technology feature Profling the dress codes of RNA-binding proteins Methods choice is broad and labs are pushing protocols to address expansive biological questions. Vivien Marx

lothes can be enabling. RNAs in a eukaryotic cell can be clothed, as some Cscientists phrase it, or unclothed1. Some RNAs in a cell, including noncoding RNAs, wear proteins that cloak certain RNA regions and expose others. RNA-binding proteins (RBPs) are powerful, versatile regulatory units. They play roles in cellular housekeeping, development, differentiation, metabolism, health and disease, a functional diversity that labs are beginning to characterize. Part of this research is profiling RBPs, for example, to learn which RNAs and proteins are connected2,3. “The proteins are what carries out the function, but the RNA is what coordinates the activity of all those proteins to put them together into a coherent unit to carry out a In a eukaryotic cell RNAs can be naked or clothed with proteins. Proteins bound to RNAs become unique function,” says California Institute RNA-binding proteins (RBPs), which are powerful gene regulators. Credit: yogysic/DigitalVision of Technology researcher Mitchell Guttman. Vectors/ Getty “Independent of the RNA, these proteins wouldn’t be able to do that role.” That, in his view, makes the RNA the “epicenter” The ‘pulldown’ material is denatured, and The wealth of RBP-profiling methods for organizing complex functions related the RNA is sequenced to identify the RNA’s calls scientists to make thoughtful choices, to RNAs such as translation, splicing, protein-binding regions. RNA-centric says Miguel Esteban, a researcher at localization. From another angle, proteins methods also apply UV-based cross-linking, Guangzhou Institutes of Biomedicine and sit at the ‘epicenter’ because they mediate and the RNA is captured from lysed cells Health. Studying RBPs with many-stepped cellular events such as enzymatic reactions with various types of baits. protocols takes much methods-related and localization, he says. RBPs and RBP Some labs seek to identify the protein troubleshooting, he says. Enrichment complexes regulate transcription and region interacting with the RNA, says strategies deliver labs a certain slant to translation, gene expression, processing Urlaub, by using mass spec after cross- their RBP data, he says. Pulling down of RNAs. linking to identify the cross-linked region RNAs and finding the proteins that interact Proteins can be the crucial players that and amino acids, down to several amino with them does not give a view of all bend, fold or squeeze the RNA, leading acids in that protein. In their experimental proteins near RBPs. to the RNA acting catalytically as in the design, he observes that some researchers Each method has experimental case of the eukaryotic spliceosome, which can be biased about protein–RNA limitations but all methods share that they is a kind of 3D RNA reaction center, says interaction: they sometimes assume that are isolation methods that enrich a target of Henning Urlaub, who runs the bioanalytical proteins lacking known RNA-binding motifs interest, says Columbia University MD/PhD mass spectrometry group at the Max don’t bind RNAs, but they do, he says. Also, student Stefanie Gerstberger, who tallied Planck Institute for Biophysical Chemistry some protein regions that bind strongly to human RBPs as part of her PhD research in Göttingen. The fact that RNAs tend RNA do not cross-link well. completed in 2016 in Tom Tuschl’s lab at to be complexed with proteins opens up Rockefeller University. With untargeted numerous possibilities for thousands of approaches “you get a very broad unspecific protein structures and domains to be picture,” she says. That picture can, for combined with presumably more than a example, be used to assess the general state thousand different RNA structures. of cells, also transcriptome-wide. Some RBP-related methods emphasize Many RBP resources exist for a data- protein capture and purification, while hunting researcher, says Gerstberger, with others highlight RNA capture, says Guttman. information from a variety of cell types Mass spectrometry helps to analyze proteins and with varying data depths. As is typical attached to the RNA. In the more protein- in biochemistry, users will see variability centric approaches, UV light cross-links the between experimental results, “so it’s often RNAs and associated proteins, which are easier and better to do the experiment from then immunoprecipitated with antibodies. scratch to control for these factors,” she says.

Nature Methods | VOL 15 | SEPTEMBER 2018 | 655–658 | www.nature.com/naturemethods 655 technology feature

Although RBP methods can be binned as RBP acts on RNA RNA acts on RBP protein-centric or RNA-centric, it takes both RNA-binding protein bins to learn about the biology of RBPs, says RNA Guttman. “I think it’s a split that comes from people’s backgrounds,” says Matthias Hentze, RNA-binding domain? who directs the European Molecular Biology RNA-binding domain Laboratory in Heidelberg and who develops and uses RBP-profiling methods. Some RNA-binding protein labs focus more on nucleic acids, while RNA others study a few proteins or take a wider proteomics approach. “Biology doesn’t care about the background of the investigator,” he Function Localization Processing Stability Localization says. “You need to consider both equally.” Modification Translation Interactions Stability

It’s two-way cross-talk: RNA-binding proteins (RBPs) can regulate RNAs, and RBPs can be regulated by Census-taking RNA. Credit: Hentze lab, EMBL. Adapted with permission from ref. 3, Springer Nature Around 5–10% of the human proteome can bind RNA. To date no census of all RNA-binding proteins has taken place; and mass spec data from RNA–protein Wild, Wild West of every possible protein in the number of known RBPs remains an cross-linking experiments, Gerstberger’s the cell is a culprit,” he says. estimate. Studies lead scientists to believe tally reached 1,542 RBPs in human cells. there are between 1,000 and 2,000 RBPs in A few more might emerge, she says, such Cross-linking mammalian cells. as proteins that bind RNA in a specific or When his lab members are cross-linking As she manually curated RBPs in human nonspecific manner. They may belong to RNAs to proteins, such as with UV light, cells, based on data of mRNA isoforms from an uncharacterized family or they might Esteban reminds them to include a non- the vertebrate genome browser Ensembl be singular members of a protein class. For cross-linked sample and to set up other an average mammalian cell under standard experimental checkpoints. This can add a culture conditions, around 2,000 RBPs might day or two to the start of an experiment, but be close to the true number of proteins that running a gel sooner can spare pain later. He bind RNA with a meaningful biological also recommends never losing sight of how purpose, says Hentze. He recommends that the more dominant proteins can overshadow when labs look across species, they watch for others. For example, RNA species in the cell species-specific differences, such as particular such as some ncRNAs are important for cell RBPs an organism might have evolved. function, “yet they represent a very small Even as RBP tallies vary, “I think it’s more percentage of the total RNA output,” he says. important that we understand the function It will be hard to pick up RNAs and coordination of gene expression of the transcribed at low levels when labs work

Cross-linking current ones,” says Gerstberger. Knowing with whole cells, says RIKEN researcher about a potential ability to bind is not Piero Carninci, given that “there are four telling enough about RBPs, she says. Much to five orders of magnitude of difference in discovery awaits related to regulation and gene expression.” Scientists can try more physiological function. targeted approaches to identify RNAs Digestion A tally of RBPs, such as one completed bound to specific compartments, he says. in an organized project, would be a valuable He and his team have been working on resource, says Esteban. Even without it, data capturing lncRNAs attached to chromatin, Immuno- precipitation about RBPs will amass through the efforts of for example. They capture these complexes many labs. Discerning the different relevance by sequencing RNA–DNA pairs that Nonspecificpecifcifiic Specific of every interaction between an RNA and associate to specific genomic locations. Then iinteractionsnteractions interactions a protein matters, he says. For example, they statistically analyze the frequency of SDS–PAGE just because a metabolic enzyme regularly interaction to explore significance. “In this Denaturnaturiingng interacts with RNA does not automatically case, sequencing introduces a measurable mean the RBP regulates metabolism; the way to count and assess frequency,” he says. RBP could play other, yet unknown, roles. As Guttman explains, cross-linking data Librarraryry ppreprep, A catalog, whether compiled as a are less noisy when UV light rather than sequencingequencini g concerted effort or by the collective effort formaldehyde is used. UV light is a zero- of many labs, is unlikely to come together distance cross-linker, meaning cross-links through the use of just one method, occur only when protein and RNA are

Transcript 1 Transcript 2 (nonspecific) says Guttman. As labs learn more about touching. “Even if it is slightly removed, it microRNAs, long noncoding RNAs will not cross-link,” he says. The solid results (lncRNAs), circular RNAs, and other poorly UV-based cross-linking delivers make it Some methods focus on characterizing the protein understood RNA classes, a catalog can be the gold standard when labs ask, “Is there in RNA-binding proteins. Credit: Guttman lab, a framework for evaluating function and interaction forming in vivo?” Cross-linking’s Caltech. Adapted with permission from ref. 2, mechanism and for exploring the RBP high degree of specificity can, however, be a Springer Nature universe. “It no longer sort of becomes the downside, such as when the RNA is not quite

656 Nature Methods | VOL 15 | SEPTEMBER 2018 | 655–658 | www.nature.com/naturemethods technology feature

and colleagues point out, over 30 years The Hentze lab and other labs are ago, labs tried to isolate the poly(A) exploring the effect of replacing oligo(dT)s RNA-bound proteome using oligo(dT) with locked nucleic acids (LNAs) for pulling Sepharose chromatography5. Since then, down polyadenylated RNA. The oligos methods developers have reached a more are, for example, instead of 20 deoxy(T)s, comprehensive approach for identifying a string of a mix of deoxy(dT)s and LNAs. RBPs with UV-based cross-linking and LNAs melt at higher temperatures than sequencing to determine RNA-binding sites. deoxy(dT)s, which lets experimenters These methods are now being used to, for be more stringent with washing steps for example, compare healthy and diseased cells purification. The more stringent the wash, the under different conditions. more unspecific hangers-on can be washed Landthaler and his team used the method away. In both his method and the one from for studying pathways involved in sensing the Landthaler lab, a “significant signal of in the right conformation when cross-linking and repairing DNA damage6. In breast ribosomal RNA” clutters pulldown results, he occurs. Another important aspect, he says, is cancer cells, they identified over 260 RBPs says, which the use of LNAs will lower. The that cross-linking tends to involve 1–5% of that show more interaction with mRNAs polyadenylated RNA stays on the beads and all RNA–protein interactions. Labs can miss in response to ionizing radiation. The team the rRNA and contaminating DNA, too, can plenty of interactions, and to get at these low- used a 4-thiouridine (4sU)-based mRNA be washed away with greater stringency. abundance interactions, researchers need to capture approach and an isotope-based Labs have long explored how to purify do experiments with many cells. labeling technique, SILAC, to label cells RNA-binding proteins with higher stringency, and help with quantification of protein says Guttman, in order to wash, keep RNA Hands on RNAs and proteins differences in samples. Integration of 4sU and protein and be rid of as many nonspecific As Gerstberger points out, RBP exploration into RNAs can vary by cell type, so labs hybridizations as possible. He points to dates back to the 1950s. Over time, high- should test nucleoside concentration and a method published in 2006 that applies throughput methods emerged such as incubation time prior to their experiments. peptide nucleic acids to achieve stringent RBP immunoprecipitation, then cDNA One approach Landthaler washing7. Guttman sees experimental array hybridization of RNA or approaches codeveloped with others is PAR-CLIP, for advantages to LNAs: they are short and have to select RNA ligands on the basis of photoactivatable-ribonucleoside-enhanced a high melting temperature, which helps with SELEX, systematic evolution of ligands by cross-linking and immunoprecipitation. higher-stringency washing. But LNAs cannot exponential enrichment. RNA-sequencing Photoreactive thionucleosides (4sU) are be made in the lab; they must be bought. and mass spec have enabled profiling of RBPs taken up by live cells, and the nucleosides “LNAs are expensive,” he says. at higher throughput and transcriptome- enhance the cross-linking between RNA RAP-MS, a method developed in the wide. Some techniques combine sequencing, and protein. By scoring thymidine-to- Guttman lab, applies biotinylated probes immunoprecipitation and sequencing leading cytidine transitions, the method reveals and an RNA antisense purification strategy. to so-called CLIP-seq approaches. RBP binding sites. The researchers note It has turned out to be a major that the method enables transcriptome- methodological advance to be able to pull wide identification of RNA-binding sites of down RNAs via polyadenylated tails. There targeted RBPs. AAAA AAAA are two methods to do so: one developed in To enable RNA capture beyond RNAs the Landthaler lab at Max Delbrück Center with polyadenylated tails, Esteban developed 4sU treatment for Molecular Medicine, and the other in the RICK, or capture of the newly transcribed AAAA AAAA 4,5 Hentze lab . RNA interactome using click chemistry. IR Irradiation The approaches were developed in parallel. The method includes RNA labeling with The two labs knew of one another’s efforts, 5-ethyluridine followed by a biotinylation methods and results, and they worked step using click chemistry. Next is bead- UV cross-linking independently and coordinated publication, based capture, after which the RNAs are AAAA says Hentze. “They’re virtually identical sequenced and the proteins analyzed by mass AAAA Cells are lysed methods; the buffers are slightly different,” he spec. He and his team are applying RICK to AAAA says. In both, there is UV-light-based cross- capture newly generated RNAs in stem cells 4sU linking of proteins with RNA in live cells, then to see how the type and patterns of RBPs Oligo(dT)-based purification RNA pulldown via the poly(A) tail, and the change under different conditions or in RBPs polyadenylated RNA is captured on oligo(dT) cells engineered to have certain mutations. Mass spectrometry beads. Next follows mass spectrometry to Capturing newly transcribed RNAs offers Cross-link identify and quantify proteins. hints about a facet of RBP function, he says. The use of oligo(dT)-coated beads for In breast cancer cells exposed to radiation, capturing RNAs has helped labs to discover Combating noise 266 proteins increased binding to mRNA

RBPs and to appreciate the scope of proteins Signal-to-noise ratios become especially AAAA that bind to RNAs, says Hentze. Labs have important when cells in different conditions AAAA tweaked the methods to, for example, are being compared, says Hentze. AAAA AAAA incorporate nucleotide analogs that are Background can hurt results and analysis. subjected to biotinylation followed by the Smaller but relevant changes risk being An approach to study the DNA damage response pulldown step, he says. overlooked. “I think it’s critical to develop using breast cancer cells. Credit: Landthaler lab, As Markus Landthaler of the Max the existing techniques in such a way that MDC Berlin. Adapted with permission from ref. 6, Delbrück Center for Molecular Medicine differences can be scored well,” he says. CSH Press

Nature Methods | VOL 15 | SEPTEMBER 2018 | 655–658 | www.nature.com/naturemethods 657 technology feature

The team uses single-stranded, biotinylated ways, another aspect that can influence 90-mer oligos, which also have high melting regulation. temperatures and can be made in the lab. RNAs can evolve the ability to bind to The hybrid is stable, says Guttman. It a particular protein surface, which means can weather denaturing and washing in “in principle any protein surface could be chaotropic agents such as 6 M urea to bring an RNA-binding surface,” says Hentze. A back what has covalently cross-linked to new awareness of how RNA and proteins the RNA. He and his team study lncRNAs influence one another is emerging. “RNA- that can be around 17,000 nucleotides long. binding proteins are not only proteins that It’s risky to use only one probe such as an bind to RNA to regulate RNA, but RNA- LNA, he says, given how fragile RNA can be. binding proteins are proteins that are bound When it breaks, a large fraction of the total by RNA to be regulated by RNA,” he says. protein bound to RNA is lost. In his lab, the He and his team have found that a protein team tiles the lncRNA with many 90-mer accessible, dynamic RNA regions in involved in autophagy is regulated by a probes so data can be captured even if the ways that can be detected by sequencing. type of ncRNA called a vaultRNA, in this RNA breaks. Another way to probe RNA structure case vtRNA1-19. Autophagy is the process Addressing purification, the Hentze lab with modifications is with DMS-MaPseq, through which cells rid themselves of debris. developed 2C, a method leveraging an old for dimethyl sulfate mutational profiling The RBP in question is an enigmRBP observation that proteins do not bind to with sequencing, a method from the identified in his lab. EnigmRBPs are enigmatic resins in silica columns, whereas RNA does8. Weissman lab at the University of in their RNA-binding, meaning it’s hard to “I think that is a really beautiful observation California at San Francisco, MIT’s Silvi identify where they are binding, says Hentze. because it makes life so much easier,” says Rouskin and colleagues. To label specific The team shows that the small ncRNA Guttman. The 2C method involves using RNA motifs in RBPs, Paul Khavari at binds the protein and regulates its function silica columns to isolate cross-linked RBPs. Stanford University School of Medicine through the RNA–protein interaction in Conventionally, labs need to validate and colleagues developed RNA–protein autophagy, he says. “That RNA–protein RBPs with immunoprecipitation, washing, interaction detection, or RaPID. It applies interaction is doing what we normally see labeling of the cross-linked RNA, elution biotinylation, another way to permit protein–protein interactions doing,” which is and autoradiography. The 2C method avoids stringent washing during RBP purification. to regulate protein function. Here, RNAs act immunoprecipitation and radiolabeling. If Many RBPs do not have recognizable as riboregulators to regulate autophagic flux. a protein is cross-linked to RNA, the RNA RNA-binding domains, says Hentze. “Clearly Such findings indicate how RBPs give cellular sticks to the resin, delivering the proteins there is a whole universe out there,” he genomes “the chance to regulate biology and bound to it. What sticks to the resin can be says. He and his team developed RBDmap biological processes in a far more direct way stringently washed, enabling better RBP to find binding domains. It can see not- than we previously knew,” he says. purification, he says. “I am very excited yet-identified regions that are enriched for Observations about the versatility of about that as a new foundation for thinking RNA binding, such as highly unstructured RBPs add an aspect to the RNA World about purification of RNA-binding proteins,” protein regions and low-complexity Hypothesis, according to which life on says Guttman. domains, he says. In RBDmap, a second Earth began with RNA, followed by the round of oligo(dT) captures the cross-linked evolution of proteins and, later, DNA. Work Shape matters RNA and protein along with a neighboring in the Hentze lab, including this latest Proteins are three-dimensional molecules, peptide, which is used to identify riboregulator and the regulatory role of and a ‘good’ RNA-binding domain might be RNA-binding sites. RNAs in RBPs more generally, all fit well exposed only under certain conditions, says Speaking more generally about RBP into the RNA World model, he says—“I Esteban. “Those domains are very difficult to techniques, Urlaub says that “we are still not could see in it, perhaps, an evolutionarily identify with bioinformatics,” he says. RNA specific and quantitative enough on the cell- early form of biological regulation.” ❐ forms 2D structures, says Guttman, because wide level.” Changes to interactions between of the thermodynamic properties of the RNAs and proteins need to be defined in Vivien Marx molecule and other aspects, and the structure terms of cell cycle and states of various cells, Technology editor for Nature Methods. will also determine which proteins bind also quantitatively. Labs want to know, for e-mail: [email protected] where. An RNA also changes conformation example, how many copies of a protein depending on its bound proteins. Structure bind to one RNA molecule and whether Published online: 31 August 2018 and function go together. Bacterial different protein domains act differently https://doi.org/10.1038/s41592-018-0117-9 riboswitches are a type of RNA that sense, on different RNAs. It remains hard to for example, cellular nutrients. In reaction to study low-complexity protein regions with References nutrients, a riboswitch changes its structure, repetitive amino acid sequences that bind 1. Singh, G., Pratt, G., Yeo, G. W. & Moore, M. J. Annu. Rev. which can shape a function such as gene to RNA, because they are nearly impossible Biochem. 84, 325–354 (2015). expression. Mammalian cells might turn out to study with mass spec, he says. 2. McHugh, C. A., Russell, P. & Guttman, M. Genome Biol. 15, 203 (2014). to have RNAs fulfilling these roles, he says. Perhaps, says Esteban, some RNAs bind 3. Hentze, M. W., Castello, A., Schwarzl, T. & Preiss, T. Nat. Rev. Studying RNA structure remains a to proteins in several different ways. “I think Mol. Cell Biol. 19, 327–341 (2018). challenge but some methods exist to help we need to change our minds a little bit,” 4. Baltz, A. G. et al. Mol. Cell 46, 674–690 (2012). 5. Castello, A. et al. Cell 149, 1393–1406 (2012). characterize RBPs, says Guttman. With he says. Some binding interactions between 6. Milek, M. et al. Genome Res. 27, 1344–1359 (2017). SHAPE, or selective 2-hydroxylacylation RNA and proteins might be “more subtle, 7. Zielinski, J. et al. Proc. Natl. Acad. Sci. USA 103, 1557–1562 (2006). analyzed by primer extension, developed delicate” than others, and these add to the 8. Asencio, C., Chatterjee, A. & Hentze, M. W. Life Sci. Alliance 1, e201800088 (2018). in the Weeks lab at the University of regulatory landscape. In addition, RNA and 9. Horos, R. et al. bioRxiv Preprint at https://doi.org/10.1101/177949 North Carolina, a small molecule marks proteins can both be modified in various (2017).

658 Nature Methods | VOL 15 | SEPTEMBER 2018 | 655–658 | www.nature.com/naturemethods