The DEAD-box helicase eIF4A

Abstract

DEAD-box helicases catalyze the ATP-dependent unwinding of RNA duplexes. They share a helicase core formed by two RecA-like domains that carries a set of conserved motifs contributing to ATP binding and hydrolysis, RNA binding and duplex unwinding. The translation initiation factor eIF4A is the founding member of the DEAD-box protein family, and one of the few examples of DEAD-box proteins that consist of a helicase core only. It is an RNA-stimulated ATPase and a non-processive helicase that unwinds short RNA duplexes. In the catalytic cycle, a series of conformational changes couples the nucleotide cycle to RNA unwinding.

eIF4A has been considered a paradigm for DEAD-box proteins, and studies of its function have revealed the governing principles underlying the DEAD-box helicase mechanism. However, as an isolated helicase core, eIF4A is rather the exception, not the rule. Most helicase modules in other DEAD-box proteins are modified, some by insertions into the RecA-like domains, and the majority by N- and C-terminal appendages. While the basic catalytic function resides within the helicase core, its modulation by insertions, additional domains or a network of interaction partners generates the diversity of DEAD-box protein functions in the cell. This review summarizes the current knowledge on eIF4A and its regulation, and discusses to what extent eIF4A serves as a model DEAD-box protein.

Introduction into DEAD-box proteins

RNA helicases catalyze the ATP-dependent separation of RNA duplexes and the structural rearrangement of RNA and RNA/protein complexes in all facets of RNA metabolism, from transcription, mRNA splicing and translation, RNA modification and transport, ribosome biogenesis and RNA/protein complex assembly in general, to RNA degradation (reviewed in refs. 1 and 2).

The largest family of RNA helicases is the so-called DEAD-box protein family that belongs to the helicase superfamily 2 (reviewed in refs. 3 and 4). DEAD-box proteins are characterized by a number of conserved motifs that are located in the same spatial arrangement on a helicase core formed by two RecA-like domains (Fig. 1). These helicase motifs line the cleft between the two RecA-like domains in the helicase core, and contribute to ATP binding and hydrolysis, RNA binding and duplex unwinding. The isolated helicase core is an RNA-stimulated ATPase, and a helicase that can only unwind short RNA duplexes.⁵ The core acts as a nucleotide-dependent molecular switch with different affinities for RNA depending on the nucleotide state.⁶ Structural studies have revealed the molecular basis for the nucleotide-regulated RNA affinities. In the absence of RNA or nucleotide, the DEAD-box helicase core adopts a variety of open conformations, in which the two RecA-like domains are separated and do not form inter-domain contacts.^7-13 Binding of RNA and ATP triggers a transition from the flexible open form to the more rigid closed conformation.^13,14 In this compact conformation, an intricate interaction network is formed between the conserved motifs and the bound RNA and ATP.^10,15,16 The nucleotide binds in the inter-domain cleft, and interacts with residues from both domains, involving residues from motifs I, II, V and VI.^10,15,16 Specificity for adenine nucleotides is conferred by a conserved glutamine in the Q-motif¹⁷ that hydrogen bonds with the Watson-Crick face of the base.^10,15,16 Similar to the nucleotide binding site, RNA binding to the helicase core involves interactions with residues from both RecA-like domains, mainly from motifs Ia, GG, Ib, IV, QxxR, V and VI.^10,15,16 ATP and RNA thus jointly stabilize the closed conformation of the helicase core.^10,13,15,16 ATP and RNA binding show a positive thermodynamic linkage in many,^6,13,18-22 but not all DEAD-box proteins.^23-25

Figure 1. RNA unwinding by DEAD-box proteins. (A) Schematic depiction of DEAD-box protein architecture (top) and structure of eIF4A-III (PDB-ID 2j0s) in complex with single-stranded RNA (yellow) and ADPNP (green). The helicase core is formed by two RecA-like domains that carry the conserved helicase signature motifs. Motifs involved in ATP binding and hydrolysis are depicted in red hues, motifs involved in RNA binding in blue hues, and motifs involved in coupling of ATP hydrolysis to duplex separation in purple. In many DEAD-box proteins, the helicase core is flanked by N- and C-terminal extensions. (B) Conformational changes in the catalytic cycle of DEAD-box proteins. In the absence of RNA or ATP, the helicase core of DEAD-box proteins exists in an open conformation, in which the cleft between the two RecA-like domains is open, and there are no inter-domain contacts formed (1). Cooperative binding of ATP and RNA stabilizes a closed conformation of the helicase core (2), with a network of interactions between the conserved helicase motifs, and with the bound ATP and RNA. A rearrangement of hitherto unidentified nature leads to the formation of an activated complex (3). The RNA duplex is locally destabilized by deformation, and the first strand can dissociate (4). The core retains the closed conformation upon ATP hydrolysis (5) and returns to the open conformation upon phosphate release (6). Opening disrupts the bipartite RNA binding site, and hence leads to release of the second RNA strand (6). After ADP dissociation, the helicase core can undergo further catalytic cycles. Reprinted from Methods in Enzymology 511, Alexandra Z. Andreou and Dagmar Klostermeier, Conformational Changes of DEAD-Box Helicases Monitored by Single Molecule Fluorescence Resonance Energy Transfer, pages 75–110, Copyright 2012, with permission from Elsevier.

A mechanistic model of duplex separation by DEAD-box proteins has emerged from a number of biochemical, biophysical and structural studies on a variety of different representatives (Fig. 1B, reviewed in refs. 3 and 26, see also references therein). RNA unwinding is achieved via a nucleotide-driven conformational cycle of the helicase core. The closure of the inter-domain cleft in the helicase core in response to ATP and RNA binding¹³ leads to assembly of the catalytic site for ATP hydrolysis, and to formation of the bipartite RNA binding site.^10,15,16 In the closed conformation, bound ssRNA is bent and its geometry is not compatible with a duplex structure. As a consequence, closure of the inter-domain cleft will lead to local destabilization of bound dsRNA. The first RNA strand is then released from an activated, closed conformer of the helicase, followed by ATP hydrolysis and product release. The rearrangement of the ATP/RNA complex to an activated form has first been proposed by Peck and Herschlag,²⁷ and is evident from two kinetic phases of nucleotide binding,²³ although the structural basis remains unclear.²⁸ Phosphate release²³ and re-opening of the inter-domain cleft²⁹ lead to disruption of the contiguous RNA binding site formed by both domains, and thereby release of the second RNA strand. Altogether, ATP hydrolysis thus resets the helicase core for subsequent catalytic cycles.^29,30 The model reconciles the observations that (1) ATP hydrolysis is not required for duplex separation,³⁰ (2) one ATP is sufficient for one unwinding event,³¹ and (3) increased duplex stability increases the number of ATP molecules hydrolyzed for unwinding.^5,31

The transitions between open and closed conformations are important checkpoints in the catalytic cycle and intimately linked to DEAD-box protein function. Thus, other factors can regulate DEAD-box protein activity by affecting these conformational changes. Inhibition of DEAD-box helicases has been demonstrated by proteins or RNA aptamers that contact both domains of the helicase core and prevent cleft closure^32,33 or re-opening.³⁴ On the other hand, activation can be achieved by stabilizing a half-open conformation from which phosphate release and nucleotide exchange is facilitated.^35,36

Very few DEAD-box proteins consist of an isolated helicase core. Examples for such minimal DEAD-box helicases are the translation initiation factor eIF4A, and the Methanococcus jannaschii DeaD protein. The majority of DEAD-box proteins, however, contains N- and C-terminal extensions that flank the helicase core, and mediate interactions with ATP, RNA or protein partners^37-43 or dimerization,⁴⁴ modulate the ATPase activity,³⁷ or contribute to strand separation.²⁸ In some helicases, insertions into the two RecA-like domains modulate the activity of the helicase core.¹¹ In addition, DEAD-box proteins can be integrated into large functional networks in vivo. The modulation of the helicase core by insertions, additional domains or interaction partners generates the diversity of DEAD-box protein functions in the cell. Nevertheless, the basic catalytic function resides within the helicase core, and the study of minimal DEAD-box proteins has been instrumental in understanding the general mechanism underlying RNA duplex stabilization by DEAD-box proteins.

eIF4A: The founding member of the DEAD-box family

The eukaryotic translation initiation factor 4A (eIF4A, DDX2; Fig. 2) is the archetypal member of the DEAD-box protein family⁴⁵ that “is characterized by a core region of 294 to 359 amino acids that shows strong sequence homology to the mouse elF4A protein, a factor involved in translation initiation.⁴⁶” It has therefore been dubbed “godfather” of DEAD-box proteins.⁴⁷ eIF4A is a minimal DEAD-box protein that only consists of the conserved helicase core carrying the characteristic conserved motifs, with a very short flanking N-terminal (~50 aa) and no C-terminal extension. As it recapitulates many of the common properties of DEAD-box proteins, eIF4A has been considered a paradigm for understanding DEAD-box protein function. It is one of the first DEAD-box proteins that have been studied in detail, leading to the discovery of the governing principles underlying DEAD-box protein function.^6,18,48,49 Biochemical experiments have demonstrated that eIF4A exhibits RNA-stimulated ATPase activity, and a basic RNA unwinding activity,⁵ confirming the concept of eIF4A constituting a minimal helicase module. A kinetic and thermodynamic framework for the nucleotide cycle in the absence and presence of RNA has been reported, and concomitant conformational changes in the catalytic cycle have been deduced from limited proteolysis experiments.^6,18 The activity of eIF4A is regulated by other translation initiation factors that stimulate its ATPase and unwinding activities and/or mediate interaction with RNA.^47,50 The first DEAD-box protein structures determined were of eIF4A domains,^7,51,52 revealing their RecA-like folds, the (largest part of) the nucleotide binding site, and interactions between conserved motifs within the domains, and of complete eIF4A from yeast.⁷ In recent years, structures of eIF4A in complex with interacting partners such as eIF4G³⁵ and Pdcd4,³² and a structural model for the eIF4A/G/H complex⁵³ have been reported. Altogether, these extensive biochemical and structural studies render eIF4A one of the best-characterized DEAD-box proteins.

Figure 2. Domain organization of the mammalian translation initiation factor eIF4A-I and its interaction factors eIF4B, eI4H, eIF4E and eIF4G. For eIF4A-I, the N-terminal RecA-like domain is depicted in red, and the C-terminal in yellow. For eIF4B, the RNA recognition motif (RRM) is shown in cyan, the DRYG repeat domain in black, and the basic domain (BD) in yellow. For eIF4H, the RRM is also depicted in cyan. The eIF4E-like domain of eIF4E is depicted in violet. For eIF4G, the binding domain for the poly-A binding protein (PABP) is indicated in turquoise, for eIF4E in green, for eIF3 in blue, for Mnk1 in purple and the two binding sites for eIF4A-I, i.e. eIF4G-M and eIF4G-C, in orange. Structurally characterized domains are depicted in cartoon representation using the same color code. Upper left: eIF4A-I in the ligand-free, open conformation (PDB-ID 1fuu), upper right: eIF4B RRM (PDB-ID: 2j76), bottom left: eIF4E in complex with eIF4G (residues 393–490; PDB-ID: 1rf8), bottom right: eIF4A-I in complex with eIF4G-M (PDB-ID: 2vso).

Different flavors of eIF4A

Mammalian and plant cells express three eIF4A proteins, eIF4A-I to -III. eIF4A-I and eIF4A-II (DDX2a, DDX2b)⁵⁴ show 90–95% sequence identity. They are differentially expressed, with eIF4A-I present in actively growing cells, and eIF4A-II in quiescent cells,^54,55 and have long been considered to be functionally interchangeable. However, although suppression of eIF4A-I upregulates eIF4A-II transcription, eIF4A-II does not rescue the eIF4A-I-deficient phenotype, indicating that their cellular functions may differ.⁵⁶ Almost all biochemical and structural studies reported to date have been performed with eIF4A-I. The third form, eIF4A-III (DDX48), is more distant from eIF4A-I and –II, with sequence identities of ~60%⁵⁷ and thus not more eIF4A-like than other DEAD-box proteins. eIF4A-III is 10-fold less abundant than eIF4A-I.⁴⁷ It does not complement eIF4A-I in translation,⁵⁸ and no indication for its participation in translation initiation has been found. Instead, eIF4A-III is part of the exon junction complex (EJC), and involved in its assembly.^59,60 The EJC coordinates pre-mRNA splicing with downstream processes such as mRNA localization,⁶¹ translation⁶² and nonsense-mediated decay (NMD).⁶³ In contrast to mammals, the yeast genome codes only for two (identical) eIF4A isoforms, Tif1 and Tif2,⁶⁴ and lacks an eIF4A-III homolog, as well as the other components of the EJC. Despite the sequence identity of ~65%, mammalian and yeast eIF4A orthologs cannot functionally substitute each other.⁶⁵ In Trypanosoma brucei, two eIF4A isoforms have been identified, an abundant cytoplasmic and a nuclear, moderately expressed isoform, possibly representing eIF4A-I and –III, respectively.⁶⁶

eIF4A-I in translation initiation

eIF4A-I is a unique component of the translation initiation factor network (reviewed in ref. 50). Translation initiation is the rate-limiting step of protein biosynthesis and is tightly regulated. Initiation starts with binding of eIF4F, a ternary complex composed of eIF4A-I, eIF4E and eIF4G⁶⁷ (Fig. 2), to the 5′-cap of eukaryotic mRNAs (Fig. 3, recently reviewed in ref. 68). eIF4E recognizes and directly binds to the 5′-mRNA cap, and eIF4G acts as a scaffold protein that interacts with eIF4A-I and eIF4E.^69,70 eIF4F binding to the 5′-cap is followed by the recruitment of the 40S ribosomal subunit. In mammals (but not in yeast), this recruitment is mediated by interactions between eIF4G and eIF3, bound to the 40S subunit.⁷¹ As a next step, the 43S pre-initiation complex scans the 5′-untranslated region (5′-UTR) toward the start codon, where the 60S ribosomal subunit is recruited, and the elongation-competent 80S ribosome is formed. The ribosome on its own has a limited capacity to unwind mRNA secondary structures,⁷² but translation of mRNAs with stable secondary structures in their 5′-UTR depends on eIF4A-I.⁷³ As an RNA-dependent ATPase⁷⁴ and a weak RNA helicase,^5,47,75 eIF4A-I is therefore thought to unwind secondary structures in the 5′-UTR⁷⁶ during scanning.^73,75,77 eIF4B or eIF4H (Figs. 2 and 3) stimulate the eIF4A-I helicase activity during scanning, and possibly stabilize generated single stranded regions by preventing reannealing.^53,78 A Brownian ratchet model for the movement of the 43S ribosomal initiation complex along the mRNA has been proposed.⁷⁹ In addition to its helicase function, eIF4A-I has been suggested to mediate displacement of proteins bound to the 5′-UTR,⁸⁰ or release of the 43S initiation complex from the 5′-cap.^50,81 Recent evidence suggests that other helicases may play additional roles in the translation initiation process, either generally (Ded1p, DHX29) or for specific mRNAs (VAS, RHA; reviewed in ref. 50).

Figure 3. Simplified model of eukaryotic translation initiation. The heterotrimeric protein complex eIF4F, which consists of the DEAD-box RNA helicase eIF4A-I, the scaffolding protein eIF4G and the cap binding protein eIF4E, binds to the m⁷G cap at the 5′-end of the mRNA. The interaction between eIF4G and the poly-A binding protein (PABP) promotes mRNA circularization. eIF4A-I unwinding activity is enhanced by the interaction with eIF4B (or eIF4H). Interaction between eIF4G and ribosome bound eIF3 may facilitate 43S ribosome complex recruitment to the mRNA. eIF4A-I, in complex with its binding partners, unwinds secondary structures in the 5′-untranslated region (UTR) of the mRNA, and allows for scanning of the 43S complex toward the AUG initiation codon.

In neurons, the BC1-RNA binds to eIF4A-I and stimulates its ATPase activity,⁸² but blocks RNA unwinding, leading to the inhibition of translation initiation.⁸³ Translation initiation is frequently deregulated in cancer cells, and a multitude of genes coding for translation initiation factors, including eIF4A-I, change their expression levels in transformed cells and cancers.⁸⁴ Altered eIF4A-I levels in melanoma cells,⁸⁵ human hepatocellular carcinoma,^86,87 lung,⁸⁸ colon and breast cancers,^89,90 and in glioma⁹¹ have linked eIF4A-I to tumorigenesis and cancer, and render it a possible target for anti-cancer therapies (reviewed in refs. 84 and 92).

eIF4A-I is a minimal helicase

The eIF4A-I helicase core carries the conserved motifs involved in ATP binding and hydrolysis, RNA binding and coupling of these activities. Mutational analysis of conserved motifs in eIF4A-I^{48,49,77,93-95} has dissected their major contributions. Motifs I and II (Walker A-, Walker B-motif/DEAD-box) have been implicated in ATP binding and hydrolysis,^48,93,94,96 the Q motif in ATP binding and specificity,⁹⁵ motif VI (HRIGRGGR) in RNA binding and motif III (SAT) in coupling ATPase and duplex separation.⁴⁸ Extensions of these studies to other DEAD-box helicases largely confirmed these functions, but showed that mutations in the conserved motifs have different effects in different DEAD-box proteins, indicating that the significance of a specific motif may be context-dependent.

Further insight into the role of conserved amino acid residues in the catalytic mechanism of DEAD-box proteins was provided by structural studies. The yeast eIF4A-I structure was the first DEAD-box protein crystal structure determined.⁷ It shows eIF4A-I in an open conformation, with the linker connecting the two RecA-like domains completely extended such that the two domains do not interact. Subsequent studies mapping the global conformation of eIF4A-I in solution using single molecule FRET³⁶ demonstrated that eIF4A-I actually adopts a more compact conformation and suggested that this “wide-open” conformation might be stabilized by crystal contacts. A structure of eIF4A-I in the closed state has not yet been reported. The network of interactions between the conserved motifs and the bound nucleotide and RNA was revealed in the crystal structure of eIF4A-III,^10,16 and of other DEAD-box proteins in complex with nucleotide and RNA.^15,28,97 The extensive interaction network between the helicase motifs rationalizes the difficulties in assigning specific functions in mutagenesis studies.

Numerous studies over the past 20 years have focused on characterizing the nucleotide cycle of eIF4A-I and its coupling to RNA binding and unwinding as a representative for the catalytic cycle of DEAD-box proteins. Both in the absence and presence of ssRNA, the affinity of eIF4A-I for ATP is lower than for ADP, which causes product inhibition.⁶ The low ATP affinity provided a first hint toward possible ATP-induced conformational changes, similar to γ-phosphate sensing by other NTPases containing Walker motifs.⁹⁸ In the absence of RNA, ATP is only very slowly hydrolyzed by eIF4A-I. In the presence of polyU-RNA, the Michaelis constant for ATP, K_M,ATP, is ~80 μM and the turnover number k_cat is 0.01–0.05 sec^{-1 6, 36}.

RNA binding to eIF4A-I occurs with low affinity, and independent of sequence.^48,49,99 Binding of ssRNA stimulates the ATPase activity of eIF4A-I.^6,27,74 From ribonuclease protection experiments, the RNA binding site of eIF4A-I was mapped to comprise 10–15 nucleotides. However, shorter RNAs already stimulate ATP hydrolysis substantially,^27,99,100 and the structure of eIF4A-III (and of other DEAD-box proteins) revealed contacts with six nucleotides. The 2´-OH is not required for RNA binding, but for the ATPase stimulation,²⁷ and it has been suggested that binding of RNA leads to a conformational change that establishes contacts with the 2′-OH. The crystal structure of eIF4A-III in complex with ssRNA and ADPNP confirmed that four of six 2′-OH groups of the RNA substrate form contacts with the protein.^10,16 In contrast to ssRNA, dsRNA binds weakly or not at all to mouse eIF4A-I,⁶ and dsRNA, hairpins or other structured RNA do not stimulate the ATPase activity.^6,99 However, binding of dsRNA has been inferred from the inhibitory effect of excess ss- and dsRNA (but not ss- or dsDNA) on RNA unwinding.¹⁰¹ Notably, this finding gave rise to a first model for RNA unwinding, in which eIF4A-I binds to an RNA duplex, and unwinds it either at the end or internally.¹⁰¹ Contrary to the observation with mouse eIF4A-I, a recent study reported ATP-dependent binding of dsRNA to yeast eIF4A-I, with dissociation constants of 70 nM in the absence, and 20 nM in the presence of ATP.¹⁰² ssRNA bound less tightly in the presence of ATP, with a K_d value of 1–2 μM, and not at all in the absence of ATP. The RNA-stimulated ATPase rate was 10-fold higher with dsRNA than with ssRNA. Interestingly, the ATPase activity was not efficiently stimulated by dsRNA when eIF4A-I was part of the eIF4F complex.¹⁰² The functional significance of the high-affinity dsRNA binding by yeast eIF4A-I is currently unclear.

Significant efforts have been made toward understanding the link between the nucleotide cycle and RNA binding. eIF4A-I binds RNA and ATP (but not ADPNP) with positive thermodynamic linkage.⁶ RNA stimulates ATP hydrolysis with an apparent K_M,_app value of 30–100 μM (per base of poly-U).^6,36 In the nucleotide cycle, eIF4A-I alternates between a high affinity-state for ssRNA when ATP is bound^6,49 and a low-affinity-state when ADP is bound.⁶ Nucleotide binding and hydrolysis are associated with conformational changes, which have been detected by limited proteolysis of different eIF4A-I/RNA/nucleotide complexes.¹⁸ Interestingly, ATPγS, which is frequently used as a “non-hydrolyzable” nucleotide analog for many ATPases, is hydrolyzed by eIF4A-I as efficiently as ATP. However, eIF4A-I is less efficiently cross-linked to RNA in the presence of ATPγS, suggesting that ATPγS induces a different eIF4A-I conformation than ATP.¹⁸ Altogether, these experiments characterized eIF4A-I as an ATP-dependent conformational switch, and established that the nucleotide cycle drives a series of conformational transitions and thus cyclic changes in RNA affinity. Changes in the mRNA affinity during the eIF4A-I nucleotide cycle have been suggested to prevent backsliding, and enforcing unidirectional 5′-to-3′-movement of the 43S complex.¹⁰³

eIF4A-I is an ATP-dependent RNA helicase.^5,75,76 ATP hydrolysis (or an associated conformational change) was initially identified as the rate-limiting step in the catalytic cycle of mouse eIF4A-I.⁶ The observation that ATPγS is hydrolyzed by mouse eIF4A-I with similar k_cat values as ATP¹⁰⁴ argues in favor of a rate-limiting conformational change associated with ATP hydrolysis. For yeast eIF4A-I, ATP hydrolysis and phosphate release are rate-limiting.³⁶ Notably, dissociation of RNA from eIF4A-I is faster than ATP hydrolysis.⁶ As a consequence, ATP hydrolysis and formation of the ADP state lead to rapid dissociation of eIF4A-I from the RNA, rationalizing why eIF4A-I is not a processive helicase.⁶

The rate of RNA unwinding increases by 30% when the RNA substrate contains ssRNA regions compared with that of blunt-end RNA,¹⁰¹ independent of the polarity of the single-stranded region. Together with the observation that eIF4A-I unwinds RNA/DNA heteroduplexes of either orientation with similar rates,⁵ this has led to the proposal of eIF4A-I being a “bidirectional helicase.” The unwinding rate depends on the thermodynamic stability of the duplex, not of its length or sequence.⁵ Longer duplexes are partially unwound by eIF4A-I, and the stability of the remaining duplex determines whether the strands dissociate (Fig. 1B and ref. 5). DNA duplexes are not unwound. ATPγS also supports RNA unwinding,¹⁰⁴ with a 10-fold reduced rate in comparison to ATP-driven unwinding. In contrast to these observations, a recent single molecule study detected hairpin unwinding by eIF4A-I with ATP, but not with ATPγS.¹⁰⁵

Stimulation of eIF4A-I helicase activity by other translation initiation factors

Although isolated eIF4A-I catalyzes ATP-dependent RNA unwinding, its helicase activity is very low. RNA unwinding by eIF4A-I is stimulated by the translation initiation factors eIF4B,⁷⁴ eIF4H, eIF4E and eIF4G.^106-108 In mammalian cells, more than 90% of eIF4A-I exist in an uncomplexed form.¹⁰⁹ The remaining 10% of eIF4A-I are part of the eIF4F complex, formed by eIF4A-I, eIF4E and eIF4G. The cellular function of free eIF4A-I is not fully understood, and it is thought that free eIF4A-I is recycled through the eIF4F complex.⁷⁷ In the eIF4F complex, the RNA-stimulated ATPase and helicase activities of eIF4A-I are increased.^75,106 In addition, the translation initiation factors eIF4B and eIF4H stimulate the helicase activity of eIF4A-I.¹⁰⁶ Based on the structure of unliganded eIF4A-I with its extended linker and the large separation of the two core domains it had been suggested that interacting factors may activate eIF4A-I by facilitating the formation of and stabilizing the closed conformation.⁷ From a number of structural and biochemical studies, the molecular mechanism of eIF4A-I activation by other translation initiation factors is beginning to emerge. The current knowledge is summarized in the following.

eIF4B and eIF4H

eIF4B is one of the least conserved translation factors, with sequence identities of 25–30% across species.¹¹⁰ Functional domains of eIF4B (70 kDa) have been identified from mutagenesis studies (Fig. 2). The N-terminal region contains an RNA recognition motif (RRM) which binds to 18S rRNA, but is dispensable for the interaction with eIF4A-I.^108,111 The C-terminal region contains a second RNA binding domain rich in arginine and serine residues (basic domain).¹¹¹ A domain rich in aspartic acid, arginine, tyrosine and glycine (DRYG domain) mediates self-association of eIF4B and interaction with the translation initiation factor eIF3.¹¹² Dimerization of wheat eIF4B^110,112 is required for RNA binding.¹¹³ So far, the only available structural information on eIF4B is a solution structure of the RRM domain which adopts a typical βαββαβ topology¹¹⁴ (Fig. 2).

eIF4H (25 kDa) shares a common core region with eIF4B, and also contains RRMs, but lacks the N- and C-terminal domains (Fig. 2). eIF4B and eIF4H are ubiquitously expressed,^115,116 but their relative expression levels vary between different tissues,¹¹⁶ suggesting that both factors regulate translation of different transcripts, or belong to different regulatory pathways.¹⁰⁸ The interaction sites of (human) eIF4B and eIF4H on eIF4A-I overlap, and their binding to eIF4A-I is mutually exclusive.¹⁰⁸ The eIF4A-I binding site has been mapped to the C-terminal 280 amino acids of eIF4B,¹⁰⁸ and to the C-terminal region of eIF4H.¹¹⁷ The C-terminal region of eIF4B on its own binds and stimulates eIF4A-I.^108,112 Despite their similar binding sites on eIF4A-I, the eIF4A-I-binding regions of eIF4B and eIF4H do not share sequence similarity.¹⁰⁸ eIF4H cannot completely substitute for eIF4B in the assembly of the translation initiation complex on mRNAs containing secondary structures,¹¹⁸ but the functional difference between eIF4B and eIF4H is not clear.

Although the stimulatory effects of eIF4H and eIF4B on eIF4A-I activity have been discovered a decade ago,¹⁰⁶ the underlying molecular mechanism and mode of interaction is still not well established. eIF4A-I and eIF4B do not co-precipitate in the absence of nucleotide and RNA,¹¹⁹ but their interaction has been observed in electrophoretic mobility shift assays.¹⁰⁸ A stable eIF4A-I/eIF4B complex has been detected in the presence of RNA and ADP·AlF_x, but complex formation requires the presence of eIF4G (ref. 119, see below), and is more efficient when the eIF4B C-terminus is deleted. The RNA-binding region in the eIF4B C-terminal tail has been implicated in stimulation of duplex unwinding by eIF4A-I in presence of eIF4G.¹²⁰ eIF4B and eIF4H stably interact with the N-terminal domain of eIF4A-I in presence of ADPNP and ssRNA,¹⁰⁸ similar to the interaction of MLN51 with eIF4A-III^10,16 (see below). eIF4H forms additional interactions with the C-terminal domain of eIF4A-I.⁵³ Interestingly, eIF4H also interacts with eIF4A-II in vitro.¹¹⁷ Both eIF4B and eIF4H increase the affinity of eIF4A-I or eIF4F for nucleotides and RNA,⁵³ and enhance the unwinding activity of eIF4A-I.^75,108 Measurements of unwinding and ATP hydrolysis rates under identical conditions demonstrated that the coupling of ATP hydrolysis with unwinding is increased in the eIF4A-I/eIF4B complex.¹²⁰

Formation of a stable eIF4A-I/eIF4B complex requires the presence of eIF4G.¹¹⁹ Surprisingly, eIF4G is not part of the complex, suggesting a catalytic role in complex assembly.¹¹⁹ Possibly, eIF4G or eIF4B binding to eIF4A-I is mutually exclusive, and the formation of the eIF4A-I/eIF4G and the eIF4A-I/eIF4B/RNA/nucleotide complex are two separate events in assembly of the translation initiation complex.¹¹⁹ Wheat eIF4B directly interacts with eIF4A-I and eIF4G.¹¹³ Mammalian eIF4G and eIF4B synergistically stimulate the ATPase activity of eIF4A-I,¹¹⁹ and eIF4B or eIF4H and eIF4G cooperatively enhance RNA unwinding.¹²⁰ The stimulatory effect of eIF4H is smaller than the effect of eIF4B.¹²⁰ Recent single molecule experiments monitoring unwinding of a hairpin by eIF4A-I detected rare unwinding events with eIF4A-I alone, but rapid and repetitive unwinding for extended periods of time when eIF4H was present.¹⁰⁵ In these experiments, eIF4H directed eIF4A-I to loop regions (> 6 bases) in the RNA substrate.¹⁰⁵ From the constant unwinding time even at low ATP concentrations, it was deduced that a single ATP is hydrolyzed per unwinding cycle.¹⁰⁵ eIF4B and eIF4H show different effects on eIF4A-I and eIF4F. eIF4A-I unwinds RNA substrates with a 5′- or 3′-ss RNA region slightly more efficiently than blunt end RNA in the absence of eIF4B or eIF4H, but much more efficiently in their presence.¹¹⁶ eIF4F, in contrast, strictly requires a ssRNA region in the presence of eIF4B or eIF4H.¹¹⁶ eIF4A-I alone preferentially binds ~17 nucleotide-RNA, whereas the eIF4A-I/eIF4H or eIF4A-I/eIF4B complexes bind 30mer oligonucleotides,¹⁰⁸ and the eIF4F complex binds oligonucleotides of > 60 bases.¹²¹ eIF4B on its own displays RNA annealing activity,¹²² but the significance of annealing in translation is not clear. Binding of eIF4B or eIF4H to RNA has been suggested to stabilize single stranded regions in the 5′-UTR,⁷⁸ or prevent reannealing,⁵³ thereby contributing to processive unwinding in 5′-to 3′-direction.

eIF4G

eIF4G functions as a scaffold protein that mediates interactions between eIF4A-I and other translation factors.^68,123 Mammalian eIF4G is a ~1600 aa (175 kDa) protein with binding sites for the poly-A-binding protein (PABP), the cap-binding protein eIF4E, eIF3, the eIF4E kinase, Mnk1 and eIF4A-I (reviewed in ref. 124 and Fig. 2). Binding of a number of these partners is cooperative, indicating that eIF4G provides a dynamic scaffold for these interactions.¹²⁴ Human eIF4G contains three consecutive HEAT repeats, helically stacked α-helical hairpins, HEAT-1 to HEAT-3. HEAT-1 is located in the middle (eIF4G-M), and HEAT-2 and HEAT-3 are in the C-terminal region (eIF4G-C). HEAT-1 and HEAT-2 constitute binding sites for eIF4A-I.¹²⁵ Yeast eIF4G (~150 kDa), lacks HEAT-2 and HEAT-3, and thus the second eIF4A-I-binding domain. Current structural information is limited to an NMR structure of a eIF4G fragment bound to eIF4E,¹²⁶ and crystal structures of eIF4G-M,¹²⁷ eIF4G-C¹²⁸ and of the eIF4A-I-eIF4G-M complex.³⁵ Moreover, a SAXS model for eIF4G-M and the eIF4G-M/eIF4A-I complex¹¹⁹ has revealed the dimensions of the elongated complex (220Å) as sufficient to span the 40S ribosomal subunit.

HEAT-1 is the primary interaction site of eIF4G with eIF4A-I, and interacts with both eIF4A-I domains.¹²⁹ Binding of HEAT-1 stimulates the eIF4A-I ATPase activity,³⁵ whereas binding of HEAT-2 modulates eIF4A-I activation, but is not essential for translation initiation.¹³⁰ Although yeast eIF4G lacks HEAT-2, it also contacts both eIF4A-I domains.^35,36

Human eIF4G-MC and eIF4G-C form complexes with eIF4A-I in the absence of nucleotide or RNA.¹¹⁹ The complex is stabilized by a high- and a low-affinity binding site (K_d values ~50 nM, ~1 μM), with HEAT-2 (eIF4G-C) providing the low-affinity binding site.¹¹⁹ F978 in eIF4G-M (equivalent to F838 of yeast eIF4G) contributes to high affinity binding. eIF4G-M stimulates the eIF4A-I ATPase by reducing K_M 10-fold, and increasing k_cat 4-fold, and promotes RNA binding.¹⁰⁷ eIF4G-C, in contrast, does not stimulate the ATPase, but instead increases K_M,ATP and K_M,RNA. For the human eIF4A-I/eIF4G complex, interactions have been mapped in NMR experiments.¹²⁹ The crystal structure of yeast eIF4A-I in complex with eIF4G-M³⁵ revealed a primary interface formed by the C-terminal RecA-like domain of eIF4A-I and the N-terminal region of eIF4G-M, and a secondary interface formed by the N-terminal RecA-like domain and the C-terminal part of eIF4G-M. In addition, the conserved W579 in the N-terminal region of eIF4G-M contributes to complex stability by packing into a pocket on the C-terminal RecA-like domain of eIF4A-I.³⁵ eIF4G-M induces a conformational change of eIF4A-I that is correlated with the stimulation of eIF4A-I activity.³⁶ In the resulting “half-open” conformation, the conserved motifs of eIF4A-I around the inter-domain cleft are pre-aligned.^35,36 Nucleotide and phosphate release,³⁶ as well as RNA dissociation⁹⁷ are stimulated from the eIF4A-I/eIF4G-M complex, explaining the overall enhancement of the eIF4A-I ATPase activity.^36,107 The tight interaction between eIF4A-I and eIF4G via the primary interface provides an anchor, such that a stable complex will be maintained even if the secondary interface is disrupted. Although the low-affinity secondary interface is not required for complex stability, both stabilization and destabilization reduce the stimulation of the eIF4A-I ATPase by eIF4G, emphasizing the required balance.³⁶ Thus, the two interfaces in conjunction allow for eIF4A-I to switch into the closed conformation upon ATP and RNA binding, while maintaining a stable interaction with eIF4G. A conformational guidance mechanism has been proposed, in which the N-terminal RecA-like domain will alternate between interactions with eIF4G in the “half-open” conformation and the C-terminal domain of eIF4A-I in the closed conformation.³⁶ When eIF4A-I does not completely re-open, the RNA affinity will not be reduced to the same extent as in the open conformation and RNA may remain bound. Interestingly, eIF4G may assist in the assembly of an eIF4A-I/eIF4B/RNA/nucleotide complex, possibly by inducing a conformational change in eIF4A-I that then allows for the interaction with eIF4B, ATP and RNA.¹¹⁹

In addition to its function as a scaffold for other proteins, eIF4G also binds RNA. Yeast eIF4G contains three RNA binding domains, RNA1-RNA3. Deletion of each one individually^102,131 still allows for tight binding of eIF4G to ssRNA and to eIF4A-I. The three RNA binding domains of eIF4G cooperatively confer 5′-end specificity to eIF4A-I when it is part of the eIF4F complex.¹⁰² Deletions reduce unwinding rates for RNA with a 5′-ssRNA overhang, and increase unwinding rates for RNA with a 3′-ssRNA overhang. Thus, it appears that the RNA binding regions on eIF4G in conjunction stimulate unwinding by eIF4A-I in the 5′-to-3′direction, and suppress 3′-to-5′-unwinding.¹⁰² Interestingly, the presence of the 5′-cap on the mRNA neither strongly affects RNA binding,^102,121 nor does it have a significant effect on unwinding rates,¹⁰² indicating that the cap/eIF4E interaction is not required for selectivity of eIF4F for the 5′-end of mRNAs. In contrast to yeast eIF4F, human eIF4F does not show a preference change to 5′-overhangs,¹²¹ suggesting that the mechanism is not universal.

The human eIF4A/eIF4G/eIF4H complex

The overall topology of the human eIF4A-I/eIF4G/eIF4H complex has been dissected in a combination of NMR, mutational and binding studies.⁵³ Interactions were identified between the N-terminal RecA-like domain of eIF4A-I and the C-terminus of HEAT-1 in eIF4G, the C-terminal RecA-domain of eIF4A-I with HEAT-1, of both eIF4A-I domains with HEAT-2, and of the C-terminal RecA-like domain of eIF4A-I and the linker between HEAT-1 and HEAT-2 in eIF4G. eIF4H forms contacts exclusively with the C-terminal RecA-domain of eIF4A-I.⁵³ Interactions with eIF4G and eIF4H modulate the nucleotide affinity of eIF4A-I. At the same time, the individual interactions are affected differentially by ATP and ADP. The interaction between eIF4H and eIF4A-I is stimulated by ATP and ADP. The presence of ATP or ADP also increases the affinity of eIF4A-I to HEAT-1, but reduces the affinity to HEAT-2 ⁵³. These results suggest a rearrangement of interactions within the eIF4A-I/eIF4G/eIF4H complex during the nucleotide cycle of eIF4A-I,⁵³ and point to a network of specific, yet highly dynamic interactions of eIF4G and eIF4H with eIF4A-I that regulate the eIF4A-I conformational cycle.⁵³

Altogether, eIF4A-I appears to act as a central hub in translation initiation that is regulated by interactions with other translation initiation factors, and integrates and amplifies the individually only subtle input signals.

Other factors that regulate eIF4A-I activity

The current knowledge on regulation and modulation of eIF4A-I activity suggests that many of its interaction partners modulate the conformational cycle. Apart from translation initiation factors, examples of other eIF4A-I regulators that target the conformational change have been described: The tumor suppressor protein Pdcd4 contacts both domains of eIF4A-I, keeping it in a non-productive conformation, and preventing the formation of the closed state.³² As a result, it interferes with RNA binding and inhibits eIF4A-I activity and cap-dependent translation initiation.¹³² Similarly, an eIF4A-I-binding RNA aptamer has been described that binds to both RecA-like domains, possibly preventing closure.³³ Pateamine A, a natural compound from the marine sponge Mycale Sp,¹³³ relieves the negative regulation by the inter-domain linker,¹³⁴ and stimulates eIF4A-I ATPase and unwinding activities by affecting the eIF4A-I conformation. Despite the stimulation of eIF4A-I activity, pateamine A inhibits translation initiation.¹³⁵ Targeting the conformational changes of DEAD-box proteins may thus be a common regulatory theme that, depending on the context, allows for inhibition or stimulation of cellular processes.

eIF4A-III in splicing: “Clamping”

In contrast to eIF4A-I and eIF4A-II, eIF4A-III (DDX48, Fig. 4) is a nuclear protein. Although eIF4A-I and eIF4A-III are similar in sequence and functional chimeras have been generated,¹³⁶ eIF4A-III cannot substitute for eIF4A-I in translation initiation. eIF4B and eIF4H do not stimulate the helicase activity of eIF4A-III,¹⁰⁸ and binding of eIF4G to eIF4A-III is without functional consequences.⁵⁸ eIF4A-III on its own does not show helicase activity.^108,137 In its best-characterized function as a component of the exon junction complex (EJC), eIF4A-III anchors the EJC on RNA.⁵⁹ The EJC core complex is formed by eIF4A-III, and the proteins MLN51 (Casc3, Barentz), and the Magoh/Y14 heterodimer¹³⁸ (Figs. 4 and 5). The EJC binds to mRNAs ca. 20–25 nucleotides upstream of the exon-exon junction during pre-mRNA splicing.¹³⁹ The EJC components accompany mRNA from the nucleus to the cytoplasm,¹³⁹ and couple splicing to downstream processes.¹⁴⁰ After assembly, the EJC interacts with Up-frameshift protein 3, (Upf3), which is one of the effectors of nonsense mediated decay (NMD) and mediates interaction of the EJC with the NMD machinery. A crystal structure of the Upf3-EJC complex¹⁴¹ revealed that Upf3 contacts the same interaction site on eIF4A-III that is contacted by eIF4G in the eIF4A-I/eIF4G complex.³⁵ Its participation in the EJC not only links eIF4A-III to NMD of mRNAs,^139,142 but also to mRNA localization,^61,63 mRNA export⁶³ and coupling of splicing and translation.⁶²

Figure 4. Domain organization of eIF4A-III and its interaction partners MLN51, Magoh, and Y14 in the exon junction complex. Domain organization of the human exon junction complex components eIF4A-III, Y14, Magoh and MLN51. The N-terminal RecA-domain of eIF4A-III is depicted in teal, the C-terminal domain in brown, the RRM of Y14 in pink, Magoh in cyan and the Btz domain of MLN51 in yellow. Bottom: Structure of the exon junction complex in cartoon representation using the same color code (PDB ID: 2j0s). Bound RNA is depicted in red.

Figure 5. eIF4A-III, the exon junction complex and its link to pre-mRNA splicing and mRNA fate. eIF4A-III is a core component of the exon junction complex (EJC) together with the Y14-Magoh heterodimer and MLN51. The complex is deposited ~25 nucleotides upstream of the exon-exon junction during splicing and remains bound during nuclear export of the RNA into the cytoplasm. The EJC plays a role in mRNA localization, translation and nonsense mediated decay (NMD). It interacts with one of the NMD effectors, Up-frameshift protein 3 (Upf3), in the nucleus. Upf3 then anchors Upf2 in the cytoplasm. The triggering step of NMD upon Upf1 binding to the EJC-Upf3-Upf2 complex is not shown. A prerequisite for activation of the NMD pathway is the translation termination at a stop codon that is 30 nucleotides upstream of an EJC.

Similar to eIF4A-I, eIF4A-III is in an open conformation in the absence of ligands.¹⁰ MLN51 and Magoh/Y14 form a stable complex with eIF4A-III in the presence of ADPNP and ssRNA,¹³⁶ and stimulate the eIF4A-III ATPase and RNA helicase activities.^136,137 Magoh and MLN51 contact different regions on eIF4A-III.¹³⁶ The eIF4A-III binding site of MLN51 has been mapped to its SELOR domain, which is necessary and sufficient for eIF4A-III binding.¹³⁶ In the EJC core complex with ADPNP, ssRNA, MLN51, Magoh and Y14 bound, eIF4A-III adopts a closed conformation similar to Vasa.^10,15,16 Magoh/Y14 interacts with the eIF4A-III C-terminal domain and the inter-domain linker.^10,16 MLN51 binds to the opposite side and contacts both the N- and C-terminal domains and the bound RNA, thereby stabilizing the RNA complex.^10,16 The binding site of MLN51 to eIF4A-III corresponds to the eIF4B/eIF4H binding site on eIF4A-I, although MLN51 has no sequence homology with either protein.¹⁰⁸ MLN51 also enhances eIF4A-I helicase activity, although less efficiently than for eIF4A-III.¹⁰⁸ The bipartite interaction of MLN51 with eIF4A-III is stabilized by Magoh and Y14.^10,16,136 In the closed conformation of eIF4A-III in the EJC core, ATP is hydrolyzed, but product release is inhibited.³⁴ As a consequence, eIF4A-III is trapped in the ADP-P_i form, tightly bound to its RNA substrate, and the EJC remains anchored to the mRNA.³⁴ Strikingly, both Magoh/Y14 binding to eIF4A-III and eIF4G binding to eIF4A-I involve both eIF4A domains, but in one case eIF4A-III is clamped in the closed conformation, preventing product release,³⁴ while in the other case the inter-domain interface is pried open to facilitate product release.³⁶

The cellular functions of eIF4A-III are not universally conserved. For example, the EJC constituents are not required for NMD in Drosophila and Caenorhabditis.^61,143,144 In yeast, NMD is independent of splicing.¹⁴⁵ Here, an eIF4A-III functional ortholog, Fal1p, interacts with the eIF4G-like protein, Sgd1p (NOM1 in humans), in the nucleolus.¹⁴⁶ The interaction between Fal1p and Sgd1p involves the corresponding residues that mediate eIF4A-I/eIF4G complex formation.¹⁴⁶ Fal1p has been linked to pre-mRNA processing¹⁴⁶ and to ribosome (18S rRNA) biogenesis,¹⁴⁷ a function that requires ATP hydrolysis. In humans, in addition to its function as a component of the EJC, eIF4A-III represses selenoprotein translation when selenium is limiting.¹⁴⁸ Upon selenium depletion, eIF4A-III expression levels increase, and binding of eIF4A-III to the selenocysteine insertion sequence in the 3′-region of the respective mRNA^148,149 leads to selective repression of transcription in a transcript-specific manner. Similar to eIF4A-I, both eIF4A-III domains contact the RNA of the selenocysteine insertion element.¹⁵⁰ Although the functions of eIF4A-III are distinct from eIF4A-I, both seem to share common principles in their action and regulation.

eIF4A: A Universal Model?

eIF4A-I is undoubtedly the best-characterized DEAD-box protein, and has been considered as a general model for DEAD-box protein function. Sequence comparisons have defined the region starting immediately upstream of the Q-motif to ~35 amino acids downstream of motif VI as the minimal functional unit of DEAD-box proteins,¹⁵¹ corresponding to eIF4A-I helicase core. The large body of biochemical, biophysical and structural data on eIF4A-I activity has established that the helicase core is a scaffold that carries the conserved motifs mediating ATP and RNA binding and their coupling required for helicase function. The helicase core acts as a nucleotide-dependent switch that alternates between open and closed conformations, with different RNA affinities at different stages of the nucleotide cycle.^6,18 The interaction with RNA is not sequence-specific.⁹⁹ Unwinding is non-processive⁵ and non-directional.⁷⁵ Studies on the helicase cores of a number of other DEAD-box proteins have largely confirmed these central features, supporting that the helicase core is a general RNA unwinding module. While the formation of the closed conformer has only been inferred for eIF4A-I, but so far not been shown experimentally, an increasing number of crystal structures, SAXS models and single molecule FRET data^{13,15,28,42,97,152} has illustrated that the ATP-driven conformational cycle is universal among DEAD-box proteins. However, the same conformational cycle can have different functional consequences, as evidenced by the different biochemical properties, cellular localization and functions of the closely related helicase cores of eIF4A-I and eIF4-III, as well as functional differences of eIF4A-I orthologs from different organisms.

Notably, as an isolated helicase core, eIF4A-I is the exception, not the rule. Most helicase modules in other DEAD-box proteins are modified, some by insertions into the RecA-like domains, and the majority by N- and C-terminal appendages. The recognition of the helicase core as a common functional element suggested a modular architecture of DEAD-box proteins. Supporting this notion, deletion of C-terminal domains from DEAD-box helicases resulted in helicase cores with eIF4A-I-like RNA-affinity,^24,25,153 and the RNA substrate specificity of the SrmB helicase has been altered by fusing its core to the RNA binding module from the DEAD-box helicase YxiN⁴¹ that confers specificity for 23S rRNA.¹⁵⁴ In contrast, efforts to construct functional chimeras by modular synthesis from a helicase core and additional domains have been less successful. A functional helicase has not been obtained by fusing eIF4A-I with flanking domains from other DEAD-box proteins.⁴⁷ An exchange of the cores and flanking domains between Ded1p and Dbp1 from yeast led to a functional helicase in one direction, but not in reverse.¹⁵¹ The limited success of these approaches demonstrates that the helicase modules and their appendages are not generally interchangeable, and implies that the molecular basis for their functional cooperation is more complex.

The activity of the canonical eIF4A-I helicase core depends on other translation initiation factors, and is thus regulated in trans, which facilitates dissection of the individual effect of these partners. Other examples of DEAD-box helicases that depend on binding partners for their activity are the helicase Dbp8, whose ATPase activity is stimulated upon binding of the helicase activator Esf2,^155,156 and the component of the degradosome, RhlB, whose ATPase activity is stimulated by RNase E.¹⁵⁷ From the extensive studies of the regulation of eIF4A-I activity within the translation initiation network, a picture for regulation of the helicase core by influencing the conformational cycle is beginning to emerge.^35,36,53,129 eIF4G stimulates the progression of eIF4A-I through the catalytic cycle by stabilizing a half-open conformation that allows for accelerated product release.^35,36 The similar conformation of the DEAD-box protein Dbp5 in complex with its activator Gle1 (and the co-activator inositol hexakisphosphate) suggests that such a regulation mechanism may be more widely used.⁹⁷ Though currently only few DEAD-box helicases have been identified that are regulated by accessory proteins, it is likely that regulation of DEAD-box protein activity by altering the conformational cycle will turn out to be a more general principle.

Altogether, eIF4A is a degenerate helicase module that recapitulates the general facets of the DEAD-box helicases. The eIF4A-like helicase core is a recurring element of DEAD-box proteins, and eIF4A can therefore be regarded as a paradigm for this enzyme class. We have undoubtedly learned a large part of the underlying principles of DEAD-box protein activity from eIF4A-I. However, the mode of action of the helicase core is context-dependent, and not universal. eIF4A relies exclusively on interactions with protein partners within a functional network. In contrast, a large repertoire of modifications of the helicase core, ranging from sequence alterations within the core to insertions and flanking domains, tailors the core properties to a specific cellular function. With the identification of novel helicase modules as part of larger enzymes, their binding partners in vivo and their RNA substrates, we are now just beginning to understand the molecular mechanisms that generate the tremendous variety of DEAD-box protein functions in their cellular context.

Perspectives

The plethora of biochemical data on eIF4A-I collected over the past 25 years has provided us with a detailed picture of the mechanism of duplex destabilization by eIF4A-I, and the governing principles of DEAD-box protein activity in general. However, the molecular basis of the functional cooperation of the DEAD-box helicase core with additional domains and interaction partners is currently not well-understood. eIF4A exerts its biological functions as part of a large functional network via a large number of interactions, many of them of transient nature. It is evident that many other DEAD-box proteins also act within functional networks in the cell, targeted by additional domains that mediate protein/protein interactions. The combination of cross-linking, structural biology and single molecule approaches will facilitate the identification and characterization of short-lived complexes of eIF4A and its binding partners. These approaches are also invaluable to identify specific interaction partners of other DEAD-box proteins, as well as their physiological RNA substrates. It will be exciting to discover the molecular basis for the extensive spectrum of cellular functions that are generated based on a simple, common module.

Glossary

Abbreviations:

ATPγS

adenosine-5′-O-(3-thio)triphosphate

ADPNP

5′-adenylyl-β,γ-imidodiphosphate

EJC

exon junction complex

FRET

fluorescence resonance energy transfer

NMD

nonsense-mediated decay

SAXS

Small angle X-ray scattering

5′-UTR

5′-untranslated region