• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. Mar 4, 2008; 105(9): 3274–3279.
Published online Feb 22, 2008. doi:  10.1073/pnas.0712235105
PMCID: PMC2265113

PDCD4 inhibits translation initiation by binding to eIF4A using both its MA3 domains


Programmed Cell Death 4 (PDCD4) is a protein known to bind eukaryotic initiation factor 4A (eIF4A), inhibit translation initiation, and act as a tumor suppressor. PDCD4 contains two C-terminal MA3 domains, which are thought to be responsible for its inhibitory function. Here, we analyze the structures and inhibitory functions of these two PDCD4 MA3 domains by x-ray crystallography, NMR, and surface plasmon resonance. We show that both MA3 domains are structurally and functionally very similar and bind specifically to the eIF4A N-terminal domain (eIF4A-NTD) using similar binding interfaces. We found that the PDCD4 MA3 domains compete with the eIF4G MA3 domain and RNA for eIF4A binding. Our data provide evidence that PDCD4 inhibits translation initiation by displacing eIF4G and RNA from eIF4A. The PDCD4 MA3 domains act synergistically to form a tighter and more stable complex with eIF4A, which explains the need for two tandem MA3 domains.

Keywords: apoptosis, eIF4G, protein, NMR, x-ray crystallography

PDCD4 is a tumor-suppressor protein that is up-regulated on induction of apoptosis (1) and down-regulated in certain aggressive tumors (2). PDCD4 is controlled by protein kinase S6K1 and the ubiquitin ligase SCFβTRCP, and its degradation is necessary for efficient protein translation in vivo, which is a prerequisite for cell growth and, consequently, for cell proliferation (3).

PDCD4 is known to bind two eukaryotic translation initiation factors eIF4A and eIF4G (46). eIF4A is an RNA helicase that works as a subunit of eIF4F, a complex composed of eIF4G and eIF4E. The helicase activity of eIF4A itself is weak but is enhanced upon binding to eIF4G (7, 8). eIF4G has two independent binding sites for eIF4A (9), one in the conserved middle domain (eIF4G-m, HEAT1/MIF4G) (Fig. 1), and the other in the adjacent second HEAT domain (eIF4G-MA3, HEAT2/MA3) (reviewed in ref. 10). NMR binding studies have shown that eIF4G-m interacts mainly with the C-terminal domain of eIF4A (eIF4A-CTD) (11), whereas eIF4G-MA3 binds to the N-terminal domain of eIF4A (eIF4A-NTD) and only weakly to eIF4A-CTD (A.M., C.S., K.A.E., and G.W., unpublished work). Mutation and deletion analysis indicates that the interaction of eIF4A with eIF4G-m is necessary for translation, whereas the interaction of eIF4A with eIF4G-c (eIF4G-MA3+HEAT3 domain) plays a modulatory role (12).

Fig. 1.
Domain architectures, sequence alignments, and the crystal structure of PDCD4 MA3-m. (A) Schematic diagram showing the domain architectures of the proteins used in this study. (B) Alignment of PDCD4 MA3-m, MA3-c, and eIF4G-MA3. Amino acids that are highly ...

PDCD4 contains two MA3 domains after an N-terminal segment of little known function (1). MA3 is a well conserved α-helical motif with typically 3–5 helical hairpins and is a subtype of HEAT domains. A single MA3 domain is found in eIF4G-c (eIF4G-MA3, Fig. 1), which has been reported to also bind eIF4A (10, 13). Recently, crystal and solution structures of mouse PDCD4's C-terminal MA3 domain (MA3-c) were reported (14, 15). MA3-c was shown to bind eIF4A, compete with eIF4G-c, and was sufficient to inhibit translation initiation. However, some questions remained. First, the function of the N-terminal MA3 domain (MA3-m) was unclear. Mutations of conserved amino acid residues in either MA3-c or MA3-m affect eIF4A binding, which implied that both PDCD4 MA3 domains have eIF4A-binding abilities and contribute in translation inhibition (16). NMR experiment revealed that MA3-c binds to eIF4A-NTD through the loop between α5 and α6 and the turn linking α3 and α4 (15), which is well conserved among MA3-c, MA3-m, and the eIF4G-MA3, also supporting a direct role for the MA3-m domain in eIF4A binding and inhibition. Despite these observations, however, the exact function of MA3-m remained unclear. Second, no structure was available for MA3-m. The sequence homology between MA3-m and MA3-c suggests a similar fold but the extent of similarity was not clear. Furthermore, it was open whether the two domains have similar or complementary function, and why PDCD4 needs two MA3 domains.

Here, we report the crystal structure of the PDCD4 MA3-m domain and its NMR-derived eIF4A-binding face. We show that the PDCD4 MA3 domains compete with the eIF4G-MA3 and with RNA for binding to eIF4A. Both the structures of the two MA3 domains and their eIF4A-binding faces are very similar. However, the two PDCD4 MA3 domains act synergistically to form a tighter and more stable complex with eIF4A, which explains the need for two tandem MA3 domains.


MA3-m Is Structurally Similar to MA3-c but Contains an Additional C-Terminal Helix.

A series of truncation mutations were carried out to identify the domain boundaries of human PDCD4 MA3-m. The region 157–302 (Fig. 1) exhibited the best NMR spectra and was thus assumed to contain the intact MA3-m domain. Backbone resonances for 142 of 158 nonprolines could be assigned by using triple-resonance NMR experiments (reviewed in ref. 17). Similarly, 114 of 127 nonproline residues of human PDCD4 MA3-c were assigned. A few short segments are invisible in the NMR spectra presumably because of line broadening due to conformational exchange. Chemical shift analysis indicates that MA3-m contains eight helices [supporting information (SI) Fig. 6A].

Both native and Se-Met MA3-m were crystallized. MA3-m (Se-Met) crystals belonged to space group P212121 with cell axes a = 37.68 Å, b = 70.06 Å, c = 110.88 Å, α = β = γ = 90° with two monomers per asymmetric unit (SI Table 1). The structure was solved at 1.7 Å resolution and is shown in Fig. 1. The domain consists of a stack of four α-helical hairpins, displaying a similar conformation as PDCD4 MA3-c (14) and eIF4G-MA3 (18). The crystal structure reveals that the folded part of MA3-m extends between residues 161 (α1) and 302 (α8) and is ≈20 residues longer than previously predicted by analogy to MA3-c (14). The C-terminal 18 residues of the MA3-m fragment (after α8) have high amino acid sequence homology to helix α9 in eIF4G-MA3. However, this region was not visible in the electron density and also appears unstructured in the NMR data (SI Fig. 6A).

Both PDCD4 MA3 Domains Bind to eIF4A-NTD but Not to eIF4A-CTD.

To test for the interaction between the two PDCD4 MA3 domains (MA3-m and MA3-c) and eIF4A domains (eIF4A-NTD and eIF4A-CTD), four separate NMR titration assays were performed. [15N,1H]TROSY-HSQC spectra were recorded for uniformly 15N-labeled MA3-m or MA3-c. Increasing amounts of nonlabeled eIF4A-NTD or eIF4A-CTD were added to each MA3 domain, and the backbone chemical shift changes were monitored. Distinct changes were observed for both MA3-m and MA3-c when eIF4A-NTD was added (Fig. 2 A and B). No significant changes in the peak positions were observed when eIF4A-CTD was added to either MA3-m or MA3-c. (Fig. 2 C and D). This indicates that both PDCD4 MA3 domains bind specifically to eIF4A-NTD but not to eIF4A-CTD.

Fig. 2.
Both PDCD4 MA3 domains bind to eIF4A-NTD but not to eIF4A-CTD. (A and C). Overlays of [15N,1H]TROSY-HSQC spectra of 15N-labeled MA3-m (150 μM) with increasing amount of nonlabeled eIF4A-NTD (A) or eIF4A-CTD (C). Concentrations of eIF4A-NTD/CTD ...

Both MA3-m and MA3-c Bind to eIF4A-NTD with a Similar Surface.

Chemical shift changes of backbone amide signals of PDCD4 MA3 domains upon adding eIF4A-NTD are summarized in Fig. 3A. The final spectra obtained in the NMR titration assays (MA3-m/-c–eIF4A-NTD ratio 1:2) were used for data analysis. To visualize the eIF4A-binding surface on both MA3 domains, we mapped the residues with distinct chemical shift changes onto the respective crystal structures. Affected residues are localized primarily in two adjacent regions, forming a contiguous surface (Fig. 3 B and C). It is located in the bent portion of helix α5 and the following loop between α5 and α6. The second region is located in the end of α3 and the connection to α4. Thus, MA3-m and MA3-c have similar binding surfaces for eIF4A-NTD. A few of the residues that showed significant chemical shift changes are located in the middle portion of α5 and α6 (L243, F264, R267 for MA3-m, I412, E431, E432 for MA3-c), which presumably reflects slight conformational changes in α5 and α6 upon eIF4A-NTD binding. MA3-m has an additional small group of residues in the loop between α7 and α8 that were affected by eIF4A-NTD binding (K283, G284). This will be further analyzed in Discussion.

Fig. 3.
Interfaces of PDCD4-MA3 domains for eIF4A-NTD binding. (A) Normalized chemical shift differences in free MA3-m (Upper) or MA3-c (Lower) and its complex with eIF4A-NTD. The horizontal dashed line represents the calculated average chemical shift perturbations. ...

PDCD4 MA3 Domains Efficiently Compete with eIF4G-c and RNA for Binding to eIF4A.

Several reports have shown that PDCD4 inhibits the eIF4A-eIF4G interaction, and recently MA3-c alone was shown to compete with eIF4G-c (14, 16). To confirm this MA3-c inhibition, and to investigate whether MA3-m also competes with eIF4G-c for eIF4A binding, NMR competition assays were used. [15N,1H]TROSY-HSQC spectra of 15N-labeled eIF4A-NTD (≈26 kDa) were recorded (SI Fig. 7A), whereas unlabeled GB1-tagged eIF4G-c and/or unlabeled PDCD4 MA3 domains were added (SI Fig. 7 B–F). When a stoichiometric amount of eIF4G-c (≈54 kDa) was added to eIF4A-NTD, a significant decrease in the signal intensities was observed, due to formation of the high molecular mass eIF4A-NTD–eIF4G-c complex (≈80 kDa, SI Fig. 7B; red, remaining peaks). Addition of an equivalent amount of PDCD4 MA3-m (SI Fig. 7C) or MA3-c (SI Fig. 7D) to the eIF4A-NTD-eIF4G-c mixture resulted in the reappearance of some peaks. When fourfold excess of MA3-m or MA3-c was added to the mixture, most of the signals recovered with a spectrum pattern virtually identical to that of the eIF4A-NTD–MA3-m/MA3-c complexes (SI Fig. 7 E–F). The recovery of the signals indicated that MA3-m and MA-c competed with eIF4G-c and formed the eIF4A-NTD:MA3-m and eIF4A-NTD–MA3-c complexes (≈43.7 kDa for MA3-m and ≈40 kDa for MA3-c).

PDCD4 is known to inhibit cap-dependent translation initiation by interacting with eIF4A. To investigate the mechanism of inhibition of eIF4A helicase activity by PDCD4, we tested whether PDCD4 competes with RNA for binding to eIF4A or forms an inactive PDCD4–RNA–eIF4A ternary complex. Because a single MA3 domain is sufficient to inhibit cap-dependent translation, and the MA3 domains bind specifically to eIF4A-NTD, we used MA3-m, eIF4A-NTD, and U6-RNA in this experiment. Again, a [15N,1H]TROSY-HSQC spectrum of 15N-labeled eIF4A-NTD was recorded as a control (SI Fig. 8A). When twofold excess U6-RNA was added, significant changes in the signals were observed, indicating the formation of eIF4A-NTD–U6-RNA complex (SI Fig. 8B, red). When fourfold excess PDCD4 MA3-m was added to the mixture, the spectrum changed again (SI Fig. 8D, red) and became virtually indistinguishable from that of the eIF4A-NTD–MA3-m complex in the absence of U6-RNA (SI Fig. 8C, red). These peak pattern changes indicate that MA3-m competed with U6-RNA and displaced it from its complex with eIF4A-NTD. As a control, we verified that neither MA3-m nor MA3-c binds to U6-RNA by NMR and fluorescent polarization, using the same concentration in both experiments (data not shown).

Intramolecular Interaction of PDCD4 MA3 Domains.

To test whether there is intramolecular interaction between MA3-m and MA3-c, we performed binding experiments with one of the MA3 domains 15N-labeled and the other unlabeled. No changes in the spectra were observed upon addition of the counterpart for both combinations (data not shown), indicating that MA3-m and MA3-c do not interact in trans. We then expressed a MA3-dual domain construct (157–449), which contains MA3-m and MA3-c. The [15N,1H]TROSY-HSQC spectrum of MA3-dual showed overall line broadening, presumably because of the large molecular size (>30 kDa), but overlays of the spectra of MA3-m, MA3-c, and MA3-dual showed some distinct chemical shift changes in peaks from both domains (SI Fig. 9A, marked by circle). The changes in the signals are mainly located at the N termini of α4, α6, and the middle of α8 for MA3-m and α1 and α2 for MA3-c (SI Fig. 9 B and C). Analysis of the electrostatic surface potential of the domains shows a broad positively charged area in MA3-m (Fig. 1D) and a broad negatively charged surface in the N-terminal region of MA3-c (14). These areas overlap with the regions where distinct chemical shift changes were observed in MA3-dual, which suggests electrostatic interaction between the two MA3 domains in cis.

Tandem MA3 Domains Bind Synergistically to eIF4A.

PDCD4 MA3-m, MA3-c, and MA3-dual were immobilized onto a sensor chip to ≈2,000 response units (RU), and various eIF4A fragments were passed over. First, binding of eIF4A fragments to immobilized MA3-dual was confirmed by flowing 100 μM eIF4A-NTD or eIF4A-CTD and 50 μM eIF4A-FL (Fig. 4A Lower). The concentration of eIF4A-FL was chosen (50 μM) to keep the total concentration of eIF4A domains constant for a fair comparison. eIF4A-FL showed the highest increase in RU for all PDCD4 MA3 domains, whereas moderate RU increase were observed for eIF4A-NTD, and no apparent binding was observed for eIF4A-CTD.

Fig. 4.
Effect of single MA3 domain versus double MA3 domains for binding to eIF4A fragments by SPR. (A) Binding analysis of eIF4A-FL and PDCD4 MA3 domains. (Upper) Sensorgrams were obtained by passing 50 μM eIF4A-FL over immobilized MA3-dual (red), MA3-m ...

The binding affinity between MA3-dual and eIF4A-FL was strikingly higher than all other combinations. MA3-dual showed 5- to 6-fold greater increase in RU for 50 μM eIF4A-FL compared with MA3-m or MA3-c, both of which have a similar moderate affinity (Fig. 4A Upper). Among the three eIF4A fragments tested, eIF4A-FL showed higher binding affinity to the single MA3 domains, but the binding affinity between eIF4A-FL and MA3-dual stands out among all (Fig. 4A Lower).

To analyze the effect of tandem MA3 domains versus single MA3 domain, a more precise binding analysis was performed by using MA3-dual and MA3-m. eIF4A-NTD and eIF4A-FL were passed over the sensor chip with concentrations ranging from 0 to 50 μM and from 0 to 20 μM, respectively (Fig. 4 C and D).

When MA3-dual and MA3-m were compared for their eIF4A-NTD-binding affinity, MA3-dual binds somewhat stronger than MA3-m (Fig. 4D). When compared for their eIF4A-FL-binding affinity, the difference was much more significant, and MA3-dual showed ≈8-fold greater RU increase at 20 μM eIF4A-FL, compared with that of MA3-m (Fig. 4C). Because relative numbers of immobilized MA3 domains are set to be equal between MA3-dual and MA3-m, the RU increase caused by binding directly corresponds to the affinity differences between the two fragments. Fig. 4D clearly shows that each MA3 domain of MA3-dual has a capacity to bind individually to different eIF4A-NTD molecules, because the RU increases for MA3-m and MA3-dual were comparable. When MA3-m is compared for its binding to eIF4A-FL and eIF4A-NTD, >3-fold increase in RU was observed for eIF4A-FL compared with eIF4A-NTD. Considering the molecular mass difference between eIF4A-FL (46.2 kDa) and eIF4A-NTD (26.6 kDa), the actual increase in the binding is ≈1.5-fold. This indicates that binding to eIF4A-FL is somewhat stronger than binding to eIF4A-NTD.


PDCD4 has been known to bind eIF4A and inhibit eIF4A-mediated translation initiation (4, 6). Based on the findings that PDCD4 prevents eIF4A binding to eIF4G-c and that PDCD4 binds to eIF4G-m in the absence of eIF4A, a possible inhibition model has been proposed, which implies two alternative inhibitory functions for PDCD4 (6): (i) PDCD4 prevents eIF4G-c from binding to eIF4A and blocks stimulation of eIF4F activity and (ii) PDCD4 binds to eIF4A and traps the inactivated eIF4A on the eIF4G-m surface, which inactivates the eIF4F complex.

Further studies revealed that PDCD4 MA3 domains play important roles in eIF4A binding and inhibition of translation (16), and the latest report showed MA3-c alone is sufficient for inhibition of eIF4A translation (14). However, the detailed functions of MA3-c and the rest of PDCD4 are still unclear. The structure of the MA3-c domain of mouse PDCD4 that has >90% homology to that of human has been solved by both x-ray crystallography and NMR (14, 15), but a structure of MA3-m has not been reported so far, and the specific functions of the two MA3 domains have been unresolved. Whether the two MA3 domains bind to eIF4A simultaneously, and how these two MA3 domains function in eIF4A binding and inhibition of translation is also not clear. In addition, PDCD4 was reported to bind eIF4G-m, but whether the PDCD4 MA3 domains are involved in the binding remains to be answered.

Structural Features of PDCD4 MA3-m and Comparison with MA3-c.

We have solved the crystal structure of the human PDCD4 MA3-m domain. MA3-m was found to consist of four helical hairpins and is longer than MA3-c. Thus, its size is intermediate between PDCD4 MA3-c and eIF4G-MA3, which have three and five helical hairpins, respectively. Superimposition of MA3-m and MA3-c clearly shows that they adopt similar secondary and tertiary structures. (SI Fig. 6C) It had been predicted in a previous report (14) that a proline residue in the middle of helix α3 (P200) might cause disruption or a bend in the helix conformation, resulting in overall structural change. Looking at MA3-m crystal structure, a slight bend in the helix α3 with the angle ≈20° was observed at the position of P200. No overall distortion of the structure was caused by P200, and the orientation of the helices is conserved.

Computation analysis of MA3 domain boundaries of human PDCD4 predicts MA3-m to be 164–275, and MA3-c to be 327–440 (SMART, simple modular architecture research tool). The region between the two domains is predicted to be a linker containing another α-helix. Upon inspection of the actual 3D structures of MA3-m and MA3-c, the observable last helix in MA3-m ends at the position L302, and the first helix in MA3-c begins at H326. There are ≈23 aa residues between the two MA3 domains, and no additional α-helix was found for this region both in the crystal structure and NMR data. This region might purely act as a flexible linker or might form an additional α-helix when MA3-m and MA3-c are tethered.

eIF4A Interaction with PDCD4 MA3 Domains.

NMR titration assays showed that both MA3 domains bind specifically to eIF4A-NTD and use homologous epitopes. (Fig. 3 B and C). This is consistent with existing mutation data: Mutations that have been reported to inhibit the interaction with eIF4A (E249 and D253 for MA3-m, D414 and D418 for MA3-c) are located within these surface areas. However, a second set of residues, whose mutation also affected eIF4A binding are located in α2 (D180 and L191 in MA3-m and D343 and L354 in MA3-c) and were not in the regions affected by eIF4A-NTD binding (16). This is consistent with the finding that the corresponding region of the human eIF4G-MA3 contacts eIF4A-CTD and not eIF4A-NTD (A.M., C.S., K.A.E., and G.W., unpublished work).

K283 and G284, which are located in the loop between α7 and α8 of MA3-m, were also affected by eIF4A-NTD binding (Fig. 3A). This loop does not exist in MA3-c and is somewhat separated from the main interaction surface. The backbone signals were significantly broadened for the C-terminal half of the loop (V286–C288) and the N-termini of α8 (A293–A295), which implies that this region is in conformational exchange. Upon eIF4A-NTD binding, the α7-α8 loop may move closer to the main interaction surface and form an additional binding interface adjacent to the main surface. This additional interface may lead to a tighter binding between MA3-m and eIF4A-NTD.

Recently, mutations in eIF4AII were reported that affect PDCD4 binding (19). As described, most of the mutated residues are buried inside of the protein and are likely to affect the structure of eIF4AII but are unlikely to be part of the binding epitope.

PDCD4 Displaces eIF4G-c and RNA from eIF4A.

A recent study has shown that MA3-c alone can efficiently compete with eIF4G-c (MA3 + HEAT3) for binding to eIF4A (14). Considering the remarkable similarity between MA3-m and MA3-c, we were intrigued by the possibility that MA3-m might also compete with eIF4G-c, and indeed it does efficiently so as revealed with NMR competition assays.

Because eIF4A-NTD has an RNA binding site, there is a possibility that PDCD4 MA3 domains occupy the same binding site and inhibit RNA binding. Using an NMR competition assay, we found that MA3-m efficiently competed with RNA, and the RNA binding to eIF4A-NTD was abolished when twofold excess MA3-m was added to eIF4A. This means that PDCD4- and RNA-binding interfaces on eIF4A-NTD are at least partially overlapped. Because PDCD4 MA3 domains and eIF4G-MA3 share the same binding interface on eIF4A-NTD, this could indicate that eIF4G-MA3 also competes with RNA. However, when eIF4G-MA3 was used instead of MA3-m, no competition was observed under the same experimental conditions (K.A.E., A.M., C.S., and G.W., unpublished data). Thus, eIF4G-MA3 has either a binding interface that overlaps with PDCD4 but does not compete with RNA, or it has a weaker affinity to eIF4A and was not potent enough to compete with RNA under the same conditions as MA3-m.

Whereas eIF4G-MA3 is dispensable for stimulation of eIF4A helicase activity and is thought to have only a modulatory function, point mutations in this domain were reported to result in 76% decrease in translation, a much more severe defect than when the entire C-terminal region of eIF4G is deleted (12). Although the reason for this phenomenon is unclear, it suggests two possible inhibitory mechanisms for the PDCD4 MA3 domains: (i) interference with the modulatory function of eIF4G-MA3 and (ii) blocking RNA from binding to eIF4A-NTD. These two possible inhibitory mechanisms are not mutually exclusive (Fig. 5C).

Fig. 5.
Model for eIF4A binding and inhibition by PDCD4. (A) Binding model between MA3-m and eIF4A-FL based on NMR and SPR analysis. The solid arrow shows the interaction between eIF4A-NTD and MA3-m. The dashed arrow shows the predicted interaction between eIF4A-CTD ...

Regions from both domains of eIF4A are thought to form a contiguous RNA-binding surface in the helicase's closed (active) conformation, and RNA and ATP bind cooperatively (20). Because the intracellular concentrations of ATP and ADP are very high (≈500 and ≈50 μM, respectively), it is likely that the predominant states of eIF4A in the cell are the two different versions of the closed conformation: ATP-bound and ADP-bound. We have previously proposed that eIF4G-m binding to eIF4A promotes the closed active conformation, and stimulates ATP and RNA binding (11). Because both PDCD4 MA3 domains compete with RNA and eIF4G-MA3, PDCD4 is presumably able to inhibit eIF4A activity by preventing RNA binding and thus formation of the active RNA-bound conformation. A previous report showed that full-length PDCD4 binds to eIF4G-m and eIF4A at the same time, which implicates an additional inhibitory function by PDCD4 (6). In this context, the interaction of PDCD4 with eIF4G-m likely serves a dual role. First, it directs PDCD4 to the eIF4E/4G/4A complex (eIF4F), which is much less abundant than free eIF4A (21) and, thus, a better target for inhibition of translation initiation. Second, it anchors PDCD4 to eIF4F and increases its effective concentration in the vicinity of eIF4G-bound eIF4A. We have performed binding assays for both PDCD4 MA3 domains with eIF4G-m, but neither MA3-m nor MA3-c showed any detectable chemical shift changes (data not shown). The N-terminal region of PDCD4 preceding MA3-m (1–157) may be responsible for binding to eIF4G-m (Fig. 5C).

Structural Model for the Mechanism of Inhibition of eIF4A Activity by PDCD4.

Although no apparent interaction was observed between the two MA3 domains in trans, distinct chemical shift differences were observed in both domains between the spectra of the free domains and MA3-dual (SI Fig. 9A). Mapping of these residues onto the structures implies a rough orientation of the two MA3 domains as shown in SI Fig. 9 B and C: There is a linker composed of ≈23 aa residues between the two MA3 domains, and this may allow MA3-c to rotate further. The oppositely charged surfaces that exist in the putative interface between MA3-m and MA3-c also indicate that the two MA3 domains actually function as a pair.

SPR data showed that MA3-m binds to eIF4A-FL ≈1.5-fold tighter than eIF4A-NTD (Fig. 4 C and D, filled squares) even though no obvious binding was observed between MA3-m and eIF4A-CTD (data not shown). We recently found that the eIF4G-MA3 binds not only to eIF4A-NTD but also weakly to eIF4A-CTD and that a mutation in α2 (D1259N) affects binding to eIF4A-CTD but not eIF4A-NTD (A.M., C.S., K.A.E., and G.W., unpublished work). The corresponding mutations in MA3-m (D180N) and in MA3-c (D343N) were reported previously to affect PDCD4 binding to eIF4A (16). Therefore, it is likely that the binding of the PDCD4 MA3 domains with eIF4A-FL is stabilized by additional interaction with eIF4A-CTD (Fig. 5A), which is too weak to be observed with the isolated eIF4A-CTD. The necessity of the two MA3 domains was shown by SPR analysis. MA3-dual displayed significantly tighter binding to eIF4A-FL compared with the single-domain MA3-m. This synergistic effect could be explained if simultaneous binding of eIF4A-NTD to one MA3 domain and of eIF4A-CTD to the other (Fig. 9B) is more favorable than both eIF4A domains binding to the same MA3 domain (Fig. 5A). Because the 1:1 complex shown on Fig. 5B Upper leaves the eIF4A-NTD-binding site on one of the MA3 domains unoccupied (as well as one putative eIF4A-CTD-binding site on the other MA3 domain), it is likely that PDCD4 binds two eIF4A molecules simultaneously. Both MA3 domains of MA3-dual have a capacity to bind an individual eIF4A-NTD (Fig. 4D), and, judging from the proposed orientation of MA3-m and MA3-c (SI Fig. 9 B and C), the eIF4A-NTD-binding site and, presumably, the eIF4A-CTD-binding site from each MA3 domains are separated enough to bind two eIF4A-FL molecules without steric hindrance.

A model for the inhibition mechanism of PDCD4 is proposed in Fig. 5C. In the absence of PDCD4, eIF4G-m and eIF4G-MA3 associate with eIF4A and RNA. PDCD4 displaces RNA and eIF4G-MA3 from eIF4A. Whether RNA remains bound to the inactive eIF4A/eIF4G/PDCD4 complex is still unresolved because both domains of eIF4A have been shown to interact with RNA, and eIF4A-CTD does not measurably bind PDCD4; it is likely that RNA remains weakly associated to the complex through eIF4A-CTD as indicated in Fig. 5C iii.

In summary, we determined the x-ray crystal structure of PDCD4 MA3-m. We found that MA3-m and MA3-c have similar structure, and NMR data revealed that both domains have similar eIF4A-binding sites. PDCD4 competes with both RNA and the eIF4G-MA3 for binding to eIF4A-NTD, thus inhibiting eIF4A activity. We also found that both PDCD4 MA3 domains bind to eIF4A-NTD, and the presence of the two MA3 domains together is essential for maximum eIF4A binding, which helps explain the presence of pairs of tandem MA3 domains in all PDCD4 homologues. We provide evidence for at least transient interactions between the two MA3 domains, but the exact interdomain orientation remains to be determined.

Materials and Methods

Vector constructions, protein expression, and purification protocols and additional experimental details are described in SI Text.

Crystallization, Data Collection, and Structure Determination.

MA3-m was crystallized by vapor diffusion against 100 mM Tris·Cl (pH 8.5), 200 mM sodium acetate, 28–30% PEG 4000, and 8% Jeffamine M-600. All datasets were collected at the Advanced Photon Source beamline 24C (NE-CAT) by using a Q4 area detector. A single-wavelength anomalous dispersion (SAD) dataset was obtained from a Se-Met MA3-m crystal at λ = 0.9793 Å; for higher resolution datasets, further datasets were obtained at λremote = 0.8551 Å. Data were processed by using HKL2000 (HKL Research). Structure determination and refinement procedures are mentioned in SI Text.

NMR Resonance Assignments, Titration, and Competition Assays.

NMR spectra were recorded at 298 K on either Bruker Avance 500-MHz, 600-MHz, or 750-MHz spectrometers, all of them equipped with cryogenic triple-resonance probes. Samples for NMR measurements typically contained 0.5–1 mM protein in buffer containing 20 mM Tris·Cl (pH 7.0), 150 mM NaCl, 2 mM DTT, 1 mM EDTA, and 10% D2O. Chemical-shift changes in titration experiments were reported as normalized values by using Δδ(H1,N15)=(ΔδH1)2+0.2(ΔδN15)2. Details of the NMR experiments are described in SI Text.

Binding Analysis by Surface Plasmon Resonance (SPR).

SPR binding analysis was carried out by using a BIAcore 3000 instrument (Biacore). PDCD4 MA3-m, MA3-c, and MA3-dual were immobilized on a research-grade CM5 sensor chip by using the amino coupling kit supplied by the manufacturer in 10 mM sodium acetate (pH 5.5). A control flow cell was subjected to activation and blocking in the same procedure but without proteins. The surface densities of immobilized proteins were ≈2,000 RU for MA3-m, MA3-c, and MA-dual. The signal generated in the control flow cell was subtracted from the experimental signals to correct for refractive index changes and nonspecific binding.

All of the binding experiments were carried out in Hepes-EP buffer (Biacore) at 25°C and with the flow rate of 20 μl/min.

Supplementary Material

Supporting Information:


We thank Drs. D. Frueh, H. Arthanari, and K. Takeuchi for help with NMR spectroscopy. This work was supported by National Institutes of Health (NIH) Grants CA68262 and GM47467 (to G.W.) and National Cancer Institute Grant K01 CA119107 (to A.M.). The crystallographic study was supported by NIH Grant RR-15301 and Department of Energy contract DE-AC02-06CH11357.


The authors declare no conflict of interest.

The atomic coordinates reported in this paper have been deposited in the Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank, www.pdb.org (PDB ID code 2RG8).

This article contains supporting information online at www.pnas.org/cgi/content/full/0712235105/DC1.


1. Shibahara K, et al. Isolation of a novel mouse gene MA-3 that is induced upon programmed cell death. Gene. 1995;166:297–301. [PubMed]
2. Jansen AP, Camalier CE, Stark C, Colburn NH. Characterization of programmed cell death 4 in multiple human cancers reveals a novel enhancer of drug sensitivity. Mol Cancer Ther. 2004;3:103–110. [PubMed]
3. Dorrello NV, et al. S6K1- and betaTRCP-mediated degradation of PDCD4 promotes protein translation and cell growth. Science. 2006;314:467–471. [PubMed]
4. Goke A, et al. DUG is a novel homologue of translation initiation factor 4G that binds eIF4A. Biochem Biophys Res Commun. 2002;297:78–82. [PubMed]
5. Kang MJ, et al. Up-regulation of PDCD4 in senescent human diploid fibroblasts. Biochem Biophys Res Commun. 2002;293:617–621. [PubMed]
6. Yang HS, et al. The transformation suppressor Pdcd4 is a novel eukaryotic translation initiation factor 4A binding protein that inhibits translation. Mol Cell Biol. 2003;23:26–37. [PMC free article] [PubMed]
7. Rozen F, et al. Bidirectional RNA helicase activity of eucaryotic translation initiation factors 4A and 4F. Mol Cell Biol. 1990;10:1134–1144. [PMC free article] [PubMed]
8. Pause A, Sonenberg N. Mutational analysis of a DEAD box RNA helicase: The mammalian translation initiation factor eIF-4A. EMBO J. 1992;11:2643–2654. [PMC free article] [PubMed]
9. Imataka H, Sonenberg N. Human eukaryotic translation initiation factor 4G (eIF4G) possesses two separate and independent binding sites for eIF4A. Mol Cell Biol. 1997;17:6940–6947. [PMC free article] [PubMed]
10. Marintchev A, Wagner G. Translation initiation: Structures, mechanisms and evolution. Q Rev Biophys. 2004;37:197–284. [PubMed]
11. Oberer M, Marintchev A, Wagner G. Structural basis for the enhancement of eIF4A helicase activity by eIF4G. Genes Dev. 2005;19:2212–2223. [PMC free article] [PubMed]
12. Morino S, et al. Eukaryotic translation initiation factor 4E (eIF4E) binding site and the middle one-third of eIF4GI constitute the core domain for cap-dependent translation, and the C-terminal one-third functions as a modulatory region. Mol Cell Biol. 2000;20:468–477. [PMC free article] [PubMed]
13. Aravind L, Koonin EV. Eukaryote-specific domains in translation initiation factors: Implications for translation regulation and evolution of the translation system. Genome Res. 2000;10:1172–1184. [PMC free article] [PubMed]
14. LaRonde-LeBlanc N, et al. Structural basis for inhibition of translation by the tumor suppressor Pdcd4. Mol Cell Biol. 2007;27:147–156. [PMC free article] [PubMed]
15. Waters LC, et al. Structure of the C-terminal MA-3 domain of the tumour suppressor protein Pdcd4 and characterization of its interaction with eIF4A. Oncogene. 2007;26:4941–4950. [PubMed]
16. Yang HS, et al. A novel function of the MA-3 domains in transformation and translation suppressor Pdcd4 is essential for its binding to eukaryotic translation initiation factor 4A. Mol Cell Biol. 2004;24:3894–3906. [PMC free article] [PubMed]
17. Ferentz AE, Wagner G. NMR spectroscopy: A multifaceted approach to macromolecular structure. Q Rev Biophys. 2000;33:29–65. [PubMed]
18. Bellsolell L, et al. Two structurally atypical HEAT domains in the C-terminal portion of human eIF4G support binding to eIF4A and Mnk1. Structure (London) 2006;14:913–923. [PubMed]
19. Zakowicz H, et al. Mutational analysis of the DEAD-box RNA helicase eIF4AII characterizes its interaction with transformation suppressor Pdcd4 and eIF4GI. RNA. 2005;11:261–274. [PMC free article] [PubMed]
20. Lorsch JR, Herschlag D. The DEAD box protein eIF4A. 2. A cycle of nucleotide and RNA-dependent conformational changes. Biochemistry. 1998;37:2194–2206. [PubMed]
21. von der Haar T, McCarthy JE. Intracellular translation initiation factor levels in Saccharomyces cerevisiae and their role in cap-complex function. Mol Microbiol. 2002;46:531–544. [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...