The RNA immunoprecipitation (RIP) has been used previously to identify targets of RBPs involved in neurological dysfunction (Buckanovich and Darnell, 1997; Napoli et al., 2008; van der Brug et al., 2008; Fernandez et al., 2015). High-throughput RIP-SEQ serves as a useful tool in determining RNA populations bound to proteins involved in local translation. RIP, similar to the protein-based immunoprecipitation procedure, has been optimized in order to preserve the RNA-protein complex during the lysing step such as gentle-freeze thawing (Keene et al., 2006; Oeffinger et al., 2007; Jain et al., 2011; Dahm et al., 2012). Other modifications in RIP protocols are to ensure that free RNAs released during lysis do not bind non-specifically to the beads or the RBP, a phenomenon that has been observed previously and contributes to the high background of some RIP experiments (Mili and Steitz, 2004). Following some of the procedures outlined by Jain and colleagues, it is possible to achieve minimal or negligible levels of background binding (Jain et al., 2011). While crosslinking with formaldehyde to bind the protein–protein or protein–RNA structures together may help in ensuring complex isolation, in some cases it may not lower the level of background binding (Penalva et al., 2004). Finally, like many other high-throughput techniques, there are limitations including epitope accessibility to the antibody, as well non-specific binding inherent with antibody-based procedures. Importantly, RIP is not able to reveal precise binding sites like other crosslinking techniques discussed below; however, it can reveal the full-length transcript of RNAs bound to the protein of interest in vivo.

Before sequencing begins, RNA obtained from cell samples are used to generate a cDNA library (Head et al., 2014). A common pitfall is degraded or impure RNA. Low quality RNA generates noisy data whose sequences can be difficult to reconstruct during the data preparation stages. Furthermore, sequencing degraded RNA leads to high variability and can impact interpretation and differential gene expression (DGE) analysis leading to the possibility of overfitting in classification analysis. The RNA integrity number (RIN, ranging from 1 to 10) is a statistical measure of RNA integrity that has been developed to assess RNA quality in a particular sample (Schroeder et al., 2006). Typically, a RIN of 7–10 is recommended for library construction unless the RNA was derived from rare tissue and obtaining higher quality RNA is costly or impossible. However, it appears possible to construct libraries and analyze data from moderately degraded RNA (RIN = 4–6) using appropriate statistical corrections (Gallego Romero et al., 2014; Sigurgeirsson et al., 2014; Cieslik et al., 2015). As an alternative to using RIN, qPCR may also be used as a quality assessment if potential transcripts are already known (Vermeulen et al., 2011). For RNA-protein interaction studies, qPCR can be used to assess the validity of target samples by using primers for known RNA targets of the immunoprecipitated RNP. This method does not assess the entire extracted RNA sample but can be a useful tool to demonstrate the success or relative quality of the RNA isolation technique.

After RNA quality control (QC), the samples typically undergo a selection stage for the removal of ribosomal RNA (rRNA). Since RNAseq will provide sequences for the most abundant RNAs in the sample, rRNAs must be removed since they comprise >90% of the RNA isolated from a cell population. There are two major strategies to approach this issue and recent work suggests that both methods introduce biases in sequence coverage (Lahens et al., 2014). However, both of these methods are currently in wide use and since no alternatives exist, efforts are ongoing to decrease bias (van Dijk et al., 2014). Since mature, processed mRNAs contain a poly(A) sequence, oligo(dT) beads can be used to select for only mature mRNAs in a technique called poly(A) selection, effectively removing the rRNA from the sample. While cheaper and of higher sensitivity compared to the alternative choice, the ribosomal RNA (rRNA) depletion technique, it will not pick out non-coding RNAs and may detect less genes overall (Cui et al., 2010; Zhao W. et al., 2014). There are still concerns of 3′ end bias for poly(A) selection sequences but recent studies suggest that this may be overcome with appropriate statistical corrections (Roberts et al., 2011). rRNA depletion utilizes beads consisting of sequences complimentary to rRNA. Currently, this method is more expensive and does not remove all rRNA (Cui et al., 2010). However, it is necessary for sequencing non-coding RNA.

Following rRNA depletion, the RNA is sheared into small fragments and converted to cDNA by reverse transcription. Unique DNA linkers are ligated onto the 3′ and 5′ ends of the isolated oligonucleotide. Special capture sites are then added on to the 3′ and 5′ ends of the molecule which allows them to be anchored onto a solid support surface during the sequencing steps. PCR amplification is often performed at this stage depending on the amount of RNA available. RNA sequencing is often multiplexed so multiple experiments can be run simultaneously. This is accomplished by attaching a barcode between the 3′ linker and the 3′ end capture site. These unique barcodes can be used to differentiate between multiple sets of samples run simultaneously (either one barcode, single index, or two, dual index). The samples are then loaded into the sequencer where they are washed over a small, thin surface known as a flow cell. The samples attach to the flow cell via the ligated capture sites. Since there are capture sites on both sides, the molecule can bend over to attach to the flow cell at both ends. A polymerase and dNTPs are introduced to amplify the DNA which is separated again into ssDNA. This process is repeated several times within an area of the flow cell to form a cluster of replicated molecules. The flow cell is then exposed to a polymerase and fluorescent dNTPs. The dNTPs are added one at a time to a cluster which fluoresces a different color depending on the incorporated base until the full sequence has been determined. The Illumina technology allows sequencing from just one side (single-read, SR) or from both ends (paired-end, PE). Paired-end runs, which increase the accuracy of genome mapping, are typically more expensive and more useful for difficult, repetitive genomic sequences or providing information about splice junctions and alternatively spliced transcripts (Williams et al., 2014). PE is not required for DGE analysis but may provide greater coverage during the genomic alignment stage. Each machine offers a maximum achievable number of output sequence reads per lane of a flow cell (with a total of eight lanes per flow cell). Therefore, the number of samples loaded per lane must be divided by the total number of possible reads to determine the number of reads each sample may receive (the HiSeq 4000 typically delivers ∼300 million total reads per lane). A standardized number of reads required per sample to successfully map to the genome, determine gene expression differences, or other experimental parameters has not yet been determined and must be considered on an experiment-to-experiment basis. Though the ENCODE consortium has released guidelines for standardized RNAseq practices they are now out-of-date given the pace the field is moving. However, some general guidelines and standards are available (Williams et al., 2014; Conesa et al., 2016).

There is no one standard analysis pipeline for RNAseq projects as they will vary given the data. Here we present a basic overview of some considerations and software for the DGE analysis process and other downstream analyses (Oshlack et al., 2010; Rapaport et al., 2013; Williams et al., 2014; Finotello and Di Camillo, 2015; Conesa et al., 2016). Also, note that CLIP studies may require specialized experimental design and analysis using specific software (Wang et al., 2015; Zhang et al., 2015). For further help and extra information, the SeqAnswers², Biostars³, and StackExchange⁴ forums are all excellent resources. A selected list of currently popular tools for RNAseq data processing and analysis are listed in Table 2.

The first major consideration before beginning RNAseq analysis is that the process requires a considerable amount of computing power. The amount of power needed will vary depending on the size and type of experiment. As a general minimum guideline: Large disk space (1–5 TB), RAM (at least 8–32 + GB, higher is preferred for large datasets), and a multi-core CPU (8 cored preferred, higher is better since it is often the source of bottleneck) are essential for fast computation; the processing can easily take days on a slower machine and will thus be unavailable for any other use during this time. GPU computing has not currently been optimized for RNAseq. Analysis may be performed using R/Bioconductor packages⁵, specialized software with graphical interfaces (e.g., Galaxy) (Afgan et al., 2016), programming and scripting languages (e.g., R, Matlab, Python, Ruby, Perl, Java etc.), and terminal commands on a UNIX-based operating system such as Ubuntu. Analysis and preprocessing is possible on Windows operating systems but may be difficult to set up because a number of bioinformatics tools have been developed for terminal/command line interfaces that are not normally compatible with the Windows environment. A Linux setup is the most convenient for analysis but it is possible to set up some tools on Windows using the Cygwin terminal. Finally, a useful source for installing bioinformatics tools on UNIX systems can be found here: https://www.biostarhandbook.com/tools/how-to-install-everything.html.

RNAseq preprocessing and analysis roughly falls into the following pipeline:

The first requirement for a protein to be synthesized in dendrites is that the mRNA coding for that protein is either targeted to the dendrites in response to activity or constitutively resides in dendrites. Some of the earliest work in this field utilized various methods such as in situ hybridization (ISH) and microarrays in conjunction with subcellular fractionation to catalog the RNAs present at the dendrite. One of the greatest advantages ISH is that it does not rely on subcellular fractionation – the transcript can be identified in fully intact tissue or cell culture. Thus, ISH and its related methods are very useful for confirming results obtained from subcellular fraction preparations. ISH has been used to characterize the localization and distribution of the RNA transcripts for the calcium/calmodulin-dependent protein kinase II alpha (CaMKII), microtubule-associated protein 2 (MAP2) and fragile X mental retardation protein (FMRP), some of the earliest localized transcripts discovered in the field (Hinds et al., 1993; Paradies and Steward, 1997; Miyashiro et al., 2003). The development of fluorescence in situ hybridization (FISH) has allowed researchers to visualize distribution patterns and quantify the amount of localization at a much higher resolution. There have been significant technical improvements made to FISH over the years which provide greater resolution and better quantification (Swanger et al., 2011). The best example of how technical improvements in FISH have changed dogma was first described by Krause and colleagues using mRNA localization in the Drosophilia embryo as a model. Prior to this study the estimated number of mRNAs that had distinct subcellular localizations in the embryo was ∼1-10%. Lecuyer et al. (2007) screened roughly 25% of the genome and found ∼ 71% of mRNAs screened had unique subcellular distribution patterns. The authors went on to suggest that since most of the transcribed mRNAs have a distinct subcellular localization these data imply that most cellular processes are mediated through mRNA localization (Lecuyer et al., 2007). Importantly, a similar finding was suggested by Schuman and colleagues in neuronal dendrites. They also employed FISH to verify high-throughput RNA sequencing data (Cajigas et al., 2012). Herein they identified 8,379 transcripts. By subtracting transcripts related to glial cells, interneurons, nuclei, blood vessels, and mitochondria, they suggest that 2,550 transcripts are localized in axons or dendrites. High-sensitivity FISH was then used to validate the localization of 50 of these transcripts. This work identified many known synaptic transcripts as well as newly discovered transcripts that had not been previously detected. In light of these findings synapse remodeling of synapses during synaptic plasticity is due to dendritic over somatic mRNA translation (Cajigas et al., 2012). Finally, a more recent developed in FISH allows for the visualization of single RNA molecules at high resolution. Singer and colleagues have developed a method that allows the tracking of single RNA molecules during translation. The researchers labeled ribosomes and mRNA molecules and correlated their signals to determine if the mRNA was undergoing active translation (Katz et al., 2016). Thus, collectively RNA sequencing data combined with high resolution FISH, for the first time, is allowing investigators to catalog mRNAs localized to site specific dendritic compartments leading to new testable hypothesis regarding memory allocation for information storage.

While FISH by itself provides subcellular localization of specific mRNAs, quantification of mRNAs in the soma versus dendritic compartments has been challenging. The development of the NanoString nCounter Gene Expression Assay has helped resolve that issue (Kulkarni, 2011). NanoString enables the detection and quantification of up to 800 mRNA molecules (without any conversion to cDNA or PCR amplification) using colored probe pairs. Cajigas et al. (2012) used NanoString technology to answer the long-standing question of whether select mRNAs segregate, with some enriched in the soma and some enriched in dendrites and axons. While this finding had been suggested by quantitative RT-PCR comparing hippocampal synaptosomal mRNA to total lysate mRNA (Raab-Graham et al., 2006) the concern of somatic contamination in the synaptosomal fraction tempered the interpretation of this finding. Still, in cases when mRNA levels are low-abundance or quick validation is a necessity, qRT-PCR remains a valid method for confirming RNAseq or microarray data, or quantifying relative abundance levels of a specific transcript between samples (Raymaekers et al., 2009).

Labeling newly synthesized proteins has been used to validate high-throughput studies investigating local translation (Buxbaum et al., 2015a,b). Since many of these techniques rely on fluorescent microscopy, it is possible to visualize and distinguish new and old proteins within a particular neuronal compartment. Surface sensing of translation (SUnSET) utilizes puromycin, a ribosome elongation inhibitor and aminoacyl-tRNA synthetase analog, to monitor the translation of de novo protein synthesis (Schmidt et al., 2009). SUnSET uses monoclonal antibodies to detect the incorporation of puromycin into the polypeptide chain during translation. Subsequent detection and visualization by either flow cytometry or fluorescent microscopy thus indicates new protein synthesis.

Single-molecule imaging of nascent peptides (SINAPS) is another alternative approach for monitoring localized transcripts in vivo which allows one to observe the real-time translation of nascent mRNA molecules at different translational stages (Wu et al., 2016). In order to strengthen the relatively weak signal of the nascent peptide and to distinguish it from background, SINAPS draws upon SunTag, another recent method that allows for the amplification of fluorescent intensity by recruiting multiple copies of GFP to a target protein (Tanenbaum et al., 2014).

The FUNCAT (fluorescent non-canonical amino acid tagging) assay is similar to the BONCAT technique described earlier which labels newly synthesized proteins using non-canonical amino acids and click chemistry to conjugate biotin. FUNCAT, however, uses fluorescently tagged amino acids to allow for the identification of newly synthesized proteins in situ by fluorescence microscopy (Dieterich et al., 2010; Tom Dieck et al., 2012; Kos et al., 2016). The drawback to FUNCAT is the fact that while it can identify whole populations of de novo synthesized proteins, it cannot identify specific ones. To overcome this difficulty, a modification was developed using a proximity ligation assay (PLA). The use of FUNCAT in tandem with PLA allows for the identification of a newly synthesized protein of interest (tom Dieck et al., 2015). FUNCAT-PLA – also called BONCAT-PLA in (Niere et al., 2016) – utilizes a biotin antibody to identify a protein of interest that has incorporated a non-canonical amino acid, signifying new protein synthesis. Another antibody is used to identify the protein of interest itself. When the secondary antibodies (pre-conjugated to a specific oligonucleotide sequence) for the two primary antibodies are in close enough proximity, they are ligated together and undergo rolling circle replication in the presence of fluorescent nucleotides to produce a signal. Alternatively, the Puro-PLA assay utilizes puromycin, a molecule that disrupts translation resulting in the release of the newly synthesized protein which can then be identified with an anti-puromycin antibody (Buhr et al., 2015; tom Dieck et al., 2015).

Recently, FUNCAT/BONCAT-PLA has been used to assess the role of mTOR complex 1 (mTORC1) in synapse modification in vivo. mTORC1 is a protein complex that regulates local dendritic translation (Tang et al., 2002; Cammalleri et al., 2003; Stoica et al., 2011). In a recent study, Niere et al. (2016) performed mass spectrometry on different subcellular fractions of neurons derived from rat cortices after an intraperitoneal injection with rapamycin, an inhibitor of mTORC1. Notably, the mass spectrometry reports changes in protein expression and does not differentiate between protein synthesis and protein stability. To differentiate between these two cellular mechanisms, they used FUNCAT/BONCAT-PLA to confirm new protein synthesis of select candidates identified by mass spectrometry (Niere et al., 2016). Indeed, FUNCAT/BONCAT-PLA demonstrated that new protein synthesis levels of Snap25 and Gap43 altered in response to treatment with the mTOR inhibitor rapamycin, and were consistent with the observed site-specific changes in protein expression indicated by mass spectrometry. Thus, new protein assays such as FUNCAT/BONCAT-PLA can detect new protein synthesis, providing mechanistic detail that mass spectrometry hints at but does not confirm.

While FUNCAT/BONCAT-PLA is useful to detect new protein synthesis it does not provide direct evidence for new protein synthesis in dendrites. For this reason, fluorescent translation reporters fused to dendritic targeting sequences of the proteins of interest still remains the best way to visualize new protein synthesis in dendrites. This approach was first developed by Schuman and colleagues (Aakalu et al., 2001). Destabilized-GFP (dGFP) was developed to address experiments that required a fluorescent reporter for proteins with transient expression (Li et al., 1998). Schuman and colleagues capitalized on the rapid turnover properties of dGFP and created a reporter consisting of cDNA coding for a myristoylated dGFP fused between the 5′ and 3′UTRs of CaMKIIα mRNA. They reasoned that the inclusion of the UTR sequences in their reporter construct would ensure that the mRNA targets the dendrite, the addition of a myristoylation sequence (myr) tethers the reporter to the membrane and thus prevents diffusion, and after photobleaching the neuron, new GFP signal detected in the dendrite is due to mRNA translation. To ensure this was the case, they continuously photobleached the soma so that any new protein synthesized in the soma would not be detected in the dendrites. One caveat that may hinder this assay is if the protein dGFP is fused to have a higher stability than dGFP, thus preventing the rapid turnover of GFP. In spite of this limitation, myrdGFP fused to the appropriate targeting sequences can be used to investigate compartment-specific translation (Aakalu et al., 2001).

An alternative approach is the usage of photoactivatable GFP-like fluorescent proteins (PAFPs) (Lukyanov et al., 2005) which overcomes the limitation of variability in mRNA stability. Kaede is a protein that allows the separation of new and old translation on the basis of UV-induced photoconversion (Ando et al., 2002). Kaede can be fused to specific proteins with the appropriate dendritic targeting sequences and used to report new translation by comparing mean puncta intensity before and after photoconversion (Leung et al., 2006; Raab-Graham et al., 2006; Leung and Holt, 2008; Banerjee et al., 2009; Workman et al., 2015). Kaede’s tetrameric structure has a distinct advantage when it comes to the tagging of ion channels which are typically tetramers (Raab-Graham et al., 2006). For example, Kaede has been used to monitor the local translation of Kv1.1, an ion channel whose translation is regulated by a microRNA, demonstrating its use and application as a detector of new protein synthesis (Sosanya et al., 2013). The Dendra2 reporter is another photoconvertible fusion protein that functions similarly to Kaede (Gurskaya et al., 2006; Chudakov et al., 2007; Wang et al., 2009; Lee et al., 2011). Dendra2 has a monomeric structure which increases functional capabilities since it can be used in protein fusion constructs. For example, Wang and colleagues fused the Dendra2 coding sequence to the 5′ and 3′UTRs of sensorin mRNA to study local translation at the sensory neuron-motor neuron junction of an Aplysia cell culture system (Wang et al., 2009). Sensorin is a peptide neurotransmitter whose UTR sequences drive the mRNA to be concentrated at synapses, thus allowing the researchers to visualize local translation specifically at the synapse. One advantage of using Dendra2 over Kaede and other similar PAFPs is that it is activated in the 488-nm laser range, well clear of the UV-activation range and thus allowing less chance of cellular toxicity. The Venus fluorescent reporter has recently been adapted to visualize the activity-dependent dendritic localization of PSD-95 (Tatavarty et al., 2012; Ifrim et al., 2015). Ifrim et al. (2015) developed a unique fusion construct in which the Venus reporter was inserted after the 5’UTR of the PSD95 mRNA sequence. Used in this way, the Venus reporter (1) fluoresces just prior to and during translation of the PSD95 open reading frame in real-time and (2) visualizes de novo protein synthesis of PSD95 specifically in the dendritic spine (Ifrim et al., 2015). Finally, another approach to monitoring local translation is the use of biarsenical fluorescent dyes such as FlAsH and ReAsH (Ju et al., 2004; Martin et al., 2005; Rodriguez et al., 2006).

TimeSTAMP is another method that can be used to monitor new protein synthesis that does not rely on either photoconversion or chemical tags (Lin et al., 2008; Lin and Tsien, 2010). An epitope tag is attached to a protein of interest which can be removed in the presence of a specific protease. Introduction of a drug prevents this process from occurring thus allowing newly synthesized proteins to retain their tags. TimeSTAMP allows the surveying of the whole brain in living animals.

Recently, a luciferase-based approach that utilizes flash kinetics has also been developed (Na et al., 2016). This approach uses a small luciferase protein derived from Gaussia princeps whose signal is dependent on the presence of coelenteraizine. Luciferase is fused to the protein of interest so that, upon translation, luciferase will react with coelenteraizine to produce a light signal. The signal decays rapidly (<9 s for the tested Arc-luciferase construct) so translating RNAs can be unambiguously detected even if coelenteraizine is still present. Thus, flashes of light signal indicate de novo protein synthesis of the protein of interest. Importantly, this technique utilizes wide-field microscopy, whereas other reporter systems usually require confocal or 2-photon imaging to detect signals above photobleaching (Na et al., 2016), making this assay accessible to a greater number of labs. Moreover, the authors who developed this assay suggest that estimation of the number of proteins synthesized in a local translation hot spot is within reach if combined with SUnSET; however, this is yet to be verified.

There are a growing number of electron microscopy (EM) studies detecting polyribosomes near spine bases that relocate in the spine head upon memory consolidation (Ostroff et al., 2010, 2014, 2017). However, new protein detection at the EM level has been difficult due to a lack of electron dense tags useful for specific protein visualization. Shu et al. (2011) has developed a small fluorescent protein, MiniSOG (mini singlet oxygen generator), that can be fused to a protein of interest and expressed in cells, tissue, and living organisms and detected both at the light level and by EM. Fluorescence photooxidation of DAB can be achieved using fluorescence and ¹O₂ from MiniSOG. It’s been suggested that spatiotemporal control of local photogeneration of ¹O₂ will rapidly inactivate proteins of interest (Shu et al., 2011), perhaps allowing one to detect site specific translation at the light and subsequently at the EM level. Although, techniques like miniSOG hold promise, detection of specific translation at the EM level is still lacking.

An additional requirement for local protein synthesis assays is to demonstrate that the synthesis occurs in the dendrites and not in the soma. In a perfect world, researchers would sever dendrites and test for the appearance of newly synthesized proteins within isolated dendritic compartments. While Martin et al. (1997) have successfully demonstrated local synthesis in processes in the absence of a soma in cultured Aplysia neurons, it has proven far more difficult in mammalian neurons. Several labs have introduced microlesions to separate the dendrite and the soma in acute hippocampal slices (Kang and Schuman, 1996; Ouyang et al., 1999; Waung et al., 2008). Kang and Schuman (1996) used this approach to show the local synthesis of BDNF occurred in the dendrites with BDNF-mediated potentiation. While this approach appears optimal, it does have its caveats. We have noted that severing dendrites in this way can induce local protein synthesis of some proteins, often confounding the results when assessing activity-dependent local protein synthesis (unpublished observation). Thus, methods have evolved that have adapted time-lapse imaging of fluorescent proteins fused to specific dendritic targeting sequences to measure de novo protein synthesis. Such measures as continuously photobleaching the soma have been used to avoid signal from the soma confounding dendritically translated fluorescent proteins (Aakalu et al., 2001). While a significant improvement over previous methods, continuous photobleaching can be toxic to the cell.

The use of photoactivated fluorescent reporters, such as Kaede, does not require constant photoconversion, in most cases, to specifically detect protein synthesis in dendrites. However, these experiments require careful controls to be able to draw this conclusion. As observed by many laboratories, locally synthesized protein synthesis occurs in hot spots of the dendrites (Aakalu et al., 2001; Job and Eberwine, 2001; Kim T.K. et al., 2013; Sosanya et al., 2015). Therefore, measuring the rate of diffusion of the protein of interest by photoconverting part of the dendrite and measuring diffusion over time (i.e., does the photoconverted protein diffuse into the unconverted dendrite) allows one to measure diffusion rates. For example, Raab-Graham et al. (2006) used such an approach by fusing Kaede to the mRNA of Kv1.1 plus its 3′UTR. It was found that Kaede-Kv1.1 did not diffuse, and stayed localized to hot spots within the dendrite. Thus, proteins presumably function close to the site of synthesis and are easily detected in dendrites when fused to photoactivatable proteins.

Additional approaches include only activating specific synapses and measuring the increase in protein expression within the activated part of the dendrite. For example, Huber and colleagues measured Arc expression after local perfusion of the mGluR agonist DHPG in distal dendrites (Waung et al., 2008). This technique relies on the fact that Arc expression occurred only in the region of mGluR activation and that protein made in the soma cannot traffic to the distal dendrites within the limited time that new Arc protein expression was detected.

Finally, one of the more thorough experimental approaches that allows one to test for function of dendritically localized mRNAs is the generation of a mouse where the dendritic targeting sequence of the protein of interest is knocked out. Mayford and colleagues created a specific knockout deleting only the dendritic targeting sequence in the 3′UTR of CaMKIIα. Importantly, ISH demonstrated that CaMKIIα mRNA was absent from the dendritic fields while still present in the soma of the CA1 hippocampal neurons in the knockout mouse relative to the wildtype litter mate. Moreover, knockout mice showed deficits in late long-term potentiation (L-LTP), a cellular model for learning and memory and at the behavioral level during a memory consolidation test (Miller et al., 2002). While these data are impressive and striking, recent data suggesting that RNA-binding proteins that compete for mRNAs may confound the interpretation of these results (Sosanya et al., 2013). The removal of a dendritic targeting sequence within the 3′UTR containing RBP motifs could possibly free up RBPs to bind to lower affinity target mRNAs which may affect synaptic plasticity and/or behavior. Still, with the development of tools like CRISPR, where making knockouts are quicker and more cost effective, site-specific knockout of proteins will facilitate the physiological relevance of localized translation of specific targets.