Gao et al.
1
2
3
4 Coupling between the DEAD-box RNA helicases
5 Ded1p and eIF4A
6
7
8
9
10 Zhaofeng Gao1, Andrea A. Putnam 1, Heath Bowers 1, Ulf-Peter Guenther1, Xuan Ye 1,
11 Audrey Kindsfather2, Angela K.Hilliker2 & Eckhard Jankowsky1
12
13
14
15 1 Center for RNA Molecular Biology & Department of Biochemistry
16 School of Medicine, Case Western Reserve University, Cleveland, OH
17 2 Department of Biology, University of Richmond, Richmond, VA
1
Gao et al.
18 Abstract
19 Eukaryotic translation initiation involves two conserved DEAD-box RNA helicases, eIF4A and
20 Ded1p. Here we show that S. cerevisiae eIF4A and Ded1p directly interact with each other and
21 simultaneously with the scaffolding protein eIF4G. We delineate a comprehensive
22 thermodynamic framework for the interactions between Ded1p, eIF4A, eIF4G, RNA and ATP,
23 which indicates that eIF4A, with and without eIF4G, acts as modulator for activity and substrate
24 preferences of Ded1p, which is the RNA remodeling unit in all complexes. Our results reveal
25 and characterize an unexpected interdependence between the two RNA helicases and eIF4G,
26 and suggest that Ded1p is an integral part of eIF4F, the complex comprising eIF4G, eIF4A, and
27 eIF4E.
2
Gao et al.
28 Introduction
29 Translation initiation in eukaryotes involves at least 12 distinct protein factors, including
30 two highly conserved DEAD-box RNA helicases, eIF4A and Ded1p (DDX3 in human) (Aitken
31 and Lorsch, 2012; Hinnebusch, 2014; Parsyan et al., 2011). DEAD-box RNA helicases,
32 whose name is inspired by a characteristic sequence motif (D-E-A-D, in single letter code), are
33 enzymes that utilize ATP to bind and remodel RNA and RNA-protein complexes (Linder and
34 Jankowsky, 2011). While S. cerevisiae eIF4A (~105 copies per cell) is roughly five times more
35 abundant in the cell than Ded1p (Firczuk et al., 2013), both helicases are essential for viability
36 (Linder and Slonimski, 1989; Giaever et al., 2002).
37 eIF4A unwinds RNA duplexes in vitro (Merrick, 2015). In the cell, the helicase interacts
38 with the large scaffolding protein eIF4G that also binds the cap-binding protein eIF4E
39 (Hinnebusch,2014; Parsyan et al., 2011). The complex containing eIF4G, eIF4E and eIF4A is
40 called eIF4F (Merrick, 2015). eIF4A binds to a MIF4G domain of eIF4G (Schütz et al., 2008),
41 and this interaction stimulates RNA- and ATP-binding, and ATP-dependent RNA unwinding
42 activity of eIF4A in vitro (Hilbert et al., 2011; Rajagopal et al., 2012). In translation initiation,
43 eIF4A is thought to be critical for steps during and subsequent to recruitment of the 40S
44 ribosome subunit to the mRNA (Parsyan et al., 2011). In S. cerevisiae, eIF4A appears to
45 promote at least one step necessary for the translation of most or all mRNAs (Sen et al., 2015).
46 Ded1p unwinds RNA duplexes in vitro as a trimer, whose formation requires the C-
47 terminus (Putnam et al., 2015). Ded1p also facilitates strand annealing, RNA structure
48 conversions, and the remodeling of RNA-protein complexes (Bowers et al., 2006; Iost et
49 al.,1999; Yang et al., 2007b; Yang and Jankowsky, 2005). Both, Ded1p and human DDX3X
50 interact with the cap binding protein eIF4E (Senissar et al., 2014; Shih et al., 2008), and with
51 eIF4G (Hilliker et al., 2011; Soto-Rifo et al., 2012). Ded1p uses its C-terminus to bind to the C
52 terminus of eIF4G, a region distinct from the eIF4A binding site in eIF4G (Hilliker et al., 2011).
3
Gao et al.
53 The interaction with eIF4G interferes with Ded1p oligomerization and thus inhibits RNA
54 unwinding by the helicase (Putnam et al., 2015). In translation initiation, Ded1p and DDX3X
55 have been implicated in mRNA recognition by eIF4F, recruitment of the 40S ribosome subunit to
56 the mRNA, the scanning process, start codon selection and joining of the two ribosomal
57 subunits (Sharma and Jankowsky, 2014). Misregulation of DDX3X expression and mutations
58 in the helicase core of DDX3X are associated with tumors, including medulloblastoma, lung-
59 and colorectal cancer and T-cell lymphoma (Bol et al., 2015; Heerma van Voss et al., 2015;
60 Jianget al., 2015; Pugh et al., 2012). DDX3 is also targeted by multiple pathogenic viruses
61 including HIV, HCV, other flaviviridae, poxviridae, and HBV (Valiente-Echeverría et al., 2015).
62 Previous data indicate genetic interactions between eIF4A and Ded1p (de la Cruz et al.,
63 1997; Senissar et al., 2014), and a recent, transcriptome-wide study in yeast shows
64 overlapping, but not identical functions of eIF4A and Ded1p in translation initiation (Sen et al.,
65 2015). While these data suggest a possible link between the two helicases, Ded1p and eIF4A
66 have been generally viewed as separately functioning factors. Despite the central biological
67 roles of the two enzymes, their level of conservation, and their implication in numerous
68 diseases, it is unknown whether and how eIF4A and Ded1p are linked biochemically and
69 physically, and if their functions are coordinated. The answer to these questions is of
70 considerable importance for understanding eukaryotic translation initiation on the molecular
71 level. However, experimentally addressing the link between the two enzymes presents
72 significant challenges. First, as RNA helicases, eIF4A and Ded1p share basic biochemical
73 properties including RNA binding, unwinding, ATP binding and ATP turnover. Therefore,
74 enzymatic readouts are similar and potentially difficult deconvolute. Second, both helicases
75 interact with eIF4G, which also binds RNA. The multitude of possible individual interactions
76 significantly complicates the analyses of coupling between the proteins. Third, interactions
77 between the proteins are likely transient or dynamic, given that both, Ded1p and eIF4A bind and
4
Gao et al.
78 hydrolyze ATP, which in turn modulates RNA binding by the helicases (Linder and Jankowsky,
79 2011). Functional analysis of dynamic multi-component assemblies is not well developed, even
80 though such RNA-protein complexes are numerous in biology.
81 Here, we have confronted the experimental challenges associated with links between
82 Ded1p, eIF4A, eIF4G, ATP and RNA. Using quantitative biochemical and enzymological
83 approaches, we show that Ded1p and eIF4A directly interact with each other and that both
84 helicases simultaneously interact with eIF4G. We devise an unbiased, comprehensive
85 thermodynamic framework for the interactions between Ded1p, eIF4A, eIF4G, RNA and ATP.
86 This framework indicates that eIF4A, with and without eIF4G, acts as modulator for activity and
87 substrate preferences of Ded1p, and that Ded1p functions as the RNA remodeling unit in
88 complexes with eIF4A. Our observations reveal an unexpected interdependence between both
89 RNA helicases and eIF4G, and suggest that Ded1p is an integral part of the eIF4F complex.
5
Gao et al.
90 RESULTS
91
92 Direct interactions between Ded1p, eIF4A and eIF4G.
93 We identified eIF4A in a screen for genes that suppress a cold-sensitive growth defect
94 conferred by DED1 mutations (ded1-tam, Hilliker et al., 2011). Overexpression of eIF4A
95 partially restored growth at non-permissive temperatures (Figure 1A), suggesting that eIF4A
96 can promote Ded1p's function, but not completely substitute for its tasks. This observation is
97 consistent with the notion that Ded1p and eIF4A have overlapping, but not identical functions in
98 translation initiation (Sen et al., 2015).
99 To further examine the connection between Ded1p and eIF4A in the cell, we tested
100 whether the eIF4A inhibitor hippuristanol exacerbated the temperature-sensitive growth defect
101 associated with the ded1-95 strain (Burckin et al., 2005). Ded1-95 bears a T408I mutation,
102 which impairs RNA binding by Ded1p in vitro (Figure 1- figure supplement 1). Hippuristanol
103 inhibits the activities of eIF4A, but not those of Ded1p (Bordeleau et al., 2006). Here,
104 Hippuristanol diminished growth of the ded1-95 strain, compared to the WT ded1 strain (Figure
105 1B), indicating that targeted inhibition of eIF4A can impact a growth defect associated with a
106 mutation in DED1. Collectively, these observations expand the scope of genetic interactions
107 between the two DEAD-box helicases. However, the data did not reveal the physical nature of
108 the link between Ded1p and eIF4A.
109 To probe physical connections between the two helicases, we performed pull-down
110 experiments with recombinant, purified proteins (Figure 1C,D). Previous data had shown that
111 both helicases bind to distinct regions of the scaffolding protein eIF4G (Hilliker et al., 2011;
112 Merrick, 2015). It was thus important to determine whether Ded1p could directly interact with
113 eIF4A, and to probe whether both proteins could interact simultaneously with eIF4G. Reactions
6
Gao et al.
114 were performed with RNases to exclude possible RNA bridging of the three RNA binding
115 proteins. EIF4A bound directly to Ded1p (Figure 1C, lanes 2). This observation indicates a
116 direct interaction between the two DEAD-box helicases that does not depend on RNA or
117 another factor. EIF4A increased the amount of eIF4G bound to immobilized Ded1p (Figure 1C,
118 compare lanes 1 and 3). As expected, Ded1p bound to immobilized eIF4G (Figure 1D, compare
119 lanes 5 and 7). Ded1p increased the amount of bound eIF4A to eIF4G (Figure 1D, compare
120 lanes 6 and 7). Together, the data reveal a direct interaction between Ded1p and eIF4A and
121 promotion of Ded1p-eIF4G binding by eIF4A. The observations show that interactions between
122 Ded1p, eIF4G, and eIF4A are not mutually exclusive and suggest that all three proteins can
123 interact simultaneously.
124
125 Complexes between Ded1p, eIF4A and eIF4G are unwinding-competent.
126 We next probed whether complexes formed between Ded1p and eIF4A alone or in
127 combination with eIF4G were biochemically active. To this end we measured ATP-dependent
128 RNA unwinding activity for the various protein combinations using an RNA substrate with a 16
129 bp duplex. This substrate was unwound by Ded1p with a much larger apparent rate constant
130 than seen for eIF4A (Figure 1E), consistent with expectations based on previous studies
131 (Rajagopal et al., 2012; Yang et al., 2007a). Nevertheless, addition of eIF4A stimulated the
132 unwinding activity of Ded1p (Figure 1E). Since the weak activity of eIF4A rules out additive
133 effects of the two helicases, the data reveal an impact of eIF4A on the helicase activity of
134 Ded1p, indicating an unwinding-competent eIF4A-Ded1p complex.
135 Addition of eIF4G to Ded1p and eIF4A further increased the unwinding activity seen with
136 only Ded1p and eIF4A (Figure 1E). In contrast, addition of eIF4G to Ded1p alone decreased
137 the rate constant for Ded1p, as previously reported (Putnam et al., 2015). The weak activity of
138 eIF4A with eIF4G precludes additive effects of Ded1p and the eIF4A-eIF4G complex
7
Gao et al.
139 (Rajagopal et al., 2012). Our data therefore suggest that Ded1p, eIF4A and eIF4G can function
140 together to unwind RNA duplexes.
141 Preparations of eIF4G contained the cap binding protein eIF4E (Hilliker et al., 2011),
142 which is required for stability of eIF4G in vitro (Dominguez et al., 2001). EIF4E had been
143 previously shown to interact with Ded1p, but it did not affect the RNA-stimulated ATPase activity
144 of Ded1p (Senissar et al., 2014). Consistent with these results, we did not detect a significant
145 impact of eIF4E on the unwinding activity of Ded1p (not shown). We therefore did not further
146 investigate the impact of eIF4E alone.
147
148 Ded1p is the unwinding module in complexes with eIF4A and eIF4G.
149 We next examined the respective contributions of eIF4A and Ded1p to the unwinding
150 activity when both enzymes were present. We first included hippuristanol in the unwinding
151 reactions (Figure 2A, Figure 2-figure supplement 1). As expected, hippuristanol decreased
152 unwinding activity for eIF4A alone and in combination with eIF4G, but did not significantly
153 impact unwinding of Ded1p alone or in combination with eIF4G (Figure 2A). However,
154 hippuristanol decreased unwinding for Ded1p in the presence of eIF4A, with and without eIF4G,
155 essentially to the level seen with Ded1p alone (Figure 2A). The data suggest that Ded1p is the
156 unwinding module in the complexes with eIF4A, but that functional integrity of eIF4A is
157 important for its impact on Ded1p. This observation provides a rationale for the exacerbating
158 effect of hippuristanol on the growth defect of the ded1-95 mutant strain (Figure 1B). To further
159 probe the role of eIF4A in complexes with Ded1p, we tested the impact of a mutation of the ATP
160 binding site of eIF4A (eIF4AE171A), which abolishes RNA unwinding, as well as ATP binding and
161 hydrolysis (Iost et al., 1999). EIF4AE171A stimulated unwinding by wtDed1p (Figure 2B).
162 However, addition of eIF4G reduced the stimulation to the level roughly seen with Ded1p alone
8
Gao et al.
163 (Figure 2B). These observations provide additional evidence that Ded1p is the unwinding
164 module in complexes with eIF4A.
165 A mutation in the ATP binding site of Ded1p (Ded1pE307A), which abolishes RNA
166 unwinding, as well as ATP binding and hydrolysis (Hilliker et al., 2011; Iost et al., 1999), did
167 not notably alter the basal unwinding activity of eIF4A alone or with eIF4G (Figure 2C). This
168 observation indicates that unwinding activity of Ded1p dictates the level of activity seen in
169 complexes with eIF4A and eIF4G. Finally, we tested the Ded1pT408I mutation (Figure 2D), which
170 showed lower activity than wtDed1p, but higher unwinding activity than eIF4A alone or in
171 complex with eIF4G (Figure 2B,C,D, compare first points). EIF4A stimulated Ded1pT408I and
172 addition of both, eIF4A and eIF4G, stimulated even more (Figure 2D). Collectively, these data
173 show that Ded1p determines the level of unwinding activity in complexes with eIF4A, further
174 supporting the notion that Ded1p is the unwinding module in the complexes with eIF4A.
175
176 The N-terminus of Ded1p is critical for interaction with eIF4A.
177 We next asked which region of Ded1p was critical for the interaction with eIF4A. We
178 removed the C-terminus of Ded1p (Ded1p1-535), which interacts with eIF4G (Hilliker et al.,
179 2011), and measured unwinding activity of Ded1p1-535 with eIF4A, eIF4G, or both (Figure 3A).
180 As expected, eIF4G alone did not notably impact Ded1p1-535,which showed lower unwinding
181 activity than wt Ded1p. EIF4A stimulated Ded1p1-535, and this stimulation was increased by
182 eIF4G (Figure 3A). These observations imply that the C-terminus of Ded1p is not critical for the
183 interaction with eIF4A, even in the presence of eIF4G.
184 We next measured the impact of eIF4A, eIF4G, or both on Ded1p117-604, which lacked the
185 N-terminus. Compared to wt Ded1p, Ded1p117-604 showed reduced unwinding activity (Figure
186 3A). EIF4G alone slightly stimulated Ded1p117-604, consistent with an interaction between eIF4G
9
Gao et al.
187 and Ded1p117-604. The slight stimulation of Ded1p117-604 by eIF4G, as opposed to the inhibition
188 seen with wtDed1p, likely reflects an increase in RNA affinity of Ded1p117-604 by eIF4G. This
189 increase in affinity is also seen with wtDed1p, but does not result in a stimulation of the wild type
190 protein under identical reaction conditions (Putnam et al., 2015). Most notably, eIF4A no longer
191 stimulated Ded1p117-604, and addition of eIF4G did not increase the activity further (Figure 3A).
192 No direct interaction between Ded1p117-604 and eIF4A was detected in pull down experiments
193 (Figure 3- figure supplement 1A). Together, the data show that the N-terminus of Ded1p is
194 important for the interaction between eIF4A and Ded1p, with and without eIF4G, and further
195 support the notion that eIF4A and Ded1p do not simply act additively.
196 In the cell, deletion of the N-terminus conferred a severe cold-sensitive growth
197 phenotype (Figure 3B), consistent with a prior report (Banroques et al., 2011). Notably,
198 overexpression of eIF4A did not alleviate the cold-sensitive growth defect of Ded1p117-604
199 (Figure 3B), as it had for another Ded1p mutation (Figure 1A). This observation emphasizes
200 the biological significance of the interaction between eIF4A and the N-terminus of Ded1p.
201 In contrast to the N-terminus, the C-terminus of Ded1p was not critical for interaction
202 with eIF4A, even in the presence of eIF4G. Given that the C-terminus is important for
203 trimerization of Ded1p (Putnam et al., 2015), our findings raised the question whether Ded1p in
204 complexes with eIF4A was still able to form trimers. We therefore examined the impact of eIF4A
205 on oligomerization of Ded1p using chemical protein-protein crosslinking (Figure 3C). Without
206 eIF4A, trimer formation of Ded1p was seen (Putnam et al., 2015). With eIF4A, no Ded1p trimer
207 was observed, and a prominent new species emerged with a molecular weight consistent with a
208 predominant one to one Ded1p-eIF4A complex (Figure 3C, Figure 3-figure supplement 1B,
209 1C). The Ded1p dimer species seen in the sample with eIF4A represents remaining Ded1p. The
210 data thus indicate that eIF4A interferes with Ded1p oligomerization, even though the C-terminus
211 of Ded1p is not critical for the complex formation with eIF4A.
10
Gao et al.
212 Collectively, results shown above and previous data on the Ded1p-eIF4G interaction and
213 Ded1p oligomerization (Putnam et al., 2015) suggest a basic model for the functional
214 architecture of the complexes formed between Ded1p, eIF4A, and eIF4G (Figure 3D). Without
215 eIF4A, Ded1p interacts via its C-terminus with eIF4G. This interaction precludes formation of a
216 Ded1p trimer (Putnam et al., 2015). The main contact between Ded1p and eIF4A involves the
217 N-terminus of Ded1p, even in the presence of eIF4G and eIF4A also interferes with
218 oligomerization of Ded1p.
219
220 A quantitative framework for the interactions between Ded1p, eIF4A, eIF4G, RNA, and
221 ATP
222 The data above indicated the existence of multiple complexes formed between Ded1p,
223 eIF4A, and eIF4G. To understand biochemical characteristics of these complexes, their
224 interdependencies, and the roles of ATP and RNA, we set out to establish a mechanistic
225 framework describing the thermodynamic stability of these complexes, their RNA and ATP
226 affinities, and their unwinding activities. To our knowledge, no comparable framework exists for
227 any dynamic complex with multiple RNA binding proteins. To devise an instructive quantitative
228 model, it was important to measure the interactions under conditions where ATP binding and
229 turnover as well as RNA binding and remodeling occurred simultaneously. These conditions are
230 best met in unwinding reactions. We therefore performed a large set of unwinding reactions
231 where we systematically varied concentrations of Ded1p, eIF4A, eIF4G, and ATP.
232 Representative data show, among other trends, scaling of the Ded1p stimulation with ATP and
233 Ded1p concentrations, and biphasic response curves with increasing concentrations of eIF4G
234 (Figure 4-figure supplements 1-4). The data indicate intricate interdependencies between the
235 three proteins, ATP and RNA.
11
Gao et al.
236 Using the measured unwinding data, we assembled a quantitative model to describe the
237 thermodynamic stability of all complexes formed between Ded1p, eIF4A, and eIF4G, their RNA
238 and ATP affinities, and their unwinding rate constants (Figure 4A). Model building, data fitting
239 strategy and assessment of model quality are described the Methods section, in Figure 4-
240 figure supplements 5,6, and Supplementary files 2,3. The model considers all 28 distinct
241 complexes formed between Ded1p, eIF4A, eIF4G, RNA and ATP (Figure 4A), plus complexes
242 formed by Ded1p during oligomerization (Putnam et al., 2015). 108 individual interactions were
243 modeled and more than 3,000 individual data points were used in the global datafit. Interactions
244 between Ded1p, eIF4A and eIF4G were modeled as one to one interactions, based on the data
245 above (Figure 3C), and on previous data characterizing the Ded1p-eIF4G interaction (Putnam
246 et al., 2015). The model faithfully recapitulates measured unwinding rate constants (Figure 4B,
247 Figure4-figure supplement 6) as well as previously reported ATP affinities for eIF4A and the
248 eIF4A-eIF4G complex (Rajagopal et al., 2012; Mitchell et al., 2010) (Supplementary file 2).
249 The model reveals the formation of a thermodynamically stable complex between
250 Ded1p, eIF4A, and eIF4G, whose stability further increases with ATP, RNA, or both (Figure 5A,
251 Supplementary file 2F). The stability of the Ded1p-eIF4A complex, although lower than for the
252 Ded1p-eIF4A-eIF4G complex, also increases with ATP, RNA, or both (Figure 5A,
253 Supplementary file 2D). In contrast, the Ded1p-eIF4G, and eIF4A-eIF4G complexes are more
254 stabilized by RNA than by ATP (Figure 5A, Supplementary file 2C-E). To independently verify
255 the calculated high stability of the Ded1p-eIF4G-eIF4A complex, we used microscale
256 thermophoresis (MST) (Seidel et al., 2013) to directly measure binding of Ded1p to a pre-
257 formed eIF4G-eIF4A complex in the absence of RNA and ATP. We measured an apparent
258 dissociation constant of K1/2 < 7 nM (Figure 5B), the experimentally accessible lower limit of
259 affinity with the available MST setup. This value is consistent with the corresponding affinity
260 calculated with our model (Supplementary file 2F). Given the high thermodynamic stability of
12
Gao et al.
261 the Ded1p-eIF4G-eIF4A complex, we suggest to consider Ded1p, like eIF4A, as integral
262 functional part of eIF4F. We note that high thermodynamic stability does not equal persistence
263 of the complex. In fact, the complex between the three proteins is unlikely to be long-lived, given
264 that both eIF4A and Ded1p constantly turn over ATP.
265 Our model further revealed that unwinding rate constants at ATP and protein saturation
266 and RNA affinities at ATP saturation are similar for the Ded1p trimer and for the complexes
267 containing Ded1p and eIF4A (Figure 5C, Supplementary file 2A). Ded1p is stimulated by
268 eIF4A at lower Ded1p concentrations because the complexes with eIF4A bind as single units
269 with hyperbolic binding isotherms, whereas Ded1p oligomerizes cooperatively, which results in
270 a sigmoidal isotherm (Figure 5D, Figure 4-figure supplement 1). Accordingly, stimulation of
271 Ded1p by eIF4A, with or without eIF4G, decreases at higher Ded1p concentration (Figure 4-
272 figure supplements 1A,B). The similar unwinding rate constants at ATP and protein saturation
273 for the Ded1p trimer, the Ded1p-eIF4A and the Ded1p-eIF4A-eIF4G complexes are consistent
274 with the notion that a Ded1p protomer acts as the unwinding unit in the complexes with eIF4A.
275 ATP turnover by Ded1p was not significantly affected by eIF4A and eIF4G (Figure 5-figure
276 supplement 1), supporting the notion that the strand separation by a Ded1p protomer is largely
277 unaffected by eIF4A, with and without eIF4G. The observations suggest that eIF4A and eIF4G
278 function to recruit Ded1p as RNA remodeling unit.
279
280 Modulation of Ded1p activity by eIF4A and eIF4G is impacted by substrate architecture.
281 Having dissected biochemical interdependencies between Ded1p, eIF4A and eIF4G, we
282 next tested whether and how the substrate architecture (lengths of duplexes, lengths and
283 orientation of unpaired tails) impacts the degree by which eIF4A, eIF4G, or both, modulate the
284 unwinding activity of Ded1p. We first performed unwinding reactions with substrates containing
13
Gao et al.
285 different overhang orientations (Figure 6A). We found a modestly enhanced preference of 3’ vs.
286 5' tailed substrates for Ded1p with eIF4A and eIF4G (Figure 6A). This observation is notable,
287 because eIF4G imparts on eIF4A alone a preference for 5’ tailed substrates (Rajagopal et al.,
288 2012) (Figure 6-figure supplement 1).
289 We next examined the impact of the length of the unpaired 3' tail (Figure 6B,C; Figure
290 6-figure supplement 2). To this end, we obtained functional affinities for the various protein
291 complexes for a substrate with 10 nt 3' overhang (Figure 6B). These parameters were
292 calculated from sets of unwinding reactions with the 10 nt tailed substrate with varying
293 concentrations of Ded1p, eIF4A, and eIF4G, analogous to the process described for the
294 substrate with the 25 nt overhang (Figure 6-figure supplement 3). We then compared the
295 functional affinities for the 10 and the 25 nt tailed substrates for each protein complex (Figure
296 6B). For Ded1p alone, affinity increased for a 25 nt vs. a 10 nt tail (Figure 6B, lower panel),
297 consistent with previous observations (Yang et al., 2007a). Addition of eIF4A, eIF4G, or both,
298 increased the relative affinity of the complexes for 25 nt vs. 10 nt tails (Figure 6B). Yet, at low
299 Ded1p concentrations, eIF4A and eIF4G stimulated Ded1p unwinding stronger for a substrate
300 with a 10 nt tail (Figure 6- figure supplement 2), compared to a substrate with a 25 nt tail
301 (Figure 6A). The stronger stimulation for substrates with shorter tails arises from the hyperbolic
302 shape of the functional binding isotherm of the Ded1p-eIF4G-eIF4A complex, compared to the
303 sigmoidal isotherm for the Ded1p trimer (Figure 6C). These data show that activity stimulation
304 can scale with the concentration of components, without a marked change in affinity.
305 Finally, we examined the impact of duplex length on the degree by which eIF4A, eIF4G,
306 or both, modulated the unwinding activity of Ded1p (Figure 6D; Figure 6-figure supplement
307 2). Since duplex length impacts only the unwinding rate constant, but not the substrate affinity
308 (Putnam et al., 2015), we calculated unwinding rate constants at ATP and enzyme saturation
309 for duplexes of different length using our thermodynamic model (Figure 4), and data collected
14
Gao et al.
310 from substrates with 13, 16, and 19 basepairs and identical 3', 25 nt overhangs (Figure 6D).
311 The shortest duplex was unwound fastest by the eIF4A-Ded1p complex, and slowest by the
312 eIF4A-eIF4G-Ded1p complex. For the longest duplex, this order was reversed (Figure 6D). The
313 scaling of the unwinding rate constants with duplex length correlates with the process of strand
314 separation (Putnam et al., 2015). The data thus indicate that both eIF4A and eIF4G directly
315 affect the strand separation step by the Ded1p protomer. EIF4A most likely increases the time
316 by which partially opened duplex stands are kept apart, while decreasing the number of
317 basepairs opened per strand separation event. In contrast, eIF4G and eIF4A together appear to
318 decrease the time by which partially opened duplex stands are kept apart, while increasing the
319 number of basepairs opened per strand separation event. For substrates containing a 5' tail, no
320 comparable scaling of the stimulation with the duplex length was seen, although the stimulation
321 was somewhat higher for the longer duplex (Figure 6A, Figure 6-figure supplement 2).
322 Collectively, our data show that substrate architecture influences the degree by which unwinding
323 activity of Ded1p is impacted by eIF4A and eIF4G, alone and in combination. Our observations
324 further reveal that eIF4A and eIF4G directly influence the strand separation step of Ded1p
325 (Figure 6D), and that substrate affinity of each protein complex scales with the length of the
326 unpaired tail (Figure 6B,C).
15
Gao et al.
327 DISCUSSION
328
329 In this study, we demonstrate direct and functional interactions between the DEAD-box
330 helicases eIF4A and Ded1p. Although both helicases have been firmly implicated in translation
331 initiation, they had not been viewed as physically connected. Our data further show that eIF4A
332 and Ded1p simultaneously interact with eIF4G to form a thermodynamically stable complex. The
333 findings reveal a previously unknown physical and functional interconnection between Ded1p,
334 eIF4A and eIF4G, and suggest that Ded1p is an integral functional part of the eIF4F complex.
335 Our data show that Ded1p, eIF4A and eIF4G, all of which bind RNA, can form four
336 discrete complexes (Ded1p-eIF4A, Ded1p-eIF4G, Ded1p-eIF4A-eIF4G, eIF4A-eIF4G) with
337 defined biochemical properties. Most importantly, Ded1p is the primary RNA remodeling unit in
338 all complexes where it is present. Compared to previously described complexes between eIF4A,
339 eIF4G, and eIF4B, the complexes with Ded1p display orders of magnitude more potent RNA
340 remodeling activity than complexes that only contain eIF4A. This observation raises the
341 possibility that remodeling of stable RNA structures in cellular 5'UTRs, which is often attributed
342 to eIF4A, might in fact be carried out by complexes that contain Ded1p.
343 Our data suggest that the various complexes readily form at physiological concentrations
344 of protein, RNA and ATP. Affinities determined with our model and reported cellular
345 concentrations of the proteins, RNA and ATP (Firczuk et al., 2013; Koç et al., 2004;
346 Zenklusen et al., 2008), allowed us to estimate the prevalence of all complexes at
347 concentrations around the physiological concentrations of Ded1p and eIF4G (Figure 7). These
348 calculations suggest that all complexes exist under physiological conditions, thus rationalizing
349 reported co-immunoprecipitation data (Senissar et al., 2014). The calculations further show that
350 at the reported cellular concentrations of Ded1p and eIF4G (~ 1 μM), most of these two proteins
16
Gao et al.
351 are bound in the Ded1p-eIF4A-eIF4G complex (Figure 7). Almost all of the remaining Ded1p
352 and remaining eIF4G is bound to eIF4A, and only little Ded1p oligomer is present.
353 Variations of both, Ded1p and eIF4G concentrations, even by a factor of only two,
354 significantly alter the prevalence of the respective complexes, with strikingly distinct functional
355 impact for Ded1p and eIF4G. For Ded1p, the changes in the complex ratio have only little
356 impact on helicase activity, since complexes with Ded1p have similar unwinding activities and
357 RNA affinities, although this depends somewhat on the duplex length (Figure 6B,D). Increasing
358 eIF4G concentrations dramatically decreases the fraction of eIF4G that is associated with
359 Ded1p and thus with potent remodeling activity. These observations suggest that Ded1p, in
360 addition to eIF4A, might be important for the function of eIF4G. The data provide a possible
361 rationale for the roughly equal cellular concentration of eIF4G and Ded1p (Firczuk et al., 2013).
362 The Ded1p-eIF4A-eIF4G complex is thermodynamically most stable, suggesting an
363 evolutionary adaptation to favor formation of this complex. For this reason, we suggest to
364 consider Ded1p, like eIF4A, as integral part of eIF4F. Our data highlight a previously unknown
365 role of eIF4A as co-factor for Ded1. This function is consistent with recently discovered
366 overlapping roles of Ded1p and eIF4A in translation initiation (Sen et al., 2015), and with the
367 partial complementation of ded1 mutations by overexpression of eIF4A (Figure 1A). Since the
368 cellular eIF4A concentration exceeds the Ded1p concentration by a factor of approximately five,
369 our findings do not exclude roles of eIF4A in translation initiation that are independent of Ded1p.
370 Our results define a basic functional architecture of the complexes between Ded1p
371 eIF4A and eIF4G (Figure 3D). Most notably, Ded1p utilizes distinct regions to bind to eIF4A and
372 eIF4G. If eIF4A is present, the Ded1p N-terminus is most critical for binding, whereas the Ded1p
373 C-terminus is critical for binding to eIF4G alone. eIF4G and eIF4A allosterically affect the
374 unwinding activity of the Ded1p protomer. In energetic terms, these effects are smaller than the
375 allosteric impact of co-factors on RNA helicases seen in other multi-component complexes,
17
Gao et al.
376 such as the TRAMP complex (Jia et al., 2011). However, Ded1p allosterically impacts eIF4A,
377 eIF4G, or both, as reflected in the marked increase in RNA affinity for the two proteins. This
378 increase in the RNA affinity most likely indicates a function of the proteins as RNA binding
379 adaptors for the Ded1p protomer that separates the strands. Unwinding activity and RNA affinity
380 in the Ded1p-eIF4A, and Ded1p-eIF4A-eIF4G complexes depends on the substrate architecture
381 to a moderate degree, but substrate preferences of Ded1p are not drastically altered by eIF4A
382 or by eIF4A-eIF4G.
383 The intricate connection between Ded1p, eIF4G and eIF4A implies that removal of
384 Ded1p or disruption of the interaction between Ded1p and eIF4A profoundly affects RNA
385 remodeling by complexes that contain eIF4A. In addition, complexes containing Ded1p are
386 more sensitive to ATP depletion and inhibition by AMP than complexes containing only eIF4A,
387 because Ded1p has weaker ATP affinity than eIF4A (Putnam et al., 2015; Putnam and
388 Jankowsky, 2013; Rajagopal et al., 2012). Complexes with Ded1p might therefore be more
389 sensitive to the metabolic state of the cell than complexes with only eIF4A (Putnam and
390 Jankowsky, 2013). Broad regulatory potential is thus likely to be associated with the interaction
391 between Ded1p to eIF4A. Consistent with this hypothesis, the Ded1p N-terminus that binds to
392 eIF4A, harbors a multitude of proven and potential sites for post-translational modifications
393 (Sharma and Jankowsky, 2014). Modification of these sites could affect the interaction with
394 eIF4A.
395 Finally, our quantitative analysis of dynamic complexes formed by multiple RNA binding
396 proteins emphasizes that effects of these proteins on each other cannot simply be described as
397 "stimulating" or "inhibiting". Because multiple sub-complexes are often simultaneously present,
398 both in vitro and in the cell, observed effects depend on the concentration of individual proteins,
399 relative to each other and in absolute molar terms, and on ATP and RNA concentrations. For
400 these reasons, frameworks such as the one developed in this study are required to adequately
18
Gao et al.
401 describe biochemical features of individual complexes, and to define the interactions between
402 all factors, especially if the complexes are not long lived. Approaches outlined in this work could
403 therefore be useful for the quantitative analysis of other dynamic complexes that contain
404 multiple RNA-binding proteins.
19
Gao et al.
405 MATERIALS AND METHODS
406
407 Recombinant Proteins
408 Ded1p and mutant Ded1p proteins
409 Wild type Ded1p with an N-terminal His6-tag (plasmid: pET-22b(+)-DED1, Supplementary file
410 1C) was expressed in E. coli (BL21) and purified and characterized as described (Yang and
411 Jankowsky, 2005).
T408I 412 The plasmid for recombinant Ded1p with N-terminal His6-tag (pEJ13, Supplementary
413 file 1C) was generated by site-directed mutagenesis of pET-22b(+)-DED1with primers X11 and
414 X12 (Supplementary file 1B). The presence of the ded1-95 allele in the transformants was
415 probed for by amplifying plasmid DNA with primers X53 and X54 and subsequent restriction
416 fragment length analysis with DdeI (New England Biolabs) according to the manufacturer’s
417 suggestions. The restriction enzyme cuts only the WT allele but leaves the ded1-95 allele uncut.
418 WT controls were included in the assay to ensure the validity of the reaction conditions.
419 Ded1pT408I was expressed in E. coli (BL21) and purified and characterized as described (Yang
420 and Jankowsky, 2005).
E307A E307A 421 The plasmid for Ded1p (N-terminal His6-tag, pET-22b(+)-DED1 , Supplementary
422 file 1C) was a gift from Dr. Patrick Linder (Geneva, Switzerland, Iost et al., 1999). Ded1pE307A
423 was expressed in E. coli (BL21) and purified and characterized as described wt Ded1p
424 described (Yang and Jankowsky, 2005).
425 To produce recombinant HA-tagged Ded1p (Ded1p-HA, N-terminal His6-tag and C-terminal
426 HA3-tag, plasmid:pET-22b(+)-His-Ded1-HA), DED1 from yeast genomic DNA was amplified by
427 PCR with primers zfDD10 and zfDD11 (Supplementary file 1B). The PCR product was gel
428 purified, digested with HindIII and PstI, and ligated into the plasmid pBlueScript II KS(+).
20
Gao et al.
429 Tandem triple HA tags were inserted with PCR amplification of the new plasmid with the primers
430 zfDD12 and zfDD13 (Supplementary file 1B). The resulting pBS II KS(+)-DED1-HA plasmid
431 was digested with NheI and HindIII and the ~2.0 kb fragment containing the coding sequence
432 for His6-Ded1p-HA3 was gel purified. The pET-22b(+)-DED1 plasmid was digested with NheI
433 and HindIII and the ~7.0 kb product containing the backbone of pET-22b(+) was gel purified.
434 The two gel purified fragments were ligated to form pET-22b(+)-His-Ded1-HA. The coding
435 sequence in the plasmid was verified by sequencing. The plasmid was transfected into E.coli
436 (BL21), and recombinant Ded1p-HA was expressed, purified, and characterized as wild type
437 Ded1p.
438 Ded1p1-535 (C-terminus deleted, pET-22b(+)-His- DED11-535) was expressed in E. coli (BL21)
439 and purified and characterized as described (Hilliker et al., 2011).
440 To produce recombinant Ded1p117-604 (N-terminus deleted) the fragment for DED1117-604 was
441 amplified with the primers zfDD14 and zfDD15 (Supplementary file 1B), using the plasmid
442 pET-22b(+)-DED1 as PCR template. The PCR product was gel purified, digested with XhoI and
443 SacI, and ligated into the pET-22b(+)-His6-thrombin-SUMO plasmid (gift from Dr. Derek Taylor,
444 Case Western Reserve University, Cleveland, OH) at the 3’-end of SUMO sequence. The
445 coding sequence in the resulting plasmid (pET-22b(+)-SUMO- DED1117-604) was verified by
446 sequencing. Subsequently, the plasmid was introduced into E.coli (DE3). Expressed His-
447 SUMO-Ded1p117-604 protein was purified through a Nickel column (PrepEaseHis-tagged protein
448 purification resin high specificity, USB), and the eluted protein was incubated with His-tagged
449 SUMO protease at 8ºC overnight in order to remove the His-SUMO-tag. The solution was briefly
450 incubated with Nickel resin to absorb the His-SUMO tag and the His-tagged SUMO protease.
451 Ded1p117-604 in the supernatant was further purified through a P11 phosphocellulose column
452 (Whatman). Protein was stored in 50 mM Tris-HCl (pH 8.0), 2 mM DTT, 1 mM EDTA, 0.1 %
453 (v/v) Triton-X 100, 30% (v/v) glycerol and 300 mM NaCl. Protein homogeneity and concentration
21
Gao et al.
454 were determined by SDS-PAGE and Coomassie Blue staining, with diluted BSA (Roche) and
455 wild-type Ded1p in known concentrations as standards.
456
457 eIF4G protein
458 Expression and purification of eIF4G [(pET41a(+)-GST-S•Tag-eIF4G1-His) and (pET21b(+)-
459 eIF4E)] were previously described (Hilliker et al., 2011).
460
461 eIF4A and mutant eIF4A proteins
462 pET-28a(+) plasmid encoding an N-terminally His6-tagged wild-type yeast eIF4A (pET-28a(+)-
463 eIF4A) was a gift from Dr. Michael Altmann (Bern, Switzerland) (Schütz et al., 2008). To
464 generate plasmid pET-28a(+)-eIF4AE171A, two pairs of primers (zfDD16, zfDD17, zfDD18 and
465 zfDD19) were used to introduce the E171A mutation into plasmid pET-28a(+)-eIF4A by the
466 SLIM-PCR procedure according to (Chiu, et al., 2004, 2008). eIF4A and eIF4AE171A were
467 purified as described (Hilliker et al., 2011), and stored in 50 mM Tris-HCl (pH 8.0), 2 mM DTT,
468 1 mM EDTA, 0.1 % Triton-X 100, 30% glycerol and 200 mM NaCl.
469 To create the plasmid for the recombinant GST-eIF4A purification, the eIF4A coding
470 sequence was amplified from pET-28a(+)-eIF4A with the primers zfDD20 and zfDD21,
471 subsequently was digested with SacII and XhoI, and ligated into the pET-41a(+) plasmid to
472 generate the pET-41a(+)-eIF4A plasmid (Supplementary file 1C). GST-His6-eIF4A was
473 expressed in E. coli (DE3). The protein was purified through a Nickel column (USB), a
474 glutathione sepharose 4B column (GE health care) and a size-exclusion NAP-25 column
475 (BioRad). Purified protein was stored in 50 mMTris-HCl (pH 8.0), 2 mM DTT, 1 mM EDTA, 0.1
476 % Triton-X 100, 30% glycerol and 200 mMNaCl.
477
22
Gao et al.
478 RNA oligonucleotides
479 RNA oligonucleotides were purchased from DHARMACON. RNAs were radiolabeled and
480 duplex substrates were prepared and characterized as described (Yang and Jankowsky,
481 2005). For sequences see Supplementary file 1A.
482
483 Complementation of ded1-tam by eIF4A overexpression
484 Yeast strains are listed in Supplementary file 4. A high copy suppressor screen of cold
485 sensitive ded1-tam allele (Hilliker et al., 2011) was performed using a yeast genomic tiling
486 library of overexpression vectors (Jones et al., 2008; GE Healthcare #YSC5103). The putative
487 suppressors included DED1 or TIF2 within the 10-kb genomic inserts. Plasmids containing
488 either wild type DED1 (pRP1555) or the mutant ded1-tam (pRP048) were transformed into a
489 DED1 yeast shuffle strain containing a chromosomal deletion of DED1 and wild type copy of
490 DED1 on a URA3-marked plasmid (yRP2799; Hilliker et al., 2011). Transformants were grown
491 on selective media before counter-selection against the URA3-marked plasmid on 5-fluoroorotic
492 acid (5-FOA; Boeke et al., 1987). These strains, now containing the newly transformed copy of
493 ded1 as an exclusive copy, were transformed with either an empty vector (pGP564; GE
494 Healthcare #YSC5034) or a plasmid overexpressing either DED1, TIF1, or TIF2 (pAKH776,
495 757, 775, or 756) on a LEU2-marked vector. Transformants were grown on selective media with
496 2% dextrose. Cells were grown to mid-log phase in selective media with 2% glucose. The
497 cultures were pin replicated at identical concentrations onto YPDA media and incubated at 30 or
498 16°C. All of the strains grew like wild type at 30°C.
499 High copy plasmids containing either TIF1, TIF2, and DED1 ORFs and roughly 500
500 nucleotides up- and downstream of the ORF were made via recombination in yeast. TIF2, and
501 DED1 genes were amplified from the genomic tiling plasmids that suppressed the ded1 mutant
502 alleles (GE Healthcare #YSC5103). The TIF1 gene was amplified from genomic DNA of the
23
Gao et al.
503 BY4741 S. cerevisiae strain. Primers (Supplementary file 1B) contained a 40 nt overhang
504 complementary to the ends of the insertion site in the vector backbone (pGP564). The vector
505 was linearized by digestion with BamHI and XmaI. Gel purified PCR product and linearized
506 vector were transformed into BY4741 and plated on media selective for the circularized vector.
507 Circularized plasmids were isolated from yeast via the Zymoprep Yeast Plasmid Miniprep II kit
508 (Zymo Research, #D2004) and transformed into DH5α competent E. coli (NEB, #C2987) to
509 amplify plasmids. Plasmid inserts were confirmed by sequencing the entire insert.
510
511 Growth measurements with Hippuristanol
512 For growth studies with hippuristanol, a plasmid with the ded1-95 allele (Ded1pT408I,
513 pEJ4, Supplementary file 1C) was created. Genomic DNA was amplified with primers X7 and
514 X8 from a yeast strain containing the WT copy of the HIS3 gene (yJC300, gift from Dr. Jeff
515 Coller, Cleveland, OH) and subsequently subcloned (Invitrogen, Carlsbad, CA). The ded1-95
516 allele was introduced into pEJ4 by site-directed mutagenesis with primers X11 and X12
517 generating pEJ6. pEJ4 and pEJ6 were linearized and used to transform BY4741 by standard
518 lithium chloride transformation yielding yeast strains yEJ1 and yEJ3, respectively. Presence of
519 the ded1-95 allele in the yeast strain was verified by amplifying genomic DNA and subsequent
520 restriction fragment length analysis as described above for pEJ13.
521 Hippuristanol was a gift from Dr. Jerry Pelletier (McGill University, Canada). WT and
0 522 ded1-95 yeast strains were grown in YPD at 30 C, diluted (V = 1.8 mL) to 0.1 OD600 and grown
0 523 for an additional 3 hours to 0.4 OD600. Cultures were shifted to 37 C, and grown for 45 min at
524 370C. Hippuristanol in DMSO was added to final concentrations of 1 µM, 5µM, 10 µM and
525 20µM. For control reactions, DMSO alone (0.2% v/v) was added. Cells were further grown for
° 526 45 min at 37 C. Growth was subsequently followed for 2 hours by measuring OD600. The growth
527 rate is reported, normalized to the growth rate of the corresponding strain without Hippuristanol.
24
Gao et al.
528
529 Complementation with DED1117-604
530 The plasmid for complementation with DED1117-604 (N-terminus deleted, plasmid pEJ10) was
531 constructed from pRP1555 (HIS3/CEN) (Hilliker et al., 2011), using primers Ded116delFt/Rs
532 and Ded116delRt/Fs (Supplementary file 1B). The plasmid insert was confirmed by
533 sequencing the insert. The ded1-null strain yRP2799 (MATa ura3Δ0 leu2Δ0 lys2Δ0 his3Δ1
534 met15Δ0 ded1::KANMX pRP1560) was transformed with HIS3-marked wild type DED1
535 (pRP1555), or DED1117-604 (pEJ10). As an additional control we also transformed yRP2799 with
536 an empty HIS3-marked plasmid (pRS413). Transformants were streaked on media lacking
537 histidine and further screened on a 5-FOA containing plate. These strains were transformed
538 with either empty vector (pGP564; GE Healthcare #YSC5034) or a plasmid overexpressing TIF1
539 (pAKH775) on a LEU2-marked vector. Transformants were grown on selective media with 2%
540 dextrose. Cells were grown to mid-log phase in selective media with 2% glucose. The cultures
541 were pin replicated at identical concentrations onto YPDA media and incubated at 30 or 16°C.
542 At 37°C, no significant growth differences between wtDED1 and DED1117-604 were observed.
543
544 Pull-down Assays
545 In vitro pull‐down measurements were essentially performed as described (Putnam et al.,
546 2015). For the pull-down assays with immobilized eIF4G , 5 µL bed volume of DynaBeads®
547 Protein G (ThermoFisher Scientific # 10003D) was pre-equilibrated with 200 µL helicase
548 reaction buffer [40 mM Tris–HCl (pH 8.0), 0.5 mM MgCl2, 2 mM DTT,0.01% (v/v) IGEPAL CA-
549 630] complemented with 100 mM NaCl (HRB-Na100 buffer) at room temperature for 30 min.
550 The pre-equilibrated beads were then incubated with 4 µg anti-S•Tag monoclonal antibody
551 (Novagen, Cat. # 71549) and 420 µg BSA (Roche) in a 25 µL reaction at 8°C overnight. Protein
25
Gao et al.
552 binding reactions (100 µL) included 300 nM S•Tag-eIF4G, 300 nM Ded1p, 1,500 nM eIF4A, 20
553 µg/mL RNases A,B (Roche), as indicated. 3 µL of the protein reaction volume was removed and
554 used as input. The remainder of the reaction volume was rotated at 8ºC overnight, and
555 subsequently mixed with the 25 µL equilibrated beads. After further incubation of one hour with
556 rotation, the beads were washed four times with RIPA buffer (50 mMTris, pH 7.5, 1% IGEPAL
557 CA-630, 0.25% sodium deoxycholate, 0.1% SDS, 200 mM NaCl, 1mM EDTA, 1mM PMSF).
558 Proteins were eluted with Laemmli buffer (Biorad), and denatured at 95°C for 3 min. Proteins
559 were resolved on the NuPAGE Novex 4-12% Bis-Tris protein gels (1.0 mm) in MOPS buffer
560 (pH7.7, Life Technology), and visualized by Western blot with primary antibodies against Ded1p
561 (α-Ded1p, polyclonal antibody, raised against full length Ded1p, dilution 1:6,000), monoclonal
562 anti- S•Tag antibody (Novagen, cat. # 71549), and anti-eIF4A polyclonal antibody (Santa Cruz,
563 prod. # sc-27227).
564 For the pull-down assays with immobilized Ded1p-HA, 1,200 nM Ded1p-HA was incubated
565 with 0.2 µg anti-HA monoclonal antibody (Roche, cat. # 11867423001, clone 3F10) and pre-
566 equilibrated DynaBeads® Protein G (Life Technologies) (5 µL bed volume) in a 100 µL reaction
567 at 8ºC for three hours. The beads were then washed three times with the HRB-Na100 buffer to
568 remove unbound Ded1p-HA protein, and subsequently mixed with 300 nM S•Tag-eIF4G, 1,500
569 nM eIF4A, 20 µg/mL RNases A,B (Roche), as indicated, in a volume of 100 µL. 3 µL of the
570 reaction volume (including beads) was removed and used as input. The remainder of the
571 reaction volume was incubated under rotation overnight at 8ºC. Beads were washed with the
572 HRB-Na100 buffer for three times. Proteins were eluted with Laemmli buffer (Biorad), and
573 denatured at 95°C for 3 min. Proteins were resolved on NuPAGENovex 4-12% Bis-Tris protein
574 gels (1.0 mm) in MOPS buffer (pH 7.7, Life Technology), and visualized by Western blot with
575 primary antibodies against Ded1p (α-Ded1p, polyclonal antibody, raised against full length
26
Gao et al.
576 Ded1p, dilution 1:6,000), monoclonal anti-S-Tag antibody, and anti-eIF4A polyclonal antibody
577 (Santa Cruz, prod. # sc-27227).
578 For the pull-down assays with GST-eIF4A or GST immobilized, 300 µL of the protein-binding
579 reaction was set up with 1500 nM GST-eIF4A or GST, 1500 nM Ded1p117-604 or Ded1p1-535, 20
580 µg/mL RNases A,B (Roche), as indicated. 30 µL of the protein-binding reaction was removed as
581 input. The reminder was incubated with rotation overnight at 8ºC in presence of 5 µg/µL BSA
582 (Roche), and was further mixed and incubated with the pre-equilibrated 40 µL (bed volume)
583 glutathione sepharose 4B resin(GE healthcare) at 8ºC for one hour. The beads were washed
584 four times with the RIPA buffer. Proteins were eluted with Laemmli Sample Buffer (Biorad), and
585 denatured at 95°C for 3 min. Proteins were resolved on the NuPAGE®Novex® 4-12% Bis-Tris
586 mini protein gels (1.0mm thick) in MOPS running buffer (pH7.7) (Life Technology), and
587 visualized either by Western blot with primary antibodies against Ded1p (α-Ded1p, polyclonal
588 antibody, raised against full length Ded1p, dilution 1:6,000), or by Coomassie Blue staining for
589 the GST-tagged proteins.
590
591 RNA Duplex Unwinding
592 Unwinding reactions were performed under a pre‐steady‐state regime as previously described
593 (Yang and Jankowsky, 2005). Briefly, Ded1p was incubated with 0.5 nM radiolabeled RNA for
594 5 min in reaction buffer, and reactions were initiated with ATP/Mg2+. Where applicable, eIF4G,
595 eIF4A, or both were included. Aliquots were removed at indicated times, and reactions were
596 stopped and applied to 15% nondenaturing PAGE. Gels were dried and RNAs visualized and
597 quantified with a Molecular Dynamics Phosphorimager and the ImageQuant 5.2 software
598 (Molecular Dynamics). Observed rate constants were calculated as described, considering
599 simultaneously occurring strand annealing (Yang and Jankowsky, 2005).
27
Gao et al.
600
601 Protein-protein crosslinking
602 In vitro protein‐protein crosslinking with formaldehyde was essentially performed as described
603 (Putnam et al., 2015),with Ded1p (0.3 µM), eIF4A (2 µM for Figure 3C, or indicated
604 concentrations in Figure 3-figure supplement 1B), RNA (2 µM, 16bp, 25 nt 3’-overhang) and
605 0.5 mM ADPNP-MgCl2 in buffer used in unwinding reactions. Formaldehyde [1% (v/v)] was
606 added, and incubated at room temperature for 30 minutes. Crosslinking was quenched with 0.5
607 mMTris-Glycine (pH 6.8). Proteins were applied to and resolved on the NuPAGENovex 4-12%
608 Bis-Tris protein gels (1.5 mm) in MOPS buffer (pH7.7, Life Technology). Proteins were
609 visualized by Western blot with polyclonal antibody against Ded1p.
610
611 Microscale Thermophoresis
612 EIF4A was labeled with Alexa Fluor 647 NHS ester (ThermoFisher Scientific # A37573)
613 according to the supplied protocol. Excess dye was removed by buffer exchange into 50 mM
614 Tris-HCl (pH 8.0), 2 mM DTT, 1 mM EDTA, 0.1 % Triton-X 100, 30% glycerol and 100 mM NaCl
615 using Zebaspin desalting columns (0.5mL, 7K MWCO, ThermoFisher Scientific # 89882). The
616 protein concentration was determined by SDS-PAGE and Coomassie Blue staining, with the
617 diluted BSA (Roche) and unlabeled eIF4A in known concentration as standards.
618 Microscale thermophoresis measurements were performed at room temperature on a
619 Monolith NT.115 system (NanoTemper Technologies) with standard treated glass capillaries
620 (NanoTemper Technologies). Buffers were identical to those used in unwinding reactions,
621 complemented with 0.5 µg/µL BSA (Roche) and 0.05% (v/v) Tween, to prevent absorption of the
622 proteins to the glass capillaries, and with 25 µg/mL RNase A (USB) to remove residual RNA.
623 20% IR-laser power was used to locally generate the temperature jump in the capillaries.
28
Gao et al.
624 Fluorescence change in the heated field was monitored with 20% LED power. Laser on and off
625 times were set at 30 seconds and 5 seconds, respectively. The eIF4A-eIF4G complex was
626 formed for at least 10 minutes prior to the addition of Ded1p. Measurements were performed
627 multiple times, averages of individual points were calculated and the resulting binding isotherm
628 was fitted with the quadratic binding equation.
629
630 Thermodynamic model and data fit
631 Kinetic modeling was performed with KINETK EXPLORER (Kintek, Austin, TX) (Johnson et al.,
632 2009a; Johnson et al., 2009b) using data obtained in this and earlier studies (Putnam et al.,
633 2015). The thermodynamic framework to describe the interactions between Ded1p, eIF4A,
634 eIF4G, RNA, and ATP was built on the previously established kinetic framework for duplex
635 unwinding by the Ded1p oligomer (Putnam et al., 2015). eIF4A and eIF4G only interact with a
636 Ded1p monomer (Figure 3). The resulting framework contains 108 binding reactions, listed in
637 Supplementary file 2. The model was constrained by considering energy conservation. That
638 is, the change in free energy for transition of one complex into another is independent of the
639 reaction pathway. In addition, we incorporated the experimentally determined RNA affinity of
640 eIF4G (Putnam et al., 2015). We further considered that the comparably weak RNA
641 independent ATPase activity of Ded1p was not affected by eIF4G and eIF4A to a measurable
642 degree (data not shown). This notion allowed us to use the Ded1p affinity for ATP in the
643 absence of RNA for multiple Ded1p complexes. Collectively, these considerations enabled us to
644 describe the entire framework in 16 groups of thermodynamically linked binding steps
645 (Supplementary file 3).
646 Ten complexes are theoretically capable of unwinding: Ded1p, eIF4A, eIF4-eIF4G, Ded1p-
647 eIF4G, Ded1p-eIF4A (with ATP bound to Ded1p, eIF4A, or both), Ded1p-eIF4A-eIF4G (with
648 ATP bound to Ded1p, eIF4A, or both). Given that eIF4AE171A (ATPase deficient) was able to
29
Gao et al.
649 stimulate unwinding by Ded1p, while Ded1pE307A (ATPase deficient) did not stimulate unwinding
650 by eIF4A (Figure 2), we considered ATP binding to eIF4A irrelevant for the unwinding activity of
651 the Ded1p-eIF4A complex. In the presence of eIF4G, ATP binding by eIF4A was required for
652 the impact on Ded1p (Figure 2). We therefore considered the Ded1p-eIF4A-eIF4G complex
653 with ATP bound to both helicases as active unwinding complex. Collectively, these
654 considerations simplified the number of unwinding complexes to six: Ded1p, eIF4A, eIF4-eIF4G,
655 Ded1p-eIF4G, Ded1p-eIF4A (with ATP bound to Ded1p independent of ATP binding to eIF4A),
656 Ded1p-eIF4A-eIF4G (with ATP bound to Ded1p and eIF4A). We did not include ATP hydrolysis
657 steps in our analysis for any complex, except for Ded1p alone, for which we had previously
658 characterized the respective steps (Putnam et al., 2015).
659 Global data fitting was performed using Kintek Global Explorer. 20 datasets for unwinding of
660 the 16 bp duplex with 3’-25 ntssRNA overhang were fit simultaneously. A dataset includes all
661 time courses for titration of 1 component (e.g., protein or ATP) with other components held
662 constant. For each complex at least one dataset was included for a titration of each component
663 (e.g., Ded1-eIF4G: 1 Ded1p titration, 1 eIF4G titration, 1 ATP titration).Datasets were as follows:
664 Ded1p-eIF4A (4 datasets), Ded1p-eIF4G (3 datasets), eIF4A-eIF4G (4 data sets), and Ded1p-
665 eIF4A-eIF4G (9 datasets). Multiple iterations were performed until the best fit to all data sets
666 was achieved. Data were fit assuming that all binding steps for Ded1p with eIF4A, and eIF4G
667 were in rapid equilibrium. Parameters for the best fit values are reported in Supplementary file
668 2. Variables not well constrained by the data are reported as limits. Errors are not always
669 symmetrically distributed around the best fit value (Figure 4-figure supplement 5). Uncertainty
670 in each parameter is presented as a range calculated at the 95% confidence interval. To
671 visually assess the quality of fit, experimental data were plotted versus values predicted with the
672 model (Figure 4B).
30
Gao et al.
673 ACKNOWLEDGEMENTS
674 We thank Dr. Jerry Pelletier (McGill Univ., Montreal) for Hippuristanol, Dr. Derek Taylor (Case
675 Western Reserve University, Cleveland) for SUMO plasmids, Dr. Michael Altmann (Univ. Bern,
676 Berne) for the plasmid containing wild-type His-tagged eIF4A, Dr. Jeff Coller (Case Western
677 Reserve University, Cleveland) for yeast strains, Dr. Yinghua Chen and the Protein Expression
678 Purification Crystallization and Molecular Biophysics Core (Case Western Reserve University,
679 Cleveland)for assistance with MST measurements, Sukanya Srinivasan (Case Western
680 Reserve University, Cleveland) for assistance with protein purification, and Drs. Michael Harris
681 and William Merrick (Case Western Reserve University, Cleveland) for helpful discussions.
682
683
684 CONFLICT OF INTEREST STATEMENT
685 The authors declare no conflicting interest.
31
Gao et al.
686 REFERENCES
687 Aitken CE, Lorsch JR. 2012. A mechanistic overview of translation initiation in eukaryotes. Nat.
688 Struct. Mol. Biol. 19, 568–576. doi:10.1038/nsmb.2303
689 Banroques J, Cordin O, Doère M, Linder P, Tanner NK. 2011. Analyses of the functional
690 regions of DEAD-box RNA “helicases” with deletion and chimera constructs tested in vivo
691 and in vitro. J. Mol. Biol. 413, 451–472. doi:10.1016/j.jmb.2011.08.032
692 Boeke JD, Trueheart J, Natsoulis G, Fink GR. 1987. 5-Fluoroorotic acid as a selective agent in
693 yeast molecular genetics. Meth. Enzymol. 154, 164–175.doi:10.1016/0076-6879(87)54076-
694 9
695 Bol GM, Vesuna F, Xie M, Zeng J, Aziz K, Gandhi N, Levine A, Irving A, Korz D, Tantravedi S,
696 Heerma van Voss MR, Gabrielson K, Bordt EA, Polster BM, Cope L, van der Groep P,
697 Kondaskar A, Rudek MA, Hosmane, RS, van der Wall E, van Diest PJ, Tran PT, Raman V.
698 2015. Targeting DDX3 with a small molecule inhibitor for lung cancer therapy. EMBO Mol
699 Med7, 648–669. doi:10.15252/emmm.201404368
700 Bordeleau M-E, Mori A, Oberer M, Lindqvist L, Chard LS, Higa T, Belsham GJ, Wagner G,
701 Tanaka J, Pelletier J. 2006. Functional characterization of IRESes by an inhibitor of the
702 RNA helicase eIF4A. Nat. Chem. Biol.2, 213–220. doi:10.1038/nchembio776
703 Bowers HA, Maroney PA, Fairman ME, KastnerB, Lührmann R, Nilsen TW, Jankowsky E.
704 2006. Discriminatory RNP remodeling by the DEAD-box protein DED1. RNA12, 903–912.
705 doi:10.1261/rna.2323406
706 Burckin T, Nagel R, Mandel-Gutfreund Y, Shiue L, Clark TA, Chong J-L, Chang T-H, Squazzo
707 S, Hartzog G, Ares M. 2005. Exploring functional relationships between components of the
708 gene expression machinery. Nat. Struct. Mol. Biol.12, 175–182. doi:10.1038/nsmb891
32
Gao et al.
709 Chiu J, March PE, Lee R, Tillett D. 2004. Site-directed, Ligase-Independent Mutagenesis
710 (SLIM): a single-tube methodology approaching 100% efficiency in 4 h. Nucleic Acids
711 Res32, e174. doi:10.1093/nar/gnh172
712 Chiu J, Tillett D, Dawes IW, March PE. 2008. Site-directed, Ligase-Independent Mutagenesis
713 (SLIM) for highly efficient mutagenesis of plasmids greater than 8kb. J. Microbiol.
714 Methods73, 195–198. doi:10.1016/j.mimet.2008.02.013
715 de la Cruz J, Iost I, Kressler D, Linder P. 1997. The p20 and Ded1 proteins have antagonistic
716 roles in eIF4E-dependent translation in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci.
717 U.S.A.94, 5201–5206.
718 Dominguez D, Kislig E, Altmann M, Trachsel H. 2001. Structural and functional similarities
719 between the central eukaryotic initiation factor (eIF)4A-binding domain of mammalian
720 eIF4G and the eIF4A-binding domain of yeast eIF4G. Biochem. J. 355, 223–230.
721 Firczuk H, Kannambath S, Pahle J, Claydon A, Beynon R, Duncan J, Westerhoff H, Mendes P,
722 McCarthy JE. 2013. An in vivo control map for the eukaryotic mRNA translation machinery.
723 Mol. Syst. Biol. 9, 635. doi:10.1038/msb.2012.73
724 Giaever G, Chu AM, Ni L, Connelly C, Riles L, Véronneau S, Dow S, Lucau-Danila A, Anderson
725 K, André B, Arkin AP, Astromoff A, El-Bakkoury M, Bangham R, Benito R, Brachat S,
726 Campanaro S, Curtiss M, Davis K, Deutschbauer A, Entian K-D, Flaherty P, Foury F,
727 Garfinkel DJ, Gerstein M, Gotte D, Güldener U, Hegemann JH, Hempel S, Herman Z,
728 Jaramillo DF, Kelly DE, Kelly SL, Kötter P, LaBonte D, Lamb DC, Lan N, Liang H, Liao H,
729 Liu L, Luo C, Lussier M, Mao R, Menard P, Ooi SL, Revuelta JL, Roberts CJ, Rose M,
730 Ross-Macdonald P, Scherens B, Schimmack G, Shafer B, Shoemaker DD, Sookhai-
731 Mahadeo S, Storms RK, Strathern JN, Valle G, Voet M, Volckaert G, Wang C, Ward TR,
732 Wilhelmy J, Winzeler EA, Yang Y, Yen G, Youngman E, YuK, Bussey H, Boeke JD, Snyder
733 M, Philippsen P, Davis RW, Johnston M. 2002. Functional profiling of the Saccharomyces
734 cerevisiae genome. Nature418, 387–391. doi:10.1038/nature00935
33
Gao et al.
735 Heerma van Voss MR, Vesuna F, Trumpi K, Brilliant J, Berlinicke C, de Leng W, Kranenburg O,
736 Offerhaus GJ, Bürger H, van der Wall E, van Diest PJ, Raman V. 2015. Identification of the
737 DEAD box RNA helicase DDX3 as a therapeutic target in colorectal cancer. Oncotarget6,
738 28312–28326. doi:10.18632/oncotarget.4873
739 Hilbert M, Kebbel F, Gubaev A, Klostermeier D. 2011. eIF4G stimulates the activity of the
740 DEAD box protein eIF4A by a conformational guidance mechanism. Nucleic Acids Res.39,
741 2260–2270. doi:10.1093/nar/gkq1127
742 Hilliker A, Gao Z, Jankowsky E, Parker R. 2011. The DEAD-Box Protein Ded1 Modulates
743 Translation by the Formation and Resolution of an eIF4F-mRNA Complex. Mol. Cell43,
744 962–972. doi:10.1016/j.molcel.2011.08.008
745 Hinnebusch, AG. 2014. The Scanning Mechanism of Eukaryotic Translation Initiation. Annual
746 Review of Biochemistry83, 779–812. doi:10.1146/annurev-biochem-060713-035802
747 Iost I, Dreyfus M, Linder P. 1999. Ded1p, a DEAD-box protein required for translation initiation
748 in Saccharomyces cerevisiae, is an RNA helicase. J. Biol. Chem274, 17677–17683.
749 Jia H, Wang X, Liu F, Guenther U-P, Srinivasan S, Anderson JT, Jankowsky E. 2011. The RNA
750 helicase Mtr4p modulates polyadenylation in the TRAMP complex. Cell145, 890–901.
751 doi:10.1016/j.cell.2011.05.010
752 Jiang L, Gu Z-H, Yan Z-X, Zhao X, Xie Y-Y, Zhang Z-G, Pan C-M, Hu Y, Cai C-P, Dong Y,
753 Huang J-Y, Wang L, Shen Y, Meng G, Zhou J-F, HuJ-D, Wang J-F, Liu Y-H, Yang L-H,
754 Zhang F, Wang J-M, Wang Z, Peng Z-G, Chen F-Y, Sun Z-M, Ding H, Shi J-M, Hou J, Yan
755 J-S, Shi J-Y, Xu L, Li Y, Lu J, Zheng Z, Xue W, Zhao W-L, Chen Z, Chen S-J. 2015. Exome
756 sequencing identifies somatic mutations of DDX3X in natural killer/T-cell lymphoma. Nat.
757 Genet. 47, 1061–1066. doi:10.1038/ng.3358
758 Johnson KA, Simpson ZB, Blom T. 2009a. FitSpace Explorer: An algorithm to evaluate
759 multidimensional parameter space in fitting kinetic data. Analytical Biochemistry387, 30–41.
760 doi:10.1016/j.ab.2008.12.025
34
Gao et al.
761 Johnson KA, Simpson ZB, Blom T. 2009b. Global Kinetic Explorer: A new computer program
762 for dynamic simulation and fitting of kinetic data. Analytical Biochemistry387, 20–29.
763 doi:10.1016/j.ab.2008.12.024
764 Jones GM, Stalker J, Humphray S, West A, Cox T, Rogers J, Dunham I, Prelich G. 2008. A
765 systematic library for comprehensive overexpression screens in Saccharomyces
766 cerevisiae. Nat Meth5, 239–241. doi:10.1038/nmeth.1181
767 Koç A, Wheeler LJ, Mathews CK, Merrill GF. 2004. Hydroxyurea arrests DNA replication by a
768 mechanism that preserves basal dNTP pools. J. Biol. Chem. 279, 223–230.
769 doi:10.1074/jbc.M303952200
770 Linder P, Jankowsky E. 2011. From unwinding to clamping - the DEAD box RNA helicase
771 family. Nat. Rev. Mol. Cell Biol. 12, 505–516. doi:10.1038/nrm3154
772 Linder P, Slonimski PP. 1989. An essential yeast protein, encoded by duplicated genes TIF1
773 and TIF2 and homologous to the mammalian translation initiation factor eIF-4A, can
774 suppress a mitochondrial missense mutation. Proc. Natl. Acad. Sci. U.S.A. 86, 2286–2290.
775 Merrick WC. 2015. eIF4F: a retrospective. J. Biol. Chem. 290, 24091–24099.
776 doi:10.1074/jbc.R115.675280
777 Mitchell SF, Walker SE, Algire MA, Park E-H, Hinnebusch AG, and Lorsch JR (2010). The 5’-7-
778 methylguanosine cap on eukaryotic mRNAs serves both to stimulate canonical translation
779 initiation and to block an alternative pathway. Mol. Cell 39, 950-962.
780 doi:10.1016/j.molcel.2010.08.021
781 Parsyan A, Svitkin Y, Shahbazian D, Gkogkas C, Lasko P, Merrick WC, Sonenberg N. 2011.
782 mRNA helicases: the tacticians of translational control. Nat. Rev. Mol. Cell Biol. 12, 235–
783 245. doi:10.1038/nrm3083
784 Pugh TJ, Weeraratne SD, Arche, TC,PomeranzKrummel DA, Auclair D, Bochicchio J, Carneiro
785 MO, Carter SL, Cibulskis K, Erlich RL, Greulich H, Lawrence MS, Lennon NJ, McKenna A,
786 Meldrim J, Ramos AH, Ross MG, Russ C, Shefler E, Sivachenko A, Sogoloff B, Stojanov P,
35
Gao et al.
787 Tamayo P, Mesirov JP, Amani V, Teider N, Sengupta S, Francois JP, Northcott PA,
788 TaylorMD, Yu F, Crabtree GR, Kautzman AG, Gabriel SB, Getz G, Jäger N, Jones DTW,
789 Lichter P, Pfister SM, Roberts TM, Meyerson M, Pomeroy SL, Cho Y-J. 2012.
790 Medulloblastoma exome sequencing uncovers subtype-specific somatic mutations.
791 Nature488, 106–110. doi:10.1038/nature11329
792 Putnam, AA, Gao Z, Liu F, Jia H, Yang Q, Jankowsky E. 2015. Division of Labor in an Oligomer
793 of the DEAD-Box RNA Helicase Ded1p. Mol. Cell59, 541–552.
794 doi:10.1016/j.molcel.2015.06.030
795 Putnam AA, Jankowsky E. 2013. AMP sensing by DEAD-box RNA helicases. J. Mol. Biol. 425,
796 3839–3845. doi:10.1016/j.jmb.2013.05.006
797 Rajagopal V, Park E-H,Hinnebusch AG, Lorsch JR. 2012. Specific domains in yeast translation
798 initiation factor eIF4G strongly bias RNA unwinding activity of the eIF4F complex toward
799 duplexes with 5’-overhangs. J. Biol. Chem. 287, 20301–20312.
800 doi:10.1074/jbc.M112.347278
801 Schütz P, Bumann M, Oberholzer AE, Bieniossek C, Trachsel H, Altmann M, Baumann U.
802 2008. Crystal structure of the yeast eIF4A-eIF4G complex: an RNA-helicase controlled by
803 protein-protein interactions. Proc. Natl. Acad. Sci. U.S.A.105, 9564–9569.
804 doi:10.1073/pnas.0800418105
805 Seidel SAI, Dijkman PM, Lea WA, van den Bogaart G, Jerabek-Willemsen M, Lazic A, Joseph
806 JS, Srinivasan P, Baaske P, Simeonov A, Katritch I, Melo FA, Ladbury JE, Schreiber G,
807 Watts A, Braun D, Duhr S. 2013. Microscale thermophoresis quantifies biomolecular
808 interactions under previously challenging conditions. Methods59, 301–315.
809 doi:10.1016/j.ymeth.2012.12.005
810 Senissar M, Le Saux A, Belgareh-Touzé N, Adam C, Banroques J, Tanner NK. 2014. The
811 DEAD-box helicase Ded1 from yeast is an mRNP cap-associated protein that shuttles
36
Gao et al.
812 between the cytoplasm and nucleus. Nucleic Acids Res. 42, 10005–10022.
813 doi:10.1093/nar/gku584
814 Sen ND, Zhou F, Ingolia NT, Hinnebusch AG. 2015. Genome-wide analysis of translational
815 efficiency reveals distinct but overlapping functions of yeast DEAD-box RNA helicases
816 Ded1 and eIF4A. Genome Res. 25, 1196–1205. doi:10.1101/gr.191601.115
817 Sharma D, Jankowsky E. 2014. The Ded1/DDX3 subfamily of DEAD-box RNA helicases. Crit.
818 Rev. Biochem. Mol. Biol. 49, 343–360. doi:10.3109/10409238.2014.931339
819 Shih J-W, Tsai T-Y, Chao C-H, Wu Lee Y-H. 2008. Candidate tumor suppressor DDX3 RNA
820 helicase specifically represses cap-dependent translation by acting as an eIF4E inhibitory
821 protein. Oncogene27, 700–714. doi:10.1038/sj.onc.1210687
822 Soto-Rifo R, Rubilar PS, Limousin T, de Breyne S, Décimo D, Ohlmann T. 2012. DEAD-box
823 protein DDX3 associates with eIF4F to promote translation of selected mRNAs. EMBO J.
824 31, 3745–3756. doi:10.1038/emboj.2012.220
825 Valiente-Echeverría F, Hermoso MA, Soto-Rifo R. 2015. RNA helicase DDX3: at the crossroad
826 of viral replication and antiviral immunity. Rev. Med. Virol. 25, 286–299.
827 doi:10.1002/rmv.1845
828 Yang Q, Del Campo M,Lambowitz AM, Jankowsky E. 2007a. DEAD-box proteins unwind
829 duplexes by local strand separation. Mol. Cell28, 253–263.
830 doi:10.1016/j.molcel.2007.08.016
831 Yang Q, Fairman ME, Jankowsky E. 2007b. DEAD-box-protein-assisted RNA structure
832 conversion towards and against thermodynamic equilibrium values. J. Mol. Biol. 368, 1087–
833 1100. doi:10.1016/j.jmb.2007.02.071
834 Yang Q, Jankowsky E. 2005. ATP- and ADP-dependent modulation of RNA unwinding and
835 strand annealing activities by the DEAD-box protein DED1. Biochemistry44, 13591–13601.
836 doi:10.1021/bi0508946
37
Gao et al.
837 Zenklusen D, Larson DR, Singer RH. 2008. Single-RNA counting reveals alternative modes of
838 gene expression in yeast. Nat. Struct. Mol. Biol. 15, 1263–1271. doi:10.1038/nsmb.1514
38
Gao et al.
839 FIGURE CAPTIONS
840
841 Figure 1
842 Direct interactions between Ded1p, eIF4A and eIF4G.
843 (A) Effect of EIF4A (TIF1,TIF2) overexpression (o/e) on growth of the DED1-tam allele. Yeast
844 with either wild type DED1 (WT), or DED1-tam (tam) as sole copy of DED1 were transformed
845 with a high copy plasmid containing either no insert (none) or TIF1, TIF2, or DED1. Serial
846 dilutions of each strain were spotted at identical concentrations on selective media and grown at
847 16°C.
848 (B) Impact of increasing concentrations of the eIF4A inhibitor hippuristanol on growth of yeast
849 with either wild type DED1 (WT), or the temperature sensitive ded1-95 (Ded1pT408I, Figure 1-
850 figure supplement 1) allele as the sole copy. Growth rates were measured at 37°C and
851 normalized to the growth rate without inhibitor. Error bars represent one standard deviation of
852 three independent measurements.
853 (C) Pull-down of purified eIF4A, eIF4G, or both, with immobilized HA-tagged Ded1p. Reactions
854 were performed in the presence of RNases. EIF4G contained an S-tag. Input (IN) lanes show 3%
855 of total reaction volume. Pull-down (PD) lanes show entire sample. Relative band intensities
856 (pull-down compared to input): reaction 1, Ded1p: 460% ± 18%, eIF4G: 31% ± 12%; reaction 2:
857 Ded1p: 413% ± 58%, eIF4A: 96% ± 55%; reaction 3, Ded1p: 592% ± 113%, eIF4G: 74% ± 9%,
858 eIF4A: 25% ± 21%. Errors represent one standard deviation of multiple independent
859 experiments.
860 (D) Pull-down of purified eIF4A, Ded1p, or both, with immobilized, S-tagged eIF4G. Reactions
861 were performed in the presence of RNases. Input (IN) lanes show 3% of total reaction volume.
862 Pull-down (PD) lanes show entire sample. Relative band intensities (pull-down compared to
863 input): reaction 5, eIF4G: 440% ± 113%, Ded1p: 14% ± 7%; reaction 6: eIF4G: 280% ± 29%,
39
Gao et al.
864 eIF4A: 3% ± 2%; reaction 7, eIF4G: 473% ± 27%, Ded1p: 33% ± 11%, eIF4A: 15% ± 5%. Errors
865 represent one standard deviation of multiple independent experiments.
866 (E) Unwinding activity of Ded1p, eIF4A, with and without eIF4G, and in combination.
867 Representative unwinding reaction with 0.5 nM RNA duplex (16 bp, 25 nt 3’-overhang, for
868 sequences see Supplementary file 1A), 0.1 µM Ded1p, 0.1 µM eIF4G, 2 µM eIF4A, 4 mM ATP
869 (0.2 % v/v DMSO). Lines represent the best fit to the integrated first order rate law.
870
871 See also Figure 1-figure supplement 1 and Supplementary file 1A.
872
873
874 Figure 1-figure supplement 1
T408I 875 Ded1p is deficient in RNA binding.
T408I ATP (A) ATP affinity of Ded1p and WT Ded1p. Apparent ATP affinities (K ) of WT Ded1p and 876 1/2
T408I 877 Ded1p for ATP at 30°C and 37°C were measured through RNA-stimulated ATPase activity
878 of the proteins with the substrate containing a 16 bp duplex and 25 nt unpaired 3' overhang
ATP (Putnam and Jankowsky, 2013). Values for K were calculated by fitting ATP titrations to 879 1/2
880 the MichaelisMenten model (Putnam and Jankowsky, 2013). kcat for ATP did not notably differ
T408I 881 between Ded1p and WT Ded1p (data not shown).
T408I RNA (B) RNA affinity for Ded1p andWT Ded1p. Apparent RNA affinities (K ) of WT Ded1p 882 1/2
T408I 883 and Ded1p for RNA at 30°C and 37°C were measured through RNA-stimulated ATPase
RNA activity of the proteins as in panel A. Values for K were calculated by fitting RNA titrations 884 1/2
885 to the Michaelis Menten model (Putnam and Jankowsky, 2013). The data indicate an RNA
T408I 886 binding defect in Ded1p .
887
40
Gao et al.
888 Figure 2
889 Ded1p is the unwinding module in complexes with eIF4A and eIF4G.
890 (A) Impact of hippuristanol on unwinding activity of Ded1p, eIF4A, with and without eIF4G, and
891 in combination. Reactions were performed as in Figure 1E with 0.2% (v/v) DMSO and 20 μM
Hippuristanol, as indicated. Left panel: Observed unwinding rate constants (k ). For 892 unw
893 representative unwinding reactions see Figure 2-figure supplement 1. Error bars represent
894 one standard deviation of at least three independent measurements. Right panel: decrease or
increase of k by hippuristanol. 895 unw
E171A 896 (B) Impact of eIF4A (100 nM) on unwinding activity of Ded1p (100 nM) with or without
897 eIF4G (100 nM), measured as in Figure 1E, but without DMSO. Left panel: Observed
unwinding rate constants (k ). Error bars represent one standard deviation of at least three 898 unw
independent measurements. Right panel: increase of k , compared to reaction without 899 unw
E171A 900 eIF4A .
E307A 901 (C) Impact of Ded1p on unwinding activity of eIF4A with or without eIF4G, measured as in
Figure 2B. Left panel: Observed unwinding rate constants (k ). Error bars represent one 902 unw
standard deviation of at least three independent measurements. Right panel: increase of k , 903 unw
E307A 904 compared to reaction without Ded1p .
T408I 905 (D) Impact of eIF4A on unwinding activity of Ded1p with or without eIF4G, measured as in
Figure 2B. Left panel: Observed unwinding rate constants (k ). Error bars represent one 906 unw
standard deviation of at least three independent measurements. Right panel: increase of k , 907 unw
908 compared to reaction without eIF4A.
909
910 See also Figure 2-figure supplement 1
911
41
Gao et al.
912
913 Figure 2-figure supplement 1
914 Reaction progress curves for Figure 2A.
915 Representative unwinding reactions for Figure 2A, with (A) Ded1p, (B) Ded1p and eIF4A, (C)
916 Ded1p, eIF4A, and eIF4G, (D) Ded1p and eIF4G, (E) eIF4A, (F) eIF4A and eIF4G. In each
917 panel, open shapes represent reactions without hippuristanol; filled shapes represent reactions
918 with hippuristanol. Lines represent the best fit to the integrated first order rate law.
919
920
921 Figure 3
922 The N‐terminus of Ded1p is critical for the interaction with eIF4A.
1-535 923 (A) Impact of eIF4A on unwinding activity of Ded1p (C‐terminus deleted, upper panels) and
117-604 924 Ded1p (N‐terminus deleted, lower panels) with or without eIF4G, measured as in Figure
2B. Left panel: Observed unwinding rate constants (k ). Error bars represent one standard 925 unw
deviation of at least three independent measurements. Right panel: increase of k , compared 926 unw
927 to reactions with Ded1p alone.
117-604 928 (B) Effect of the deletion of the N‐terminus of Ded1p (Ded1p ), without or with EIF4A (TIF1)
929 overexpression (o/e), on growth in yeast at 30°C and 16°C.
930 (C) Formaldehyde crosslinking of Ded1p without and with eIF4A in the presence of RNA (16 bp,
931 25 nt 3’‐ overhang) and ADPNP. Mobilities of the Ded1p trimer, dimer, monomer, and the
932 Ded1p‐eIF4A complex are indicated. Proteins were visualized by western blot (α‐Ded1p).
933 (D) Schematic model for the basic functional architecture of Ded1p‐eIF4A and
934 Ded1p‐eIF4A‐eIF4G complexes. Ovals represent the helicase cores of Ded1p (red) and eIF4A
42
Gao et al.
935 (green). C‐ and N‐termini of Ded1p and eIF4G are marked. The dotted lines between the N-
936 termini in the Ded1p trimer indicate interactions between the N-termini that contribute to
937 oligomrization.
938
939 See also Figure 3-figure supplement 1 and Supplementary file 1A.
940
941
942 Figure 3-figure supplement 1
943 Direct interaction between eIF4A and the amino terminus of Ded1p.
117-604 1-535 944 (A) Pull-down of purifed Ded1p or Ded1p , with immobilized GST-eIF4A.
117-604 1-535 945 (1500 nM GST-eIF4A or GST, 1500 nM Ded1p or Ded1p , and 20 µg/mL RNases A at
946 8°C). Ded1p was probed with the α-Ded1p antibody. Deletion of the N-terminus eliminates the
947 interaction between Ded1p and eIF4A. Input (IN) lanes show 10% of total reaction volume. Pull-
948 down (PD) lanes show entire sample. The relative band intensities (pull-down compared to input)
949 for the reaction with GST-eIF4A and Ded1p1-535: GST-eIF4A: 160% ± 17%, Ded1p1-535: 3% ± 1%.
950 Errors represent one standard deviation of multiple independent experiments. No signal for
951 Ded1p117-604 was detected in the pull-down.
952 (B) Formaldehyde crosslinking of Ded1p, with increasing concentrations of eIF4A, in the
953 presence of RNA (16 bp, 25 nt 3’‐ overhang) and ADPNP. Mobilities of the Ded1p trimer, dimer,
954 monomer, and the Ded1p‐eIF4A complex are indicated. Proteins were visualized by western
955 blot (α‐Ded1p). Increasing concentrations of eIF4A diminish the Ded1p trimer and dimer species.
956 (C) Formaldehyde crosslinking of Ded1p1-535 without and with eIF4A in the presence of RNA
957 (16bp, 25 nt 3’-overhang) and ADPNP. This experiment was performed to directly verify the
958 weak interaction between Ded1p and eIF4A detected by pull-down in panel (A). Deletion of C-
959 terminus of Ded1p diminishes formation of the Ded1p trimer (Putnam et al., 2015). Mobilities of
43
Gao et al.
960 the Ded1p dimer, monomer, and the Ded1p-eIF4A complex are indicated. Ded1p was visualized
961 by western blot (α-Ded1p).
962
963
964 Figure 4
965 Thermodynamic framework for the interactions between Ded1p, eIF4A and eIF4G.
966 (A) Graphic representation of the framework. Ovals represent the complexes containing Ded1p
967 (D), eIF4A (A), eIF4G (G), RNA (R), ATP (T). D3 indicates the Ded1p timer (Putnam et al.,
968 2015). Black lines mark a transition from one complex to another through binding of one of the
components. Colored squares indicate the calculated equilibrium dissociation constant (K ) for 969 1/2
970 the transition, as shown in the legend at the lower right. For visual reasons, the framework does
971 not depict specific states for ATP-bound eIF4A. Equilibrium constants for all individual
972 transitions are listed in Supplementary file 2.
973 (B) Observed apparent unwinding rate constants versus those calculated with the model in
974 panel (A). Measured and calculated rate constants from different reaction conditions are listed
975 without reference to the specific reaction conditions, solely to visualize the overall agreement
976 between observed and calculated values. For fitting parameters see Figure 4-figure
977 supplement 5 and Supplementary file 2. The black line represents a diagonal, R2 is the
978 correlation coefficient for a linear fit of the shown data to this line. Error bars mark the 95%
979 confidence interval for the observed values.
980
981 See also Figure 4-figure supplements 1-6 and Supplementary files 2,3.
44
Gao et al.
982 Figure 4-figure supplement 1
983 Impact of the Ded1p concentration on the modulation of its unwinding activity by eIF4A
984 and eIF4G.
985 All unwinding reactions were performed as in Figure 2B at the indicated Ded1p, eIF4A and
eIF4G, and ATP concentrations. Apparent rate constants (k ) represent the mean of at least 986 unw
987 three independent measurements. Error bars indicate one standard deviation.
988 (A‐C) Upper panels: Impact of eIF4A (A), eIF4A and eIF4G (B) and eIF4G alone (C) at
989 increasing Ded1p concentration at 4 mM ATP. Lines represent the best fit to a binding isotherm.
Note the different scale of the y‐axis in (C). Lower panels: Fold increase or decrease of k at 990 unw
991 each Ded1p concentration over the reaction with Ded1p alone.
992 (D) Hill coefficients calculated from Ded1p binding isotherms (A-C) in the presence of the
993 indicated combinations of eIF4G and eIF4A.
994
995
996 Figure 4-figure supplement 2
997 Impact of eIF4A concentration on the modulation of Ded1p unwinding activity by eIF4A
998 and eIF4G.
999 All unwinding reactions were performed as in Figure 2B at the indicated Ded1p, eIF4A and
eIF4G, and ATP concentrations. Apparent rate constants (k ) represent the mean of at least 1000 unw
1001 three independent measurements. Error bars indicate one standard deviation.
1002 (A‐C) Impact of increasing concentrations of eIF4A at fixed Ded1p concentration (A), and fixed
1003 Ded1p and two different eIF4G concentrations (B,C) at 4 mM ATP. Lines represent the best fit
1004 to a binding isotherm.
45
Gao et al.
1005 (D) Impact of increasing concentrations of eIF4A on unwinding in the absence and presence of
1006 eIF4G, without Ded1p. The corresponding curve in the presence of only Ded1p is shown as
1007 dotted line for reference.
1008
1009
1010 Figure 4-figure supplement 3
1011 Impact of eIF4G concentration on the modulation of Ded1p unwinding activity by eIF4A
1012 and eIF4G.
1013 All unwinding reactions were performed as in Figure 2B at the indicated Ded1p, eIF4A and
eIF4G, and ATP concentrations. Apparent rate constants (k ) represent the mean of at least 1014 unw
1015 three independent measurements. Error bars indicate one standard deviation.
1016 (A‐C) Impact of increasing concentrations of eIF4G at fixed Ded1p concentration at 4 mM ATP,
1017 and increasing concentrations of eIF4A, as indicated. Lines represent a trend, dashed lines
1018 mark the maximal observed unwinding rate constant.
1019 (D) Impact of increasing concentrations of eIF4G on Ded1p (100 nM, no eIF4A) activity in the
1020 presence of 4 mM ATP. Note the scale of y and x-axis.
1021
1022
1023 Figure 4-figure supplement 4
1024 Impact of ATP concentration on the modulation of Ded1p unwinding activity by eIF4A
1025 and eIF4G.
1026 All unwinding reactions were performed as in Figure 2B at the indicated Ded1p, eIF4A and
eIF4G, and ATP concentrations. Apparent rate constants (k ) represent the mean of at least 1027 unw
1028 three independent measurements. Error bars indicate one standard deviation.
46
Gao et al.
1029 (A‐C) Upper panels: Impact of ATP on reactions with fixed concentrations of Ded1p and eIF4A
1030 (A), Ded1p, eIF4A, and eIF4G (B), and Ded1p and eIF4G (C). Lines represent the best fit to a
1031 binding isotherm. Note the different scale of the y‐axis in (C). Lower panels: Fold increase or
decrease of k by eIF4A, eIF4G, or both, over the reaction with Ded1p alone. 1032 unw
1033 (D) Upper panels: Impact of ATP on the unwinding reaction with Ded1p and the ATP-binding
E171A 1034 deficient eIF4A (black solid line, open diamonds). The reactions with Ded1p and WT eIF4A
1035 (dotted line, filled diamonds) (A) and with Ded1p alone (dotted line, no symbols) (A) are shown
E171A as reference. Lower panels: Fold increase or decrease of k by WT eIF4A, or eIF4A , over 1036 unw
1037 the reaction with Ded1p alone. (E) Upper panels: Impact of ATP on the unwinding reaction with
E171A 1038 Ded1p, eIF4G and the ATP-binding deficient eIF4A (black line, open squares). The
1039 reactions with Ded1p, eIF4G and WT eIF4A (dotted line, filled squares) (B) and with Ded1p
1040 alone (dotted line, no symbols) (B) are shown as reference. Lower panels: Fold increase or
E171A decrease of k by eIF4G with WT eIF4A or eIF4A , over the reaction with Ded1p alone. 1041 unw
1042 (E) Upper panels: Impact of ATP on the unwinding reaction with Ded1p, eIF4G and the ATP-
E171A 1043 binding deficient eIF4A (black line, open squares). The reactions with Ded1p, eIF4G and
1044 WT eIF4A (dotted line, filled squares) (B) and with Ded1p alone (dotted line, no symbols) (B)
are shown as reference. Lower panels: Fold increase or decrease of k by eIF4G with WT 1045 unw
E171A 1046 eIF4A or eIF4A , over the reaction with Ded1p alone.
1047
1048
1049 Figure 4-figure supplement 5
1050 Relative Chi square values (Χ2) for variations of two representative parameters.
1051 Left panel: Ded1p binding to pre-formed eIF4A-eIF4G complex; right panel: ATP binding to
1052 Ded1p-eIF4G-eIF4A-RNA complex) . The two examples were chosen to illustrate a case where
47
Gao et al.
1053 both, increase and decrease in the parameter (K1/2) relative to the optimal value results in a
1054 poorer fit (left panel), and for a case where only a decrease of the parameter, but not an
1055 increase result in a poorer fit (right fit). Cases for the latter scenario are restricted to ATP
1056 binding and marked in the Supplementary file 2 with an ">" sign. Relative Χ2 represent the Χ2
1057 obtained for the entire thermodynamic model for the parameter on the X axis divided by the
1058 smallest (optimal) Χ2. For the optimal parameter, the relative Χ2 = 1. Dotted lines mark the 95%
1059 confidence interval. This interval is given for each parameter in the Supplementary file 2.
1060 Values for Χ2 represent the quality of the fit for the varied parameter to the entire
1061 thermodynamic model (Figure 4A).
1062
1063
1064 Figure 4-figure supplement 6
1065 The thermodynamic model adequately describes observed data for each subset of
1066 complexes.
Correlation between observed and calculated unwinding rate constants (k ) for subsets of the 1067 unw
2 1068 complexes, as indicated. R is the correlation coefficient of a linear fit of the data in each panel
1069 to a diagonal (black line). Error bars mark the 95% confidence interval for the calculated data.
1070 For fitting parameters see Supplementary file 2.
1071
1072
1073
1074
48
Gao et al.
1075 Figure 5
1076 Functional parameters of complexes with Ded1p, eIF4A, or both.
1077 (A) Free energies ΔG° for complexes formed between Ded1p, eIF4A, eIF4G, RNA (25 nt 3‘-
1078 overhang, 16 bp duplex) and ATP in the combinations shown, calculated with the model in
1079 Figure 4A.
1080 (B) MST measurement of Ded1p binding to eIF4A-eIF4G (30 nM eIF4G, 36 nM Cy5-labeled
1081 eIF4A, increasing concentrations of Ded1p as indicated, 25 μg/mL RNase A, 23°C). Datapoints
1082 show the change in fluorescence signal at the indicated Ded1p concentrations, normalized to
1083 the signal without Ded1p. Error bars indicate one standard deviation of multiple independent
1084 measurements. Datapoints were fitted with the quadratic form of the binding isotherm (K'1/2 < 6.6
1085 nM).
(max) (C) Unwinding rate constants (k , ATP and protein saturation, RNA: 25 nt 3' overhang, 16 1086 unw
(RNA) bp duplex) and RNA affinity (K , ATP saturation) for complexes formed between Ded1p, 1087 1/2
1088 eIF4A, eIF4G, and, for comparison, for Ded1p and eIF4A, calculated with the model in Figure
1089 4A.
1090 (D) Unwinding rate constants (RNA: 25 nt 3' overhang, 16 bp duplex) for the Ded1p timer, the
1091 Ded1p‐eIF4A, and the Ded1p‐eIF4A‐eIF4G complex as function of the Ded1p concentration
1092 (ATP saturation, eIF4 A = 2000 nM, eIF4G = 100 nM), calculated with the model in Figure 4A.
1093
1094 See also Figure 5-figure supplement 1.
1095
1096
1097 Figure 5-figure supplement 1
1098 ATPase activity of complexes formed between Ded1p, eIF4A, and eIF4G at RNA and ATP
1099 saturation.
49
Gao et al.
1100 ATPase reactions were performed with Ded1p (100nM), eIF4A (2000 nM), eIF4G (100 nM), as
1101 indicated, and with RNA (16bp, 25nt 3’-overhang, 40 µM) and ATP (4 mM). Initial rates for the
1102 ATPase reactions (V0) under these conditions represent ATP turnover numbers (equivalent to
1103 kcat). Values were measured as described (Putnam and Jankowsky, 2013).
1104
1105
1106 Figure 6
1107 Impact of the substrate architecture on the functional parameters of complexes with 1108 Ded1p, eIF4A, and eIF4G
1109 (A) Impact of eIF4A, eIF4G, or both, on the unwinding activity of Ded1p with 5' and 3' tailed
1110 substrates (16 bp, 25 nt 3’- or 5’-tail), and blunt-end, shown schematically at the left (numbers
1111 indicate nucleotides in unpaired overhang and basepairs in duplex region). Reactions were
performed as in Figure 2B. Sliding point graphs show observed unwinding rate constants (k , 1112 unw
1113 average of at least three independent measurements), error bars mark one standard deviation.
Bar graphs mark increase or decrease of k , compared to the reaction with only Ded1p for 1114 unw
1115 each substrate.
(B) Upper panel: Affinities (K ) for RNA substrates with 10 nt and 25 nt unpaired regions 3' to 1116 1/2
1117 the duplex (16 bp) for Ded1p, eIF4A, eIF4G alone and in combination as indicated, calculated
1118 with the model shown in Figure 4A, and with data obtained with two substrates containing
1119 either 10 or 25 nt overhangs. For modeling parameters see Figure 6-figure supplements 2-3,
Supplementary file 2 and Materials and Methods. Lower panel: Fold change in K between 1120 1/2
1121 the substrates with 10 and 25 nt unpaired regions.
1122 (C) Upper panel: Apparent unwinding rate constants (kunw, ATP = 4 mM, eIF4A = 100 nM, eIF4G
1123 = 100 nM) for Ded1p alone and for the Ded1p‐eIF4A‐eIF4G complex on RNA substrates with 10
1124 or 25 nt overhangs, 3' to a 16 bp duplex, as function of increasing Ded1p concentration,
1125 calculated with the model shown in Figure 4A, and with data obtained with two substrates
50
Gao et al.
containing either 10 or 25 nt overhangs. Lower panel: Fold change in k for reactions with 1126 unw
1127 Ded1p‐eIF4A‐eIF4G, compared to reactions with Ded1p alone for both substrates.
1128 (D) Unwinding rate constants at ATP and protein saturation for RNA substrates with 25 nt 3'
1129 overhangs and 13, 16, and 19 bp duplexes, for Ded1p, the Ded1p‐eIF4A complex and the
1130 Ded1p‐eIF4A‐eIF4G complex, calculated with the model shown in Figure 4.
1131
1132 For modeling parameters see Figure 4-figure supplements 5,6; Figure 6-figure supplement
1133 2, Supplementary files 2,3 and Methods.
1134
1135 See also Figure 6-figure supplements 1,3
1136
1137
1138 Figure 6-figure supplement 1
1139 Stimulation of unwinding activity of eIF4A by eIF4G with an RNA with a 5' tail.
1140 Unwinding reactions were performed in buffer used for all other unwinding reactions at 26°C
1141 with 100 nM eIF4A, 300nM eIF4G, 3 mM ATP, 10 nM RNA (13 bp, 25 nt overhang, 5' to duplex)
1142 and 3 μM unlabeled top strand (13 nt), made from DNA. For sequences see Supplementary
1143 file 1A.
1144
1145
1146 Figure 6-figure supplement 2
1147 Substrate architecture affects the modulation of Ded1p activity by eIF4A and eIF4G.
1148 (A‐I) Impact of eIF4A, eIF4G, or both on the unwinding activity of Ded1p with different
1149 substrates, shown schematically at the left (numbers indicate nucleotides in unpaired overhang
1150 and basepairs in duplex region). Reactions were performed as in Figure 2B. Sliding point
51
Gao et al.
graphs show observed unwinding rate constants (k , average of at least three independent 1151 unw
1152 measurements), error bars mark one standard deviation. Bar graphs show increase or decrease
of k , compared to the reaction with only Ded1p for each substrate. 1153 unw
1154
1155
1156 Figure 6-figure supplement 3
Correlation between observed and calculated unwinding rate constants (k ) for the 16 1157 unw
1158 bp substrate with 10 nt 3' tail.
2 1159 R is the correlation coefficient for the linear fit of the data to a diagonal (black line). Error bars
1160 mark the 95% confidence interval for the calculated data.
1161
1162
1163 Figure 7.
1164 Calculated prevalence of complexes formed between Ded1p, eIF4A and eIF4G at
1165 physiological protein concentrations. Fraction of Ded1p (left) and eIF4G (right) present in the
1166 various complexes with eIF4A, eIF4G, or both, at three different Ded1p and eIF4G
1167 concentrations around their reported physiological concentrations of ~ 1 μM (RNA = 6 μM, ATP
1168 = 2.5 mM). Relative abundance of the complexes were calculated with the binding parameters
1169 obtained from the thermodynamic framework (Figure 4A, Supplementary file 2).
1170
1171 See also Figure 4 and Supplementary file 2.
52