Biochemical and Kinetic Characterization of the RNA Helicase Activity of Eukaryotic Initiation Factor 4A*

Eukaryotic initiation factor (eIF) 4A is the prototypic member of the DEAD box family of proteins and has been proposed to act as an RNA helicase to unwind secondary structure in the 5′-untranslated region of eukaryotic mRNAs. Previous studies have shown that the RNA helicase activity of eIF4A is dependent on the presence of a second initiation factor, eIF4B. In this report, eIF4A has been demonstrated to function independently of eIF4B as an ATP-dependent RNA helicase. The biochemical and kinetic properties of this activity were examined. By using a family of RNA duplexes with an unstructured single-stranded region followed by a duplex region of increasing length and stability, it was observed that the initial rate of duplex unwinding decreased with increasing stability of the duplex. Furthermore, the maximum amount of duplex unwound also decreased with increasing stability. Results suggest that eIF4A acts in a non-processive manner. eIF4B and eIF4H were shown to stimulate the helicase activity of eIF4A, allowing eIF4A to unwind longer, more stable duplexes with both an increase in initial rate and maximum amount of duplex unwound. A simple kinetic model is proposed to explain the mechanism by which eIF4A unwinds RNA duplex structures in an ATP-dependent manner.

Initiation of protein synthesis in mammalian systems is a complex process in which the 40 S and 60 S ribosomal subunits are joined with mRNA and initiator methionyl-tRNA (Met-tRNA i ) 1 to form a translationally competent 80 S initiation complex (for reviews on translation, see Refs. [1][2][3]. Prior to the formation of an 80 S complex, a 48 S pre-initiation complex composed of the mRNA, 40 S ribosomal subunit, initiator methionyl-tRNA, and several initiation factors (eIFs) must be formed. The first major event in the formation of this 48 S complex is the binding of eIF2, GTP, and Met-tRNA i to the 40 S ribosomal subunit, which is facilitated by eIF1A (1). The second major event is the binding of mRNA to this 40 S ribosomal subunit. This is a key regulatory step in the initiation of protein synthesis and involves eIF3, eIF4A, eIF4B, eIF4F, and possibly eIF4H, a novel initiation factor demonstrated to interact with RNA (1,3,4).
Eukaryotic mRNAs are typically recognized via their 7-methylguanosine (m 7 G) cap structure by eIF4E, the small subunit of eIF4F (5). eIF4F also contains a large subunit (170 kDa), eIF4G, and a subunit that is eIF4A (thus eIF4A may exist as part of eIF4F or in free form) (6). Following binding of the eIF4F to the 5Ј end of the mRNA, initiation factors eIF4A and eIF4B interact with the mRNA and disrupt secondary/ tertiary structure existing in the 5Ј-untranslated region (UTR) through the helicase activity of eIF4A (7)(8)(9). This facilitates the binding of the 40 S ribosomal subunit to the 5Ј end of the mRNA and allows for subsequent scanning by the 40 S subunit to the initiating AUG codon and correct placement of the Met-tRNA i . Although the general sequence of events is known, the molecular mechanism by which these steps take place is still poorly understood.
eIF4A is the prototypic member of the DEAD box family of proteins. These proteins are highly conserved from bacteria to humans and are involved in a wide range of cellular processes including translation, ribosome biogenesis, and mRNA splicing (for reviews, see Refs. 10 and 11). Since these proteins have been implicated in a variety of cellular functions that require unwinding of RNA secondary structures, it has been proposed that these proteins act as RNA helicases. This is further supported by the fact that all DEAD box proteins studied in vitro exhibit an RNA-dependent ATPase activity and share amino acid sequence similarity to DNA helicases (11)(12)(13). Whereas several DEAD box (and related DEAH and DExH box) proteins have been shown to unwind RNA duplex structures in vitro (8, 9, 14 -17), relatively little is known about the biochemical properties of these proteins or the molecular mechanism of their unwinding activity.
eIF4A has been previously characterized as an ATP-dependent RNA-binding protein and an RNA-dependent ATPase and is thought to unwind secondary structures in the 5Ј-UTR of mRNAs (7, 18 -21). Prior studies with eIF4A have shown that an additional initiation factor, eIF4B, is a strict requirement for RNA duplex unwinding activity of eIF4A in vitro (8,9,22,23). It was hypothesized, however, that eIF4A may be able to unwind RNA duplex structures independently of other initiation factors based on the following: (i) eIF4A displays inherent ATP-dependent RNA binding and RNA-dependent ATPase activities in the absence of other initiation factors; (ii) the sequence of eIF4A contains the common motifs associated with DEAD box (and related DEAH and DExH box) proteins, and in several cases it has been demonstrated that these proteins are capable of unwinding RNA duplexes in vitro (14 -17, 24); (iii) eIF4A alone has been shown to alter secondary and/or tertiary structure in reovirus mRNAs and also to have a limited ability in disrupting duplexes formed by annealing short DNA oligonucleotides to globin and reovirus mRNAs (25). Furthermore, previous studies of the helicase ac-tivity of eIF4A used concentrations of duplex RNA that were roughly 4 orders of magnitude below the reported K act 2 value needed to stimulate the ATPase activity of eIF4A (9,19,20,22,23), and thus failure to observe helicase activity may have been the result of the poor affinity of eIF4A for the RNA duplex substrate. In addition, the duplex employed in these studies had a relative stability (⌬G) of Ϫ25.7 kcal/mol, which may have been too stable a duplex for eIF4A to unwind independent of eIF4B or other initiation factors. Therefore, it was reasoned that eIF4A may possess an inherent RNA duplex unwinding activity in the absence of eIF4B or other initiation factors given either increased concentrations of duplex substrate or a duplex substrate of lesser stability.
To directly examine these possibilities, as well as to understand the mechanism of RNA duplex unwinding by DEAD box proteins in general, an investigation was undertaken to examine the biochemical and kinetic properties of the helicase activity of eIF4A by using a family of RNA duplexes of limited stability. Evidence is presented that, in contrast with previous studies, eIF4A can act in the absence of eIF4B as an ATP-dependent helicase. The biochemical and kinetic parameters of the RNA duplex unwinding activity of eIF4A are examined, and a mechanism is proposed by which eIF4A unwinds RNA duplexes. Finally, preliminary results are given concerning how this activity is affected by eIF4B, eIF4F, and eIF4H.

Materials
Reagents were purchased from the following suppliers: rabbit reticulocyte lysate from Green Hectares, Oregon, WI; ATP, ADP, ADPNP, and bovine serum albumin from Sigma; [␥- 32
DNA and RNA Oligonucleotides-DNA oligonucleotides were synthesized by the Case Western Reserve University Molecular Biology Core Laboratory, purified using an oligonucleotide purification cartridge and stored in double-distilled H 2 O. RNA oligonucleotides were synthesized by Cybersyn, purified by polyacrylamide gel electrophoresis, lyophilized, and resuspended in double-distilled H 2 O. Quantitation of each oligonucleotide was performed by UV spectroscopy, and a value of 33 g per 1 A 260 was used in determining concentration. Integrity and proper size of each oligonucleotide was assessed by 32 P end labeling each oligonucleotide (described below) and running on a denaturing (7 M urea) 20% polyacrylamide gel with known size standards.
Transcription of RNA-1-RNA-1, a 50-nucleotide RNA oligonucleotide (see Table I), is synthesized by in vitro transcription using T7 RNA polymerase. The template for transcription is composed of the following synthetic DNA oligonucleotides: 5Ј-GAATTTAATACGACTCACTATA-G-3Ј and 3Ј-CTTAAATTATGCTGAGTGATAT*CCCCTCT(TTTTG) 5 AT-CGTGGCATTTCGTGCG-5Ј. (* denotes the transcription start site.) Transcription reactions are performed using Ambion's Megashortscript TM transcription kit per the manufacturer's instructions. Integrity and proper size of the transcript was confirmed by running a sample on a denaturing (7 M urea) 20% polyacrylamide (19:1 bis:acrylamide) gel. The reaction was then made to be 50 l total and passed over a Bio-Rad P6 spin column to remove salts and unincorporated nucleotides. The recovered volume was phenol-extracted once and ether-extracted three times. Concentration of RNA-1 was measured by using Molecular Probes Ribogreen TM RNA quantitation assay kit, using rRNA of a known concentration to generate a standard curve. picomoles of RNA-10,  -11, -12, -13, -14, or -15 oligonucleotides (see Table I) and 10 pmol of [␥-32 P]ATP (specific activity 3000 Ci/mmol, 10 mCi/ml) were combined with 10 units of T4 polynucleotide kinase (10 l final volume), and reactions were performed as described in Ref. 28. Percent incorporation of the label into these oligonucleotides is 90 -95%, and specific activities of the oligonucleotides are on the order of 1 ϫ 10 6 cpm/pmol.

P End Labeling of Oligonucleotides-Forty
Helicase Substrates-Duplexes used in the helicase reactions are made by combining RNA-1 and 32 P-labeled RNA-10, -11, -12, -13, -14, or -15 oligonucleotide in a 1.25:1 ratio, respectively. This excess of RNA-1 is to ensure that a majority of the labeled oligonucleotide is hybridized to the RNA-1. The complementary strands are combined in 1ϫ hybridization buffer (1ϫ Tris-EDTA plus 100 mM KCl), and the concentration of duplex was 0.5 pmol/l. Samples were heated to 95°C for 5 min then slow cooled to 4°C over 90 min (0.1°C/5 s) using a programmable thermocycling instrument. Under these conditions, approximately 85-90% of the labeled oligonucleotide hybridizes to the RNA-1. The duplex was stored at Ϫ20°C and diluted in 1ϫ hybridization buffer to 0.02 pmol/l for use in the helicase assay.
Helicase Assay-Unwinding of duplex substrates was monitored by following displacement of the radiolabeled strand from the duplex. In general, 20-l reactions contained 20 mM HEPES, pH 7.5 (final pH 7.2), 70 mM KCl, 2 mM dithiothreitol, 1 mg/ml bovine serum albumin, and 1 mM magnesium acetate. The concentration of ATP was 1 mM; duplex concentration was 1.7-1.8 nM (34 -36 fmol), and eIF4A concentration was 0.4 M (8 pmol), unless stated otherwise. Concentrations of eIF4B (monomer), eIF4F, and eIF4H (when applicable) were 0.2 M. Time courses of reactions without added eIF4A were performed to quantitate the background unwinding, or thermal melting, 3 of each duplex in the absence of protein. All reactions were mixed in siliconized tubes to minimize the amount of material sticking to tube walls and kept on ice throughout the assembly process. Reactions were incubated at 35°C (unless otherwise stated) for the duration of the time course. In most circumstances, eIF4A was the last component added to the reaction, just before incubation at 35°C; however, equivalent results were obtained when either duplex or ATP was added as the last component. Reactions were terminated by adding 5 l of a solution containing 50% glycerol, 2% SDS, 20 mM EDTA, and 0.01% bromphenol blue and xylene cyanol dyes. The products of unwinding reactions were analyzed by separation of displaced strand (labeled RNA-10, -11, -12, -13, -14, or -15 oligonucleotide) from duplex species by electrophoresis on 12% native polyacrylamide gels (19:1 bis:acrylamide) for 1.5 h at 200 V at 4°C in 1ϫ Tris borate-EDTA buffer. Gels were pre-electrophoresed at 4°C for 20 -30 min. Gels were scanned directly using an Ambis radioanalytical scanner for 30 -90 min. Gels were then exposed to Kodak X-Omat AR film at Ϫ80°C for 30 -120 min using intensifying screens.
Quantitation of the Helicase Assay-Quantitation of cpm in duplex and displaced/single-stranded RNA bands was performed using the Ambis software. The degree of unwinding for each reaction was performed by measuring the percent of cpm in duplex and displaced/singlestranded RNA bands using the following formulas: % duplex ϭ cpm duplex/(cpm duplex ϩ cpm displaced strand) and % displaced strand ϭ cpm displaced strand/(cpm duplex ϩ cpm displaced strand). This method of quantitation was used to account for slightly different yields in total cpm due to variations in pipetting and gel loading steps. The total yield in cpm among different reactions did not vary by more than Ϯ10% of the average cpm value of all the reactions in the given experiment. Furthermore, all reagents and protein preparations used in the assay were judged to be free of RNase contamination. A single reaction without (Ϫ) eIF4A was incubated at 4°C during the course of each experiment, and the amount of duplex in this control reaction (typically 34 -36 fmol) was normalized to 100%, and the amount of (background) single-stranded RNA (typically 4 -6 fmol) was normalized to 0%. All other reactions were scaled to these values and then converted to femtomoles of duplex or single-stranded RNA species. Incubation of the minus eIF4A sample at 4°C indicates the amount of starting material (duplex RNA) present at the start of the reactions. Monitoring the disappearance of duplex RNA (or the appearance of single-stranded RNA) at 35°C versus the control reaction at 4°C allows for the measurement of both eIF4A-dependent unwinding and thermal melting over the time course. Furthermore, incubation at 4°C (t ϭ 0 and t ϭ 30 min) indicates that the amounts of duplex and single-stranded RNA species are preserved during the time course of the reaction.
Data Treatment-Time courses of unwinding RNA duplexes are the average of at least three separate experiments. All plots were fit using Prism software by GraphPad. Duplex stability (⌬G) values for RNA duplex substrates were calculated for 35°C using the nearest neighbor method of analysis as described by Turner et al. (29). The unwinding reaction may be approximated by the pseudo-first order rate equation V obs ϭ kЈ [duplex], where V obs is the observed rate of duplex unwinding (or the rate of single-stranded RNA formation) by eIF4A in fmol/min, kЈ is the pseudo-first order rate constant, and [duplex] is the concentration of duplex at time t. kЈ is a macroscopic pseudo-first order rate constant composed of terms including the concentration of eIF4A (which is constant), and microscopic constants reflecting eIF4A binding to ATP, eIF4A⅐ATP binding to the RNA duplex, the ATP hydrolysis rate constant, and the duplex unwinding rate constant. In the time course experiments, ATP is saturating ( Fig. 2A), and it is assumed that there is only one eIF4A molecule per unwinding event due to the poor affinity of eIF4A single-stranded RNA (20,21). Since V obs also equals d[ssRNA]/ dt (the appearance of single-stranded RNA with respect to time), then d[ssRNA]/dt ϭ kЈ [duplex], which when integrated and evaluated yields [ssRNA] t ϭ [duplex] 0 (1 Ϫ e ϪkЈt ) ϩ c, where [ssRNA] t ϭ the amount of single-stranded RNA formed at time t (i.e. amount of duplex unwound), [duplex] 0 ϭ the initial amount of duplex at t ϭ 0, kЈ ϭ pseudo-first order rate constant, and c ϭ amount of single-stranded RNA (background) at time 0. This equation describes a single-phase exponential curve, and kinetic data were fit to this equation. A single phase exponential equation describes the data obtained since [eIF4A] Ͼ Ͼ [duplex], and a decrease in the amount of duplex reflects a decrease in the rate of the reaction. Observed initial rates were calculated by linear regression using the linear portion (0 -2 min) of each plot. Initial rates were used to describe the relationship between unwinding and duplex stability because (i) the reaction is being performed under conditions of limiting substrate (duplex) and does not reflect steady-state kinetics and (ii) correction of the forward rate of unwinding due to the reverse (reannealing) reaction is not needed, as its rate is negligible (less than 5% of the forward unwinding rate), and the amount of single-stranded species which could reanneal is quite small in the initial phase (0 -2 min) of the unwinding reaction. The observed initial rates for the thermal melting of duplexes were calculated as above and subtracted from the observed initial rates of eIF4A-dependent duplex unwinding.

RESULTS
Design of the Helicase Substrates-As noted in the Introduction, it was hypothesized that eIF4A may display an ATP-dependent helicase activity in the absence of accessory initiation factors (eIF4B and eIF4F). To test this hypothesis, a family of RNA duplex substrates of limited stability was designed that contained an unstructured 5Ј single-stranded region of 31 nucleotides, followed by a duplex region of 10 -15 base pairs. The long strand, designated RNA-1, includes a short sequence to optimize transcription by T7 polymerase, followed by five A 4 C repeats that assume an unstructured conformation. This is followed by an arbitrary sequence of 18 nucleotides designed to minimize the formation of any intramolecular secondary struc-ture. Short oligonucleotides of 10 -15 nucleotides, designated RNA-10 to RNA-15, with complementary sequences to RNA-1 are then hybridized to RNA-1 to form duplexes of increasing length and stability. The characteristics of these substrates in terms of sequence, length, and stability are summarized in Table I.
The nearest neighbor method of Turner et al. (29) was used to calculate the relative stability (⌬G value) of duplex substrates. Although assay conditions used in this report differ from the conditions used by Turner et al. (29) to determine stability, it is emphasized that the objective of these experiments is to focus on the relative differences in stabilities among different duplexes (⌬⌬G values) and the rate of unwinding by eIF4A, rather than the determination of the absolute ⌬G value of each duplex. Thus, even though the calculated stabilities of the duplex substrates may not entirely reflect their correct absolute ⌬G values, they provide a reliable scale for correlating the helicase activity of eIF4A with relative increases/decreases in duplex stability.
It was reasoned that the initial rate of unwinding of RNA duplexes by eIF4A should decrease as the length and stability of the duplex region is increased, to a point where eIF4A is unable to unwind duplexes of a given stability. Whereas the helicase activity of eIF4A has been characterized as bi-directional (9), a 5Ј single-stranded region was used as this best represents the 5Ј-UTR of mRNAs undergoing translation initiation. Previous studies have estimated an RNA-binding site size of 15-20 nucleotides for eIF4A, which is also required for maximal stimulation of the ATPase activity of eIF4A (19,30); thus a single-stranded region of 31 nucleotides ensures a sufficient binding site size for eIF4A.
eIF4A Acts Independently of Other Initiation Factors as an ATP-dependent RNA Helicase-Preliminary results showed that under conditions used in the helicase assay (see "Methods"), eIF4A was able to unwind the RNA-1/RNA-10 duplex in an ATP-dependent fashion in the absence of eIF4B. Typical raw experimental data of the unwinding reaction using the RNA-1/RNA-10 duplex is shown in Fig. 1A. It is apparent that the labeled RNA-10 oligonucleotide is released (unwound) from the RNA-1 strand in the presence of eIF4A and ATP (Fig. 1A, lanes labeled ϩ4A, ϩATP) but not in the presence of eIF4A when ATP was omitted (Fig. 1A, lanes labeled ϩ4A, ϪATP). In contrast, the RNA-1/RNA-15 duplex was barely unwound by eIF4A under these same conditions (Fig. 1B, lanes labeled ϩ4A,  ϩATP). This difference is addressed below and under "Discussion." Unwinding of the RNA-1/RNA-10 substrate was not seen in the presence of the non-hydrolyzable ATP analog, ADPNP (data not shown). This unwinding activity is due to eIF4A and not a contaminant in the reticulocyte lysate preparation, as recombinant human eIF4AI purified from Escherichia coli displays the same activity (data not shown).
Once eIF4A was shown to possess an ATP-dependent heli-  case activity, the biochemical properties of this activity were investigated using the RNA-1/RNA-11 duplex. This duplex was chosen as it displays the highest ratio between the initial rates of eIF4A-dependent unwinding and background thermal melting and thus allows for the most accurate measurement of differences in the unwinding activity of eIF4A as reaction conditions are changed. Addition of ATP⅐Mg 2ϩ to the reaction showed saturation at 1-2 mM ( Fig. 2A). A K m value for ATP⅐Mg 2ϩ in the helicase reaction is calculated to be 440 M, which is in good agreement with the K m value of 330 M for ATP in the RNA-dependent ATPase reaction (21). It is stressed, however, that this is an apparent K m , as duplex substrate under these conditions is subsaturating. Titration of Mg 2ϩ (using 1 mM ATP) into the reaction shows optimal helicase activity at 1 mM or when Mg 2ϩ is equimolar to ATP (Fig. 2B). This suggests that excess Mg 2ϩ inhibits the helicase activity of eIF4A, probably due to enhanced stabilization of the duplex by free Mg 2ϩ . The helicase activity was optimal under a pH range of 6.4 -7.0 (Fig. 2C), and the optimal concentration of monovalent cation (KCl) was 50 mM (Fig. 2D). The helicase activity of eIF4A was inhibited by ADP, as a 50% inhibition of the initial rate of unwinding was seen with a 1/10 molar ratio (100 M ADP) of ADP to ATP (data not shown).
Effects of protein and substrate concentration on the helicase activity of eIF4A were investigated using the RNA-1/RNA-12 substrate, as this allowed accurate measurement of both increases and decreases in the initial rate of the reaction as a function of eIF4A or duplex concentration. Addition of eIF4A into the reaction displayed a linear response in initial rate of unwinding over the range tested (0 -1.6 M, 0 -32 pmol, data not shown). Furthermore, increasing the concentration of protein also increased the maximal amount of duplex that is able to be unwound (data not shown and Fig. 6). Titration of duplex into the reaction displayed a linear response in initial rate over the range tested (0 -1.8 nM, 0 -36 fmol, data not shown).
Kinetic Characterization of the Helicase Activity of eIF4A-Whereas many DEAD box proteins have been shown to display helicase activity in vitro, the molecular mechanism of RNA duplex unwinding is poorly understood. In order to begin elucidating this mechanism, simple kinetic experiments using the duplexes listed in Table I were performed. It was reasoned that longer, more stable duplex structures should take longer to unwind, and thus the initial rate of unwinding by eIF4A of these substrates should decrease as the length and stability is increased. Therefore, time course experiments using RNA-1/ RNA-10 through RNA-1/RNA-15 were performed. Fig. 1A demonstrates the rapid unwinding of the RNA-1/RNA-10 duplex (lanes labeled ϩ4A, ϩATP), whereas in contrast, Fig. 1B shows that the RNA-1/RNA-15 duplex is barely unwound by eIF4A in the presence of ATP (Fig. 1B, lanes labeled ϩ4A, ϩATP). Raw data time points from each duplex substrate were quantitated and plotted as the amount of duplex unwound (fmol) versus time. The unwinding reaction may be described by a pseudofirst order kinetic model and the data fit to a single-phase  3). Observed initial rates of the unwinding reaction were calculated using the linear portion (0 -2 min) of each time course (Fig. 3). Initial rates of the thermal melting (unwinding of duplex in the absence of eIF4A) of each duplex were also measured (data not shown) and subtracted from eIF4A-dependent initial rates of unwinding. It is clear from these plots that the initial rate of eIF4A-dependent unwinding decreases as the stability of the duplex substrate increases. Furthermore, the maximum amount of substrate unwound during the time course also decreased as stability of the duplex increased. The values of the initial rates of thermal melting and eIF4A-dependent unwinding and maximum percentage of duplex unwound for each substrate are summarized in Table II.
In order to establish a relationship between the initial rate of unwinding and duplex stability, the ln of the initial rates of both eIF4A-dependent unwinding and thermal melting were plotted versus the stability of the duplex (⌬G in kcal/mol, Fig.  4). It can be seen in Fig. 4 that a direct relationship exists between the rate of eIF4A-dependent unwinding and duplex stability. It was expected that plotting the ln(initial rate of eIF4A-dependent unwinding) versus ⌬G would yield a straight line with a negative slope due to the following: the thermodynamics of duplex unwinding can be given by ⌬G ϭ ϪRT ln K eq ϭ ϪRT ln k 1 /k 2 where k 1 is the rate constant for unwinding/ melting and k 2 is the rate constant for hybridization (31). 4 Since k 2 changes little under the experimental conditions of the assay described here, ⌬G ϭ ϪRT ln k 1 ϩ C, where C ϭ RT ln k 2 . Since the initial rate is directly proportional to k 1 , then ln(initial rate) versus ⌬G should also yield a straight line with a negative slope, and this is what is observed in Fig. 4. Furthermore, the slopes of eIF4A-dependent unwinding (10 -15 base pairs) and thermal melting (10 -12 base pairs) are approximately equal, suggesting that the rate of eIF4A-dependent unwinding is related to the rate of thermal melting. First it should be noted that thermal melting of the 13, 14, and 15 base pair duplexes is essentially undetectable using these assay conditions. This is due to the fact that for the 13, 14, and 15 base pair duplexes, equilibrium between duplex species and single-stranded species has been achieved at t Ϸ0 and that the T m values of these duplexes are greater than the reaction temperature of 35°C (41.5, 43.1, and 46.3°C, respectively). Second, extrapolation for predicted ln(initial rate of thermal melting) for these more stable duplexes yields values that are too small to be distinguished from background. Thus, the value of the rate of thermal melting for the 12 base pair duplex represents the lower limit of measurable unwinding/melting by this experimental method. The observation that the slopes describing the processes of eIF4A-dependent unwinding and thermal melting are parallel provides insight into a model for the unwinding activity of eIF4A (see "Discussion").
Whereas the result that the initial rate of unwinding by eIF4A decreases with increasing duplex stability was expected, the fact that maximum amount of duplex unwound also decreased was not. Under conditions used ([eIF4A] Ͼ Ͼ [duplex]), it was expected that the unwinding of all duplexes should go to completion (90 -100% unwound) since the initial rates of unwinding by eIF4A for each duplex were greater than those for thermal melting. One would predict that it should simply take longer for eIF4A to completely unwind the more stable duplex substrates, which have a slower initial rate of unwinding. In order to investigate this observation, several further experiments were performed. Since the kinetic unwinding curves resemble approaches to equilibrium (Fig. 3 and Table II), it was postulated that the curves simply demonstrated an equilibrium between eIF4A unwinding and subsequent reannealing of displaced radiolabeled strands to the RNA-1 strand. However, the rates of reannealing were measured independently and found to be less than 5% of the initial rate of eIF4A-dependent unwinding for each duplex tested (data not shown). In order to clarify this, experiments were performed that included a 10fold excess of a DNA oligonucleotide complementary to the radiolabeled RNA strand. The DNA oligonucleotide would then "capture" the labeled RNA strand once it was displaced from the RNA-1 by eIF4A. This would prevent reannealing and, subsequently, "pull" the reaction to completion. Control experiments in which the RNA-1/RNA-12 duplex was fully melted by heating to 95°C showed that the excess complementary DNA oligonucleotide captured the RNA-12 strand efficiently and completely, with 4 James G. Wetmur, personal communication.  almost no reannealing to the RNA-1 strand (data not shown). However, inclusion of this capture strand in the helicase reaction using the RNA-1/RNA-12 duplex did not increase the maximum amount of duplex that eIF4A was able to unwind (Ϸ50%, data not shown) and shows that reannealing is not responsible for the unwinding reactions not reaching completion. The possibility was then considered that eIF4A may becoming inactive over the time course of the experiment. To test this hypothesis, eIF4A was incubated for 30 min under usual reaction conditions without duplex substrate. When duplex was added, it unwound with the same initial rate and reached the same plateau value as determined without preincubation of eIF4A (data not shown). Furthermore, a duplex addition experiment was performed in which eIF4A was allowed to unwind the RNA-1/RNA-15 duplex for 15 min, and then the RNA-1/ RNA-11 duplex was added (Fig. 5). The added RNA-1/RNA-11 duplex was unwound to the same maximal value (Ϸ90%) and with a similar initial rate to what had been determined for this duplex alone (Fig. 5). Again, this indicated that eIF4A was not becoming inactive over the time course of the experiment.
It was then postulated that the unwinding reaction may be approaching equilibrium due to a second process in which eIF4A is "stalled" in the process of unwinding and that a build-up of these stalled, non-productive complexes prevented the helicase reaction from reaching completion. Furthermore, it was reasoned that the amount of these putative non-productive complexes formed should be directly proportional to the length or stability of the duplex. An attempt was made to isolate these non-productive complexes by nitrocellulose filter binding assays and gel shift experiments. In the filter binding assays, relatively little eIF4A-dependent retention of duplexes (5% or less) was observed, and there was no detectable difference in retention among the less stable duplexes, more stable duplexes, or even single-stranded RNA (RNA-15) (data not shown). Gel shift experiments were performed in which no SDS or EDTA was used in the quench solution, and gels were run under reaction buffer conditions. Results yielded no evidence of a shifted band corresponding to the postulated stalled complex containing eIF4A (data not shown). This is in agreement with previous results obtained by Sonenberg and co-workers (32), who were also unable to detect a complex of eIF4A and duplex RNA in similar experiments. This hypothesis was further tested by the addition of an equal amount (8 pmol) of eIF4A to the reaction after the maximal value of unwinding the RNA-1/ RNA-12 duplex by eIF4A had been reached (15 min time point, Fig. 6). It can be seen in Fig. 6 that addition of 8 pmol eIF4A to obtain 0.8 M eIF4A total gave the same final maximal amount of duplex unwound (Ϸ70%) as starting the reaction with 0.8 M eIF4A. If the entire amount of unwound duplex was sequestered in non-productive complexes, then addition of more eIF4A to the reaction should have had little or no effect on the maximal amount of duplex unwound. This result suggests that the duplex is not sequestered in non-productive complexes and that this is not the reason for the reactions not going to completion. The results in Fig. 5 and Fig. 6 also implicitly demonstrate that the reaction is not falling short of completion due to a drop in the concentration of ATP from ATP hydrolysis or inhibition of the reaction due to a concomitant increase in ADP concentration, as both the addition of eIF4A or duplex increase the final maximal amount of substrate unwound.
Activation Energy of the Helicase Reaction-As a means to examine the thermodynamic nature of the helicase activity of eIF4A, kinetic measurements were taken at different temperatures. Time courses of eIF4A-dependent unwinding using the RNA-1/RNA-11 substrate at 35, 30, and 25°C were performed (data not shown). Pseudo-first order initial rate constants were calculated as described under "Methods" and plotted versus 1/T (in Kelvin) (data not shown). The slope from this Arrhenius plot (r 2 value of 0.998) yields a value of 24.7 kcal/mol as the activation energy for the unwinding reaction.
Effects of eIF4B, eIF4F, and eIF4H on Helicase Activity of eIF4A-Previous investigation of the helicase activity of eIF4A, eIF4B, and eIF4F showed that eIF4A or eIF4F (which contains eIF4A as a subunit) were able to function as ATP-dependent RNA helicases only in the presence of eIF4B (eIF4B has no helicase or ATPase activity by itself) (8,9,22,23). However, since the above results have demonstrated that eIF4A is capable of unwinding RNA duplexes in the absence of eIF4B, it was necessary to reinvestigate the effect of eIF4B on the activity of eIF4A and to ascertain if eIF4F was also able to function as an RNA helicase in the absence of eIF4B. These experiments were extended to study the effect eIF4H on the helicase activity of eIF4A, since eIF4H has been shown to stimulate the ATPase activity of eIF4A in a manner similar to eIF4B (4). Furthermore, eIF4H contains an RNA recognition motif (RRM) that is 45% identical to the RRM in eIF4B that is required for maximal FIG. 4. ln(initial rate of duplex unwinding or melting) versus duplex stability. The graph illustrates that a linear relationship exists between the ln(initial rate of eIF4A-dependent duplex unwinding) and duplex stability (q) and a linear relationship exists between the ln(initial rate thermal duplex melting) and duplex stability (E). See text for details of this relationship. Note that the slopes of eIF4A-dependent unwinding (10 -15 base pairs) and thermal melting (10 -12 base pairs) are approximately equal, suggesting that the rate eIF4A-dependent unwinding is related to the rate of thermal melting. stimulation of the helicase activity of eIF4A (4,33).
To this end, the helicase assay was performed with 0.2 M eIF4B (monomer), eIF4F, or eIF4H, both in the presence and absence of ATP. Then the assay was performed with the addition of 0.4 M eIF4A to each factor in the presence and absence of ATP. The RNA-1/RNA-15 duplex was used as a substrate since it was barely unwound by eIF4A alone and thus allowed measurement of any enhancement of the helicase activity of eIF4A by these other factors. Fig. 7A shows the raw data obtained when eIF4A, eIF4B, eIF4F, or eIF4H are incubated with the RNA-1/RNA-15 duplex in the presence of ATP (lanes 3-6, respectively), and when eIF4A is added to eIF4B, eIF4F, or eIF4H in the presence of ATP (lanes 8 -10, respectively). Fig.  7B shows the quantitation of the unwinding reactions for these combinations tested. As seen before, eIF4A alone (lane 3) shows a small amount of unwinding activity toward the RNA-1/ RNA-15 duplex (Ϸ5.6 fmol, 16%). eIF4F alone (lane 5) shows approximately the same magnitude (Ϸ7.0 fmol, 21%) of unwinding as free eIF4A, indicating that it is able to function as a helicase in the absence of eIF4B. When eIF4F was added to eIF4A (lane 9), a simple additive effect (Ϸ13.2 fmol, 40.7%) was seen between these two factors. Both eIF4B and eIF4H (lanes 4 and 6, respectively) showed minimal background unwinding of this duplex in the presence of ATP (Ϸ2.9 fmol, 8.4%, 3.2 fmol, 9.2%, respectively). When eIF4B was added to eIF4A (lane 8), a 3-fold stimulation in duplex unwinding (Ϸ16.9 fmol, 49.5%) was observed. eIF4H (lane 10) was also able to enhance the helicase activity of eIF4A, showing a 4.5-fold stimulation (Ϸ23.7 fmol, 69.2%) of duplex unwinding. These results indicate that (i) eIF4A, as a subunit of eIF4F, is also capable of unwinding RNA duplexes in the absence of eIF4B and (ii) either eIF4B or eIF4H is capable of enhancing the helicase activity of eIF4A. Furthermore, preliminary results indicate that not only is the maximum amount of duplex unwound increased by the addition of these factors, but in the case of eIF4B and eIF4H, the initial rate of unwinding is increased 5-10-fold (data not shown).

DISCUSSION
The results presented here demonstrate that eIF4A is able to function as an ATP-dependent RNA helicase independently of other initiation factors. This was accomplished using eIF4A purified from rabbit reticulocyte lysates (greater than 95% homogenous) and a family of RNA duplex substrates that contained a 5Ј single-stranded unstructured region of 31 nucleotides followed by a duplex region of increasing length (10 -15 base pairs) and stability (⌬G ϭ Ϫ16.1 to Ϫ24.7 kcal/mol, see Table I). These results are in contrast with previous studies showing that either eIF4A or eIF4F required eIF4B for unwinding activity (8,9,22,23). Sonenberg and co-workers (9, 22, 23) employed a single RNA strand of 40 bases that contained a 10-nucleotide region of alternating G and C residues, making this RNA self-complementary. This 10-base pair G:C duplex has a calculated stability (⌬G) of Ϫ25.7 kcal/mol. It was postulated that the failure to observe helicase activity by eIF4A alone could have been due to the poor affinity of eIF4A for the RNA duplex substrate (K d Ϸ 125 M) (21), or that previous studies may have used an RNA duplex with too great a stability for eIF4A alone to unwind.
Therefore, experiments were initially designed to have approximately the same reaction conditions as those used previously, except that an RNA duplex substrate of lesser stability (RNA-1/RNA-10, ⌬G ϭ Ϫ15.7 kcal/mol) was used. As is evident in Figs. 1 and 3 and Table II, the RNA-1/RNA-10 duplex is unwound rapidly and to completion, whereas increasing the stability of the duplex lowers both the initial rate of eIF4A-dependent unwinding and the maximum amount of duplex able to be unwound. From these results, it is concluded that the ability to demonstrate that eIF4A can act as an RNA helicase in the absence of eIF4B is due to the difference in duplex stability. It should be noted that the stability of the duplex used by Sonenberg and co-workers (9, 22, 23) (⌬G ϭ Ϫ25.7 kcal/mol) is above the upper limit of our most stable duplex (RNA-1/RNA-15, ⌬G ϭ Ϫ24.7 kcal/mol), which was barely unwound by eIF4A alone under the reaction conditions used in this study. Further comparison of the data in this report with that described previously shows that the rate of unwinding RNA duplexes of eIF4A is dependent on duplex stability rather than simply length of the duplex, as eIF4A was unable to unwind an RNA duplex consisting of 10 G:C base pairs (⌬G ϭ Ϫ25.7 kcal/mol) but is able to efficiently unwind 10 -12 base pair duplexes consisting of both A:U and G:C base pairs (⌬G ϭ Ϫ16.1 to Ϫ21.4 kcal/mol, Table II). In addition, it was demonstrated that eIF4A is most active as a helicase when the concentration of Mg 2ϩ is approximately 1 mM (in the presence of 1 mM ATP, Fig.  2B), and lower concentrations of Mg 2ϩ decrease the helicase activity of eIF4A (Fig. 2B). Since the previous studies used only 0.5 mM Mg 2ϩ (in the presence of 1 mM ATP), this probably also contributes slightly to the differences in observations between this investigation and those performed previously.
These results suggest that, by itself, eIF4A is an inefficient, non-processive helicase. This finding is consistent with the fact that eIF4A alone is a relatively slow ATPase, maximally hydrolyzing ATP on the order of 3 per min (21). In addition, studies by Lorsch and Herschlag (21,34) demonstrated that the association and dissociation rates of RNA and ATP for eIF4A are very fast compared with the ATP hydrolysis event, which evokes a conformational change in eIF4A that likely allows for RNA duplex unwinding. Thus, under reaction conditions used in this study (0.4 M eIF4A, 1.7-1.8 nM RNA duplex), the ATPase/helicase activity of eIF4A is most likely a single turnover event for any one eIF4A⅐RNA substrate complex (calculations indicate that only 1-2% of the RNA duplex would be bound to protein at any given time). These results strongly suggest that eIF4A is only able to unwind a minimal number of base pairs during the slow ATP hydrolysis step before dissociating from the duplex substrate. The non-processive nature of eIF4A under these conditions is further supported by the fact that addition of an excess of poly(A) after initiation of the reaction abrogates any further unwinding of substrate (data not shown).
The results from the experiments reported in this article combined with the results of Lorsch and Herschlag (21, 34) suggest a model for the helicase activity of eIF4A that has eIF4A unwinding a limited number of base pairs, which may vary depending on the sequence's contribution to stability. This model is summarized schematically in Fig. 8 and depicts two kinetic steps of unwinding; the first step is the ATP-dependent minimal duplex unwinding by eIF4A, and the second step consists of the remaining base pairs of the duplex unwinding at the rate of thermal melting (see below). Thus, under the reaction conditions used, eIF4A acts as a helicase by unwinding only a few base pairs of the duplex, and this substrate subsequently behaves as a shorter, less stable duplex in a thermodynamic sense. This minimal unwinding is sufficient to completely destabilize the RNA-1/RNA-10 and RNA-1/RNA-11 duplexes, and unwinding of these substrates is rapid and complete (Fig. 8, path c). However, with longer, more stable duplexes (12-15 base pairs), this minimal unwinding is insufficient to completely destabilize these substrates (Fig. 8, path d). Thus, as soon as eIF4A dissociates, only a fraction of these partially unwound duplexes continue to unwind (thermally), whereas the other fraction of partially unwound duplexes reanneal. This is supported by the fact that the initial rate of unwinding decreases for more stable substrates. Although these results do not allow the explicit calculation of a specific number of base pairs unwound per binding event, it is hypothesized that it is probably small, on the order of 4 -5 base pairs. This would be in agreement with the kinetic data obtained for the step size of unwinding by the UvrD helicase (35) and is also supported by the structural data obtained for the hepatitis C NS3 helicase (the sequence of eIF4A is 44% identical and 63% similar to the NS3 helicase sequence in conserved motifs), which has a binding cavity occupied by 5 nucleotides (36).
If the helicase activity of eIF4A makes a longer (more stable) duplex behave like a shorter (less stable) duplex with respect to the rate of thermal melting, then this model would predict that the slope of the ln(initial rate of thermal melting) versus duplex stability would be equal to the slope of the ln(initial rate of eIF4A-dependent unwinding) versus duplex stability. For example, the 14-base pair duplex unwound by eIF4A would subsequently behave thermally as the 10-base pair duplex. This relationship is what is observed in Fig. 4; the slopes characterizing these two processes (eIF4A-dependent unwinding and thermal melting) are the same. This fact supports the model proposed above in which eIF4A unwinds a minimal number of base pairs in a non-processive fashion and then the duplex proceeds to thermally melt at a rate dependent on the number of base pairs remaining. The actual "apparent" change in duplex length would be dependent on eIF4A concentration, which yields increased rates of strand separation with an increase in eIF4A concentration. For the eIF4A concentration used in these studies (0.4 M), it would appear that, on average, eIF4A has unwound about 4 -5 base pairs (⌬G Ϸ 8 kcal/mol). Furthermore, increasing the concentration of eIF4A would shift the entire curve of the ln(initial rate of eIF4A-dependent unwinding) versus duplex stability up due to the linear increase in initial rate, while the slope would remain unchanged.
Models of translation initiation hold that eIF4A, eIF4B, eIF4F, and eIF4H all act at the "mRNA activation" step in protein synthesis. The results in Fig. 7 demonstrate that eIF4F is also able to function as a helicase in the absence of eIF4B and suggest that eIF4F is approximately twice as active on a molar basis as eIF4A, since the molar amount of eIF4F (0.2 M) was half that of eIF4A alone (0.4 M). This finding is in corroboration with previous data showing that eIF4F is more active in poly(A) binding and ATPase activity than eIF4A on a molar basis (19,37); however, it cannot be ruled out that this difference in helicase activity may simply reflect a difference in the percentage of active molecules in eIF4A and eIF4F preparations. Including eIF4A with eIF4F in the helicase reaction showed a simple additive effect between these two proteins (Fig. 7). The results of adding eIF4B or eIF4H to eIF4A indicate that either of these proteins are able to stimulate the helicase activity of eIF4A, allowing a faster initial rate of unwinding and a greater maximal value of duplex unwound (data not shown and Fig. 7). Previous studies have shown that eIF4B's stimulation of the ATPase activity of eIF4A is primarily through increasing the affinity (Ϸ250-fold) of eIF4A for singlestranded RNA, whereas a small (25%) increase was seen in the maximum rate of the reaction (20). Thus, by analogy, we suggest that the ability of eIF4A to rapidly and efficiently unwind more stable duplex structures (RNA-1/RNA-15) in the presence of eIF4B is also in part due to eIF4B increasing the affinity of eIF4A for the substrate. Increased affinity for RNA would also explain why eIF4A is approximately twice as active as a helicase when part of the eIF4F complex, as the eIF4G subunit has RRM motifs that would allow eIF4F to be more tightly bound to the duplex than eIF4A alone. Thus, while it was demonstrated that eIF4B and eIF4H stimulate the helicase activity of eIF4A, further experimentation will need to be performed in order to ascertain if this is due to an increased affinity for substrate, a stimulation of the rate of unwinding, conferring processivity to the activity of eIF4A, or a combination of these possibilities.
The observation that the maximal amount of duplex unwound decreases with increasing duplex stability was unexpected. Experimentation has demonstrated that this decrease in the maximal value unwound was not due to the following: (i) strands displaced by eIF4A subsequently reannealing, (ii) eIF4A becoming inactive over the time course, (iii) the rate of eIF4A-dependent unwinding slowing/stopping due to a decrease in ATP concentration or a concomitant increase in ADP concentration over the time course of the reaction, and (iv) the more stable duplexes were being sequestered into non-productive complexes (see "Results"). It was observed, however, that the maximum amount of duplex unwound may be correlated to the initial rate of unwinding. This is evident in the unwinding profiles of duplexes with different stabilities (Fig. 3), and furthermore, increases in the initial rate of the reaction due to increases in eIF4A concentration also lead to an increase in the maximal value of duplex unwound (Fig. 6, compare 0.4 M eIF4A to 0.8 M eIF4A). This effect is also seen when the initial rate decreases due to a lowering of reaction temperature, and subsequently, a decrease in the maximal value of unwound duplex is observed (data not shown). Thus, the maximum amount of duplex unwound is positively correlated to the initial rate of unwinding; however, the reason for this relationship will require further experimentation.
In conclusion, this report demonstrates that eIF4A is able to function in the absence of eIF4B as an RNA helicase. The initial rate of this unwinding activity and the maximal amount of duplex unwound is dependent upon the stability of the duplex. Our model suggests a single ATP hydrolysis event allows eIF4A to unwind a limited number of duplex base pairs. eIF4F is also able to function as an RNA helicase in the absence of eIF4B. eIF4B and eIF4H are capable of enhancing the helicase activity of eIF4A, and experimentation is currently being performed to address how these additional factors affect the kinetic parameters of this helicase activity. Overall, this study provides biochemical and kinetic characterization of the unwinding activity of eIF4A that may offer a basic framework for understanding the helicase mechanism of DEAD box proteins in general.