Introduction

Protein-RNA interactions play a significant role in gene expression and its regulation. A recent study has reported a census of 1542 non-redundant human RNA binding proteins (RBP) which interact with all known RNA types [1]. The study also reports that RBPs form 7.5 % of all the coded proteins in human proteome. The RNA binding domains (RBD) in RBPs can recognise the RNA molecule in sequence dependent as well as independent manner [2]. There are many large protein-RNA complexes including ribosome, RNA polymerase and spliceosome, which bind to the target RNA as well as catalyses various reactions. Many RBPs have multiple RBDs that can independently recognise its partner RNA [3, 4]. Detailed analysis of RBDs reveals the inherent flexibility and plasticity of the binding surfaces [5]. These domains have higher specificity and affinity towards RNA, which are achieved via linkers that are disordered extensions of RBPs [6]. Moreover, it is likely that oligomerisation and the combinatorial modular architecture of RBPs enhances their specificity and affinity to its partner RNA [7]. In most of these protein-RNA interactions, conformational multiplicity is observed in either or both the partners [8]. In such cases, the RBDs and the partner RNA undergo binding-induced folding that finally gives rise to a biologically functional protein-RNA complex. In a recent computational study, Varadi et al. [9] has reported the abundance of disordered regions in RBPs and the significance of its conformational multiplicity in RNA binding. They not only found that most of the interface residues making direct contacts with RNA are disordered in nature, but also claim that these residues are highly conserved compared to the underlying sequence. A proteomic and bioinformatics study showed that in comparison to the whole cell extract, disordered proteins are overrepresented in the cell nucleus [10]. Moreover, from the Gene Ontology analysis it is also revealed that most of the transcription factors are enriched in disordered regions.

In the last two decades, many experimental data showed that several proteins or regions of proteins are intrinsically disordered (IDP: Intrinsically Disordered Proteins; IDR: Intrinsically Disordered Regions) in nature under native, functional condition [11, 12]. A correlation analysis between the ‘functional keywords’ and ‘IDPs function’ in Swiss-Prot [13] revealed that IDPs are mainly involved in cell cycle, gene expression, signalling events and regulations; whereas, structured proteins are found majorly correlated with keywords related to enzymatic catalysis [14]. It is the structural pliability of IDPs that makes them an important player in various interactions within the living system. An extensive study on the functional roles of IDPs have come up with five different classes: effectors, assemblers, entropic, scavengers and display sites [15]. The increased interaction potential of the IDPs is achieved by various functional features. The IDRs are involved in many low-affinity and high-specificity interactions for example, cell-signalling and regulation, which are achieved via functional elements like small linear motifs (SLiMs), molecular recognitions features (MoRFs) and low complexity regions (LCR) [16]. SLiMs are short stretches of amino acids (3–10 residues long) involved in complex formation mostly by regulating the low-affinity interactions [17]. The compact binding of them to the surface of globular proteins promotes their multiple-occurrence in IDP [17–19]. On the other hand, MoRFs (10–70 residues long) are generally longer than SLiMs and are involved in specific protein–protein interactions [20, 21]. There are regions in IDPs called LCRs with repetitive amino acid residues along the sequence [22]. These motifs undergo disorder to order transition after binding to its partner, and in the unbound form, they are mostly biased to the conformation they adopt upon binding [23]. Considering the multi-functionality, the mode of interactions involving IDPs is always an interesting topic of research in protein science. A comparative study between globular proteins and IDPs revealed that the latter have unique molecular principles of interactions [24]. It was found that compared to the globular proteins, IDPs have larger surface area as well as interface area per residue. The exposed segments of IDPs forming the interface with its partner molecules are enriched in hydrophobic residues. Moreover, it was also observed that hydrophobic–hydrophobic interactions are dominant over polar–polar interactions in molecular recognition involving IDPs. Hydrophobic residues of IDPs have a typical tendency to be present at the exposed interacting segments and avoid getting collapsed in forming a regular folded structure. Based on these observations, it was suggested that IDPs follow ‘inside-out’ folding mechanism where the partner binds to the interacting segments of the IDP mainly involving hydrophobic contacts, which further promotes folding [24].

Structural disorder can be studied by many experimental techniques such as X-ray crystallography, NMR spectroscopy, circular dichroism (CD) or small-angle X-ray scattering. While NMR can give reliable consensus, X-ray fails to give information about disordered regions, raising confusion whether the failure to resolve is because of the technical issues (missing electron density can arise from failure to solve the phase problem, crystal defects or even from unintentional proteolytic removal during protein purification) or because of the inherent property of IDPs. On the other hand, CD with a combination of far-UV and near-UV gives an idea whether a protein has disordered regions [12, 25, 26]. Intriguingly, thousands of the structures in PDB are now known to contain disordered regions [27].

Flexibility of binding partners is an indispensable feature in protein-RNA recognition. The involvement of IDRs in such recognition process facilitates RBPs to attain alternate conformations upon binding with the RNA. In this review, we are focusing on the structural basis of the polymorphic conformations involved in RBP-RNA interactions. This polymorphic conformations lead to the structural plasticity of the interface, thereby influencing the recognition, which eventually regulates different cellular processes.

RNA recognition involving IDPs

Protein-RNA recognition is essential for post-transcriptional gene regulation [28]. The recognition can be both sequence and structure dependent [29]. Besides, recognition may involve induced-folding of protein, RNA or both [8, 30]. The diversified mode of RNA binding is a major characteristic of the recognition, and the low energy requirement of the RNA to deform and unfold facilitates the process [31]. The major benefit of a protein domain having structural flexibility is its potential to mould into a binding surface according to its partner molecule, and likewise can facilitate to bind multiple partners [30]. This has been observed in a complex between TruB and its partner RNA. Here, the thumb loop in TruB is disordered, which undergoes disorder to order transition and gains helical conformation upon binding with the RNA (Fig. 1) [32]. RBPs have characteristic sequence features, which particularly bind to single stranded or double stranded RNA. Some of the very common RNA binding domains are zinc finger [33], RNA recognition motif (RRM) [34] and K homology domain [35]. RGG/RG motif is a SLiM, which is also found abundantly in RBPs [36]. This RGG motif is found to have high binding affinity towards G-rich RNA sequences [37]. The RG-rich regions take part in RNA metabolic processes via both selective and non-selective binding [7]. A very recent computational analysis showed that the nucleiome of archae, bacteria and eukaryota have abundance of disordered regions [38]. Disordered regions promote the formation of ribo-nucleoproteins, which can further act as assembly domains [39]. The flexibility owing to the polymorphic conformations in RBD is a highly facilitating factor for RNA recognition promoting IDP-RNA interactions in various cellular processes [40].

Fig. 1
figure 1

Superposed structure of TruB with its partner RNA in bound (coloured in blue cartoon, PDB ID: 1R3E) and unbound (coloured in green cartoon, PDB ID: 1R3F) conformations [32]. The disordered thumb loop (red dashed lines) of TruB undergoes conformational transitions and become ordered upon binding with its partner RNA

Full size image

RNA chaperones

One of the major concerns in RNA folding is the kinetically trapping of alternate RNA conformers. This can be resolved by the RNA chaperones, which make non-specific interactions with the RNA thereby assisting the RNA folding [8]. In many cases, it has been found that the protein domains crucial for the function of RNA chaperone are disordered in nature [15, 31, 41]. These disordered regions render interaction multiplicity along with higher momentum of interaction with the partner molecule. A comparative study between protein and RNA chaperones revealed that the occurrence of disordered regions in RNA chaperones are much higher than that in protein chaperones [41]. This study also suggested a probable mechanism of entropy transfer, which assists the misfolded RNA to come out of the local energy minima trough and rearrange to attain the favourable conformation. The flexible domains of the RNA chaperones interact with the RNA molecule and undergo disorder to order transition, which in turn provides a thermodynamic advantage to the trapped RNA molecule to get a minimum energy conformation. The folding of RNA is enthalpy driven and followed by the thermodynamic equation:

$$\Delta G = \Delta H - T\Delta S$$

The chaperone molecule binds to the RNA and transfers its entropy to the RNA molecule making the overall process entropy driven. This brings an overall change in ΔG, which is favourable for spontaneous folding of RNA. A prominent example of such mutual induced-fit phenomenon is observed in the binding of ribosomal protein L5 of Xenopus oocyte with 5S rRNA [42]. Here, the flexible N and C terminal regions of the protein, enriched with non-polar amino acids, increase the interacting surface area by making multiple contacts. Ribosomal protein S12 is also found to assist the folding of phage T4 group I intron, and is assumed to stabilise the correct conformation of the RNA in due of its flexible binding region [43]. The disordered N terminal of prion protein, huPrP, also shows chaperoning function in nucleic acid annealing, viral RNA dimerisation and binding to complementary tRNA [44].

Ribosomal assembly

The interactive features of IDPs like MoRF, SLiM and LCR provide them multiple binding competence, which promotes them to involve in macromolecular assemblies [45]. Polymorphic conformation translated into structural flexibility provides an advantage in complex formation by reducing the steric hindrances [16]. One such higher-order complex is the ribosome assembly, which involves many RBPs having IDRs. Genome wide study has revealed the widespread prevalence of unstructured regions in ribosomal proteins [46]. The flexible extended regions of ribosomal proteins penetrate into the core of the ribosomal subunits, where it is assumed that they undergo disorder to order transition and facilitates the ribosomal RNA (rRNA) folding [47]. This is exemplified in Fig. 2a, which shows the small subunit protein S6 and S11 have defined secondary structures in the periphery of the ribosomal complex but interact with the core region mainly through long disordered extensions [48]. The binding of disordered regions of ribosomal proteins with its partner RNA in the ribosomal assembly is facilitated by the electrostatic interaction between the positively charged regions of the proteins and the negatively charged RNA backbone [46]. High glycine content, one of the characteristic features of IDPs, permits the extension of ribosomal proteins to be flexible and interact with the rRNA. This is evident in almost all the small subunit proteins, and 50 % of the large subunit proteins [49]. Figure 2b shows that the small subunit ribosomal protein, S12, interacts with mRNA with its flexible linkers. Structural analysis of the globular domains of the assembly indicates the presence of unfamiliar morphologies of the ordered domains that may be attributed to conformational polymorphism. Further, Nussinov’s plot analysis probed that many globular domains are formed as a result of binding-induced folding mechanism [46, 50]. Eukaryotic core ribosomal proteins, L4, L22, L23, L29, those make the polypeptide exit tunnel has long extended regions which stretches up to the border of the large subunit. Intruding flexible extensions of L24e crosses the 60S subunit and penetrates into the 40S subunit [51]. Some of the very long extensions are also seen in L2, L3 and S12, which can reach up to the peptidyl transferase domain and participate in some vital interactions with the ribozyme [52]. Apart from taking part in the major events of translation, some of the ribosomal proteins (S1, L1 and L4) are also associated with extra ribosomal functions like maintaining stability among various ribosomal components and preventing apoptosis by reducing the nuclear stress [53].

Fig. 2
figure 2

Ribosomal proteins of 30S subunit (PDB ID: 1N34) [48]. a Small subunit proteins S6 and S11 are shown in yellow cartoon and other small subunit proteins are shown in green surface. b Small subunit protein S12 (shown in yellow cartoon) is interacting with mRNA (shown as a red fragment) through its disordered extension, the other small subunit proteins are shown in green surface.

Full size image

Multi domain linkers

Many RBPs utilise multiple domains to recognise its partner RNA. This multi-domain binding involves cooperative interactions between the RBDs and dictates the dynamics of conformational polymorphism in RNA binding [54]. These interacting domains of RBPs are often linked by a flexible stretch of polypeptides called linkers or spacers. The NMR relaxation data and missing residues or high B-factor in X-ray crystallographic data of the linkers indicate their intrinsic disorderedness [6]. These linkers play a crucial role in determining the sequence specificity of RBPs. Such interplay has been observed in the complex between the zinc finger domains of TIS11d and the class II AU-rich element (ARE) [55] (Fig. 3). Here, the two finger motifs are spaced with an 18 residue linker region. The highly conserved residues of the linkers, residing between the two finger motifs, make significant contacts with the U-rich motif of the RNA and stabilises the TIS11d-ARE complex. Recent studies by Barik et al. [56] showed that salt-bridges play an important role in protein-RNA recognition by contributing to the binding affinity. Moreover, they also showed that stacking interactions involving protein side chains and nucleotide bases also contribute to the recognition process. The aromatic residues phenylalanine and tyrosine actively interact with the uracil to make stacking interactions [56, 57]. This phenomenon is observed in TIS11d-ARE complex. Here, the C-terminal linker in TIS11d interacts closely with ARE RNA and is stabilised by a slat bridge between the OE1 of Glu195 and 2′ OH of U1 base (Fig. 3). Besides, stacking interaction is also observed between the side chain of Phe214 and the U2 base. In some cases, the length of the linker is expected to be conserved better than the sequence as the length regulates the conformational dynamics of the interactions [6]. Leepar et al. [3] showed that the RRM domains of Hrp1 and Rna15 protein of yeast form a ternary complex with the pre-mRNA segment during mRNA processing. The solution structure of this complex reveals the role of the linker that forms a helical conformation between two RRM domains of Hrp1. This structured linker, arises due to the conformational polymorphism, enhances contacts with the RNA and stabilises the ternary complex. Apart from stabilising the protein-RNA interaction, the linker of dsRBD of ADAR2 is assumed to facilitate its interaction with the stem-loop pre-mRNA by exhibiting fly-casting mechanism as shown in Fig. 4 [4]. Such mechanism contributes to the increased binding affinity of adjacent domains, which, otherwise has weak interaction with the RNA [58].

Fig. 3
figure 3

The RBD of TIS11d bound to AU-rich motif of RNA (PDB ID: 1RGO) [55]. The linker between the RBDs of TIS11d is closely interacting with the RNA, which is stabilised by stacking interaction between F214 (yellow sticks) and U2, and a salt bridge (shown in yellow dashed lines) between E195 and U1. The protein is shown in green cartoon and the RNA is shown in red sticks

Full size image
Fig. 4
figure 4

ADAR2 bound to dsRNA showing multiple-domain interaction regulated by a linker in between (PDB ID: 2L3J) [4]. The protein is shown in green cartoon and the RNA is shown in red stick

Full size image

Spliceosome complex

Among the eukaryotic gene regulation machineries, spliceosome is one major complex that involves extensive protein-RNA interactions. This ribonucleoprotein complex works on precursor mRNA to splice the introns and create the functional open reading frame [59]. Structural investigation on splicing factors has revealed the presence of regions showing polymorphic conformations, which participate in various protein–protein and protein-RNA interactions [60]. A study on human proteome also predicted abundant disorderedness in human spliceosomal proteins [61]. The splicing factors, SR proteins, are one of the most important components of metazoan gene expressions. They have one or two RRM domain at the N terminal and an arginine/serine rich RS domain at the C terminal [62]. Sequence analysis study involving charge hydropathy classification, cumulative distribution function analysis and disorder prediction on SR proteins showed that their amino acid sequence exhibits similar properties to that of IDPs [63]. While the RRM domain is mainly responsible for recognising the specific RNA [64], RS domain interacts with other splicing factors [65]. Furthermore, protein-RNA crosslinking studies has reported that the RS domain bound to splicing enhancers make significant contacts with the pre-mRNA branch point and promote pre-spliceosome assembly formation [66]. This RS domain is vital for the functioning of the splicing factor, and was found completely disordered according to the sequence based studies [63]. This disorderedness is also confirmed by the biophysical study where the CD spectrum shows random coil characteristics of this domain [67]. SR proteins play a very important role in the large spliceosome complex by regulating the splicing phenomenon. Due to the limited structural information of splieceosome complex [68], the molecular and structural mechanism of RNA binding by the flexible RS domain is still elusive; however, the extensive presence of conformational diversity implies their significant role in binding [63].

Viral RNA

Viruses in their host cells are one of the most dynamic living organisms on earth. They constantly adapt itself to the hosts’ environment, come up with the mechanism to evade the hosts’ immune system, propagate and evolve. They are also found to harbour many proteins having highly polymorphic conformations, which performs multiple functions including binding viral RNA. The hepatitis C virus (HCV) codes for a core protein (HCV-C), which has two N terminal binding domains and a signal peptide [69]. Domain 1 has a 150 residue long hydrophilic region that is essential for the assembly of nucleocapsid-like particles (NLPs) and RNA binding; while domain 2 plays a role in HCV-C interaction with lipid droplets [70]. Computational analysis as well as biophysical experiments such as far UV CD spectra and NMR spectroscopy revealed the random-coil conformation of the N-terminal of HCV-C [71]. The chaperoning activity was found to be active even after heat denaturation, which further supports the disorderedness of the HCV-C [72]. Thereby, it was assumed that the flexibility in the terminal region is the facilitating factor in binding assisted folding of the viral RNA in the process of viral genome packaging. In the nucleocapsid protein (NC) of severe acute respiratory syndrome coronavirus (SARS-CoV), stretches of flexible linkers are found in between the two structured domains, NTD (N-terminal domain) and CTD (C-terminal dimerisation domain) [73]. The NC protein binds to the RNA in cooperative manner. Although the structural mechanism of RNA binding by NC protein is still unknown, it has been assumed that the flexible linker between the domains makes significant contacts on multiple RNA sites [73]. The human immunodeficiency virus type-I (HIV-I) encodes a trans-activator of viral transcription named Tat protein [74]. It interacts with kinases and transcription factors thereby facilitating the viral RNA transcription elongation process. The assembly of the elongation complex B is regulated by the interaction of Tat with transactivation response region TAR located at the 5′ end of the viral transcript. Amino acid sequence analysis showed that Tat has high net positive charge and lacks hydrophobic residues, indicating its disordered nature. This has also been established by CD and NMR studies [75]. Disorder to order transition that can be attributed to the conformational polymorphism is observed in anti-termination N proteins of bacteriophages (like P22 and λ) when it interacts with the boxB RNA motif of the viral mRNA [76]. NMR experiments revealed that the N protein is completely disordered and upon binding to boxB RNA, only the N terminal region (N peptide) gets structured and gains stability [77, 78]. This phenomenon is also supported by the molecular dynamic simulation study on P22 N protein and boxB RNA. Here, Bahadur et al. [79] showed that the electrostatic field of the RNA has a favourable influence on the coil-to-a-helix transition of the N peptide. This is evident in Fig. 5, where the N peptide of P22 N protein is disordered in the unbound form but attains a helical conformation upon binding with the boxB RNA.

Fig. 5
figure 5

The disordered N-peptide is transformed into helical conformation upon binding with the boxB RNA (PDB ID: 1A4T) [79]. The peptide is shown in red and the RNA is shown in grey

Full size image

Human diseases associated with IDP-RNA interactions

Protein-RNA interaction, being one of the most abundant cellular phenomenon, is associated with a number of human diseases [80]. IDPs are also annotated with diseases as they are part of various cell signalling and regulatory pathways [81]. Recent analysis shows that a number of RBPs encoded by FET (FUS/TLS, EWS and TARF15) genes have low complexity regions, and are found to play significant role in DNA damage response [82]. Several point mutations were found within the FET, which are associated with protein aggregation resulting in neurodegenerative diseases including amyotrophic lateral sclerosis (ALS) and frontotemporal lobe degeneration (FTLD). A computational study on disease causing mutations revealed that a considerable section of such mutations are associated with the disordered regions [83]. These mutations were found to affect the post-translational modifications, macromolecular assembly and other regulatory processes. The study also shows that the mutations are majorly responsible for the disorder to order transition in the IDRs, hence curbing the structural flexibility and limiting its binding ability. This is evident from the frequent mutations found in the Arg-rich RGG/RG motifs in fused sarcoma protein (FUS), a RBP associated with neurodegenerative disorder ALS [84]. Another autosomal recessive disease, spinal muscular atrophy (SMA) is caused by mutations in the SMN1 (survival of motor neuron) protein [85]. The central region of SMN1 forms the Tudor domain, which shows conformational polymorphism in solutions. Mutations found in the Tudor domain of the protein interfere with its binding to the RG rich domains of various snRNPs, which finally causes neuronal apoptosis [86, 87]. FMRP (fragile X mental retardation protein) is a RBP that exhibits chaperoning activity and has disordered regions [88, 89]. Suppression of this protein is the major cause for the fragile X syndrome, a X-linked disorder. The FMRP is found to recognise U-rich RNA sequences, and interact with various mRNAs along with non-coding miRNA and siRNA [90–93]. The nucleocapsid protein of HIV-1 (NCp7) is another disordered protein having RNA chaperoning function [94]. NCp7 interacts with the viral RNA and coats it by forming oligomers [95]. Based on the degree of RNA occupancy, NCp7 perform activities from assembly of virus particle to genome packaging. The polymorphic conformations of NCp7 allow it to perform such array of activities, which makes it a vital factor for proper viral replication and propagation of the disease [95].

Conclusion

Growing number of atomic structures in PDB facilitate the study of RBP-RNA interactions. Diversity of structural rearrangement of RNA induces the conformational transition to its partner RBP. Mutual folding of RNA and protein in ribonucleoprotein complexes is a common phenomenon and is probably ubiquitous. A recent analysis suggested that the IDPs are proteins waiting for their partner to bind and acquire a particular conformation [96]. Presently, a significant number of atomic structures of RBPs in bound and unbound conformations is available in the PDB [97]. These structural information shows that polymorphic conformation is intrinsic to RBP. In this study, we propose that polymorphic conformation of RBPs promote structural flexibility, which considerably influences the conformational dynamics of RBP-RNA interaction. Indeed, RNA recognition by RBPs is more than a simple handshake and is associated with a variety of conformational flexibility in both the partners. Hence, the lack of structural integrity, camouflaged in conformational polymorphism, is highly favoured in this kind of molecular recognition. Although, the PDB covers only about 15 % of the known protein-RNA complexes of human RBP, many more structural information is required in near future to make a complete repertoire of RBP-RNA interactions, which eventually can lead us to a better understanding of human diseases associated with this recognition process.