ADP enhances the allosteric activation of eukaryotic elongation factor 2 kinase by calmodulin

Significance

Translation of an mRNA message by the ribosome represents one of the most energy-consuming processes in a eukaryotic cell, necessitating its downregulation under energy stress. The primary mechanism of suppressing translational elongation is by a reduction in the ribosome affinity of the GTPase eukaryotic elongation factor 2 (eEF-2) through specific phosphorylation. This covalent eEF-2 modification is uniquely catalyzed by the calmodulin-activated eEF-2 kinase (eEF-2K). It has been suggested that eEF-2K indirectly senses the cellular energy state through the master sensor, AMP-activated protein kinase, activated upon depletion of energy reserves. Here, we suggest a direct energy-sensing role for eEF-2K mediated by ADP whose engagement with the kinase at a unique site leads to enhanced sensitivity toward its allosteric activator, calmodulin.

Abstract

Protein translation, one of the most energy-consumptive processes in a eukaryotic cell, requires robust regulation, especially under energy-deprived conditions. A critical component of this regulation is the suppression of translational elongation through reduced ribosome association of the GTPase eukaryotic elongation factor 2 (eEF-2) resulting from its specific phosphorylation by the calmodulin (CaM)-activated α–kinase eEF-2 kinase (eEF-2K). It has been suggested that the eEF-2K response to reduced cellular energy levels is indirect and mediated by the universal energy sensor AMP-activated protein kinase (AMPK) through direct stimulatory phosphorylation and/or downregulation of the eEF-2K-inhibitory nutrient-sensing mTOR pathway. Here, we provide structural, biochemical, and cell-biological evidence of a direct energy-sensing role of eEF-2K through its stimulation by ADP. A crystal structure of the nucleotide-bound complex between CaM and the functional core of eEF-2K phosphorylated at its primary stimulatory site (T348) reveals ADP bound at a unique pocket located on the face opposite that housing the kinase active site. Within this basic pocket (BP), created at the CaM/eEF-2K interface upon complex formation, ADP is stabilized through numerous interactions with both interacting partners. Biochemical analyses using wild-type eEF-2K and specific BP mutants indicate that ADP stabilizes CaM within the active complex, increasing the sensitivity of the kinase to CaM. Induction of energy stress through glycolysis inhibition results in significantly reduced enhancement of phosphorylated eEF-2 levels in cells expressing ADP-binding compromised BP mutants compared to cells expressing wild-type eEF-2K. These results suggest a direct energy-sensing role for eEF-2K through its cooperative interaction with CaM and ADP.

Protein translation is one of the most energy-consumptive processes in a eukaryotic cell (1), necessitating its robust regulation, especially under nutrient and/or energy-deprived conditions. While the regulation of initiation, the first of three broad stages of protein synthesis (2), has been studied extensively (3, 4), there is a growing appreciation that the second stage, elongation, which involves the sequential incorporation of amino acids into a lengthening polypeptide chain, is also highly regulated (5). In addition to fashioning the stress adaptation response (6), the regulation of elongation contributes to translational reprogramming associated with processes such as synaptic plasticity and memory formation (7) under normal cellular conditions. Tight control of translational elongation rates also ensures the maintenance of protein quality by optimizing transit times to limit cotranslational misfolding and aggregation (8).

A critical step in translational elongation involves translocation of the nascent chain from the ribosomal A-site to the P-site to accommodate an incoming aminoacyl-tRNA. This process is mediated through the hydrolysis of GTP by the ribosome-associated GTPase, eukaryotic elongation factor 2 (eEF-2) (2). Modulation of the phosphorylation state of eEF-2 represents the principal mechanism of regulating elongation (9). eEF-2 phosphorylated at a specific site, Thr-56, is compromised in its ability to engage the ribosome (10–12) resulting in a reduction in global translational elongation rates while simultaneously altering the relative rates of translation of specific mRNA signals (7). This covalent modification on eEF-2 is uniquely catalyzed by a serine/threonine kinase, eEF-2 kinase (eEF-2K) (13–15), a member of the α-kinase family of atypical protein kinases (16, 17). Aberrant eEF-2K activity correlates with a variety of disease states, e.g., progressive neurodegenerative conditions like Alzheimer's-related dementia (18) and Parkinson’s disease (19), and several cancers through the promotion of tumorigenesis (20, 21), invasion, and metastasis (22).

The process by which eEF-2K is activated deviates substantially from other CaM-regulated kinases (23, 24). eEF-2K is allosterically activated by calmodulin (CaM) (14, 15, 25), whose binding promotes autophosphorylation at the primary stimulatory site, T348 (26, 27). eEF-2K activity is modulated by additional inputs, including Ca²⁺ ions (25, 28), pH (29, 30), and secondary phosphorylation events mediated by energy/nutrient-sensing kinase pathways (20, 31). Despite more than three decades of robust cell-biological and biochemical analyses, the precise mechanisms by which these disparate signals integrate at eEF-2K to regulate translational elongation through eEF-2 remain poorly defined.

We recently identified a truncated construct that represents the minimal functional module of eEF-2K (eEF-2K_TR; SI Appendix, Fig. S1), that is similarly activated by CaM as the wild-type enzyme and retains the ability to phosphorylate eEF-2 efficiently in cells (32). We successfully crystallized and determined the structure of T348-phosphorylated eEF-2K_TR in its complex with CaM (CaM•peEF-2K_TR) (33). This structure, which revealed the presence of an “activation spine” linking the C-lobe of CaM (CaM_C) to the kinase active site (33) (SI Appendix, Fig. S2), provided insight into the mechanism of the allosteric activation of eEF-2K by CaM. However, this structure did not contain any nucleotides bound at the active site [despite crystallization in the presence of the slowly hydrolyzable ATP analog, adenylyl-imidodiphosphate (AMP-PNP)], providing little insight into nucleotide recognition by eEF-2K. To define the nucleotide recognition mode of eEF-2K and thereby characterize a critical functional state, we determined the structure of the CaM•peEF-2K_TR complex in the presence of the nucleotides, ATP, and ADP. While ATP is expectedly engaged to the active site in this structure, the most significant and unexpected finding is the presence of ADP bound at a pocket formed at the interface of the N-lobe of the kinase domain (KD_N) and CaM_C. Informed by this structure and specific biochemical and cell biological analyses, we demonstrate that ADP functions as an activator of eEF-2K by stabilizing CaM within the active CaM•peEF-2K complex. This activating role of ADP becomes critical in sustaining eEF-2K activity and maintaining optimal levels of phosphorylated eEF-2 to suppress translational elongation under energy stress and diminished cellular reserves of ATP. Thus, our present study identifies a direct energy-sensing role of eEF-2K that supplements its previously suggested indirect role through the AMP-activated protein kinase (AMPK) (20).

Results

In our efforts to streamline procedures to produce high-quality crystals of the CaM•peEF-2K_TR complex consistently, we modified our previously used crystallization conditions (33). Specifically, our earlier structure was determined using micro-batch crystallization in the presence of a significant excess of Mg²⁺ over Ca²⁺ (~300 mM vs. ~3 mM in the crystallization buffer); we refer to that structure as high-Mg²⁺ (PDB: 7SHQ) for brevity. For the present work, we successfully crystallized CaM•peEF-2K_TR using vapor diffusion without excess Mg²⁺ (~1.5 mM) in the crystallization buffer. This change in methodology enabled the routine production of quality crystals for structural studies. Using this approach, we determined the structure of the CaM•peEF-2K_TR complex under low Mg²⁺ conditions in the absence of nucleotides (referred to as the nucleotide-free structure, NF; see Table 1 for details on data collection, refinement, and statistics). The NF structure provides an appropriate reference for a nucleotide-bound structure (NB; see Table 1) that was also determined under similar conditions but in the presence of the nucleotides, ATP, and ADP.

Table 1.

Data collection and refinement statistics^*

	NF	NB
Wavelength (Å)	0.9793	0.9201
Space Group	P21	P1
Unit cell a, b, c (Å)	77.527 58.901 88.394	59.160, 83.352, 88.978
Unit cell α, β, γ (º)	90.000 109.675 90.000	65.350 90.031 86.473
Resolution range (Å)	73.001–2.553 (2.597–2.533)	80.836–2.222† (2.576–2.222)
Total reflections	150,222 (7,209)	139,877 (7,183)
Unique reflections	24,680 (1,228)	38,423 (1,921)
Mean(I)/σ(I)	10.3 (2.3)	3.5 (1.7)
Completeness (spherical)	99.9 (100)	50.6 (7.0)
Completeness (ellipsoidal)†	-	89.1 (57.7)
Multiplicity	6.1 (5.9)	3.6 (3.7)
CC1/2	0.996 (0.838)	0.980 (0.800)
Rmerge (all I+ & I−)	0.112 (0.796)	0.174 (0.509)
Rmeas (all I+ & I−)	0.123 (0.875)	0.205 (0.596)
Rpim (all I+ & I−)	0.070 (0.506)	0.107 (0.307)
Reflections used in refinement	24,649	38,398
Reflections used for Rfree	1,153	1,912
R work	0.2169	0.2083
R free	0.2315	0.2294
Number of nonhydrogen atoms	4,849	9391
Macromolecules	4,739	8960
Ligands	3	122
Solvent	107	309
Protein residues	612	1124
RMS (bonds)	0.003	0.004
RMS (angles)	0.57	0.7
Ramachandran favored (%)	95.93	97.62
Ramachandran allowed (%)	3.9	2.29
Ramachandran outliers (%)	0.17	0.09
Rotamer outliers (%)	5.65	1.77
Clash score	5.94	3.47
Average B-factor (Å2)	64.49	36.11
Macromolecules	64.75	36.35
Ligands	49.2	35.48
Solvent	53.36	29.3
PDB accession code	8FO6	8FNY

^*Statistics for the highest resolution shell are shown in parenthesis.

^†Diffraction limits; principal axes of ellipsoid fitted to diffraction cut-off surface:

2.690 Å; 0.7549, 0.3929, 0.5251; 0.550a^* + 0.451b^* + 0.703c^*

3.244 Å; −0.1467, 0.8815, −0.4488; −0.119a^* + 0.992b^* − 0.048c^*

2.222 Å; −0.6392, 0.2618, 0.7231; −0.472a^* + 0.231b^* + 0.851c^*

Comparison of the high-Mg²⁺ and NF structures reveals substantial global differences (SI Appendix, Fig. S3A) with weak/missing density for several regions of the CaM N-lobe (CaM_N) in the latter. This, together with its generally elevated B-factors, suggests that CaM_N is not well-ordered within the NF complex (SI Appendix, Fig. S3B). In the high-Mg²⁺ structure, CaM_N contacts the β6–β7 loop (K151-N165) of the kinase in a Mg²⁺-loaded “closed” state (34), being stabilized in this configuration through numerous crystal contacts (33). The absence of these contacts in the NF structure leads to the disengagement of CaM_N from the complex and results in a high degree of disorder in the β6–β7 loop that, like CaM_N, can only be partially resolved (SI Appendix, Fig. S3A). Further, the linker between CaM_N and CaM_C, which also shows elevated B-factors (SI Appendix, Fig. S3B), adopts a helical conformation reminiscent of that seen in free Ca²⁺-loaded CaM (35). However, an inspection of those parts of CaM_N that could be modeled with some degree of reliability suggests that it adopts a Ca²⁺-free closed conformation similar to that in apo-CaM (36) rather than a Ca²⁺-loaded “open” form (37) (SI Appendix, Fig. S3C). The relative orientation between the KD and C-terminal domain (CTD) appears slightly altered between the high-Mg²⁺ and NF structures, with the latter adopting a more closed configuration relative to the former (SI Appendix, Fig. S3A), suggesting some degree of conformational plasticity between the KD and the CTD.

Crystals corresponding to the NB complex (CaM•peEF-2K_TR bound to ATP and ADP) displayed moderately anisotropic diffraction and were processed as such (see Materials and Methods for details). The NB structure contains two heterodimers in the asymmetric unit arranged in an antiparallel fashion, with the KD of one protomer contacting the CTD of the other (SI Appendix, Fig. S4). Given that this 2:2 geometry shows more optimal agreement with previously determined unambiguous crosslinks (SI Appendix, Table S1) (32) compared to a single heterodimeric 1:1 assembly (represented by our previous high-Mg²⁺ (33) and the present NF structures), it is likely that a similar arrangement is populated in solution. The ordered parts of the peEF-2K_TR protomers within this assembly adopt identical conformations with some differences in the loop regions, most notably in the remaining R-loop near pT348 (SI Appendix, Fig. S5). In each case, ATP is engaged to the active site, while ADP is found to be stably engaged to a remote pocket at the interface of CaM_C and KD_N (Fig. 1A and SI Appendix, Fig. S6). The density corresponding to CaM_N is weak/diffuse, preventing an unambiguous chain trace and suggesting that the orientation of this domain is highly heterogeneous within the NB complex. The orientation of CaM_C is identical within the two protomers (rmsd of 0.12 Å over 66 Ca atoms for the 81 to 146 region).

Fig. 1.

Structure of the nucleotide bound (NB) CaM•peEF-2K_TR complex. (A) The NB structure is shown in ribbon representation; the CaM-targeting motif (CTM, cyan), the kinase domain (KD, lime), and the C-terminal domain (CTD, orange) are highlighted. Also shown is the C-terminal (CaM_C, dull yellow) lobe of CaM; the dynamic N-terminal lobe of CaM (CaM_N) is not resolved in the structure. The nucleotides, ATP bound at the kinase active site, and ADP bound at the basic pocket (BP) are depicted as spheres. (B) Key interactions of ATP bound at the peEF-2K_TR active site are illustrated with the peEF-2K_TR sidechains colored brown. The positions of selected sidechains, of the nucleotide-free (NF, in yellow) complex, whose conformations are altered (indicated by the cyan ovals and arrows) upon ATP binding, are shown together with the sidechains of key active site residues (labeled in the lighter font). (C) The P- and the β10-αD loops in the NB complex (brown) are slightly more closed (indicated by the red arrows) relative to the N/D-loop compared to that in the NF complex (yellow). The bound ATP is shown in stick representation.

An intact, i.e., unhydrolyzed, ATP molecule engages the peEF-2K_TR active site and is stabilized through interactions involving both its adenosine and triphosphate moieties. The adenine ring stacks with the aromatic sidechain of Y236 and forms hydrogen bonds with the mainchains of H230 and I232 and the sidechains of K170 and E229 (Fig. 1B); the ribose O₄′ and O₅′ interact with the C146 sidechain (SI Appendix, Table S2). The triphosphate moiety makes numerous hydrogen bonds, notably involving the mainchain and sidechain of R144. However, no density corresponding to the catalytically essential Mg²⁺ ion is found at the active site of either protomer. This absence of this cation results in a degree of conformational variability within the triphosphate moiety of ATP (SI Appendix, Fig. S7) and some differences in the interaction patterns in the two protomers (SI Appendix, Table S2). Based on the Mg²⁺•ATP-bound structure of the KD of the related myosin heavy chain kinase A (MHCK-A, PDB: 3LMI) (38), Q276 is expected to coordinate Mg²⁺ and the γ–phosphate of ATP. However, in the NB structure, Q276 is significantly more distant from ATP (with Q276, CD–ATP,γP distances of 6.8 and 6.4 Å; SI Appendix, Fig. S7) than the corresponding Q758 in MHCK-A (4.9 Å). In this conformation, the Q276 sidechain is stabilized through hydrogen bonds with the sidechain of K238 (also the Y236 sidechain for protomer B), and the mainchains of D284 and the putative catalytic base, D274 (33). The basic K238 sidechain, through its Nζ position, appears to fulfill an interaction with Q276 expected of the missing Mg²⁺, being located at a spatial position that is roughly similar to that of the divalent ion in MHCK-A; this position is occupied by a water molecule in the structure of the MHCK-A KD bound to AMP-PCP that also lacks Mg²⁺ (PDB:3LLA; SI Appendix, Fig. S7). A comparison of the active site geometries of the NB and NF structures reveals few differences except for a small shift in the backbone of the A142-R144 segment on the P-loop and I232 on the β10-αD loop (Fig. 1B), resulting in slight closures of the P- (relative to the N/D-loop) and the β10-αD loops upon ATP binding (Fig. 1C). The sidechains of R144 and Y236, that now contact ATP, are also displaced relative to their orientations within the NF complex (Fig. 1B).

As noted earlier, we made the surprising discovery of ADP bound at a pocket created through the interaction between CaM_C and KD_N at their mutual interface upon complex formation (SI Appendix, Fig. S8A). This pocket, which we term the basic pocket (BP), is located behind the active site (that engages ATP; see Fig. 1A) on the opposite face of the kinase and lined with sidechains of several basic residues from eEF-2K_TR (R114, R116, K151, H230) and CaM (Arg90). Unlike in the case of ATP discussed above, ADP appears to occupy the BP in a well-defined conformation. ADP interacts with KD_N utilizing both its adenosine moiety and its phosphate backbone, with the latter being involved in a somewhat larger number of interactions (SI Appendix, Table S3 and Fig. 2A; also shown schematically in SI Appendix, Fig. S8B). The α–phosphate of ADP forms hydrogen bonds with the sidechains of R114 and S153 and with the N154 mainchain; the β–phosphate forms hydrogen bonds with the sidechains of R114, K151, Y167, and H230. In addition, both the α–and β–phosphates form numerous hydrogen bonds with the sidechain of Arg90 on CaM. The adenine ring of ADP stacks against the guanidinium moiety of R116 in a configuration conducive to cation–π interactions (39). The methyl groups of L125 form hydrophobic contacts with the adenine ring of ADP, with the latter also forming hydrogen bonds with the Q211 sidechain and the Y117 mainchain. The ribose moiety stacks against the aromatic ring of F155 and forms hydrogen bonds with the guanidino group of R116. ADP also makes additional superficial contacts with the ⁹⁶Gly-Asn-Gly⁹⁸ segment of the Ca²⁺-binding loop on the third EF-hand of CaM.

Fig. 2.

Interactions that stabilize ADP at the BP in the NB complex. (A) Key interactions that stabilize ADP at the BP are indicated. The sidechains of peEF-2K_TR and CaM are colored brown and cyan, respectively, with the corresponding residues labeled using 1-letter and 3-letter codes. (B) Conformational differences between the NF (peEF-2K_TR in yellow, CaM in pink) and NB (peEF-2K_TR in brown and CaM in cyan) complexes at the BP and adjoining regions are shown. The key CTM anchor residue W85 is indicated for reference; ADP is shown in stick representation. The β6—β7 loop is also indicated with its incomplete chain trace in the NF structure denoted by the ‘*’. The right panel shows an expansion of the region indicated by the black rectangle on the left panel. Regions that show significant local differences between the NF and NB structures are indicated by the red arrows. The sidechains of F155 and Arg90 (both labeled in red), which are only partially resolved in the NF structure (suggesting disorder), become ordered in the NB complex. The ADP molecule is omitted for visual ease.

Relative to the NF structure, the eEF-2K_TR CaM-targeting motif (CTM) in the NB complex undergoes a small rigid body displacement within CaM_C (Fig. 2B). No other substantial differences in overall structure are seen. However, several local conformational differences in and around the BP are apparent upon comparing the NF and NB structures. As noted above, the β6—β7 loop that is disordered in the NF complex becomes ordered in the NB structure. S153, N154, and F155 (whose sidechain is not fully resolved in the NF structure; Fig. 2B) at the N-terminus of this loop make stabilizing contacts with ADP. In addition, the sidechain of Arg90 on CaM_C that is only partially resolved in the NF structure (Fig. 2B) becomes ordered in the NB structure through its numerous contacts with the phosphate backbone of ADP (SI Appendix, Table S3). No other substantial changes in the orientations of the ADP-contacting eEF-2K_TR sidechains are evident upon comparing the NF and NB structures (Fig. 2B).

Given the unexpected finding of ADP engaged at the BP, we sought to determine its affinity for this site in the context of the complex between CaM and unphosphorylated full-length wild-type eEF-2K using isothermal titration calorimetry (ITC) measurements. A saturating concentration of the ATP-competitive inhibitor, A484954 (IC₅₀ = 0.28 μM) (40), was used to prevent the competing engagement of ADP at the active site. The measurements yield a K_D value of 38 ± 9 μM (averaged over independent measurements performed in triplicate; see Fig. 3A for a representative thermogram). Next, to test the functional role, if any, of the interaction, we measured the influence of increasing amounts of ADP (0 to 900 μM) on the ability of the active complex between wild-type full-length eEF-2K (peEF-2K) and CaM to phosphorylate a fluorescent peptide substrate, Soxtide (41). Our results illustrate a steady reduction in the concentration of CaM required to achieve a given phosphorylation level toward lower CaM concentrations with increasing ADP (SI Appendix, Fig. S9A). Indeed, the concentration of CaM required for half-maximal substrate phosphorylation (K_CaM), obtained from fits to Eq. 1 (Materials and Methods) at various concentrations of ADP, shows a substantial reduction (Table 2) from 16.4 nM (in the absence of ADP) to 1.9 nM (in the presence of 900 μM ADP), with no significant changes in corresponding maximal phosphorylation rates (k_obs,max). A fit of the variation of the value of K_CaM with ADP concentration using Eq. 2 (Fig. 3B and Materials and Methods) yields the corresponding ADP sensitivity (K_ADP = 80 ± 14 μM). In contrast to ADP, AMP has no measurable effect on K_CaM (SI Appendix, Fig. S9B).

Fig. 3.

Influence of ADP on the CaM-mediated activity wild-type eEF-2K. (A) A representative thermogram for the interaction of ADP with the CaM•eEF-2K complex is shown. The active site was blocked using the ATP-competitive inhibitor A484954. The K_D represents the value obtained by a fit to a single site model (the stoichiometric constant was fixed to 1) for the specific thermogram shown. (B) Variation of the CaM sensitivity (K_CaM) of the activity of wild-type peEF-2K against the Soxtide substrate with ADP concentration (C_ADP) is shown. Circles and error bars indicate the average and SD values, respectively, obtained from fits of the observed phosphorylation rates (k_obs) with CaM concentration at a fixed C_ADP value (SI Appendix, Fig. S9A). The red lines indicate 95% confidence bounds of the fit. (C–F) Changes in k_obs with CaM concentration in the absence (filled circles) or in the presence of a saturating amount of ADP (900 μM; open circles). Data are shown for (C) full-length wild-type (WT) peEF-2K and the corresponding BP mutants: (D) R114A, (E) R116A, and (F) H230Q. The circles and error bars indicate mean values and corresponding SDs over two replicates for each data point.

Table 2.

Influence of ADP on the CaM-mediated activity of full-length wild-type eEF-2K and corresponding BP mutants^*

Enzyme	kobs,max (sec−1)		KCaM (nM)
	No ADP	900 μM ADP	No ADP	900 μM ADP
WT	0.20 ± 0.01	0.18 ± 0.01	16.4 ± 3.6	1.9 ± 0.7
R114A	0.26 ± 0.01	0.29 ± 0.01	4.9 ± 1.2	5.6 ± 1.3
R116A	0.24 ± 0.01	0.20 ± 0.01	7.9 ± 2.2	5.2 ± 1.7
H230Q	0.18 ± 0.01	0.12 ± 0.01	46.5 ± 11.9	37.9 ± 9.0

^*Parameters determined by simultaneous fits of datasets acquired in duplicate for each species at a given ADP concentration to Eq. 1. The errors correspond to the SEs obtained from the corresponding nonlinear least squares fits.

To further assess the origin of this ADP-induced effect, we compared the influence of CaM on the activity of wild-type peEF-2K with those of the BP mutants, R114A, R116A, and H230Q, in the absence or presence of a saturating concentration of ADP (900 μM; Fig. 3 C–F). As illustrated in Fig. 2A (shown schematically in SI Appendix, Fig. S8B, and listed in SI Appendix, Table S3), the R114 and H230 sidechains contact the α– and β–phosphates of ADP engaged at the BP. R116 forms stabilizing contacts with the adenine ring of ADP and forms hydrogen bonds with its ribose moiety. These residues are universally conserved in eEF-2K (SI Appendix, Fig. S10). Indeed, in contrast to wild-type peEF-2K which shows nearly a ninefold reduction in K_CaM, none of the mutants show altered CaM-sensitivity in the presence of ADP (Table 2). The k_obs,max is not significantly affected by the presence of ADP for either wild-type eEF-2K or the BP mutants (Table 2). Taken together, these data suggest that the engagement of ADP at the BP, as illustrated by the structure of the NB complex, enhances the affinity of peEF-2K toward CaM.

Next, to test the validity of our findings in the cellular context, we challenged MCF10A eEF-2K^−/− expressing either wild-type eEF-2K or one of two BP mutants, R114A and R116A, with stimuli known to result in increased eEF-2K activity and enhanced eEF-2 phosphorylation (26). Two specific stimuli, starvation (treatment with Dulbecco’s phosphate buffered saline) or addition of the glycolysis inhibitor, 2-deoxyglucose (2-DG), were tested, and the levels of eEF-2 phosphorylation were compared to those in the corresponding untreated cells. As expected, enhanced levels of eEF-2 phosphorylation are seen for cells expressing wild-type eEF-2K under both conditions relative to untreated cells. However, for cells expressing the corresponding R114A or R116A mutants, the levels of phosphorylated eEF-2 upon treatment with 2-DG (that reduces cellular ATP levels by inhibiting glycolysis) are indistinguishable from those in untreated cells (Fig. 4). However, starvation, that represents an extreme stress condition and induces a more elaborate adaptive response (42), does result in enhanced levels of eEF-2 phosphorylation in cells expressing the BP mutants compared to untreated cells. These findings suggest that the BP represents a unique regulatory site within eEF-2K that plays a crucial role in fashioning the cellular response to specific stimuli, e.g., the suppression of glycolysis, by maintaining optimal levels of phosphorylated eEF-2.

Fig. 4.

eEF-2 phosphorylation in cells expressing wild-type eEF-2K or specific BP mutants. (A) eEF-2K^−/− MCF10A cells expressing wild-type eEF-2K (WT) or specific BP mutants (R114A or R116A) were assessed for eEF-2K activity through eEF-2 phosphorylation (on Thr-56) levels following no treatment (NT; control), starvation with DPBS (STRV), or incubation with the glycolysis inhibitor, 2-deoxyglucose (2-DG). A representative immunoblot is shown on the top panel. Pan-actin is used as loading control. (B) eEF-2 phosphorylation levels in cells expressing wild-type eEF-2K or the R114A or R116A mutants under specific conditions are indicated. Data are normalized to the untreated WT eEF-2K sample. Data represent three independent experiments (N = 3); all data points are shown; the corresponding mean and SDs are indicated by the horizontal and vertical lines, respectively. Statistical significance, obtained through a two-way ANOVA and Tukey post hoc analysis, for relevant pairs are indicated. P value markers: < 0.0001 (****), 0.0001 to 0.001 (***), 0.001 to 0.01 (**), 0.01 to 0.05 (*).

Discussion

In our continuing efforts to elucidate the mechanisms underlying the CaM-mediated activation and activity of eEF-2K using eEF-2K_TR as a structural and functional proxy, we obtained the structure of the CaM•peEF-2K_TR complex bound to ATP and ADP. In this structure, ATP, in its intact, unhydrolyzed form, is bound to the peEF-2K_TR active site engaging most of the residues previously predicted to be involved in its recognition (33). However, the catalytically essential Mg²⁺ ion is absent. The lack of Mg²⁺ at the active site can be attributed to the abundance of F^- ions in the crystallization buffer and the low solubility of MgF₂. Surprisingly, CaM_C bound to the peEF-2K_TR CTM contains two ions assigned to Ca²⁺, as was the case in the high-Mg²⁺ structure (33), despite the similarly low solubility of CaF₂.

As described above, we made the unexpected discovery of ADP bound to the BP that forms at the interface between the eEF-2K_TR KD_N and CaM_C. Our biochemical and cell biological experiments suggest that this interaction stabilizes CaM within its complex with eEF-2K (reducing K_CaM) without a significant effect on the nature of its active state (no substantial change in k_obs,max, a proxy for k_cat). The impact of this interaction is to enhance eEF-2K activity by boosting the formation of the active CaM•peEF-2K complex at lower CaM concentrations. A K_ADP (80 ± 14 μM) value for the CaM•peEF-2K complex is comparable in magnitude to the measured K_M (66 ± 1 μM) for ATP (26). This suggests roughly equal affinities of ATP (ATP•Mg²⁺) and ADP for the active site and the BP (in the context of the CaM•peEF-2K complex), respectively. Based on measurements of the nucleotide affinities for the KD of the related MHCK-A, we expect ADP to preferentially engage the BP of the CaM•peEF-2K complex (43), and given the measured positive cooperativity of binding between ADP and CaM, we expect the ADP occupancy of an “incomplete” BP to be significantly lower (a reduction in affinity by ~10-fold may be estimated) in the free kinase. Based on the geometry of the BP (SI Appendix, Fig. S8A), it is unlikely that it can efficiently accommodate ATP (in fact, none is seen bound at this site in the present structure). Additionally, AMP, that could presumably engage the BP, albeit without the stabilizing influence of the numerous interactions involving the absent β—phosphates, does not enhance the CaM affinity of the complex, even at levels far exceeding its cellular concentration (SI Appendix, Fig. S9B). We expect that this engagement with ADP at the BP becomes crucial in sustaining eEF-2K activity and restricting translation under energy-deprived conditions with decreased cellular levels of ATP but stable (or increasing) levels of ADP, as is the case upon the inhibition of glycolysis.

It is notable that a large majority of the key eEF-2K BP residues involved in contacting ADP are distinct from their spatially equivalent counterparts in the KDs of the related MHCK-A (38) and TRPM7 (44) (SI Appendix, Fig. S8B). Additionally, since eEF-2K is the only α-kinase known to be activated by CaM, and its interaction with CaM is necessary for a “complete” BP (SI Appendix, Fig. S8A), it is highly likely that the ADP-mediated regulation mechanism discovered here is unique to eEF-2K.

It is now well established that eEF-2K senses the cellular energy state and suppresses elongation by phosphorylating eEF-2 under energy-deprived conditions, thereby promoting energy homeostasis (20, 45, 46). However, this sensing role was suggested to be an indirect one (45, 47) mediated by AMPK, the master sensor of the cellular energy state (48). In this scenario, AMPK activates eEF-2K by directly phosphorylating it at a secondary R-loop stimulatory site, S398 (45). However, our preliminary results suggest that phosphorylation of eEF-2K on S398 does not substantially alter its enzymatic properties in vitro (49). To maintain high levels of phosphorylated eEF-2 during energy deprivation, the proposed ADP-sensing mechanism of eEF-2K to sustain its activity could complement other operative mechanisms, e.g., the AMPK-mediated downregulation of mTORC1 suggested to inhibit eEF-2K by adversely affecting its interaction with CaM (50) or the recently discovered role of the γ-subunit of AMPK in sequestering the phosphatase PPP6C thereby preventing the dephosphorylation of eEF-2 (51). Further, it has been suggested that, during muscle contractions, the enhancement of phosphorylated eEF-2 levels and the resultant suppression of protein synthesis are mediated by up-regulated eEF-2K activity in an AMPK-independent fashion (52). It is possible that under these circumstances, the proposed ADP-sensing mechanism could serve to enhance the stability of CaM within the active complex to drive eEF-2K activity.

Given the cellular concentration of ADP (0.3 to 1.3 mM) (53) and the apparent affinity of ADP for the complete BP in the context of the CaM•peEF-2K complex (SI Appendix, Fig. S8A), one may expect significant occupancy of the BP by ADP in the absence of any other modulatory effects. It is therefore likely that this ADP-mediated regulation also contributes to the basal activity of eEF-2K as suggested by the reduction in the levels of phosphorylated eEF-2 in untreated cells expressing the ADP-binding-compromised eEF-2K mutants compared to those expressing the wild-type enzyme (Fig. 4). In this scenario, the upregulation due to ADP binding competes with inhibitory signals mediated by pathways such as mTORC1 (54, 55) to maintain optimal basal activity. In the mutants, where the ability to bind ADP is compromised, the influence of the inhibitory effects is enhanced and leads to a reduction in basal eEF-2K activity and lower levels of phosphorylated eEF-2.

It is possible that the BP itself could be a site for the push-pull effect described above. Given that several interactions at the BP involve the phosphate backbone of ADP (Fig. 2A and SI Appendix, Table S3), it is tempting to speculate that it could also host a phosphorylated R-loop residue in competition with ADP. Such an engagement of the BP by a phosphorylated residue would, through the adjoining R-loop segment, create a steric barrier for the association of CaM_C with KD_N, thereby disrupting the activation spine (SI Appendix, Fig. S2) and inactivating eEF-2K. An inhibitory mechanism such as the one proposed would be operative under normal cellular conditions when ATP is in substantial excess of ADP, and nutrients are plentiful. The R-loop contains several inhibitory phosphorylation sites that could potentially target the BP (SI Appendix, Fig. S11A). Many of these are housed on the segment of the R-loop that was removed to generate the eEF-2K_TR construct (SI Appendix, Figs. S1 and S9) to facilitate its crystallization. Indeed, eEF-2K_TR, while similarly activated by CaM as the wild-type enzyme, represents a hyperactive form in cells (32). This feature is partially attributable to the absence of several regulatory phosphorylation sites that could modulate the association of eEF-2K with CaM, an interaction that is the principal driver of kinase activity (26). Potential sites for this negative regulatory control could include S359, targeted through the mTOR pathway [a pathway that is active under normal cellular conditions and disrupted by AMPK under energy stress (56, 57)], in addition to p38δ signaling (58). Additionally, S366, a target of the p70 S6 and p90^RSK kinases (59) could also potentially engage the BP. The expected flexibility of the R-loop around these sites would allow them to access the BP without substantially disrupting the ordered structural elements of eEF-2K (SI Appendix, Fig. S11A). Indeed, our preliminary analyses suggest that a saturating amount of ADP can reverse the effects of eEF-2K inhibition induced by p38δ treatment in vitro (SI Appendix, Fig. S11B). Whether this is the result of ADP outcompeting phosphorylated residue/s engaged at the BP, or a consequence of some other regulatory mechanism/s, requires further analysis. Thus, it remains to be seen whether the BP represents a general regulatory site that could host a phosphorylated residue, other metabolites, or a site that is exclusive to ADP.

Our discovery of the structural and regulatory features of ADP as an activator of eEF-2K provides additional practical opportunities. It has been shown that the eEF-2K/eEF-2 nexus represents a key feature in the adaptation of cancer cells to nutrient deprivation (20), making eEF-2K activity a suitable target for therapeutic intervention (60). As mentioned above, the BP is formed exclusively through the interaction between CaM and eEF-2K providing it with physiochemical features that are likely unique to this particular kinase. The BP, therefore, represents a specific druggable pocket to modulate the CaM/eEF-2K interaction and, thereby, the activity of eEF-2K.

Materials and Methods

Protein Expression and Purification.

eEF-2K_TR and CaM were expressed and purified, the former was selectively phosphorylated at T348 (to generate peEF-2K_TR), and the 1:1 heterodimeric CaM•peEF-2K_TR complex was generated using previously described protocols (32, 61). Wild-type full-length eEF-2K (62) was expressed using a codon-optimized construct using essentially the same protocol used for the expression and purification of eEF-2K_TR. The BP mutants, H230Q, R114A, and R116A, were made using site-directed mutagenesis with the bacterial expression plasmid, pET32a Trx-His₆-eEF-2K (62), or the mammalian expression plasmid, pCDNA3 eEF-2K (26).

Crystallization and Structure Determination.

Crystals were obtained by vapor diffusion at room temperature using an Art Robbins Intelli-Plate 96-3 LP. Crystals of the NB complex were obtained by mixing, in a 1:2 ratio (0.2 μL volumes), an 11.0 mg/mL stock solution of the CaM•peEF-2K_TR complex in a buffer containing 20 mM Tris (pH 7.5), 100 mM NaCl, 1.0 mM TCEP, 3.0 mM CaCl₂, 1.5 mM MgCl₂, and 1.0 mM ATP with a solution containing 100 mM Bis-tris propane (pH 7.4), 100 mM NaF, 20.5 % w/v PEG-3350. Data were collected on the AMX beamline at the NSLS-II light source (Brookhaven National Labs). Crystals for the NF complex were obtained by mixing, in a 2:1 ratio (0.2 μL volumes), a 10.3 mg/mL stock solution of the CaM•peEF-2K_TR complex in the same buffer as above but without ATP and a solution containing 100 mM Bis-tris propane (pH 5.9), 200 mM NaF, and 17.6 % w/v PEG-3350. Data were collected at the FMX beamline at NSLS-II.

Data were processed using the autoPROC toolbox (63), with the NB crystals that showed moderately anisotropic diffraction being integrated using ellipsoidal shells (64). The initial models were built by introducing our previously deposited structure of the CaM•peEF-2K_TR complex (high-Mg²⁺; PDB: 7SHQ) (33) after removing CaM_N as input for Phaser (65). The models were further manipulated and refined using Coot (66), phenix.refine (67), and ISoLDE (68). The ATP and ADP ligands were built using eLBOW (69) and fitted using phenix.ligandfit; the metal ions were assigned using the CheckMyBlob server (70).

ITC Measurements.

For the isothermal calorimetry measurements, samples of CaM and unphosphorylated full-length eEF-2K were prepared in a buffer containing 25 mM N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid (HEPES) (pH 7.5), 50 mM KCl, 0.1 mM ethylenediaminetetraacetic acid (EDTA), 0.1 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), and 1 mM tris(2-carboxyethyl)phosphine (TCEP). Each sample was further concentrated using spin columns after adding 5 mM CaCl₂ and 10 mM MgCl₂, 60 mM of A-484954 [0.13% ²H-labeled dimethyl sulfoxide (DMSO)] and degassed before the experiments. Titrations, in triplicate, were performed at 25 °C on an iTC200 (Malvern) by loading 24.1 μM eEF-2K, 36.1 μM CaM, 60 μM A-484954 in the cell, and 754 μM ADP, 36.1 μM CaM, and 60 μM A-484954 in the syringe. The protein concentrations were assessed using the absorbance at 280 nm. Nineteen data points were collected by injecting 2 μL in 4 sec, with 180-s delays with stirring at 750 rpm using a reference power of 10 μCal/s. A reference isotherm was collected by loading the same titrant in the syringe, while the protein-free buffer containing 60 μM A-484954 (40) was loaded into the cell. Data were fitted with a one-site model restricting the number of sites (N) to 1 using the Origin-7 software provided by the manufacturer.

Kinase Assays.

All assays were performed in activity buffer [25 mM HEPES pH 7.5, 50 mM KCl, 10 mM free MgCl₂, 150 μM CaCl₂, 20 μg/mL BSA, 100 μM EGTA, and 2 mM dithiothreitol (DTT)] at 30 °C in a 384-well plate to a final volume of 60 μL. A 40% serial dilution of CaM (0 to 1.5 μM) into activity buffer was performed across 14 wells. A solution with peEF-2K (3 nM) and Soxtide (10 μM) (71) containing ADP at several concentrations (0, 62.5, 125, 250, 500, and 900 μM) in activity buffer was added to each well. The mixture was incubated at 30 °C for 10 min before initiating the reaction by adding 1 mM ATP. Product turnover was monitored through fluorescence (excitation at 360 nm, emission at 482 nm) using a Synergy H4 plate reader (BioTek). Control assays were performed using 0 nM or 23 nM eEF-2K. The fluorescence data were converted to the k_obs values using the procedures described by Devkota et al. (71) and then to the fractional activity (SI Appendix, Fig. S9A). Data were collected in triplicate for all ADP concentrations except for 900 μM, that was acquired in duplicate. K_CaM values at a given ADP concentration were obtained by a combined fit of all data measured at a particular ADP concentration to Eq. 1.

$$\begin{array}{c}{k}_{\mathrm{o}\mathrm{b}\mathrm{s}}=k_0+k_{\mathrm{o}\mathrm{b}\mathrm{s},\mathrm{m}\mathrm{a}\mathrm{x}}\frac{\left(C_{\mathrm{E}}+C_{\mathrm{C}\mathrm{a}\mathrm{M}}+K_{\mathrm{C}\mathrm{a}\mathrm{M}}\right)-\sqrt{(C_{\mathrm{E}}+C_{\mathrm{C}\mathrm{a}\mathrm{M}}+K_{\mathrm{C}\mathrm{a}\mathrm{M}}{)}^2-4C_{\mathrm{E}}{C}_{\mathrm{C}\mathrm{a}\mathrm{M}}}}{{2C}_{\mathrm{E}}}.\hfill \end{array}$$ [1]

k₀, k_obs, and k_obs,max represent the basal rate (in the absence of CaM), that observed at a CaM concentration of C_CaM, and the maximal rate, respectively. The concentration of eEF-2K (C_E) was fixed to 3 nM. The K_CaM values (and corresponding SEs) obtained from the nonlinear least squares fits to Eq. 1 at each concentration of ADP (C_ADP) were then fitted to Eq. 2 to obtain the apparent K_ADP.

$$K_{\mathrm{C}\mathrm{a}\mathrm{M}}=K_{\mathrm{C}\mathrm{a}\mathrm{M},0}-\frac{K_{C\mathrm{a}\mathrm{M},\mathrm{s}\mathrm{a}\mathrm{t}}{C}_{\mathrm{A}\mathrm{D}\mathrm{P}}}{K_{\mathrm{A}\mathrm{D}\mathrm{P}}+C_{\mathrm{A}\mathrm{D}\mathrm{P}}}.$$ [2]

K_CaM,sat and K_CaM,0 represent the affinities at saturating ADP concentration and that in its absence (i.e., for C_ADP = 0; fixed to the value obtained from the corresponding fit to Eq. 1, above), respectively. An additional set of analyses were carried out with wild-type eEF-2K to measure K_CaM with and without 1 mM AMP (SI Appendix, Fig. S9B). All data analyses were carried out using Prism 9 software.

To compare the influence of K_CaM values for wild-type peEF-2K with those of the corresponding R114A, R116A, and H230Q mutants, the same protocols as above were used with 5 nM (wild-type, R114A, and R116A) or 3 nM (H230Q) enzyme, in the presence and absence of 900 μM ADP, with varying concentrations of CaM (0 to 625 nM for WT, R114A, and R116A, 0 to 1,008 nM for H230Q). Datasets collected at a particular ADP concentration (in duplicate) were fitted simultaneously to Eq. 1 to obtain the corresponding K_CaM and k_obs,max values shown in Table 2.

Cell Culture.

MCF10A eEF-2K^−/− cells (Sigma-Aldrich) were maintained as described previously (26). Cells were plated onto 6-well plates (0.15 × 10⁶ cells/well). Twenty-four hours after plating, the cells were transfected with 0.1 μg/well of wild-type eEF-2K pcDNA3 FLAG HA, R114A eEF-2K pcDNA3 FLAG HA, or R116A eEF-2K pcDNA3 FLAG HA vector (cloned from plasmid 10792; Addgene) using Lipofectamine 3000^® (Invitrogen), following the manufacturer-provided protocol. Transfected cells were incubated for 24 h to allow protein expression and then either lysed untreated or treated with various stimuli before lysis. Treated cells were washed twice with PBS (pH 7.4) and either incubated with 25 mM 2-DG in complete media or with Dulbecco phosphate-buffered saline (DPBS, Corning) for 2 h. Following treatment, cells were washed twice with ice-cold PBS (pH 7.4) (Invitrogen) and lysed in M-PER™ lysis buffer (Thermofisher Scientific) supplemented with Halt™ protease and phosphatase inhibitor single-use cocktail (Thermo Scientific). Lysates were cleared via centrifugation at 15,000 g for 15 min, and protein concentrations were determined using a Bradford assay (Bio-Rad).

Western Blot Analysis.

From each sample, 40 μg of protein aliquots was subjected to SDS-PAGE and analyzed by western blotting using previously described protocols (27). Images were quantified using ImageJ software, and Thr-56-phosphorylated eEF-2 was normalized to pan-actin as a loading control. The statistical significance of the changes noted under various conditions was determined using a two-way ANOVA followed by a Tukey post hoc analysis using Prism 9 software.

The following commercial antibodies were used in the study: peEF-2 (Thr-56) (Catalog number 905-775-100; Assay Designs); eEF-2K (C-12) (Catalog number: sc-390710;Santa Cruz Biosciences); eEF-2 (Catalog number: 2332; Cell Signaling); anti-actin, clone C4 (Catalog number: MAB1501; Millipore); and IRDye^® 800CW Goat anti-Rabbit (Catalog number: 926-32211; LiCor) and IRDye^® 680RD Goat anti-Mouse (Catalog number: 926-68070; LiCor).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The atomic coordinates and structure factors for the NF and NB complexes have been deposited into the Protein Data Bank (PDB) with accession codes 8FO6 (72) and 8FNY (73), respectively. Corresponding information for additional structures discussed in the manuscript may be found under the cited accession codes in the PDB. All other data may be found in the manuscript and accompanying SI Appendix.