Coilin: The first 25 years

Abstract

Initially identified as a marker of coiled bodies (now Cajal bodies or CBs), the protein coilin was discovered a quarter of century ago. Coilin is now known to scaffold the CB, but its structure and function are poorly understood. Nearly devoid of predicted structural motifs, coilin has numerous reported molecular interactions that must underlie its role in the formation and function of CBs. In this review, we summarize what we have learned in the past 25 years about coilin's structure, post-transcriptional modifications, and interactions with RNA and proteins. We show that genes with homology to human coilin are found in primitive metazoans and comment on differences among model organisms. Coilin's function in Cajal body formation and RNP metabolism will be discussed in the light of these developments.

Coilin History

Cajal bodies (CBs) were discovered at the beginning of 20^th century by Spanish cytologist Ramón y Cajal as structures in neuronal nuclei stained with silver. The bodies were then rediscovered and named coiled bodies in the 1960's due to their specific appearance in electron microscope images; they were later renamed Cajal bodies in honor of their discoverer.¹ CB research accelerated with the discovery of autoantibodies present in autoimmune patient sera, which were shown to react with a protein of apparent molecular weight 80 kD (named p80-coilin) that specifically localized to CBs.² The human protein was characterized and its cDNA cloned.³ The term “coilin” first appeared in a paper published a year earlier, in which Ivan Raska and his co-workers named the protein after coiled bodies and used anti-coilin autoimmune sera to analyze the relationship between CBs and nucleoli.⁴

Soon after human coilin was discovered, the Xenopus laevis homologue was identified and shown to specifically localize to sphere organelles within the germinal vesicles of X. laevis oocytes.⁵ Matera and colleagues cloned mouse, rat and zebrafish coilin genes and revealed the conservation of N- and C-terminal regions across these species (Fig. 1). Coilin's amino acid sequence is not well conserved and is characterized by low complexity and unstructured regions. Hence, it took considerable effort and time to identify genes for coilin in plants⁶ and fruit flies,⁷ despite prior evidence for the existence of CBs in these organisms.^8,9 Meanwhile, the genomes of a wide variety of organisms have become available. A current search for coilin homologues revealed that predicted proteins with significant sequence homology to human coilin can be found in as primitive an organism as Trichoplax adherens (Fig. 1), which is one of the simplest metazoans. Surprisingly, human coilin is more similar to the predicted Trichoplax coilin than it is to D. melanogaster coilin.

Figure 1.

Coilin is highly conserved among metazoans. Predicted coilin amino acid sequences (RefSeq) for the indicated species were aligned with coilin proteins identified by homology search in Cephalochordates (Lancelet), Cnidarians (Hydra) and Placozoa (Trichoplax). Alignments of conserved (A) N-terminal and (B) C-terminal regions are shown, demonstrating the greatest divergence of Drosophila melanogaster coilin from the other species. (C) Cladogram shows numbers of annotated and/or predicted coilin proteins per clade together with hypothetical coilin proteins found by homology search (numbers in brackets).

Coilin protein domains

Evolutionary conservation at the amino acid level is highest within coilin's N- and C-termini. The middle portion is not conserved and low complexity region and RG box found in human coilin appear to be specific for vertebrates. The first N-terminal 92-amino acid domain was shown to self-interact and to be essential for proper targeting of coilin to CBs.^10-12 The central part contains 2 nuclear localization signals and a putative nucleolar localization sequence,¹² which might explain coilin's affinity toward the nucleolus, which was noticed already 25 years ago.⁴ The conserved C-terminal domain folds into a Tudor domain like structure.¹³ Several Tudor domains were found to bind methylated amino acids but direct tests did not show any interaction of coilin Tudor domain with monomethyl-lysine, trimethyl-lysine and dimethyl-arginine.¹³ However, the C-terminal domain was shown to interact with Sm proteins, which contain symmetrically dimethylated arginines.^14-16 Coilin interaction is stronger when Sm proteins are purified from eukaryotic cells rather than bacterially expressed, suggesting that posttranscriptional modifications might play a role in Sm protein-coilin interactions.^15,16 Thus, the precise molecular mechanism of coilin-Sm protein binding is still unknown.

The search for coilin function

Coilin loss of function has been analyzed in Arabidopsis thaliana, Drosophila melanogaster, Danio rerio and Mus musculus. It was initially disappointing that coilin gene disruptions in plants and flies did not detectably affect viability or fertility.^6,7 Yet coilin depletion has strong effects on the viability of vertebrate embryos. In mouse, gene disruption leads to a dramatic loss of homozygote pups in matings of heterozygotes, and coilin^−/− mice are significantly less fertile.^17,18 In zebrafish embryos, coilin depletion through morpholino injection was lethal within the first 24 hours of development and was accompanied by reduced levels of snRNPs and spliced mRNAs.¹⁹ Fish embryos were rescued upon injection of spliceosomal small nuclear ribonucleoprotein particles (snRNPs), which suggests that coilin's essential function in embryos is to promote macromolecular assembly of snRNPs.¹⁹ This conclusion agrees with early findings that snRNAs and snRNP proteins are concentrated in CBs.^20,21

Spliceosomal snRNPs are assembled and recycled in CBs.^22-28 Mathematical modeling and measurement of snRNP kinetics in CBs suggest that concentration of snRNPs in the CB increases snRNP assembly rate by a factor of 10.^29,30 Similarly, co-localization of snRNAs and scaRNAs, which guide snRNA modifications, in CBs might enhance snRNA modification efficiency. However, in D. melanogaster, scaRNAs function in the absence of coilin, showing that concentration of snRNA and scaRNAs in CBs is not essential in flies.^31,32 Recent work showed that coilin directly binds snoRNAs³³ and that snoRNAs shuttle via CBs before they reach their nucleolar destination.^33,34 Therefore it is tempting to speculate that snoRNP assembly, like snRNP assembly, takes place in CBs and that coilin plays an important role as well.

Based on these observations, we suggest that coilin promotes RNP biogenesis by acting as a chaperone of nuclear small non-coding RNAs. This might not be essential in differentiated cells or cells lines cultured in vitro, but may become critical in rapidly developing embryos. A clear example is zebrafish, where embryos depleted of coilin and lacking CBs are unable to complete embryogenesis, display splicing defects and reduced numbers of mature snRNPs.^19,35 Because zebrafish embryos require splicing of zygotically transcribed pre-mRNAs,³⁶ these combined data lead to the interpretation that coilin supports the high embryonic demand for splicing by promoting rapid snRNP assembly in CBs. This explanation could easily also apply to mouse embryos. We suggest that the viability of a small number of homozygous coilin^−/− mice reflects the chance that a sufficient number of embryonic cells survive on maternal stores of snRNPs. In other systems, like plants or insects, differences in mechanisms of development might determine whether coilin's function is essential or not. For example, insects are characterized by a syncytial blastoderm that may confer a survival advantage for embryos. If snRNP assembly is slower in D. melanogaster colin null mutants, perhaps blastoderm nuclei are able to share the snRNP deficit among all of the cells of the blastula during the critical period when rapid splicing is needed. An additional or auxiliary possibility is that more cell death is tolerated in insect embryos in general. Given that conserved coilin domains in D. melanogaster diverged significantly from vertebrate coilin (Fig. 1), it is also possible that flies have evolved a mechanism for snRNP assembly that does not depend on coilin as strongly as in vertebrates.

Additional data connecting coilin, CBs and snRNPs were added this year when Novotny et al. showed that incomplete or defective snRNPs are anchored to coilin and CBs. This finding suggests that coilin is part of a quality control mechanism that proofreads final snRNP assembly.³⁷ Again, this function might not be essential under normal conditions but becomes important when snRNP assembly is perturbed or when transcription and splicing rates are high, producing a large quantity of mono-snRNPs that require recycling and reassembly. Coilin-dependent concentration of mono-snRNPs in CBs would increase their assembly rate and at the same time prevent incomplete snRNPs from entering splicing reaction.

Coilin and CBs were suggested to be important for telomerase assembly and telomere maintenance. Early experiments detected telomerase RNA in CBs, and CBs associate with telomerase during S-phase.^38-42 Interestingly, telomerase RNA localization to CBs seems to be human-specific because telomerase RNA was not found in mouse CBs.⁴³ Telomerase RNA is retained in CBs via protein WRAP53, which binds the CAB box sequence found in telomerase RNA and many scaRNAs.^39,44-47 WRAP53 interacts with coilin, providing the mechanistic link between telomerase and coilin.^48,49 Telomerase RNA localization to CBs was suggested to be important for telomerase function and mislocalization of telomerase RNA from CBs correlates with reduced telomerase activity.^45,50,51 In addition, coilin depletion by siRNA inhibits association of telomerase with telomeres.^52,53 However, the role of coilin or CBs in telomerase assembly is unclear. Loss of coilin could be overcome by overexpression of telomerase,⁵² and human cancer cells lacking coilin do not exhibit any inhibition of telomerase activity or defects in telomere lengths.⁵⁴ Why did coilin depletion by siRNA, which is never complete, reveal defects in telomere maintenance, when a genetic coilin knockout did not show any telomere shortening? While RNAi mediated knockdowns are transient and cells are assayed 2-3 days after RNAi treatment, knockout requires prolonged period of clone selection. Perhaps only clones that successfully adapted to coilin deletion survive. The situation is similar to mouse, where some coilin knockout embryos die during embryogenesis while surviving pups are viable despite reduce size and fertility.¹⁸

The interaction between coilin and WRAP53 could also explain why coilin is found at sites of DNA damage, and CBs are sensitive to UV irradiation.^55,56 While coilin function during DNA repair is unclear, WRAP53 was recently shown to be critical for DNA double-strand break repair.⁵⁷ Independent work found coilin associated with DNA repair factor Ku.^58-60

Coilin could be also involved in higher order chromatin organization albeit molecular details of this function remain elusive. Genes encoding snRNAs, snoRNAs and histone mRNAs were found closely associated with CBs in cytological assays.^61-67 Coilin association with snRNA and histone genes was recently re-examined by genome-wide ChIP-Seq, which detected coilin peaks at most of these genes.³³ CBs contain pre-U2 snRNA⁶⁸ and contact between CBs and chromatin is maintained by active snRNA gene transcription.^61,69 CBs may be assembled on or transported toward active snRNA genes.⁷⁰ This transcription-dependent contact between snRNA genes and CBs was recently confirmed by the finding that coilin ChIP signals on snRNA genes is cell cycle specific, with the highest signal in S-phase when snRNA genes are actively transcribed.³³ Similarly, coilin binding to U7 snRNA brings histone genes to a close proximity of CBs.^33,71 Taken together, these findings suggest that coilin brings several gene loci together within the 3D space of the nucleus and thereby function in global organization of chromatin.

Toward Cajal body formation

Coilin is an indispensable structural component of CBs, because coilin depletion results in CBs disintegration in all tested organisms.^6,7,17,19 The domain critical for CB formation is the N-terminal 92 amino acids, which is necessary for coilin self-interaction, coilin targeting to CBs, and de novo CB formation.^10-12 However, the ability of the N-terminus to form CBs is modulated by the C-terminus and even subtle differences in C-terminal sequence between mouse and human coilin affect CB nucleation.⁷² The C-terminus interacts with snRNP-specific Sm proteins^14-16 and snRNP expression level is a factor important for CB formation.⁷³ We recently showed that increased concentration of unassembled or defective snRNPs triggers formation of CBs in cells that are normally devoid of CBs.³⁷ This suggests that incomplete snRNPs enhance coilin self-interaction properties and induce formation of CBs through an unknown mechanism.

Recent studies have suggested that low-complexity domains, multivalent interactions and RNA are important factors that favor liquid phase separation and the formation of cellular bodies, like nucleoli, that are not delimited by lipid bilayers.^74-78 Does CB formation follow a phase separation model? Direct evidence is lacking, though hints exist. CB formation in primary fibroblasts is temperature dependent,⁷⁹ consistent with the phase separation model.⁸⁰ Coilin contains short stretches of serines and lysines in the central low complexity region, but their role in CB formation has not been tested. Finally, the role of RNA in CB formation has been documented.^33,81,82

Coilin also directly interacts with many CB proteins (Fig. 2) and is therefore widely considered as a molecular hub, bridging otherwise distinct CB components and thereby “gluing” them together via numerous protein-protein interactions.³⁵ Artificial tethering of individual CB components to chromatin induces CB formation, which is consistent with a model that CBs are formed via protein-protein and protein-RNA interactions.⁸³ snRNP-specific proteins are highly active in this assay for CB formation, which is consistent with a role for snRNPs in the context of naturally occurring CBs.^37,73

Figure 2.

Coilin interaction partners and the formation of CBs. (A) Schematic representation of human coilin protein showing key structural features: Tudor domain (purple), RG-box (orange), bipartite nuclear localization signal (blue), cryptic nucleolar localization signal (light blue) and self-interaction domain (green). Black bars signify parts of coilin known to interact with indicated proteins/RNA.^{12,14,15,33,37,48,58,84-86,101-104} (B) The current hypothesis of nuclear body assembly suggests that protein and RNA components (colored shapes) tend to have more interactions (double arrows) among themselves as compared to other molecules (blue circles) in the environment. When the components reach a critical concentration locally, liquid phase separation occurs and creates the nuclear body.

Coilin: A non-canonical RNA binding protein

Despite the fact that coilin does not contain a canonical RNA binding motif, its X. laevis coilin was shown to interact with poly-U and poly-G in vitro. It was noted by the same authors that the N-terminal self-interaction domain may contain a degenerate RNA recognition motif.⁸⁴ Later, Makarov et al. showed that plant coilin binds U1 snRNA in vitro but the RNA binding domain was not unambiguously determined. The authors also showed that the U1 snRNA induces multimerization of Arabidopsis coilin in vitro, suggesting that RNA binding modulates coilin self-interaction.⁸⁵ Finally, human coilin was shown to bind RNA in vitro and co-precipitated several RNAs from cell extracts, though it was unclear whether this interaction was direct^86,87; these studies suggest that the RNA binding domain is localized in the central, least conserved part of coilin.⁸⁷

Direct evidence that coilin binds RNA in vivo was recently provided by iCLIP, in which 0 Å crosslinking with UV light followed by immunopurification of coilin-RNA adducts permitted sequencing of bound RNAs.³³ This study identified hundreds of short ncRNA including snRNA, snoRNA and telomerase RNA, as coilin binding partners in vivo. Interestingly, coilin prefers RNA stem-loop regions for binding.³³ This may explain why previous work investigating coilin binding to homopolymer RNAs revealed low affinity interactions. Interestingly, extensive coilin-RNA interactions revealed by iCLIP agrees with original findings: CBs were first observed in neuronal sections by silver staining, which deposits on RNA-rich structures.⁸⁸ In addition, coilin was first discovered as a target of autoantibodies, which often recognize ribonucleoprotein complexes or nucleic acid binding proteins (snRNPs, hnRNP proteins, histones, Ro, La, SR proteins etc.).

Post-transcriptional modifications

Many aspects of coilin function are modulated by posttranscriptional modifications. Arginines in the RG box can be symmetrically methylated. In the dimethylated state, arginines are bound by the SMN Tudor domain and this interaction mediates association of nuclear gems and CBs.^14,89,90 Cells lacking 5′-methylthioadenosine phosphorylase, a key enzyme of the methionine salvage pathway, have reduced methylation activity and coilin is re-localized to nucleoli.⁹¹ However, this result has to be cautiously interpreted because 2 methyltransferases, PRMT5 and PRMT7, that symmetrically methylate arginines in Sm proteins, are important for snRNP biogenesis⁹² and ongoing snRNP biogenesis may be critical for CB formation.^93,94

Coilin is also heavily phosphorylated and phosphorylation controls CB disassembly at mitosis, proper CB assembly after mitosis and coilin localization.^12,79,95,96 Perhaps contributing to cell cycle variation in CB number, cdk2/cyclin E and the nuclear phosphatase PPM1G phosphorylate and dephosphorylate coilin in vitro.^97,98 It was suggested that the interplay between coilin phosphorylation and dephosphorylation affects coilin self-association and interaction with snRNPs, SMN and RNAs.^16,98-100

Future perspective

During the first 25 years, we learned that coilin is a tricky protein to deal with. Coilin often proves insoluble when expressed or otherwise purified, and little structural information exists. Coilin depletion always impacts CB size and number, yet experiments in model organisms have yielded mixed phenotypes ranging from embryonic lethality in zebrafish to no apparent phenotype in flies and plants. Future establishment of additional direct approaches to analyzing coilin structure and function will be instructive. It is encouraging that global mapping of protein and RNA interaction partners did not identify surprisingly novel proteins or RNAs,³³ raising optimism that coilin's interactors are known and molecular function(s) will be revealed soon. A screen for synthetic interactions in model organisms might provide great insight, alongside bottom-up approaches in vitro. We also need to devise experiments that would identify the physical principles that hold CBs together.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

David Stanek was supported by a Fulbright scholar fellowship, the Academy of Sciences of the Czech Republic (RVO68378050) and the Czech Science Foundation (P305/12/G034).