Nicholas Hornstein†1, 2,
- Daniela Torres†3, 4,
- Sohani Das Sharma1,
- Guomei Tang5,
- Peter Canoll4Email author and
- Peter A. Sims1, 6, 7Email author
© The Author(s). 2016
Received: 25 February 2016
Accepted: 10 June 2016
Published: 5 July 2016
Ribosome profiling has emerged as a powerful tool for genome-wide measurements of translation, but library construction requires multiple ligation steps and remains cumbersome relative to more conventional deep-sequencing experiments. We report a new, ligation-free approach to ribosome profiling that does not require ligation. Library construction for ligation-free ribosome profiling can be completed in one day with as little as 1 ng of purified RNA footprints. We apply ligation-free ribosome profiling to mouse brain tissue to identify new patterns of cell type-specific translation and test its ability to identify translational targets of mTOR signaling in the brain.
Ribosome profiling allows genome-wide measurements of ribosomal occupancy with single-nucleotide resolution . Using deep sequencing as a readout for protein synthesis, the technique has enabled the discovery of previously unannotated open reading frames (ORFs) [1–4] and provided new insights into the mechanisms of translation initiation and elongation, localized translation , and the signaling pathways underlying translational control [7, 8]. In addition, ribosome profiling has been applied in many cellular contexts, including yeast , bacteria , primary mammalian cells , and complex tissues , to assess the role of translational control in basic physiological processes and its dysregulation in diseases like cancer.
In addition to characterizing the performance of ligation-free ribosome profiling, we applied our technique to assess cell type-specific translational regulation in the murine brain. The brain harbors a broad diversity of cell types, including astrocytes, oligodendrocytes, microglia, glial progenitors, endothelial cells, and many different types of neurons that likely control translation through different signaling pathways. In addition, many neuron-specific transcripts are translated locally in dendrites and translational control has been shown to play a key role in memory [16–19]. We took advantage of a recently reported database of neural cell-specific gene expression  to identify patterns that indicate cell type-specific regulation of translation. As an orthogonal validation for neuron-specific genes, we used the RiboTag system  to purify and identify actively translated transcripts from excitatory neurons in the cortex of Camk2a-Cre/RiboTag mice. Finally, we used our technique to identify the genes controlled by mammalian target of rapamycin (mTOR) signaling in the brain by conducting ribosome profiling on the brains of mice treated with AZD-8055, an ATP-competitive inhibitor of mTOR that crosses the blood–brain barrier [22, 23].
A ligation-free protocol for ribosome profiling
Ribosome profiling is more complicated than conventional RNA-Seq because the ribosome-protected mRNA footprints are short (~30 nucleotides) and lack poly(A) tails, which are often used as handles for either isolation or reverse transcription of eukaryotic mRNA. Previously established protocols for ribosome profiling address this problem by single-stranded ligation of a universal adapter to the 3′ end of mRNA footprints to facilitate reverse transcription, which incorporates a longer adapter into the 5′ end of the resulting cDNA . Intramolecular ligation (circularization) of the cDNA effectively attaches a universal adapter to the 3′ end of the cDNA to enable PCR enrichment of the library . Alternatively, a second ligation reaction can be used to attach an adapter to the 3′ end of the cDNA. These ligation reactions are notoriously inefficient and require excess adapter, which is typically removed by gel purification and subsequent overnight precipitation of the product . These multi-step procedures and intermediate purification steps require multiple work days, are intrinsically lossy, and, therefore, require relatively high input .
To address these issues, we have applied the template-switching approach to library construction that has been successfully implemented in other low-input RNA sequencing protocols such as single-cell RNA-Seq [24–26]. Specifically, we have adapted a newly developed version of the SMARTer library construction technology (Clontech) for ribosome profiling (Fig. 1a). We first polyadenylate dephosphorylated RNA footprints using RNA poly(A) polymerase, similar to the earliest reported protocol for ribosome profiling . We then reverse transcribe the polyadenylated footprints using an enzyme with template-switching activity. In a template-switching reaction, the reverse transcriptase (RT) first extends a primer (in this case oligo(dT) linked to a universal sequence on its 5′ end) to produce cDNA. Once the RT reaches the end of the RNA template, the terminal transferase activity intrinsic to the RT adds a low complexity sequence to the 3′ end of the cDNA in a non-template-directed fashion. The reaction is carried out in the presence of a second universal sequence adapter that is 3′ terminated with a low-complexity sequence, which hybridizes to the tail added to the cDNA by the RT. Upon hybridization of this second sequence adapter, the RT switches templates and copies the second adapter onto the 3′ end of the cDNA. As a result, both 5′ and 3′ universal adapters are simultaneously added to the cDNA in a single reaction without single-stranded ligation or intermediate purification steps. We then deplete the resulting product of rRNA using complementary oligonucleotides  and enrich the deep sequencing library by PCR.
Comparison of ligation-free ribosome profiling with conventional ribosome profiling
We used ligation-free ribosome profiling to measure genome-wide translation in the forebrains of adult mice. Unlike fragments generated in RNA-Seq, ribosome footprints map to the transcriptome with a three-nucleotide periodicity due to the characteristic translocation interval of the ribosome as it translates codons . To verify that the RNA libraries generated using our technique originate from ribosome footprints, we computed the power spectrum of the 5′ mapping positions of RNA fragments (Fig. 1b). As expected, the data are highly periodic with a characteristic frequency of ~0.33 nucleotides−1, similar to what has been observed for conventional ribosome profiling . In addition to three-nucleotide periodicity, ribosome profiling also exhibits a characteristic gene body distribution. The majority of reads are expected to map to the coding sequences (CDSs) of transcripts, whereas relatively few should map to the untranslated regions (UTRs) . Many genes have been shown to contain unannotated upstream ORFs (uORFs) and so we also expect that more reads will map to the 5′ UTRs than the 3′ UTRs, which are largely depleted of ribosomes. As shown in Fig. 1c, ligation-free ribosome profiling reads map to the transcriptome with the expected gene body distribution.
To further validate the technique, we compared these results with our previously reported mouse forebrain data that we generated using conventional ribosome profiling . Figure 1d shows that the ribosome footprint counts for each gene across the two data sets are highly correlated. We also compared the gene detection efficiency, saturation properties, and library complexities of the two data sets. We note that in our previously reported experiment with conventional ribosome profiling, we used more input monosomal RNA for library construction than in the current experiment with ligation-free ribosome profiling. In Fig. 1e, f, we use downsampling analysis to show that the two data sets are quite similar in terms of both the number of genes detected and number of unique ribosome footprints detected, respectively, at a given sequencing depth. These results imply that the library complexities produced by the two protocols are highly comparable.
In order to determine the sensitivity of both conventional and ligation-free ribosome profiling, we generated libraries from a defined 34-base RNA oligonucleotide at five input levels ranging from 0.01 to 100 ng. We constructed Illumina libraries from each dilution using the convention ribosome profiling protocol described by Ingolia et al.  and the ligation-free protocol described here. We then assessed our yield for each dilution using an Agilent Bioanalyzer (Additional file 1: Figure S1). We found that the ligation-free method is more sensitive and able to generate detectable libraries from less than 1 ng of input. For both methods we were able to generate quantifiable libraries; however, we were only able to generate libraries at 10 and 100 ng of input when using the conventional protocol with nine PCR cycles. In contrast, we were able to generate detectable libraries at all concentrations tested when using the ligation-free protocol with nine PCR cycles. We note that the 10 and 100 ng input libraries made with the ligation-free protocol exhibit over-amplification as evidenced by a broader product length distribution at higher-than-expected molecular weights. To directly compare all of the samples, we kept the number of PCR cycles constant and note that lower cycle numbers could be used to avoid over-amplification of higher input libraries with the ligation-free protocol. In addition, we note higher cycle numbers may result in sufficient library yields for the conventional protocol at lower concentrations, although this could result in increased amplification bias.
Cell type-specific translation in the brain
One of the key metrics obtainable from ribosome profiling experiments is the translation efficiency (TE), which can be computed for each gene as the ratio of its ribosome footprint density to its expression level measured by RNA-Seq . TE is proportional to the number of ribosomes per transcript averaged over all copies of a given gene.
We validated our TE measurements by performing quantitative PCR (qPCR) on a set of highly translated (Syt1, Snap25) and lowly translated (Trpv6, Tgfb1, Pkd1) genes based on our ribosome profiling data. We first used sucrose gradient fractionation to separate mRNAs based on the number of bound ribosomes and collected fractions. We then used qPCR to assess the relative abundance of each gene in each fraction (Additional file 2: Figure S2). Several complications are associated with directly comparing qPCR data obtained from polysome profiles and ribosome profiling data. While the majority of transcripts for a highly translated gene may appear in polysomes with more than five ribosomes per transcript, resolution constraints make it difficult to accurately measure the number of bound ribosomes for each fraction, particularly for heavier polysomes. Furthermore, calculating TE based on log ratios without correcting for cytosolic mRNA levels has been previously shown to produce an inaccurate estimation of TE . While it is difficult to quantitatively compare TE calculated from next-generation sequencing with that obtained from qPCR, we found that the highly translated genes probed are clearly shifted to heavier polysomes compared with the lowly translated genes probed. For example, we found that the maximum abundance of the highly translated genes Syt1 and Snap25 were in the seventh and ninth polysome fractions (greater than five ribosomes per transcript), respectively (Additional file 2: Figure S2). However, the maximum abundances of Trpv6, Tgfb1, and Pkd1, all of which are lowly translated, were in the fourth and fifth fractions (two or three ribosomes per transcript).
We also compared our ligation-free ribosome profiling and RNA-Seq data with a previously published whole-brain mass-spectrometry data set obtained from a mouse of similar genetic background and age . We found that our ribosome profiling data were better correlated with protein abundance in the brain than our corresponding RNA-Seq measurements (Additional file 3: Figure S3). Hence, some of the difference in the explained variance may be attributable to the contribution of translation regulation on protein expression. This result is consistent with previously published observations in yeast in which mass spectrometry, RNA-Seq, and ribosome profiling were compared .
A recent study by Zhang and colleagues  produced RNA-Seq expression profiles from seven different cell types in the brain by sorting or immune-panning, including astrocytes, neurons, oligodendrocyte progenitor cells (OPCs), newly formed oligodendrocytes, myelinating oligodendrocytes, microglia, and endothelial cells. We used this data set to compute cell-type enrichment scores proportional to the specificity with which each gene is expressed in each cell type (see “Methods”). We then divided the transcriptome into ten gene sets evenly binned by TE and conducted gene set enrichment analysis (GSEA) against rank-ordered lists of cell-type enrichment scores for each cell type . This analysis allowed us to systematically associate genes with varying degrees of cell type specificity and TE. The normalized enrichment score (NES) for each GSEA is shown in the heatmap in Fig. 2b (with bin-by-bin and cell type-by-cell type statistical analysis in Additional file 4: Figure S4), which reveals several interesting patterns. First, we found that microglial genes generally exhibit low TEs. Because we are studying the brains of healthy mice, these microglia are presumably not in an activated state. Previous studies have shown that protein synthesis-associated pathways are upregulated in microglia in certain disease contexts  and so these results could be dependent on genotype or other activating conditions such as injury or an inflammatory stimulus. Conversely, neurons, when considered as a broad group, exhibit the highest degree of variation in TE among their cell type-specific genes. As shown in Fig. 2b, most neuronal genes are either very highly or very lowly translated, suggesting that neuronal genes are under a relatively high degree of translational regulation in comparison with other cell types in the brain.
Translational control is well-known to play an important role in neuronal function and memory formation. Structurally, neurons are highly complex cells that make extensive use of local translation to efficiently modulate protein expression far from the soma . To validate our observation that neuronal genes are highly translationally regulated, we used the RiboTag system to isolate polysomal mRNAs from a specific neuronal subtype, namely excitatory neurons that express Camk2a. As shown in Fig. 2c, the RiboTag mouse harbors a modified ribosomal protein L22 (Rpl22) gene with a floxed terminal exon followed by a second copy of the terminal exon with a triple hemagluttinin tag (HA-tag) . We crossed the RiboTag mouse with a mouse that expresses Cre recombinase under the control of the Camk2a promoter to produce mice which express HA-tagged ribosomes in Camk2a-expressing cells. Figure 2d shows that, as expected, the HA-tag is expressed exclusively in neurons, marked here by the pan-neuronal marker NeuN (Rbfox3). Hence, we can isolate polysomes from homogenized brain tissue of Camk2a-RiboTag mice and purify mRNA–ribosome complexes that originate from Camk2a-expressing neurons by immunoprecipitation (IP) of the HA-tag (Fig. 2c). We obtained RNA-Seq expression profiles from both homogenized brain tissue and immunoprecipitated polysomes of two Camk2a-RiboTag mice. We compared the expression levels of each gene in the immunoprecipitated and homogenate profiles and observed that canonical markers of excitatory neurons were enriched by IP, whereas markers of other cell types in the brain, including inhibitory neurons, were depleted by IP (Fig. 2e). We then repeated the GSEA described above with TE gene sets and genes rank-ordered based on their enrichment by RiboTag IP. This analysis recapitulated the results found for neuronal genes derived from purified neurons in that genes specific to Camk2a-expressing neurons, and not just neurons in general, appear highly translationally regulated (Fig. 2b). A subset of genes expressed in these neurons exhibit relatively high TE, while the remaining exhibit relatively low TE. Not only do these results provide an orthogonal validation of our GSEA based on pan-neuronal gene expression, they also show that the pattern holds for a specific subtype of excitatory neurons in the cerebral cortex.
Finally, these data reveal a simple developmental trend in the oligodendrocyte lineage. Oligodendrocytes, which are primarily responsible for enwrapping neuronal axons with myelin sheaths, are a unique cell type in that their progenitor cells (OPCs) are widely distributed in the adult brain, where they actively proliferate and differentiate to generate new myelinating oligodendrocytes. Hence, we can detect gene expression and translation from different stages of oligodendrocyte development within homogenized brain tissue. Based on our analysis, OPC-specific genes are translated more efficiently than those of either newly formed or mature, myelinating oligodendrocytes, which exhibit the lowest TE of the three. As shown in our statistical analysis in Additional file 4: Figure S4, the comparison between OPCs and myelinating oligodendrocytes is very significant for highly translated genes, as is the comparison between newly formed oligodendrocytes and myelinating oligodendrocytes. While one might expect myelinating oligodendrocytes to be less translationally active in comparison with OPCs because they are post-mitotic, their primary role in the brain is to produce large amounts of myelin, which is comprised mainly of proteins and lipids. Nonetheless, we found that most myelin genes have low TE compared with the overall median in the brain (log2(TE) = −0.02), including Mog (−0.15), Mbp (−0.51), Mobp (−1.42), and Mag (−0.28), with the exception of the transmembrane protein Plp1, which has a TE of 1.02. Hence, despite the importance of protein synthesis to the function of myelinating oligodendrocytes, translation of oligodendrocyte-specific genes is relatively inefficient.
uORFs and 5′ UTRs in the brain
We also sought to determine how more general features of the 5′ UTR affect translation efficiency of the corresponding CDS in the brain. Figure 4c is a heatmap that simultaneously displays the relationships between CDS TE and both the length and GC content of the 5′ UTR across the transcriptome. Figure 4d, e display these relationships independently. In general, longer 5′ UTRs are associated with low TE and both high and low GC content are associated with low TE. Previous studies have shown that genes with highly structured 5′ UTRs are less abundant at the protein level in yeast , which is consistent with the reduced TE associated with long, GC-rich 5′ UTRs observed here.
Translational targets of mTOR in the brain
A common application of ribosome profiling is the identification of translational alterations in response to perturbations such as drug treatment or stress. Cells have evolved elegant mechanisms for regulating the translation of specific genes, often through the interaction of signaling molecules with translation factors that control TE through specific cis-regulatory elements in mRNA. We sought to further test the efficacy of our ligation-free ribosome profiling method in the context of this important application by identifying the translational targets of mTOR signaling in the brain.
mTOR plays a crucial role in the translational control of ribosomal proteins and protein factors involved in translation initiation and elongation . Many of these genes contain a terminal oligopyrimidine (TOP) motif in their 5′ UTRs through which translational control is thought to be mediated . Multiple studies have used ribosome profiling to show that mTOR inhibition causes a coherent decrease in the TEs of the TOP motif-containing genes in cell culture [7, 8]. mTOR is an important drug target in multiple neurological disorders . For example, rapalog inhibitors of mTOR have been shown to mitigate seizures in certain contexts . We sought to determine whether mTOR controls the same set of target genes in brain.
We used ligation-free ribosome profiling to compare genome-wide TEs in mice treated with AZD-8055 and vehicle-treated mice. We then conducted a differential TE analysis comparing the treated and untreated conditions to identify genes with significant translational alterations (see “Methods”). Figure 5b shows that, overall, the amplitude of the observed alterations in TE are much larger than those found at the level of transcription alone. In addition, Fig. 5b shows that all of the canonical TOP motif-containing genes exhibit reduced TE in the brains of mice treated with the mTOR inhibitor AZD-8055. Furthermore, most of these TE changes are highly significant based on our differential translation analysis (Fig. 5c). Overall, we found 37 genes with significant TE reduction after treatment and fold change amplitudes greater than 2. Of these 37 genes, 25 were in the list of canonical TOP motif-containing genes . Of the remaining 12 genes, all but one are ribosomal proteins and all 12 genes clearly contain TOP motifs (Additional file 7: Table S2). Not only do these results further validate our ligation-free ribosome profiling technique, they also demonstrate rapid and widespread translational control of the TOP motif-containing genes by mTOR in the brain only 1 h following administration of an inhibitor.
We have demonstrated a new approach to library construction for ribosome profiling and used it to show new cell type-specific patterns of protein synthesis in the brain. Through the use of template switching, we bypassed several inefficient and time-consuming steps associated with conventional ribosome profiling, such as ligation, and eliminated almost all gel purification steps. Using ligation-free ribosome profiling, we can construct libraries from as little as 1 ng of purified RNA footprints and the resulting library complexity and gene detection efficiency are comparable to those of conventional ribosome profiling. Furthermore, due to the elimination of several enzymatic and precipitation steps, the amount of time required to perform library construction with ligation-free ribosome profiling is as little as one day following isolation of RNA footprints.
Although ligation-free ribosome profiling offers the advantages described above, conventional ribosome profiling has some advantages in terms of resolving ribosome footprints. Both the 3′ and 5′ ends of ligation-free ribosome profiling reads are associated with low complexity sequences. Specifically, the 3′ end is poly(dA) and the 5′ end is another low complexity sequence. This complicates precise determination of the ribosome footprint insert sequence, a problem that is resolved by ligation of specific sequence adapters in the conventional library construction protocol. Nonetheless, for the purposes of measuring translation efficiency and other metrics presented here, this shortcoming does not pose a major issue.
Using ligation-free ribosome profiling, we have shown that genes expressed in specific cell types exhibit distinct distributions of translation efficiency in the brain. Interestingly, most neuron-specific genes have either relatively high or low translation, implying that they are under a high level of translational regulation. We validated these findings in Camk2a-expressing neurons using the RiboTag system, which allows isolation of polysomal mRNA from specific cell types. At the level of GOs, neuron-specific genes involved in synaptic function are efficiently translated as a group compared with, for example, neuron-specific ion channels. We also found that genes associated with three stages of oligodendrocyte differentiation exhibited different translation efficiencies. OPC-specific genes were translated more efficiently than genes specific to newly formed oligodendrocytes, while fully differentiated, myelinating oligodendrocyte-specific genes had the lowest translation efficiency of the three stages. We have also determined the relationship between CDS translation efficiency and the GC content and length of 5′ UTR sequences in the brain. In general, long, GC-rich 5′ UTRs are associated with low translation efficiency, consistent with the notion that genes containing highly structured 5′ UTRs are lowly translated. Finally, we observed widespread translational repression of genes containing the TOP motif in response to mTOR inhibition. Our treatment window was just 1 h, suggesting that these alterations comprise the earliest effects of competitive mTOR inhibition in the brain.
Taken together, the above results provide convincing evidence that ligation-free ribosome profiling allows rapid and quantitative translational profiling, even in complex tissues like the mammalian brain. We anticipate that the simplified procedure described here will expand the use of ribosome profiling and may enable new, low-input or larger-scale applications.
Camk2a-RiboTag mouse model
Camk2a-cre mice (JAX ID 005359) have the mouse calcium/calmodulin-dependent protein kinase II alpha (Camk2a) promoter driving Cre recombinase expression in the forebrain, specifically in principal excitatory neurons. Camk2a-cre mice were crossed to RiboTag mice (JAX ID 011029) which contain a conditional knock-in allele where exon 4 of the ribosomal protein L22 (Rpl22) is flanked by loxP sites, followed by an identical exon tagged with three repeated hemagglutinin epitope coding sequences (HA-tag). The resulting Camk2a-cre-RiboTag cross expresses the HA-tagged Rpl22 protein in principal excitatory neurons. Camk2a-cre heterozygotes were crossed to homozygous RiboTag mice and genotyped with primers for Cre (GCG GTC TGG CAG TAA AAA CTA TC (transgene), GTG AAA CAG CAT TGC TGT CAC TT (transgene), CTA GGC CAC AGA ATT GAA AGA TCT (internal positive control forward), GTA GGT GGA AAT TCT AGC ATC ATC C (internal positive control reverse)) and for RiboTag (GGG AGG CTT GCT GGA TAT G (forward), TTT CCA GAC ACA GGC TAA GTA CAC (reverse)).
Previous reports have shown that recombination with the Camk2a promoter-driven cre begins during the third postnatal week and is completed by the fourth postnatal week; therefore, we chose to use mice that were 3 months old for all experiments .
Drug delivery and tissue collection
AZD-8055 (Selleckchem) was dissolved in Captisol and diluted to a final Captisol concentration of 30 % (w/v). A single dose of AZD-8055 was administered by oral gavage (100 mg/kg). Vehicle consisted of 30 % captisol and was also delivered by oral gavage. Camk2a-cre-RiboTag mice were sacrificed 1 h after AZD-8055 or vehicle administration; two mice were used per condition. Cervical dislocation was performed and the right frontal lobe of the brain was collected and snap-frozen in liquid nitrogen prior to polysome extraction. The remaining brain lobes were fixed in 4 % paraformaldehyde for 48 h and embedded in paraffin for histological analysis.
Fixed brains were embedded in paraffin and tissue sections (5 μm) were used for staining. To remove excess paraffin, slides were immersed in xylene then rehydrated by incubation in 100, 95, and 75 % ethanol. Slides were washed in phosphate-buffered saline (PBS) then water. For antigen retrieval 10 mM citrate buffer (pH 6.0) was heated and slides were immersed for 20 minutes, followed by PBS washes. Sections were then permeabilized with 0.5 % Triton-X100 in PBS for 15 minutes, blocked in 5 % goat serum for 1 h, and incubated with primary antibodies overnight at 4 °C. Sections were washed three times in PBS and incubated with AlexaFluor-conjugated secondary antibodies (1:1000, Invitrogen) for 1 h at room temperature and counterstained with DAPI. Stained tissue sections were imaged using a Nikon TE2000 epifluorescence microscope.
The following primary antibodies were used for immunofluorescence and western blotting: mouse monoclonal anti-HA.11 ascites (1:500, Biolegend #901515), rabbit anti-pS6 S240/244 (1:500, Cell Signaling #2215), rabbit anti-NeuN (1:500, Cell Signaling #12943), rabbit anti-pS6 S235/236 (1:1000, Cell Signaling #2211), rabbit anti-S6 (1:1000, Cell Signaling #2217), rabbit anti-β-actin (1:1000, Cell Signaling #4970S). The following secondary antibodies were used for immunofluorescence and western blotting: goat anti-rabbit Alexa 488 (1:1000, Invitrogen #A11008) and goat anti-mouse Alexa 568 (1:1000, Invitrogen #A11031).
Western blot analysis
Tissue was collected 1 h after vehicle or AZD-8055 administration (20 mg/kg or 100 mg/kg AZD-8055). The right frontal brain lobe was lysed from male mice that were 12 weeks old. Tissue was lysed in 1 mL cell extraction buffer (Invitrogen #FNN10011) supplemented with protease (Sigma #P7626) and phosphatase inhibitors (Sigma#P5726, #P0044) with a Dounce homogenizer. Lysate was centrifuged and the supernatant was collected for total protein quantification. Total protein (30 μg) was loaded to a NuPAGE 4-12 % Bis-Tris gel and subject to gel electrophoresis according to the manufacturer’s instructions (Invitrogen #NP0321BOX). Bands were detected by fluorescent imaging using the Typhoon imaging system.
Tissue processing for RNA
Snap frozen tissue samples (5 mg) were homogenized at 4 °C with a Dounce homogenizer in 1 mL of polysome lysis buffer (20 mM Tris-HCl pH 7.5, 250 mM NaCl, 15 mM MgCl2,1 mM DTT, 0.5 % Triton X-100, 0.024 U/ml TurboDNase, 0.48 U/mL RNasin, and 0.1 mg/ml cycloheximide). Homogenates were centrifuged for 10 minutes at 4 °C, 14,000 × g. The supernatant was removed and used for the isolation of ribosome footprints, total RNA, and polysome immunoprecipitation (IP). SUPERase-In (0.24U/mL) was added to the lysate used for polysome IP to prevent RNA degradation.
Lysate (100 μL) was used as the input, from which RNA was extracted using the RNeasy Mini Kit (Qiagen). The remaining lysate was used for indirect IP of polysomes. We coupled 15 μL of mouse monoclonal anti-HA.11 (ascites, Biolegend) to lysate with rotation at 4 °C for 4 h. We used 150 μL of protein G-coated Dynabeads (30 mg/mL, Life Technologies) and washed them with 600 μL polysome lysis buffer three times. The conjugated lysate was then added to protein G-coated Dynabeads and incubated with rotation at 4 °C overnight. Beads were then washed three times with 500 μL of polysome lysis buffer. RNA was extracted from magnetic beads with polysome release buffer (20 mM Tris-HCl pH 7.3, 250 mM NaCl, 0.5 % Triton X-100, 50 mM EDTA) four times for 5 minutes each (140 μL × 4). RNA from the pooled supernatants (560 μL) was then extracted with the RNeasy Mini Kit (Qiagen) and RNA integrity was assessed using a Bioanalyzer (Agilent).
RNA sequencing libraries
RNA samples were provided to the Columbia Sulzberger Genome Center for poly(A)-selection and RNA-Seq using the Illumina TruSeq kit. A total of four RNASeq libraries were generated for AZD-treated and vehicle control mice. RNASeq libraries were generated from matched samples used in ligation-free ribosome profiling experiments. Four additional libraries were sequenced from non-ribosome profiling matched samples; two total input samples and two matched HA-IP samples.
Polysome profiling and qPCR validation
The left frontal lobe, contralateral to the portion used to generate a ligation-free ribosome profiling library, was conserved and used to generate qPCR data from polysome profiles. The tissue sample was lysed with a Dounce homogenizer, as previously described, and fractionated with a 15–50 % sucrose gradient at 37,000 RPM for 3.5 h. Polysome profiles were obtained and RNA was extracted from fractions using an RNA Clean and Concentrator column (Zymo). cDNA was generated with a high-capacity RNA to cDNA kit (Life Technologies). qPCR was performed on each fraction with five probes representing genes with either high or low TE as found by ribosome profiling: SYT1 (Mm00436858_m1), SNAP25 (Mm01276449_m1), TGFB1 (Mm01178820_m1), PKD1 (Mm00465434_m1), and TRPV6 (Mm00499069_m1) (ThermoFisher). TaqMan Universal Master Mix (Life Technologies) was used to setup qPCR reactions and a Bio-Rad CFX-96 was used to amplify and read plates. All experiments were performed in triplicate. CQ was determined for each sample and an average CQ number was calculated for each set of triplicates. CQ numbers were converted using abundance = 21−CQ and the highest value for each gene normalized to 1.
These values were then plotted according to the polysome peak from which they were obtained.
Ribosome Profiling Sensitivity Measurement
A 34-base RNA oligo, ‘AUGUACACGGAGUCGAGCUCAACCCGCAACGCGA[Phos]’, was purchased from Sigma and used to generate conventional and ligation-free ribosome profiling libraries. Conventional libraries were generated using the protocol described in Ingolia et al.  using the primers described in Gonzalez et al. . The template oligo was serially diluted to the following concentrations; 100 ng, 10 ng, 1 ng, 0.1 ng and 0.01 ng. Following dephosphorylation, both conventional and ligation-free construction schemes were used to attempt to generate libraries at each concentration. For the final PCR step for all libraries in both protocols, PCR was restricted to 9 cycles with 90 % of the remaining material. Samples were diluted as necessary and assessed with a High-Sensitivity DNA Bioanalyzer Chip (Agilent).
Poly(A) tailing of size selected fragments
Ribosomal footprints were isolated with a sucrose cushion, size-selected, and dephosphorylated as previously described [2, 11]. Following dephosphorylation of size-selected footprints, we determined the concentration of input material using a Bioanalyzer (RNA 6000 Pico Chip, Agilent Technologies). We found that quantification with a Bioanalyzer was more accurate than with a RNA Qubit or Nanodrop due to the presence of Glycoblue (Ambion) as a precipitant. We used a newly developed kit for small RNA library construction (SMARTer® smRNA-Seq Kit for Illumina®, Clontech catalog number 635030) to generate ligation-free ribosome profiling libraries. Between 1 and 5 ng of size-selected material was used as input and diluted with water to a total volume of 7 μL. Ensuring that reagents remained on ice, polyadenylation mix was prepared by combining 7 μL of RNA input with 2.5 μL of mix 1, which includes poly(A) polymerase. After adding the polyadenylation mix, samples were incubated for 5 minutes at 16 °C. Following incubation, samples were immediately placed on ice to ensure the poly(A) tailing reaction did not continue.
Reverse transcription and template switching
Proceeding from the previous step within 5 minutes, samples were allowed to cool for 1 minute on ice. A 3′ smRNA dT primer (1 μL) was added to each tube and mixed by pipetting. Samples were incubated for 3 minutes at 72 °C and then transferred to ice for 2 minutes. During this incubation step, RT master mix was prepared. The RT master mix consisted of 6.5 μL smRNA mix 2, 0.5 μL RNase inhibitor, and 2 μL PrimeScript RT and 9 μL was added to each sample and mixed by pipetting. Samples were placed in a thermocycler pre-heated to 42 °C and incubated at 42 °C for 1 h followed by a 10-minute incubation at 70 °C to heat-inactivate the enzyme.
Ribosomal RNA depletion
Ribosomal RNA (rRNA) was depleted from samples with a subtraction oligo pool as described previously . Briefly, the subtraction oligo pool consists of several dozen short biotinylated oligos complementary to rRNA fragments that commonly contaminate mammalian ribosome profiling libraries. Following hybridization, the oligos are removed with magnetic streptavidin beads. We combined 10 μL of the previous RT reaction with 2 μL of the subtraction oligo pool and mixed. The mixture was heated to 100 °C for 90 s in a thermocycler. Following heating, the mixture was placed into a 100 °C heatblock and allowed to cool to 37 °C. Upon reaching 37 °C, the mixture was removed from the heatblock and incubated for 15 minutes at 37 °C in a thermocycler. While the depletion mixture incubated, 37.5 μL myOne Strepavidin C1 DynaBeads (Invitrogen) were prepared for each sample. Streptavidin beads were washed three times with an equal volume of 1× polysome buffer. Following the final wash, beads were split into 25 μL and 12.5 μL aliquots. After removing the polysome buffer from the 25 μL aliquot of beads, the depletion mixture was added to the beads and the resulting mixture was incubated for 15 minutes at 37 °C. The depletion mixture was then recovered from the beads using a magnet and added to the second, 12.5 μL aliquot of beads. The resulting mixture was incubated for 15 minutes at 37 °C. Ensuring no beads were carried over, the depleted RT reaction was then recovered using a magnet.
PCR library amplification
The SeqAmp DNA polymerase included in the SMARTer® smRNA-Seq Kit (Clontech) was used to amplify cDNA from the depleted RT reactions. For the experiments reported, we used the low-throughput primer set from Clontech (catalog number 634844) but have also had success using Clontech’s high-throughput primers (included in the SMARTer® smRNA-Seq Kit). PCR reactions were incubated for 1 minute at 98 °C followed by 12 cycles of a two-step protocol of 98 °C for 10 s and 68 °C for 10 s.
Purification of libraries
Purification is necessary due to the presence of primers and other contaminants from upstream reactions. Furthermore, it is critical to ensure reduction of a non-product secondary peak ~25 nucleotides smaller than the product peak. The secondary peak increases linearly with PCR cycle number and is inversely related to total input used. Because the secondary peak is similar to the expected peak size from ribosome profiling and can interfere with sequencing, it is essential to ensure that it is at least less than half the size of the product-peak. We performed two rounds of purification with AMPure XP beads (Beckman Coulter) at a 1.8× and 1.2× ratio (due to differences in product size, the ratio must be altered when used with the high-throughput primer set).
Validation of ribosome profiling libraries
We used the Qubit dsDNA High-Sensitivity kit (Life Technologies) to quantify libraries prior to pooling. Libraries were evaluated for the presence of primer and secondary peak with the High-Sensitivity Bioanalyzer DNA chip (Agilent Technologies). In order to fully remove primers and to reduce the contribution of the aforementioned no-insert secondary peak, some libraries require an additional round of 1.2× or 1.0× AMPure XP bead cleanup. Sequencing was performed on a NextSeq 500 desktop sequencer with a 75 cycle high-output kit (Illumina). We obtained between 20 and 50 million demultiplexed, pass-filtered, single-end reads for each sample.
Bioinformatic analysis of ribosome profiling and RNA-Seq libraries
Each read contains a G-rich region from terminal transferase activity, followed by a ribosome footprint and a poly(A) tail. The first 5 and last 20 bases of each read were removed with fastx_trimmer from the FASTX Toolkit. Because the poly(A) tail can appear at different points in the read, stretches of “AAAAAAAA” at the 3′ end of reads were removed with fastx_clipper; reads shorter than 15 bases after trimming and clipping were discarded. Contaminating rRNA reads were removed by mapping all reads to a rRNA reference library with Bowtie2, allowing for one error and outputting reads which did not align . Reads which did not map to the rRNA reference were aligned to the genome and transcriptome with TopHat2 without looking for novel junctions. Following mapping, read counting was performed with HTSeq set in interstrict mode. We obtained between four and ten million reads uniquely mapped to the CDS per ribosome profiling sample. RNA-Seq data were sequenced and analyzed as previously reported . We obtained between nine and ten million reads uniquely mapped to the CDS per RNA-Seq sample and 17–19 million reads uniquely mapped to exons.
Calculation of unique fragments
The number of unique fragments was calculated for both methods of ribosome profiling with Picard Tools downloaded from the Broad Institute. Picard Tools was used in MarkDuplicates mode and was run using files downsampled from the original.bam file output from TopHat that was previously generated for each sample. Downsampling was performed with fastq-sample from the fastq-tool suite. Following sorting and indexing with SamTools, the number of unique fragments was determined with Picard Tools.
Analysis of translational activity and RiboTag enrichment
To analyze differential translation efficiency between the control and AZD-treated samples, we used the recently reported RiboDiff algorithm with the CDS-mapping RNA-Seq and ribosome profiling reads as input . RiboTag enrichment scores were calculated from two RiboTag IP experiments and two homogenate experiments. RiboTag enrichment scores were calculated for each gene by first normalizing counts found in RiboTag and homogenate samples by size factors generated from DESeq2. Following normalization, enrichment scores were calculated by dividing normalized RiboTag counts by normalized homogenate counts.
Translation efficiency was also calculated on a per-sample basis by normalizing ribosome profiling and RNA-Seq counts by size factors from DESeq2 and dividing ribosome profiling counts by RNA-Seq counts. We thresholded downstream analyses by removing genes that had less than 37 counts in ribosome profiling and RNASeq data. When the TE of both samples in a group was used, the threshold was increased to 75 counts.
Cell type-specific specific lists
This resulted in seven cell type-specific enrichment scores between −0.5 and 0.5 for each gene. This value was later recalculated without including newly formed oligodendrocytes as a cell type (in order to improve enrichment among the remaining cell types due to significant overlap between myelinating and newly formed oligodendrocytes). These cell-type enrichment rank lists were later used in gene set enrichment analysis (GSEA) and to define which genes were most associated with specific cell types. Cell type-specific genes were defined as having an enrichment score greater than 0.2.
Gene set enrichment analysis
In order to determine the role of translational regulation in cell type-specific genes, we performed a GSEA with software downloaded from the Broad Institute . In all instances of GSEA we performed a “classic” GSEA analysis in pre-ranked mode. Gene sets were constructed from previously calculated and thresholded TE values for each sample individually and for combined samples as described above. Between 10,201 and 9904 genes (difference due to previously mentioned thresholding) were ranked based on their TE calculated from untreated RiboTag brains into bins. Equal sized bins spanning 0.75 TE units were constructed around the median and populated with genes based on their TE rank. This was then used as the gene set input for GSEA for each sample.
Rank lists were then generated for each pairwise combination of cell types composed of calculated comparison scores for each gene. GSEA was run with the same settings as before using the previously generated gene sets based on TE scores. False discovery rate (FDR)-corrected p values are plotted in Additional file 3: Figure S3.
As a secondary means of displaying the cell type-specific translational landscapes we observed, we generated lists of cell type-specific GOs. In order to calculate the enrichment of cell type-specific genes in GOs, a list of 1400 GOs taken from the iPAGE database  was used to create gene sets where each set was a single ontology. NES for the enrichment of cell type-specific genes in individual ontologies were produced using this gene set in conjunction with previously generated rank lists comprised of enrichment scores (one for each cell type). A GO was defined as being enriched in an individual cell type if the NES for that cell type was at least three units higher than the next highest NES for that GO. Median TE was calculated for genes within enriched ontologies and plotted.
5′ UTR analysis
The number of ribosome profiling and RNA-Seq reads mapped to the 5′ UTR were counted with HTSeq-counts set to region-interstrict mode for each matched sample. Cell type-specific genes were defined for this analysis as having a previously calculated enrichment value greater than 0.2. The fraction of cell type-specific genes with 5′ UTR ribosomal density was calculated as the percentage of cell type-specific genes with at least one ribosomal footprint in the 5′ UTR region. Upstream AUG sequences were identified with a custom python script and defined as any AUG sequence found within the 5′ UTR region of a gene in genes with 5′ UTR density. The median TE was calculated for cell type-specific genes as well as for the subgroups of cell type-specific genes with 5′ UTR density and containing uAUG and genes containing 5′ UTR density without uAUG. The weighted average of 5′ UTR length for each gene was calculated using isoform abundance information from Cufflinks. Cufflinks was quantified against a reference transcript annotation and otherwise run with default settings. GC content of 5′ UTRs was calculated in the same manner using isoform abundance information from Cufflinks. Genes were sorted into bins defined by GC content and length and the median TE was calculated. The significance of the change in TE due to 5′ UTR GC content and 5′ UTR length was calculated using the Mann–Whitney U test.
cDNA, complementary DNA; CDS, coding sequence; GO, gene ontology; GSEA, gene set enrichment analysis; HA, hemagglutinin; IP, immunoprecipitation; mTOR, mammalian target of rapamycin; NES, normalized enrichment score; OPC, oligodendrocyte precursor cell; ORF, open reading frame; PBS, phosphate-buffered saline; PCR, polymerase chain reaction; qPCR, quantitative PCR; RT, reverse transcription; TE, translation efficiency; TOP, terminal oligopyrimidine; uORF, upstream open reading frame; UTR, untranslated region
The authors acknowledge technical assistance from and valuable discussions with Dr. Christian Gonzalez and both Rebecca Solomon and Roxanne Ko in the Columbia Sulzberger Genome Center. They also acknowledge Drs. Nathalie Bolduc and Andrew Farmer from Clontech Laboratories for technical assistance and discussion as well as for sharing reagents.
NH was supported by grant F31NS089106 from NIH/NINDS. DT was supported by grant F31CA200375 from NIH/NCI. GT was supported by grant K01MH096956 from NIH/NIMH. PAS was supported by grant K01EB016071 from NIH/NIBIB. PAS and GT were supported by grant W81XWH-15-1-0112 from DOD and grant 345915 from the Simons Foundation. PAS and PC were supported by grant R03NS090151 from NIH/NINDS.
Availability of data and materials
The dataset supporting the conclusions of this article is available on the Gene Expression Omnibus under accession GSE78163.
NH, DT, SDS, PC, and PAS designed experiments. NH conducted ligation-free ribosome profiling, sensitivity, and qPCR experiments and DT and SDS processed the mouse brain tissue and conducted RiboTag RNA-Seq experiments. DT conducted the animal experiments, including the AZD-8055 treatment experiments. DT generated the tissue immunofluorescence data. GT generated the Camk2a-RiboTag mouse. NH and PAS analyzed the deep sequencing data. NH, DT, PC, and PAS wrote the paper. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Ethical approval for experiments involving animals was granted by Columbia’s Institutional Animal Care and Use Committee under protocol number AC-AAAN5600.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Ingolia NT, Ghaemmaghami S, Newman JR, Weissman JS. Genome-wide analysis in vivo of translation with nucleotide resolution using ribosome profiling. Science. 2009;324(5924):218–23.View ArticlePubMedPubMed CentralGoogle Scholar
- Ingolia NT, Lareau LF, Weissman JS. Ribosome profiling of mouse embryonic stem cells reveals the complexity and dynamics of mammalian proteomes. Cell. 2011;147:789–802.View ArticlePubMedPubMed CentralGoogle Scholar
- Brar GA, Yassour M, Friedman N, Regev A, Ingolia NT, Weissman JS. High-resolution view of the yeast meiotic program revealed by ribosome profiling. Science. 2012;335(6068):552–7.View ArticlePubMedGoogle Scholar
- Calviello L, Mukherjee N, Wyler E, Zauber H, Hirsekorn A, Selbach M, Landthaler M, Obermayer B, Ohler U. Detecting actively translated open reading frames in ribosome profiling data. Nat Methods. 2016;13(2):165–70.Google Scholar
- Lareau LF, Hite DH, Hogan GJ, Brown PO. Distinct stages of the translation elongation cycle revealed by sequencing ribosome-protected mRNA fragments. Elife. 2014;3:e01257.View ArticlePubMedPubMed CentralGoogle Scholar
- Jan CH, Williams CC, Weissman JS. Principles of ER cotranslational translocation revealed by proximity-specific ribosome profiling. Science. 2014;346(6210):1257521.View ArticlePubMedPubMed CentralGoogle Scholar
- Thoreen CC, Chantranupong L, Keys HR, Wang T, Gray NS, Sabatini DM. A unifying model for mTORC1-mediated regulation of mRNA translation. Nature. 2012;485(7396):109–U142.View ArticlePubMedPubMed CentralGoogle Scholar
- Hsieh AC, Liu Y, Edlind MP, Ingolia NT, Janes MR, Sher A, Shi EY, Stumpf CR, Christensen C, Bonham MJ, et al. The translational landscape of mTOR signalling steers cancer initiation and metastasis. Nature. 2012;485(7396):55–U196.Google Scholar
- Oh E, Becker AH, Sandikci A, Huber D, Chaba R, Gloge F, Nichols RJ, Typas A, Gross CA, Kramer G, et al. Selective ribosome profiling reveals the cotranslational chaperone action of trigger factor in vivo. Cell. 2011;147(6):1295–308.Google Scholar
- Gonzalez C, Sims JS, Hornstein N, Mela A, Garcia F, Lei L, Gass DA, Amendolara B, Bruce JN, Canoll P, et al. Ribosome profiling reveals a cell-type-specific translational landscape in brain tumors. J Neurosci. 2014;34(33):10924–36.Google Scholar
- Ingolia NT, Brar GA, Rouskin S, McGeachy AM, Weissman JS. The ribosome profiling strategy for monitoring translation in vivo by deep sequencing of ribosome-protected mRNA fragments. Nat Protoc. 2012;7(8):1534–50.View ArticlePubMedPubMed CentralGoogle Scholar
- Guo H, Ingolia NT, Weissman JS, Bartel DP. Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature. 2010;466(7308):835–40.View ArticlePubMedPubMed CentralGoogle Scholar
- Weinberg DE, Shah P, Eichhorn SW, Hussmann JA, Plotkin JB, Bartel DP. Improved ribosome-footprint and mRNA measurements provide insights into dynamics and regulation of yeast translation. Cell Rep. 2016;14(7):1787–99.Google Scholar
- Luo GX, Taylor J. Template switching by reverse transcriptase during DNA synthesis. J Virol. 1990;64(9):4321–8.PubMedPubMed CentralGoogle Scholar
- Zhu YY, Machleder EM, Chenchik A, Li R, Siebert PD. Reverse transcriptase template switching: a SMART approach for full-length cDNA library construction. Biotechniques. 2001;30(4):892–7.PubMedGoogle Scholar
- Cho J, Yu NK, Choi JH, Sim SE, Kang SJ, Kwak C, Lee SW, Kim JI, Choi DI, Kim VN, et al. Multiple repressive mechanisms in the hippocampus during memory formation. Science. 2015;350(6256):82–7.Google Scholar
- Kelleher 3rd RJ, Govindarajan A, Jung HY, Kang H, Tonegawa S. Translational control by MAPK signaling in long-term synaptic plasticity and memory. Cell. 2004;116(3):467–79.View ArticlePubMedGoogle Scholar
- Kandel ER. The molecular biology of memory storage: a dialogue between genes and synapses. Science. 2001;294(5544):1030–8.View ArticlePubMedGoogle Scholar
- Davis HP, Squire LR. Protein synthesis and memory: a review. Psychol Bull. 1984;96(3):518–59.View ArticlePubMedGoogle Scholar
- Zhang Y, Chen K, Sloan SA, Bennett ML, Scholze AR, O’Keeffe S, Phatnani HP, Guarnieri P, Caneda C, Ruderisch N, et al. An RNA-sequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex. J Neurosci. 2014;34(36):11929–47.Google Scholar
- Sanz E, Yang L, Su T, Morris DR, McKnight GS, Amieux PS. Cell-type-specific isolation of ribosome-associated mRNA from complex tissues. Proc Natl Acad Sci U S A. 2009;106(33):13939–44.View ArticlePubMedPubMed CentralGoogle Scholar
- Chresta CM, Davies BR, Hickson I, Harding T, Cosulich S, Critchlow SE, Vincent JP, Ellston R, Jones D, Sini P, et al. AZD8055 is a potent, selective, and orally bioavailable ATP-competitive mammalian target of rapamycin kinase inhibitor with in vitro and in vivo antitumor activity. Cancer Res. 2010;70(1):288–98.Google Scholar
- Pike KG, Malagu K, Hummersone MG, Menear KA, Duggan HM, Gomez S, Martin NM, Ruston L, Pass SL, Pass M. Optimization of potent and selective dual mTORC1 and mTORC2 inhibitors: the discovery of AZD8055 and AZD2014. Bioorg Med Chem Lett. 2013;23(5):1212–6.Google Scholar
- Islam S, Kjallquist U, Moliner A, Zajac P, Fan JB, Lonnerberg P, Linnarsson S. Characterization of the single-cell transcriptional landscape by highly multiplex RNA-seq. Genome Res. 2011;21(7):1160–7.Google Scholar
- Ramskold D, Luo S, Wang YC, Li R, Deng Q, Faridani OR, Daniels GA, Khrebtukova I, Loring JF, Laurent LC, et al. Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nat Biotechnol. 2012;30(8):777–82.Google Scholar
- Picelli S, Faridani OR, Bjorklund AK, Winberg G, Sagasser S, Sandberg R. Full-length RNA-seq from single cells using Smart-seq2. Nat Protoc. 2014;9(1):171–81.View ArticlePubMedGoogle Scholar
- Larsson O, Sonenberg N, Nadon R. Identification of differential translation in genome wide studies. Proc Natl Acad Sci U S A. 2010;107(50):21487–92.View ArticlePubMedPubMed CentralGoogle Scholar
- Sharma K, Schmitt S, Bergner CG, Tyanova S, Kannaiyan N, Manrique-Hoyos N, Kongi K, Cantuti L, Hanisch UK, Philips MA, et al. Cell type- and brain region-resolved mouse brain proteome. Nat Neurosci. 2015;18(12):1819–31.Google Scholar
- Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545–50.Google Scholar
- Chiu IM, Morimoto ET, Goodarzi H, Liao JT, O’Keeffe S, Phatnani HP, Muratet M, Carroll MC, Levy S, Tavazoie S, et al. A neurodegeneration-specific gene-expression signature of acutely isolated microglia from an amyotrophic lateral sclerosis mouse model. Cell Rep. 2013;4(2):385–401.Google Scholar
- Steward O, Schuman EM. Protein synthesis at synaptic sites on dendrites. Annu Rev Neurosci. 2001;24:299–325.View ArticlePubMedGoogle Scholar
- Arribere JA, Gilbert WV. Roles for transcript leaders in translation and mRNA decay revealed by transcript leader sequencing. Genome Res. 2013;23(6):977–87.View ArticlePubMedPubMed CentralGoogle Scholar
- Dvir S, Velten L, Sharon E, Zeevi D, Carey LB, Weinberger A, Segal E. Deciphering the rules by which 5′-UTR sequences affect protein expression in yeast. Proc Natl Acad Sci U S A. 2013;110(30):E2792–801.Google Scholar
- Meyuhas O. Synthesis of the translational apparatus is regulated at the translational level. Eur J Biochem. 2000;267(21):6321–30.View ArticlePubMedGoogle Scholar
- Wong M. Mammalian target of rapamycin (mTOR) pathways in neurological diseases. Biomed J. 2013;36(2):40–50.View ArticlePubMedGoogle Scholar
- Wong M. A critical review of mTOR inhibitors and epilepsy: from basic science to clinical trials. Expert Rev Neurother. 2013;13(6):657–69.View ArticlePubMedGoogle Scholar
- Tsien JZ, Chen DF, Gerber D, Tom C, Mercer EH, Anderson DJ, Mayford M, Kandel ER, Tonegawa S. Subregion- and cell type-restricted gene knockout in mouse brain. Cell. 1996;87(7):1317–26.Google Scholar
- Zong Y, Karaletsos T, Drewe P, Sreedharan VTT, Singh K, Hans-Guido W, Ratsch, G. RiboDiff: detecting changes of translation efficiency from ribosome footprints. bioRxiv. 2015. doi:10.1101/017111.
- Goodarzi H, Elemento O, Tavazoie S. Revealing global regulatory perturbations across human cancers. Mol Cell. 2009;36(5):900–11.View ArticlePubMedPubMed CentralGoogle Scholar