Structural characterization of the Z RING-eIF4E complex reveals a distinct mode of control for eIF4E

Abstract

The eukaryotic translation initiation factor eIF4E, a potent oncogene, is highly regulated. One class of eIF4E regulators, including eIF4G and the 4E-binding proteins (4E-BPs), interact with eIF4E using a conserved YXXXXLΦ-binding site. The structural basis of this interaction and its regulation are well established. Really Interesting New Gene (RING) domain containing proteins, such as the promyelocytic leukemia protein PML and the arenaviral protein Z, represent a second class of eIF4E regulators that inhibit eIF4E function by decreasing eIF4E’s affinity for its m⁷G cap ligand. To elucidate the structural basis of this inhibition, we determined the structure of Z and studied the Z-eIF4E complex using NMR methods. We show that Z interacts with eIF4E via a novel binding site, which has no homology with that of eIF4G or the 4E-BPs, and is different from the RING recognition site used in the ubiquitin system. Z and eIF4G interact with distinct parts of eIF4E and differentially alter the conformation of the m⁷G cap-binding site. Our results provide a molecular basis for how PML and Z RINGs reduce the affinity of eIF4E for the m⁷G cap and thereby act as key inhibitors of eIF4E function. Furthermore, our findings provide unique insights into RING protein interactions.

The eukaryotic translation initiation factor eIF4E is a key effector of epigenomic regulation. eIF4E coordinately and combinatorially modulates the expression of genes involved in proliferation and survival (1, 2). eIF4E acts at two levels of gene expression: cap dependent translation and mRNA export (3, 4). For both of these processes, eIF4E must bind the 7-methyl-guanosine cap (m⁷G cap) on the 5′ end of the mRNA. eIF4E expression levels are highly elevated in several human cancers, including breast, prostate, and some leukemias, and elevated levels of eIF4E are a marker of poor prognosis in patients (5, 6). Notably, both the mRNA export and translation functions of eIF4E are dysregulated in cancer (2). Targeting eIF4E-m⁷G cap-binding activity in a phase II trial in leukemia patients led to clinical benefit (7), which further underscores the importance of understanding the regulation of this eIF4E activity at the molecular level. eIF4E is regulated by many key effectors that can be grouped into three classes: (i) those containing a conserved eIF4E binding motif such as eIF4G; (ii) those containing a Really Interesting New Gene (RING) domain, e.g., arenaviral Z proteins; (iii) others, such as the viral protein linked to the genome (VPg) (8).

The first class of eIF4E regulators uses a conserved helical binding motif (YXXXXLΦ, where X is any residue and ϕ is any hydrophobic residue) to bind the dorsal surface of eIF4E. The dorsal surface is distal (> 25 Å) to the cap-binding site of eIF4E (9). Despite this, association of eIF4G or the eIF4E binding proteins (4E-BPs) increase the affinity of eIF4E for the m⁷G cap (10). In the case of eIF4G, this could increase translational efficiency of eIF4E sensitive mRNAs while for the 4E-BPs, this could repress translation by simultaneously sequestering both eIF4E and the cognate mRNA from the translational machinery (4). Other proteins belonging to this group, PRH and HOXA9, were found to use this strategy to modulate the nuclear activities of eIF4E (11, 12). However, their binding to eIF4E had only minor effects on eIF4E cap affinity (12). To date the YXXXXLΦ proteins (specifically, eIF4G and 4E-BPs) are the only class of eIF4E regulators for which there is structural information.

The second category of eIF4E regulators is defined by the presence of a RING motif. This group includes RINGs from the promyelocytic leukemia protein PML, human homologue of drosophila ariadne (HHARI), and the arenaviral Z proteins from lymphocytic choriomeningitis virus (LCMV) and Lassa virus (LASV) (13–15). Biochemical studies showed that PML and Z use their RING motifs to bind and regulate eIF4E (14, 16). In contrast to 4E-BPs and eIF4G, PML and Z reduce the cap-binding activity of eIF4E by about 100-fold. PML is a key cellular regulator of the oncogenic activities of eIF4E and understanding how it regulates eIF4E at the atomic level is of great interest. PML’s ability to inhibit eIF4E function is closely tied to its ability to impair cap binding and thus the mRNA export activity of eIF4E (1, 2, 17, 18). Additionally in vitro, PML (a small percentage being cytoplasmic) and Z impair eIF4E-dependent translation through their RING domains (16). Importantly, PML and Z do not appear to influence expression levels or stability of eIF4E (16), and in this context they function independently of any ubiquitination activity normally associated with RINGs.

The structural basis underlying the effects of PML and Z RINGs on eIF4E was unknown. Further, the features for recognition of RINGs with proteins other than ubiquitin conjugating enzymes (Ubc) had not been reported. To address these issues, we used NMR methods to study the structural aspects of the Z-eIF4E interaction as a prototypic example of RING-eIF4E interactions. We report the structure of an arenaviral protein, Z, and demonstrate that eIF4E recognition by the RING domain of Z differs substantially from the recognition of ubiquitin conjugating enzymes (Ubc) by RINGs. In particular, residues within the first zinc-binding site make key contacts with eIF4E. A structural comparison of Z, PML, and RINGs acting as ubiquitin ligases suggested molecular underpinnings for their different interactions. In addition, Z binds to regions of the dorsal surface of eIF4E that differ from those used by eIF4G and the 4E-BPs. This leads to differential effects on the conformation of the distal cap-binding site of eIF4E. Our findings provide the foundation for further studies aimed at elucidating the biological consequences of RING-eIF4E interactions, unique RING-protein interactions, and for interactions of arenavirus and host cell proteins.

Results and Discussion

Solution Structure of the Z RING Protein.

As a first step in understanding RING-eIF4E interactions, we solved the solution structure of the 99 residue Z protein from LASV. We chose LASV Z because of its better solution behavior than PML RING (19). This Z construct associates with eIF4E and reduces cap binding to the same extent as previously used constructs derived from LCMV Z (Fig. S1), making it an ideal candidate to investigate RING-eIF4E interactions by NMR.

The RING domain of Z (residues 30 to 70) forms a well-defined structure comprising a helix (residues 50–58), two short antiparallel β-sheets (β1: residues 42–44, 47–49; and β2: 63–64, 69–70) and two loop regions: 35–39 (loop 1) and 58–62 (loop 2). For residues 30 to 70, the rmsd with respect to the mean coordinate positions is 0.39 ( ± 0.06) and 0.91 ( ± 0.23) Å for backbone and heavy atoms, respectively (Fig. 1B and Table S1). These residues are relatively rigid on the ns/ps timescales, while residues in loop 1 have some degree of motion (Fig. 1C). In contrast, the N terminus (1–29) and C terminus (71–99) exhibit very low heteronuclear NOE values, consistent with these being disordered.

Fig. 1.

Solution structure of the RING domain of Z (residues 26 to 72 are shown). (A) Ribbon representation. The helix and the two β-sheets are colored green and blue, respectively. Cys and His residues coordinating the zinc atoms (yellow spheres) are shown in orange. (B) Superposition of the final 10 energy-minimized structures for residues 30 to 70. The structures are superimposed for minimal mutual deviation of the backbone atoms (Cα, N, and C). (C) Backbone {¹H}-¹⁵N heteronuclear NOE values for Z.

Two zinc atoms bind the RING in a typical cross-brace fashion (Fig. S2A). Interestingly, when Z was compared to the structurally diverse family of RINGs using the DALI program (20), no significant superimposition was obtained (DALI score < 3.0) due to its unique conformation around site II. Specifically, its topology differs from most other RINGs by the location of the first two zinc ligating residues within site II being arranged on either side of the first small β-strand (β1). Further, Z has a second small β-sheet (β2) also around site II (Fig. 1A). A comparison of the structure of Z and PML RING domains revealed that these are quite homologous around site I, in which certain aromatic and charged side chains have nearly identical positions (Fig. S3). The rmsd between PML (residues 9–16, 28–32) and Z (residues 31–38, 49–53) is 1.9 Å (Fig. S3F), which is striking given the low identity between these proteins (Fig. S2C). Loop 1 in PML and Z adopts a more open conformation than found in other RINGs such as BRCA1 or Cbl, neither of which bind eIF4E (16). This conformational difference is likely important for eIF4E recognition (see below).

Identification of a Unique Mode of RING Interaction.

To map the Z binding surface for eIF4E we employed 2D ¹H-¹⁵N heteronuclear single quantum correlation (HSQC) NMR chemical shift mapping and transferred cross saturation (TCS) experiments (21) (Fig. 2). Note that throughout this study we used cap-free eIF4E. Upon addition of eIF4E, ¹⁵N Z undergoes relatively large chemical shift changes and/or line broadening (in red and green, respectively, in Fig. 2A–C) for residues around site I and on the underside of the RING. In particular, residues in loop 1 of site I were most affected, suggesting these are at the interacting surface with eIF4E (Fig. S4 and Fig. 2A–C). Interestingly, in the free form, loop 1 was somewhat less ordered than the rest of the Z RING (Fig. 1B and C), which may be important for facilitating local rearrangement upon eIF4E binding. Also, the N and C termini of Z are not perturbed by eIF4E addition consistent with previous studies where deletion of these regions did not affect the Z-eIF4E interaction (16).

Fig. 2.

Z binding surface for eIF4E. (A) Z ¹H-¹⁵N HSQC chemical shift perturbations mapped onto the Z structure upon addition of eIF4E. Residues are color-coded according to the extent of chemical shift perturbation from white (no variation) to red (large variation). Residues are colored in dark green due to severe peak broadening, or black, due to very weak intensity in the reference spectrum. Zinc-binding residues and zinc atoms are shown in purple. (B,C) Per-residue plot of amide chemical shift perturbations for ratio 1∶5 (Z/eIF4E). Backbone (B) and side chain (C) are shown. Residues that undergo substantial line broadening in the NMR spectrum are colored in light green, while residues with peak broadening beyond detection at 1∶1 molar ratio are in dark green and are shown as an arbitrary chemical-shift change of 0.3 (backbone) and 0.2 ppm (side chains). (D) Plot of the intensity ratios of the cross-peaks in the TCS experiment for the backbone (blue) and the Asn/Gln/Trp side chain (red dots) resonances for Z. (E) GST pull-down assay between purified eIF4E and GST-Z proteins (wild type and mutants). Z I and Z II indicate mutation of zinc-binding site I (C31A/C34A) or site II (C64A/C67A) of Z. Equal loading of Z-GST was verified by Coomassie staining.

We performed TCS experiments (21) to distinguish chemical shift perturbations arising due to ligand binding from conformational changes distal to the interaction surface. The TCS experiment showed large reductions in signal intensity for residues around site I of Z, particularly loop 1 (F30, K32, S33, W35, and the Hδ side chain of N38), confirming chemical shift mapping that these residues are at the complex interface (Fig. 2D). Mutations W35A, F36A, N38A, or K39A within loop 1 of Z also abrogated its association with eIF4E (Fig. 2E) further underlying the importance of these residues. Consistent with previous data, mutation of site II did not reduce association with eIF4E (Fig. 2E) (16). In contrast, mutation of site I or addition of EDTA, both of which unfold Z, abrogated its association with eIF4E [(16), Fig. 2E]. This is consistent with mutational and biophysical studies indicating that the integrity of site I, but not site II, is required for folding of PML and Z RINGs (16). Studies with BRCA1 and PML RINGs suggested that zinc binding is anticooperative with site I being the higher affinity site (22). This previous study suggested that the pliability in site II may enable RINGs to bind different partners as a function of zinc levels.

Altogether, our studies indicate that the region comprising residues 30 to 39 interacts with eIF4E. This region is highly conserved among known Z proteins (Fig. S2B). Significantly, the Z recognition site for eIF4E does not utilize a YXXXXLΦ or any α-helix motif as used by eIF4G and the 4E-BPs. Thus, this represents a previously undescribed type of eIF4E binding domain.

BRCA1 and Cbl RINGs bind Ubcs using a shallow mainly hydrophobic groove formed by the central helix and zinc sites (Fig. S5A) (23). In the comparative region of Z and PML, residues K32, H47, R63, and K68 form a positively charged groove (Fig. 2 and Fig. S5B) and association with eIF4E is mediated via a surface formed by the residues around loop 1, which is distinct to the region used by other RINGs to interact with Ubcs. The conformation of site I is conserved between PML and Z, and differs from that observed in BRCA1 and Cbl. These analyses reveal that the molecular underpinnings of RING-Ubc and RING-eIF4E interactions are markedly different (see Fig. S5 for details).

Mapping of the Interaction Site on eIF4E.

To characterize the binding site of eIF4E for Z, we monitored perturbations in the NMR signal of ¹⁵N-eIF4E as a function of Z addition (Fig. S6) and also performed TCS studies of the ²H, ¹⁵N-labeled eIF4E/unlabeled Z complex. The largest chemical shift perturbations were found for residues in helices 1 and 2 of eIF4E, together with residues preceding the initial β-strand (Fig. 3A). These were accompanied by significant line broadening, some of which were broadened beyond detection (green in Fig. 3A and B), e.g., the indole NH of W73. The TCS experiments confirmed the chemical shift perturbation mapping with residues H37, V69, E70, W73, A74, and W130 exhibiting the largest intensity changes (Fig. 3C). Interestingly, no ordering of the N-terminal eIF4E arm was observed upon Z addition, in contrast to the eIF4G-eIF4E-cap ternary complex in yeast (24). These results indicate that Z binds to a region of eIF4E that only partially overlaps with the YXXXXLΦ class of regulators (see below).

Fig. 3.

eIF4E binding surface for Z. (A) Chemical shift perturbation of backbone and ¹⁵N resonances (color-coded as in Fig. 2A) between eIF4E alone and in complex with Z. Key amino acid side chains are displayed. (B) Per-residue plot of chemical shift perturbations for backbone amides at a ratio 1∶5 (eIF4E/Z). The inset indicates the perturbations for the Trp side chains. Residues that are substantially broadened or disappear upon complex formation are indicated by light and dark green bars, respectively. N and C termini are indicated as Nt and Ct. (C) Plot of the intensity ratios of the cross-peaks in the TCS experiments for the backbone (blue) and Trp side chain (red dots) resonances of eIF4E.

Modeling the eIF4E:Z Complex.

Attempts to use residual dipolar couplings to obtain the relative orientation of Z to eIF4E were unsuccessful given the complete insolubility of the complex in multiple alignment media. Further, the intermediate NMR exchange regime observed for the eIF4E-Z complex meant that it was not possible to obtain intermolecular NOEs due to loss of the resonances specifically at the interaction surface. Therefore, we generated a model of the eIF4E:Z complex with the restraint-driven docking program Haddock (25), using chemical shift perturbation, TCS, and mutagenesis data (Figs. 2, 3, and 4C). The Z-eIF4E complex was a good candidate for this strategy as neither protein undergoes large-scale conformational changes upon complex formation as seen by the limited extent of chemical shift perturbation, consistent with previous circular dichroism studies (15, 16). The resulting model showed a globally consistent orientation of Z with eIF4E (Fig. S7). Here, W35 and F36 of Z interact with the hydrophobic cavity between helices 1 and 2 of eIF4E. Also, residues N38 and K39 on Z interact with H37 and E70 on eIF4E, consistent with mutational data (Figs. 2E and 4C).

Fig. 4.

Binding interfaces of eIF4E with Z (A) or eIF4Gp from NMR. The width of the tube shows the TCS intensity between eIF4E and Z (A) or the intensity of NOEs observed between eIF4E and eIF4Gp (B) (27). Chemical shift perturbations upon binding to Z or eIF4Gp are in red. (C) GST pulldown (for Z, Top) and His pulldown (for the eIF4G peptide, Bottom) assays. Equal loading for GST/MFH fusion proteins was verified by Coomassie staining. (D) Residues mutated on eIF4E are shown in D. The amino acid sequence of eIF4Gp is KKQYDREFLLDFQFMPA (human eIF4GII) in the NMR titration (B) and LEEKKRYDREFLLGFQFIF (human eIF4GI) in the His pulldown assay (C).

How Does Z Decrease Affinity of eIF4E for the Cap?

The RING class of eIF4E partners were the first class observed to reduce the affinity of eIF4E for the cap (26). In contrast to Z, eIF4G binding increases the affinity of eIF4E for the cap (27). Binding footprints of Z and the eIF4G peptide (eIF4Gp) from our NMR data indicate that they bind overlapping but distinct surfaces on eIF4E (Fig. 4A and B). Z binding is more centralized to helix 1, whereas eIF4Gp binding is more toward the top of the dorsal surface. To confirm these observations, we carried out pulldown experiments using a series of mutations in eIF4E to preferentially modulate eIF4Gp vs. Z binding (Fig. 4C and D). Note, all mutants studied were structured as observed by NMR and/or CD (Fig. S8A and C). We observed that mutations in the shallow groove between helices 1 and 2 (Y76A or W130F) substantially reduce Z binding but not binding to eIF4Gp. Conversely, mutation of residues toward the top of the dorsal surface (Q40A, K65A, V69G, D144A, and, to a lesser extent, D147A) reduced binding to eIF4Gp whereas these mutations did not substantially affect the binding of eIF4E to Z, although there was a slight reduction for Q40A. E70A and W73A mutants reduce both Z and eIF4Gp binding. Importantly, equal amounts of bait proteins were immobilized on the beads (Figs. 2E and 4C), and the same amounts of soluble partners were added (Fig. S8B). In summary, eIF4E uses specific features on the dorsal surface to interact with structurally distinct partners.

We investigated whether the distinct binding footprints underlie the differential effects on cap binding by analyzing chemical shift perturbations (Fig. 5A and B and Fig. S9) and line broadening (Fig. S10) in eIF4E for the two complexes. Interestingly, the chemical shift for the indole nitrogen of W130 was altered and substantially broadened in the Z complex, but not in the eIF4Gp complex (Fig. 5B). Further, W130 is at the Z interaction site according to TCS data. W130 is adjacent to β3 and β5 strands, which back onto the cap-binding site, and thus its perturbation could modulate cap binding (Fig. 5B and Fig. S10A). Y76, a residue adjacent to W130, underwent substantial line broadening in the wild-type eIF4E-Z complex, and its mutation abrogated binding for Z. Interestingly, many residues in the cap-binding site were perturbed by either addition of Z or mutations of Y76 or W130 (Fig. 3 and Figs. S10 and S11). Notably, mutation of Y76 or W130 did not abrogate eIF4Gp binding. Thus, altering the Y76-W130 interaction could be the initial step in a chain of events communicating Z binding to the cap-binding site. Interestingly, W130 was proposed to be part of an allosteric track important for communication between the dorsal surface and the cap-binding site (28). Given the differing binding surfaces for Z and eIF4Gp, communication to the cap-binding site may go through different allosteric routes and thereby yield different outcomes.

Fig. 5.

(A) An overlay of ¹⁵N-¹H HSQC spectra of eIF4E showing residues in the core and the cap-binding site without (cyan), with Z (purple), or with eIF4Gp (red). (B) Chemical shift differences for the eight Trp indoles between the apo-eIF4E and either eIF4E/eIF4Gp (red) or eIF4E/Z (purple) complexes.

Residues W102 and W56 are important for binding cap. Studies of apo eIF4E and the eIF4E-cap complex showed that W102 rotates into the cap-binding site, whereas W56 and its adjacent loop moves as a hinge and together these tryptophans sandwich the m⁷G cap (27). Neither Z nor eIF4Gp perturbed W56 or W102 indole resonances. In our previous studies of apo-eIF4E, the backbone resonances corresponding to W56 and W102 were masked by residues in the flexible N-terminal arm of eIF4E. To overcome this, we designed a shorter eIF4E (s-eIF4E) lacking residues 4 to 26. In the s-eIF4E/Z complex, small chemical shift perturbations are observed for the backbone resonances of the W56-containing loop, while the W102-containing loop, is unaffected. Additionally, Z binding leads to substantial line broadening of residues in the phosphate-binding region (relative to other residues in the complex, e.g., W166 or E105) suggesting that this region is undergoing conformational exchange. These include R157, A158, and C89 (Fig. 5). In the eIF4E/eIF4Gp complex, residues in the phosphate-binding region (D90, R157, K162) also undergo substantial conformational changes, and residues W102, E105, and W166 backbone resonances (but not Trp side-chain resonances) exhibit small shifts suggesting a small conformational change in this region (Fig. 5A). In general, residues exhibit sharper linewidths in the eIF4Gp complex (e.g., L81, Y91) than in the Z complex or for apo-eIF4E (Fig. 5, Fig. S10B–D). Thus, Z binding appears to induce conformational exchange in the cap-binding site leading to reduced cap affinity, whereas eIF4Gp binding reduces mobility and prestructures eIF4E into a higher affinity state (27).

Conclusions

No structures of any arenavirus proteins were reported prior to these studies. Z folds as a RING domain with a unique topology around site II, but exhibits structural similarity with PML RING around site I. Our studies show that this site forms the binding surface for eIF4E, specifically around loop 1, and represents a unique eIF4E binding site with no homology to that used by eIF4G and the 4EBPs. The structural similarity of Z and PML at this site combined with previous mutational data for PML (16) suggests that PML likely uses a similar strategy to bind and regulate eIF4E. Our studies also provide unique insights into RING recognition (Fig. S5C and D).

Compared to the eIF4G-eIF4E complex, Z binds eIF4E on a distinct region of the dorsal surface centered on helix 1. Further, Z and eIF4G differentially alter residues at the distal cap-binding site, providing a possible mechanism for the different effects of Z and eIF4G on cap affinity. Our results suggest that perturbing the interaction between W130 and Y76 leads to conformational changes in the cap-binding site including increased conformational exchange (Fig. 6). The increased conformational exchange of eIF4E upon addition of Z, taken together with the substantial line broadening observed, suggests that dynamically driven allostery could be an important contributor to the effect of Z on the cap affinity of eIF4E. Note that it is also possible that multiple allosteric routes could be used simultaneously.

Fig. 6.

Communication from the dorsal surface to the cap-binding site of eIF4E. Key amino acid side chains are displayed in stick mode (backbone amides displayed in Fig. 5A are in green). The backbone is color-coded according to chemical shift changes in the eIF4E/Z complex.

Z is important to many viral processes including regulation of virus gene expression, virion assembly, and budding. Z is likely a pleiotropic mediator of arenavirus-host interactions given it interacts not only with other viral proteins but also with a variety of host proteins other than eIF4E, including PML and PRH (11, 29, 30). The mechanisms by which pathogenic arenaviruses overcome the host innate defense response remain to be determined, but appear to involve a complex network of cellular interactions including plasmocytoid dendritic cells (pDC) (31). Interferon regulatory factor 7 (IRF7) contributes to the control of type I interferon expression by pDCs, which plays an important role in the host antiviral response (32). Translation of IRF7 mRNA is influenced by the 4E-BP/eIF4E ratio (32), raising the intriguing possibility that the Z-eIF4E interaction contributes to altered IFN production in arenavirus-infected pDCs. The generation of recombinant LCM viruses with Z mutations will facilitate studies aimed at determining the importance of Z/eIF4E interactions in the context of the natural course of virus infection and reveal undiscovered aspects of arenavirus biology.

Materials and Methods

Expression and Purification.

Z and eIF4E were expressed and purified as described (27, 33) and confirmed by mass spectrometry (see Fig. S12). Site-directed mutants for Z and eIF4E were prepared using the Quick-Change mutagenesis kit (Stratagene). The N-terminal truncated eIF4E (Δ4-26 eIF4E, or s-eIF4E) was generated by a PCR cloning method. All constructs were verified by sequencing.

NMR Spectroscopy and Structure Calculations.

NMR samples typically contained 0.2 mM protein in 93% H₂O/7% D₂O containing 20 mM phosphate buffer (pH 7.2), 200 mM NaCl, 10 μM tris(2-carboxyethyl)phosphine, 50 μM ZnSO₄, and 0.02% sodium azide. NMR experiments were acquired at 600 MHz on a Varian INOVA spectrometer equipped with an HCN cold probe at 20 °C. NMR experiments for assignment of resonances were reported (33). Three-dimensional ¹⁵N-edited and ¹³C-edited NOESY spectra (100 ms mixing time) were acquired at 600 and 800 MHz. Heteronuclear ¹H-¹⁵N NOE spectra were recorded in an interleaved manner (34). Data were processed with NMRPipe (35) and analyzed with Sparky (36).

TCS experiments (21) were performed using a molar ratio of 10∶1 for [100%-²H, ¹⁵N] eIF4E to unlabeled Z, and 4∶1 for [100%-²H, ¹⁵N] Z to unlabeled eIF4E. Protons were saturated at 0.9 ppm using a 1.5 s train of EBURP pulses of 20 ms separated by 1 ms and a relaxation delay of 1.5 s. The saturation frequency was shifted 25,000 Hz upfield for the off-resonance experiments. To evaluate the effect of the residual aliphatic protons within the labeled protein, and possible effects of spin diffusion from the high amide protein content, the TCS experiments were also carried out under identical conditions but without the unlabeled protein.

Distance restraints were obtained from 3D ¹⁵N-edited and ¹³C-edited NOESY spectra. A total of 483 manually assigned distance restraints were classified according to peak intensities. Also, 30 experimental ϕ dihedral angles were obtained from values of ³JH_Nα coupling constants derived from an HNHA experiment (37). Hydrogen bonds were determined from a lyophilized sample of ¹⁵N-labeled Z dissolved in D₂O and ¹H-¹⁵N HSQC experiments recorded at 10-min intervals at 20 °C. Amide protons were considered resistant to exchange and thus involved in hydrogen bonding, if they were visible in the second HSQC. For each hydrogen bond, two distance restraints were applied for H^N(i)-O(j) and N(i)-O(j). Structures were calculated as described in SI Text. The quality of structures obtained was assessed with PROCHECK (38) and Molprobity (39). Graphic representations of 3D structures were performed using MOLMOL (40).

NMR Titrations.

All titrations were conducted with protein concentrations at or below 0.2 mM. The molar ratios were calculated to be 1∶0.3, 1∶0.6, 1∶1, 1∶2, and 1∶5 for both ¹⁵N-Z/eIF4E and ¹⁵N-eIF4E/Z titrations. Two-dimensional ¹H-¹⁵N-HSQC NMR spectra were acquired for each titration point. Chemical shift perturbations for each resonance were calculated using the equation (41).

Pulldown Assays.

Purified eIF4E was incubated with the fusion protein bound to either glutathione sepharose (GST-Z pulldown) or to Ni-NTA-agarose beads (His-eIF4Gp pulldown) in 0.5 ml of binding buffer [PBS supplemented with 250 mM KCl and 0.5% (wt/vol) NP-40] for 1 h at room temperature while tumbling. Equivalent loading of the fusion proteins or eIF4E inputs (wild type and mutants) were verified by Coomassie blue staining. After incubation, beads were washed 3 times with 1 ml of washing buffer [PBS supplemented with 500 mM KCl and 1% (wt/vol) NP-40]. The presence of eIF4E was determined by Western blotting. In the His-eIF4G pulldown, the carrier protein MFH with a six-histidine tag was used for the control (42).