ERK1/2 interaction with DHPS regulates eIF5A deoxyhypusination independently of ERK kinase activity

SUMMARY

This study explores a non-kinase effect of extracellular regulated kinases 1/2 (ERK1/2) on the interaction between deoxyhypusine synthase (DHPS) and its substrate, eukaryotic translation initiation factor 5A (eIF5A). We report that Raf/MEK/ERK activation decreases the DHPS-ERK1/2 interaction while increasing DHPS-eIF5A association in cells. We determined the cryoelectron microscopy (cryo-EM) structure of the DHPS-ERK2 complex at 3.5 Å to show that ERK2 hinders substrate entrance to the DHPS active site, subsequently inhibiting deoxyhypusination in vitro. In cells, impairing the ERK2 activation loop, but not the catalytic site, prolongs the DHPS-ERK2 interaction irrespective of Raf/MEK signaling. The ERK2 Ser-Pro-Ser motif, but not the common docking or F-site recognition sites, also regulates this complex. These data suggest that ERK1/2 dynamically regulate the DHPS-eIF5A interaction in response to Raf/MEK activity, regardless of its kinase function. In contrast, ERK1/2 kinase activity is necessary to regulate the expression of DHPS and eIF5A. These findings highlight an ERK1/2-mediated dual kinase-dependent and -independent regulation of deoxyhypusination.

Graphical Abstract

In brief

Becker et al. determine the cryo-EM structure of the DHPS-ERK2 complex to show that ERK2 hinders DHPS active site entrance through an atypical interface. Raf/MEK/ERK pathway activation decreases DHPS-ERK1/2 interaction, thereby facilitating increased DHPS-eIF5A association. Kinase function is unnecessary for ERK1/2 to mediate this regulation.

INTRODUCTION

The Raf/MEK/extracellular regulated kinase 1/2 (ERK1/2) pathway is a critical mitogen-activated protein kinase pathway that regulates cell proliferation, survival, and differentiation.^1,2 The Raf/MEK/ERK pathway is activated by various mitogen stimulations that receptor tyrosine kinases or G-protein-coupled receptors relay to turn on the Ser/Thr kinases Raf (A-Raf, B-Raf, C-Raf [also called Raf-1]). Active Raf, in turn, phosphorylates and activates the dual-specificity kinase MEK1 and its homolog MEK2 (collectively referred to as MEK1/2), which in turn phosphorylate and activate the ubiquitously expressed Ser/Thr kinase ERK1 and its homolog ERK2 (collectively referred to as ERK1/2) by sequentially phosphorylating Thr202/Tyr204 of ERK1 and Thr185/Tyr187 of ERK2.³ ERK1 and ERK2 are >84% identical at the amino acid level and have highly overlapping functions under most physiological conditions, accounting for most, if not all, effects mediated by MEK1/2. Therefore, MEK1/2 and ERK1/2 are usually inseparable in addressing their physiological effects. ERK1/2 phosphorylate diverse proteins containing the Ser/Thr-Pro signature, although Pro-Xaa-Ser/Thr-Pro is a preferred substrate signature.^4,5

ERK1/2 interact with their substrates, regulators, and scaffolds through distinct motifs present in these proteins, which determines the specificity of ERK1/2 signaling. For example, many ERK1/2-interacting proteins contain a conserved docking domain (D-domain) that consists of two basic residues followed by a cluster of hydrophobic residues or the F-site that contains the Phe-Xaa-Phe-Pro signature.^6,7 The D-domain interacts with the common docking (CD) site of ERK1/2 that consists of a few acidic residues followed by a cluster of hydrophobic residues, while the F-site binds to the hydrophobic F-site recruitment site (FRS) of ERK1/2.^6,7 Albeit less well characterized, the Ser-Pro-Ser (SPS) motif can facilitate ERK1/2 binding to importin-7 and, subsequently, nuclear localization upon phosphorylation by casein kinase II.^8,9 ERK1/2 regulate diverse cellular processes, including transcription, translation, adhesion, migration, and metabolism in various intracellular compartments.² While these functions of ERK1/2 are mediated mainly by their kinase activity, it is also known that ERK1/2 can mediate non-kinase effects.¹⁰ For example, kinase-deficient mutants of ERK1/2 can promote the activation of poly(ADP-ribose) polymerase, MAP kinase phosphatase-3, and topoisomerase IIα.^11-13 We also demonstrated that active site-, but not activation loop-, disabled ERK1/2 can facilitate Raf/MEK-induced growth arrest responses.¹⁴

Deoxyhypusine synthase (DHPS) is an enzyme that catalyzes hypusination of the eukaryotic translation initiation factor 5A (eIF5A).^15-17 Hypusination is a highly specific post-translational modification detected only in eIF5A.^18-20 Two sequential enzyme reactions mediate eIF5A activation. First, DHPS conjugates the 4-aminobutyl moiety of spermidine to a specific lysine residue of eIF5A to produce deoxyhypusinated eIF5A (eIF5A^Dhp).^21-23 Deoxyhypusine hydroxylase (DOHH)-mediated hydroxylation of deoxyhypusine residue follows this step, irreversibly producing hypusine.^24,25 These biochemical modifications enable eIF5A to promote translational elongation of the polypeptide chain by resolving stalling caused by disruptive sequences such as polyproline motifs and sequential charged residues.^26-28 Inhibition of hypusination suppresses the proliferation of different cell types, including normal as well as tumor cells,¹⁹ while upregulated expression of DHPS, DOHH, and eIF5A is detected in various cancers, including pancreatic ductal adenocarcinoma, lung adenocarcinoma, glioblastoma, and colorectal carcinoma.^29-34 Although kinases such as protein kinase C and casein kinase II have been shown to induce DHPS phosphorylation,^35,36 the effect of ERK1/2 on this enzyme and the hypusination cascade is unclear. In this study, we visualized a physical interaction between DHPS and ERK1/2 and investigated the underlying molecular mechanisms thereof.

RESULTS

ERK2 binds DHPS at the eIF5A-DHPS interface and covers the active site entrance

We identified DHPS from a tandem affinity purification screening using ERK2 as the bait (STAR Methods) and validated their physical interaction by performing co-immunoprecipitation (coIP) in cells and in vitro binding assays (Figures 1A and 1B). To understand the molecular basis of DHPS’s interaction with ERK1/2, we performed single-particle cryoelectron microscopy (cryo-EM) to visualize the architecture of the DHPS-ERK2 complex, which was reconstituted in vitro using heterologously expressed and purified recombinant proteins. Preliminary processing of the collected dataset and inspection of initial two-dimensional (2D) classes indicated that ERK2 could interact with DHPS in various stoichiometries. Because the active form of DHPS is a tetramer, up to 1:4 complex stoichiometry (i.e., 1 DHPS tetramer to 4 ERK2 molecules) is feasible. Indeed, a detailed analysis of our cryo-EM dataset (Figure S1) revealed six distinct 3D classes reflecting all possible complex modalities (Figure 1C). Imposing symmetry during the refinement of the symmetric complexes had little impact on the map quality apart from the 1:4 complex, which was reconstructed from a significantly smaller number of particles (Figure S2A; Table S1). A microscale thermophoresis (MST)-based approach recapitulated up to 1:4 stoichiometry in solution (Figure S2B).

Figure 1. ERK2 binds DHPS at the eIF5A-DHPS interface and covers the active site entrance.

(A) Western blot detection of DHPS in the IP fraction prepared using an ERK1/2-specific antibody from a total lysate of HEK293 cells.

(B) Pull-down assay showing the in vitro interaction between C-terminal six histidine (HIS)-tagged DHPS (HIS-DHPS) and untagged ERK2. C-terminal HIS-tagged PopB (PopB-HIS) is the negative control. See also the supplemental information for tandem affinity purification and mass spectrometry (MS) data.

(C) Cryo-EM maps reflecting all possible DHPS-ERK2 complex modalities.

(D and E) Cryo-EM map of the 1:1 complex post-processed with DeepEMhancer (D) and corresponding molecular model showing that ERK2 binds at the DHPS active site entrance (E).

(F) Complex interface as per DIMPLOT analysis. See also Figure S1 for cryo-EM workflow and Table S1 for map and model statistics, as well as Figure S2 for maps reconstructed with symmetry imposed, stoichiometry validation in solution, and detailed views of the interface.

ERK2 binding did not induce any global conformational change in the DHPS structure, suggesting that all four binding sites are equivalent and independent. The resolution of the map of a fully saturated complex did not allow for model building. However, the highest-resolution map of the complex reflecting 1 DHPS tetramer bound to 1 ERK2 molecule enabled us to solve the structure of DHPS-ERK2 at 3.50 Å resolution, revealing that the interaction interface on the DHPS comprises the surface residues directly surrounding the entrance to the active site (Figures 1D and 1E; Table S1). The main part of the DHPS-ERK2 interface engages the same DHPS region that we recently showed to be involved in eIF5A binding,¹⁷ although it does not encompass the amino acids located deeper within the tunnel leading to the DHPS active site (Figure 1F). This interface appears to also engage parts of an α helix (residues 173–195, chain B) and the end of another α helix (residues 240–251, chain B), which may interact with the ERK2 activation loop as well as a loop covering DHPS amino acids 272–275 (chain A; Figure S2C). On the ERK2 side, our data showed that parts of an α helix (residues 233–244) and its preceding loop, which extends slightly toward the DHPS active site, mediate complex formation (Figure S2C). This interaction also seems to involve an end of an α helix with the beginning of the following loop (residues 263–268) and the activation loop containing the TEY motif (Figure S2C). Moreover, the DHPS-ERK2 interaction causes the C-lobe of ERK2 to locate over the entrance to the DHPS active site, while the ERK2 N-lobe hangs over the distal DHPS protomer-chain B (Figure 1E).

Single hotspot governs the DHPS-ERK2 interaction

To verify the ERK2-DHPS interaction interface identified with cryo-EM, we performed a structure-guided site-directed mutagenesis and an MST analysis measuring the dissociation constants of interaction against a wild-type partner. We found that the K_d values for DHPS-E180A and DHPS-D243A mutants are almost ten and five times higher than the values for wild-type DHPS, respectively (Figures 2A and 2B, top, and S3A). On the ERK2 side, the K_d values for H232A and R191A mutants were over ten and almost four times higher, respectively, than those for wild-type ERK2 (Figures 2A and 2B, bottom, and S3B). Albeit with a weaker impact, K231, N238, H239, and Y187 of ERK2 also participated in complex formation (Figures 2A and 2B, bottom, S3B, and S3C). Subsequently, our structural data revealed that R191 and H232 of ERK2 interact via hydrogen bonds with E180 and D243 in DHPS chain B, respectively, to form a clear hotspot in this interface (Figure 2C). These data suggest that E180 and D243 of DHPS and H232 and R191 of ERK2 are essential residues for complex formation. Surprisingly, this topological arrangement is quite similar to the DHPS-eIF5A complex because H232^ERK2 and R191^ERK2 correspond to H51^eIF5A and K55^eIF5A, respectively (Figure 2D), and D243^DHPS is also critical for DHPS-eIF5A binding, as we previously demonstrated.¹⁷ However, in contrast to the DHPS-eIF5A complex, two DHPS protomers do not equally contribute to the formation of the DHPS-ERK2 complex, as both E180 and D243, which pair with the ERK residues, are located in a single protomer of DHPS (chain B in the cryo-EM structure).

Figure 2. A single hotspot governs the DHPS-ERK2 interaction, and ERK2 binds preferentially to enzymatically active DHPS.

(A) MST data (mean ± SD, n = 4) for the alanine scanning of the DHPS-ERK2 interface from the DHPS (top) and ERK2 (bottom) sides.

(B) Mapping the effect of DHPS (top) and ERK2 (bottom) mutations on the K_d. The visualizations are color coded with the scale in the middle and based on the source data provided in Figure S3.

(C) Two crucial interactions revealed by structure-guided site-directed mutagenesis. See Figure S3 for snapshots of additional interactions involved in the DHPS-ERK2 interface.

(D) Structural details of DHPS-ERK2 complex highlighting the residues involved in core interactions, which are also used in the DHPS-eIF5A interface.

(E) MST data (mean ± SD, n = 4) assessing the effect of NAD, spermidine (SPD), and NADH on the binding of wild-type ERK2 and DHPS. The exact values of measured dissociation constants are shown on the right.

ERK2 binds preferentially to the enzymatically active DHPS

We previously showed that NAD binding affects the structural rigidity of DHPS.³⁷ Because NAD is an essential cofactor for DHPS, we determined how it affects DHPS’s interaction with ERK2. When tested within the range of physiologically relevant concentrations, NAD could lower the K_d of the DHPS-ERK2 interaction by a factor of five (Figure 2E), likely through stabilizing the DHPS structure in a favorable state for ERK2 binding. We also found that NADH, which the DHPS enzymatic cycle generates, can affect the DHPS-ERK2 interaction to an extent similar to NAD (Figure 2E). Therefore, this cofactor might regulate the DHPS-ERK2 interaction independently of its redox state. In contrast, spermidine, the substrate for DHPS catalysis, did not by itself affect DHPS-ERK2 binding, although it augmented the NAD effect by further decreasing the K_d value to ~1 μM (Figure 2E). These data suggest that ERK1/2 might preferentially bind to catalytically active DHPS.

ERK1/2 do not phosphorylate DHPS

DHPS contains two putative ERK1/2 substrate motifs (Ser/Thr-Pro) at Thr202 and Ser233, respectively; thus, we determined whether DHPS is a substrate for ERK1/2 using an antibody specific to the ERK1/2-phosphorylated substrate signature–phosphorylated Ser/Thr-Pro. This antibody did not detect any signal associated with ΔRaf-1:estrogen receptor (ER) activation in our IP-western blot analyses of the pull-down fractions of endogenous and exogenous DHPS from HEK293 cells (Figures 3A and S4A). This antibody detected only a weak signal that matched DHPS and was present irrespective of ERK activity. However, in our positive control assay, this antibody detected a strong signal in the pull-down fractions of the ERK1/2 substrate ribosomal S6 kinase 1 (RSK) upon ΔRaf-1:ER activation (Figure S4B).

Figure 3. ERK1/2 do not phosphorylate DHPS, and their interaction with DHPS is negatively regulated by their activation.

(A) IP-western blotting determining whether ERK1/2 phosphorylate DHPS. The IP fractions (left) were pulled down by a DHPS-specific antibody and its control normal immunoglobulin (Ig)G from total lysates (right) of HEK293-ΔRaf:ER cells treated with 1 μM 4-hydroxytamoxifen for 4 h for ΔRaf:ER activation (Raf act.) and then immunoblot analyzed using an antibody specific to the phosphorylated Ser/Thr-Pro motif (ERK1/2 substrate signature). β-Tubulin is the control for equal protein loading. See also Figure S4 for additional data.

(B) ATPγS-based in vitro kinase assay conducted using recombinant active ERK2 (pERK2) and DHPS. Myelin basic protein (MBP) is the positive control for pERK2-mediated phosphorylation. Thiophosphorylated proteins were immunoblot detected using an anti-thiophosphate antibody. See STAR Methods for details.

(C) Structural details of the DHPS-ERK2 complex validating that DHPS is not an ERK2 substrate. Potential ERK2 phosphorylation sites on DHPS Thr202 (green) and Ser233 (blue) in the two closest chains, as well as ERK2 catalytic Lys54 (gray), are shown as spheres. The DHPS α helix preventing the ERK2 catalytic cleft from accessing the DHPS Thr202 is marked as well (solid cyan).

(D) IP-western blotting of total lysates of HEK293-ΔRaf:ER cells infected with the lentivirus expressing C-terminal HA-tagged DHPS (DHPS-HA) and subjected to 4 h tamoxifen-induced ΔRaf:ER activation (Raf act.). pHAGE is the control empty virus.

(E) IP-western blotting of total lysates of A375 cells infected with aforementioned lentiviral constructs and treated with 100 nM AZD6244 (MEK inhibitor) for 24 h. (D and E) Shown on the right is the densitometry of western blot signals normalized to HA signals. β-Actin is the loading control.

Numeric data are presented as mean ± SEM of biological duplicates (B) and triplicates (D and E). **p < 0.005 (by two-tailed Student’s t test).

Next, we conducted an in vitro kinase assay using recombinant DHPS, TEY-phosphorylated active ERK2, and a bio-orthogonal ATPγS analog that ERK2 can use to thiophosphorylate its substrates.³⁸ Thiophosphorylated proteins in this reaction can be specifically detected by a thiophosphate ester-specific antibody.³⁸ In our positive control assay based on myelin basic protein (MBP), a known ERK1/2 substrate,³⁹ this antibody detected a strong signal matching the size of MBP only in the presence of active ERK2 (Figure 3B, left two lanes). However, it did not recognize any signal from DHPS specifically associated with active ERK2, although it revealed a weak non-specific signal irrespective of the presence of active ERK2 (Figure 3B, right two lanes). Together, these data show that ERK1/2 do not phosphorylate DHPS.

In further support of these results, our structural analysis revealed that, although not buried deep within the protein structure and possibly accessible to ERK2, Thr202^DHPS is located at the interface formed by the DHPS protomers, while ERK2 interacts with DHPS on the other side of its tetrameric assembly, which causes both sites to be spatially separated (Figure 3C). Any conformational changes of this complex bringing Thr202^DHPS and the ERK2 catalytic cleft closer are also unlikely, as access to this residue is guarded by a major DHPS surface helix (residues 173–196; Figure 3C). On the other hand, Ser233^DHPS is buried deep within the protein core, making it unavailable for interaction with the active site of any kinase. Any conformational changes of DHPS exposing this residue to the protein surface and enabling its phosphorylation also seem implausible, as they would require a disruption of the essential structural organization of the DHPS (Figure 3C). Moreover, although the ERK2 catalytic cleft is rotated toward DHPS, it is located far enough from the DHPS surface to disprove DHPS as an ERK2 substrate (Figure 3C).

DHPS’s interaction with ERK1/2 is negatively regulated by ERK1/2 activation

We were able to detect ERK1/2 in the IP fractions of exogenously expressed C-terminal hemagglutinin (HA)-tagged DHPS (DHPS-HA) pulled down by an HA-specific antibody from the lysates of HEK293 and A375, a BRAF^V600E-mutated human melanoma line exhibiting constitutively high MEK/ERK activity (Figures 3D and 3E). Interestingly, we found that 4 h ΔRaf-1:ER activation in HEK293 cells substantially reduced total ERK1/2 in the DHPS-HA pull-down fraction, as determined by an antibody that can recognize both unphosphorylated and phosphorylated ERK1/2, although the ratio of phosphorylated ERK1/2 to total ERK1/2 in this IP fraction increased (Figure 3D). Conversely, AZD6244, an advanced MEK1/2-specific inhibitor,⁴⁰ notably increased total ERK1/2 in the DHPS-HA pull-down fraction from A375 cells (Figure 3E). Together, these data suggest that a physical interaction exists between ERK1/2 and DHPS, which is negatively regulated by the Raf/MEK/ERK pathway.

The activation loop and the SPS motif of ERK1/2, but not an intact active site, are necessary for regulating ERK1/2-DHPS interaction

Given the sensitivity of the DHPS-ERK1/2 interaction to Raf/MEK activity, we sought to determine the molecular mechanism by which the pathway regulates the DHPS-ERK1/2 interaction. First, we examined how the phosphorylation-defective TEY site mutation in the activation loop of rat ERK2, which replaces Thr183 and Tyr185 (corresponding to Thr185 and Tyr187 in human ERK2) with Ala and Phe, respectively (hereinafter referred to as ERK2-TY/AF), and the active-site-disabling mutations, which replace Lys52 and Asp165 (corresponding to Lys54 and Asp167 in human ERK2, respectively) with Arg (ERK2-K52R) and Ala (ERK2-D165A), affect ERK2’s interaction with DHPS in the absence or presence of ΔRaf-1:ER activity. As determined by IP-western blotting, these ERK2 mutants interacted with DHPS-HA without a notable difference in HEK293 cells before ΔRaf-1:ER activation (Figure 4A). After ΔRaf-1:ER activation, in stark contrast, ERK2-TY/AF’s interaction with DHPS persisted, whereas ERK2-K52R and ERK2-D165A exhibited substantially decreased interactions with DHPS (Figure 4A). Since ERK2-K52R and ERK2-D165A have intact activation loops, these data suggest that the activating conformational change, but not the kinase activity, of ERK1/2 is necessary for Raf/MEK to regulate DHPS-ERK1/2 complex formation. Of note, consistent with our alanine scanning analysis above, H230A (corresponding to His232 in human ERK2) mutagenesis abrogated the ability of ERK2-D165A and wild-type ERK2 to interact with DHPS, as determined by IP (Figures 4A and S4C).

Figure 4. The activation loop and the SPS motif of ERK1/2, but not an intact active site, are necessary for regulating the ERK1/2-DHPS interaction.

(A) IP-western blotting of HEK293-ΔRaf:ER cells co-infected with lentiviruses expressing C-terminal HA-tagged DHPS (DHPS-HA) and N-terminal HIS-tagged ERK2 mutants and then subjected to 4 h tamoxifen-induced ΔRaf:ER activation (Raf act.). TY/AF (left), K52R (left), D165A (middle), and D165A/H230A (middle) indicate ERK2-T183A/Y185F, ERK2-K52R, ERK2-D165A, and ERK2-D165A/H230A, respectively. pHAGE is the control virus. Densitometry (right) indicates western blot signals normalized to HA signals. See Figure S4C for ERK2-H230A control data.

(B) IP-western blotting of HEK293-ΔRaf:ER cells co-infected with lentiviruses expressing DHPS-HA and N-terminal HIS-tagged ERK2 constructs and then subjected to 4 h ΔRaf:ER activation. ERK2, D316/D319A, and Y261N indicate wild-type ERK2, ERK2-D316/D319A, and ERK2-Y261N, respectively. Densitometry (right) indicates western blot signals normalized to HIS signals.

(C) IP-western blotting of HEK293-ΔRaf:ER cells co-infected with lentiviruses expressing DHPS-HA and N-terminal HIS-tagged ERK2 constructs and then subjected to 4 h ΔRaf:ER activation. ERK2, APA, and EPE indicate wild-type ERK2, ERK2-S244/246A, and ERK2-S244/246E, respectively. Densitometry (right) indicates western blot signals normalized to HIS signals. β-Tubulin is the loading control.

Numeric data are mean ± SEM of biological triplicates. *p < 0.05, ***p < 0.001, and ****p < 0.0001 (by two-tailed Student’s t test).

(D) MST data (mean ± SD, n = 4) comparing wild-type DHPS binding to wild-type ERK2 (WT) and ERK2-S246/248E (EPE). The exact values of measured dissociation constants are shown below.

(E) Structural details of DHPS-ERK2 complex revealing DHPS’s interaction with the TEY site, but not the SPS, FRS, and CD sites, of ERK2. See Figure S4D for further details of TEY site interaction.

The specificity of ERK1/2 signaling is regulated through different domains and motifs in these kinases, including CD, FRS, and SPS.^6-9 Accordingly, we determined the significance of these regulatory sites for the ERK1/2-DHPS interaction using different ERK mutants. Our IP-western blotting revealed that the CD-groove-defective ERK2-D316/319A (corresponding to D318/321A in human ERK2) mutant or the FRS-defective ERK2-Y261N (corresponding to Y263N in human ERK2) mutant can interact with DHPS without any notable difference compared with the wild-type ERK2 in HEK293 cells (Figure 4B). However, interestingly, we found that the ERK2-APA mutant in which Ser244 and Ser246 of the SPS motif (corresponding to Ser246 and Ser248 in human ERK2) were replaced with the phosphorylation-defective Ala exhibited a low but notable interaction with DHPS in the presence of ΔRaf-1:ER activity, although it interacted with DHPS at similar levels to wild-type ERK2 when the stimulation was absent (Figure 4C). Conversely, the phosphomimetic mutant ERK2-EPE could not interact with DHPS, regardless of ΔRaf-1:ER activity status (Figure 4C). However, on the contrary, recombinant ERK2-EPE bound to DHPS as efficiently as wild-type ERK2 in the MST-based study using purified proteins (Figure 4D). Besides, our structural data indicated that the SPS site does not directly interact with DHPS (Figure 4E). Therefore, we speculate that the SPS site regulates the ERK1/2-DHPS interaction in cells via a mechanism involving as-yet-un-identified regulators. Our structural data also indicated that the ERK1/2 active site, CD groove, and FRS do not directly interact with DHPS, further supporting their insignificance for the ERK1/2-DHPS interaction (Figure 4E). Notably, structural data indicated that the TEY site indeed takes part in the interaction, particularly through Tyr187 (corresponding to Tyr185 in rat ERK2) (Figure 4E), and that this site is located within an electro-negative patch on the DHS surface (Figure S4D). These data are consistent with DHPS-ERK complex dissociation upon phosphorylation of the TEY motif. The importance of Tyr187 is further addressed below.

ERK1/2 regulate DHPS-eIF5A interaction independently of kinase activity

Given the data above that ERK1/2 do not phosphorylate DHPS, although these proteins physically interact with each other in a Raf/MEK-regulated manner, we sought to identify what aspect of DHPS function is regulated by non-kinase ERK1/2 activity. To address this, we determined how ΔRaf-1:ER activity affects the interaction between DHPS and its substrate eIF5A. Our IP-western blotting revealed that 24 h ΔRaf-1:ER activation increased the interaction between the ectopically expressed DHPS-HA and eIF5A while concomitantly decreasing the interaction between DHPS-HA and ERK1/2 in HEK293 cells (Figure 5A). Conversely, 24 h AZD6244 treatment decreased the interaction of DHPS-HA with eIF5A while concomitantly increasing its interaction with ERK1/2 in A375 cells (Figure 5B). When the physical interactions of endogenous proteins were analyzed by IP-western blotting, 24 h AZD6244 treatment consistently decreased the interaction between DHPS and eIF5A while increasing the interaction between DHPS and ERK1/2 in A375 cells (Figure 5C) and MIA PaCa-2, a KRAS^G12C-mutated human pancreatic cancer line (Figure S5A). Similar effects were also observed after 5 min epidermal growth factor (EGF) stimulation of HeLa cells (Figure S5B). Of note, eIF5A did not interact with ERK1/2 (Figure 5C). These data demonstrate an inversed correlation between the DHPS-eIF5A interaction and the DHPS-ERK1/2 interaction in the face of ERK1/2 activity.

Figure 5. ERK1/2 regulate the DHPS-eIF5A interaction independently of kinase activity and competitively inhibit eIF5A deoxyhypusination in vitro.

(A) IP-western blotting of total lysates of HEK293-ΔRaf:ER cells infected with the DHPS-HA virus and subjected to 4 h tamoxifen-induced ΔRaf:ER activation (Raf act.). pHAGE is the control virus. Densitometry indicates coIP signals normalized to HA signals (right).

(B) IP-western blotting of total lysates of A375 cells infected with the DHPS-HA virus and treated with 100 nM AZD6244 for 24 h. An equal volume of DMSO was used as the vehicle control for AZD6244. Densitometry indicates coIP signals normalized to HA signals (right).

(C) IP-western blotting of total lysates of A375 cells treated with 100 nM AZD6244. eIF5A (top) and ERK2 (middle) were immunoprecipitated from the same inputs (bottom). Densitometry below shows co-immunoprecipitated DHPS signals normalized to primary IP signals.

(D) IP-western blotting of HEK293-ΔRaf:ER cells infected with the DHPS-HA virus and subjected to 24 h tamoxifen-induced ΔRaf:ER activation (Raf act.) in the presence or absence of ERK1/2 inhibitors acting through different mechanisms. SCH-772984 inhibits the ERK1/2 activation loop and catalytic site. LY-5214996 inhibits the catalytic site only. An equal volume of DMSO was used as the vehicle control for these inhibitors. Densitometry (right) indicates coIP signals normalized to HA signals. See also Figure S5 for additional data.

(E) Time course in vitro eIF5A deoxyhypusination assay determining DHPS inhibition by ERK2. Western blotting of deoxyhypusinated eIF5A was used as the assay readout. Images of representative blots forming different assay conditions are on the right, and a graph with densitometry is on the left. Control stands for reaction without ERK2. BSA was used as a control for molecular crowding effects.

(F) Dose-dependent ERK2 inhibition of eIF5A deoxyhypusination in the in vitro assay, similar to that described in (E). Western blotting of deoxyhypusinated eIF5A was used as the assay readout. A representative western blot image is shown above the graph with densitometry.

Data are mean ± SD (A–C, E, and F) or mean ± SEM (D) of biological duplicates (B and C) or triplicates (A and D–F). *p < 0.05 and **p < 0.005 (by two-tailed Student’s t test).

To further verify whether ERK1/2 can negatively regulate DHPS-eIF5A interaction by competing for DHPS, we examined the effects of two contrasting ERK1/2-specific inhibitors. SCH-772984 targets the catalytic site and the activation loop of ERK1/2, while LY-3214996 targets only the catalytic site.⁴¹ In HEK293 cells, both inhibitors effectively blocked ΔRaf-1:ER-induced RSK phosphorylation without a notable difference, which confirmed their high efficacy in inhibiting ERK1/2 (Figure 5D, input). Consistent with their expected mechanisms, SCH-772984 substantially inhibited ΔRaf-1:ER-induced ERK1/2 phosphorylation, whereas LY-3214996 did not affect that phosphorylation (Figure 5D, input). Strikingly, in the presence of SCH-772984, ΔRaf-1:ER activity could neither decrease the DHPS-ERK1/2 interaction nor increase the DHPS-eIF5A interaction (Figure 5D, IP). However, in stark contrast, LY-5214996 did not effectively inhibit the effects of ΔRaf-1:ER activity on these protein interactions, although it mildly inhibited the ΔRaf-1:ER-induced DHPS-eIF5A interaction (Figure 5D, IP), showing the significance of ERK1/2 activation status, but not kinase activity, for these ΔRaf-1:ER effects.

Consistent with this, overexpression of eIF5A decreased the interaction between DHPS and ERK1/2 in HEK293 cells in which ΔRaf-1:ER was not activated, a condition that promotes the DHPS-ERK1/2 interaction (Figure S5C). We also verified that the activation-loop-defective ERK2-TY/AF mutant, whose affinity to DHPS is constitutive and not regulated by ΔRaf-1:ER (Figure 3A), can compete with eIF5A for DHPS under a condition in which ΔRaf-1:ER is activated (Figure S5D). Together, these data strongly suggest that ERK1/2 regulate the DHPS-eIF5A interaction via physical interaction, for which activating phosphorylation of their activation loop, but not kinase activity, is critical.

ERK2 competitively inhibits eIF5A deoxyhypusination in vitro

We determined how ERK2 affects DHPS-catalyzed reactions by performing an in vitro eIF5A deoxyhypusination assay, wherein DHPS, eIF5A, and ERK2 are the only proteins present, while spermidine and NAD were also supplemented in the assay buffer (a detailed description available is in STAR Methods). The assay was performed in the absence of DOHH; thus, only eIF5A^Dhp could be formed, which we detected using an antibody recognizing both eIF5A^Dhp and hypusinated eIF5A (eIF5A^Hyp). Western blot analysis of the reaction product indicated that ERK2 can competitively inhibit the DHPS-mediated reaction (Figure 5E). We previously demonstrated that, although the DHPS tetramer possesses four binding sites for eIF5A, only one eIF5A molecule binds to the tetrameric assembly during each catalytic cycle.^17,42 Consequently, three out of the four active site entrances remain unoccupied. While investigating the mechanism by which ERK2 inhibits eIF5A deoxyhypusination, we observed that the inhibition rate did not linearly correlate with the increase in ERK2 concentration. Instead, inhibition occurred abruptly only after ERK2’s concentration surpassed a threshold that could fully saturate the complex (Figure 5F), while gradual inhibition was seen only in the concentration range in which the transition from the predominance of the 1:3 complex to the 1:4 complex state was expected (Figure S5E). Therefore, the competition between eIF5A and ERK2 for DHPS observed in cellulo cannot be thoroughly explained only by the interaction between these proteins alone. It might rather reflect a complex regulation that involves additional cellular components.

ERK1/2 regulate cellular levels of DHPS and eIF5A dependent on kinase activity

Observing the non-kinase effect of ERK1/2 on the DHPS-eIF5A interaction, we asked whether ERK1/2 can regulate DHPS and eIF5A at an additional level. In addressing this, our time course analysis revealed that ΔRaf-1:ER activation increased DHPS and eIF5A protein levels in HEK293 cells within 24 h (Figure 6A). These increases were correlated with increased western blot signals detected by an antibody that recognizes both eIF5A^Dhp and eIF5A^Hyp—its selective affinity to these two derivatives is unknown.⁴³ Consistent with these data, ectopic expression of BRAF^V600E or a constitutively active MEK1 was sufficient to increase the levels of DHPS, eIF5A, and eIF5A^Dhp/Hyp in HEK293 cells (Figures 6B and S6A). Conversely, treatment with the B-Raf inhibitor PLX4032 or the MEK1/2 inhibitor AZD6244 decreased these proteins, and their mRNA levels, in the BRAF^V600E tumor cells and A375 and SK-MEL-28 cells (Figures 6C, S6B, and S6C). Of note, although PLX4032 and AZD6244 decreased DHPS and eIF5A levels repetitively in multiple biological replicates, eIF5A^Dhp/Hyp levels varied across the experiments (Figure 6C, densitometry). Similarly, RNA interference of ERK1 and ERK2 decreased eIF5A more consistently than eIF5A^Dhp/Hyp in A375, ΔRaf-1:ER-activated HEK293, and LNCaP cells (Figures 6D, S6D, and S6E). eIF5A^Dhp/Hyp basal levels fluctuated during the time course analysis and hindered the precise determination of ERK knockdown effects on the modification, as shown in A375 cells (Figure 6D). We also observed that that ERK1 knockdown decreased eIF5A levels as effectively as ERK2 knockdown despite its inferior contribution to ERK1/2 activity in ΔRaf-1:ER-activated HEK293 cells (Figure S6D) and that ΔRaf-1:ER activation was not sufficient to increase eIF5A and eIF5A^Dhp/Hyp levels in LNCaP cells despite the necessity of ERK1/2 for basal eIF5A expression in these cells (Figure S6E). These data suggest that (1) ERK1/2 regulate not only the DHPS-eIF5A interaction but also their expression, presumably at mRNA levels; (2) different cell types differentially respond to Raf/MEK/ERK activity; and (3) cellular levels of eIF5A hypusination are determined by a complex regulation involving Raf/MEK/ERK signaling and as-yet-unknown additional mechanisms.

Figure 6. ERK1/2 regulate cellular levels of DHPS and eIF5A dependent on kinase activity.

(A) Western blotting of total lysates of HEK293-ΔRaf:ER cells subjected to tamoxifen-induced ΔRaf:ER activation for indicated periods to determine time-dependent MEK/ERK effects on DHPS and eIF5A. eIF5A^D/H collectively indicates deoxyhypusinated eIF5A and hypusinated eIF5A. Densitometry (right) of western blot signals at 48 h was normalized to β-actin.

(B) Western blotting of total lysates of HEK293 cells infected with lentiviral pHAGE expressing N-terminal HA-tagged wild-type MEK1 (MEK1wt) and constitutively active MEK1 (MEK1ca) for indicated periods to determine time-dependent MEK/ERK effects on DHPS and eIF5A. Densitometry (right) of western blot signals was normalized to β-actin.

(C) Western blotting of total lysates of A375 cells treated with 1 μM PLX4032 or 100 nM AZD6244 for 48 h to determine how BRAF or MEK1/2 inhibition affects DHPS and eIF5A in BRAF^V600E tumor cells. An equal volume of DMSO was used as the vehicle control. Densitometry (right) of western blot signals was normalized to β-actin. See also Figure S6B for additional data.

(D) Western blotting of total lysates of HEK293 cells infected with lentiviral pLL3.7 expressing short hairpin RNA (shRNA) targeting ERK1 (shERK1) or ERK2 (shERK2) for 2 days and then switched to fresh media for indicated periods to determine ERK knockdown effects on DHPS and eIF5A. Densitometry (right) of western blot signals was normalized to β-actin. See also Figures S6D and S6E for additional data.

(E) Western blotting of total lysates of HEK293-ΔRaf:ER cells subjected to 24 h tamoxifen-induced ΔRaf:ER activation in the presence or absence of ERK1/2 inhibitors acting through different mechanisms. SCH-772984 inhibits the ERK1/2 activation loop and catalytic site. LY-5214996 inhibits the catalytic site only. Data are mean ± SD of biological triplicates, where *p < 0.05, **p < 0.005, and ***p < 0.001 (by two-tailed Student’s t test).

Next, we determined how inhibition of the ERK1/2 catalytic site, alone (by LY-3214996) or in combination with inhibition of activation loop phosphorylation (by SCH-772984), affects ΔRaf-1:ER-induced increases in DHPS and eIF5A protein levels in HEK293 cells. Contrary to their contrasting effects on the ERK1/2-regulated DHPS-eIF5A interaction, both SCH-772984 and LY-3214996 blocked the ΔRaf-1:ER-induced increases in DHPS and eIF5A protein levels at similar levels in cells (Figure 6E). In these cells, both inhibitors blocked ΔRaf-1:ER-induced RSK phosphorylation, but only the dual-mechanism inhibitor SCH-772984 inhibited ΔRaf-1:ER-induced ERK1/2 phosphorylation (Figure 6E). These data suggest that ERK1/2 require their kinase activity to regulate the cellular levels of DHPS and eIF5A while utilizing a non-kinase function to regulate the DHPS-eIF5A interaction in cells.

DISCUSSION

This report demonstrates a novel molecular interplay between the Raf/MEK/ERK pathway and the DHPS-eIF5A hypusination pathway. Our data show that ERK1/2 directly interact with DHPS, and this interaction decreases in cells when the Raf/MEK/ERK pathway is active. Strikingly, ERK activation loop phosphorylation is critical to this regulation, but its kinase activity is unnecessary. Mechanistically, our single-particle cryo-EM analysis suggests that ERK binding locks the entrance to the active site of DHPS to inhibit eIF5A hypusination and that two amino acid pairs govern the ERK2-DHPS interaction, which resembles eIF5A-DHPS complex formation. Moreover, our data demonstrate that DHPS does not contain a motif with an ERK1/2 substrate signature within reach for the catalytic center of the kinase, and, indeed, ERK1/2 do not phosphorylate DHPS. Based on these observations, we propose that the DHPS-eIF5A interaction is a new target for regulation by the Raf/MEK/ERK pathway, and this regulation occurs through a non-kinase ERK1/2 activity.

The CD groove and FRS of ERK1/2 recognize the D-domain and FXFP motif, respectively, in their binding partners.^6,7 Interestingly, DHPS contains a D-domain signature (Figure 7A), which was shown to be necessary for DHPS to interact with ERK1/2.⁴⁴ However, our cryo-EM structure of the DHPS-ERK2 complex clearly shows this is not the case. The D-domain signature in DHPS is not fully accessible for a physical interaction because its conserved hydrophobic residues Ile99 and Leu101 are not protruded to the surface. Of note, the D-domain in RSK1, a bona fide ERK1/2 substrate, interacts with ERK2 CD and is connected by a flexible linker to the phosphorylation target domain located underneath the catalytic cleft of ERK1/2 (Figure 7B). Using this structure as a reference, our cryo-EM results suggest that ERK2 interacts with DHPS not through the putative D-domain but through a completely different interface. The DHPS interaction surface is in proximity to ERK2 FRS (Figure 7A), in a similar way to the mechanism described for ERK2 recognition of the astrocytic phosphoprotein PEA-15 (Figure 7C). Of note, ERK2’s interaction with PEA-15 via FRS appears to be atypical, as PEA-15 lacks the consensus motif FXFP.⁴⁵ Similarly, our data show that the DHPS-ERK2 interaction is not mediated by any of the crucial FRS residues, nor does DHPS contain the FXFP motif. However, there is also a difference. Apart from engaging ERK2 FRS, PEA-15 also interacts with ERK2 CD through a typical ball-and-chain-type D-domain motif (Figure 7C). We did not detect this type of interaction in the DHPS-ERK2 structure (Figure 7A). Therefore, our study demonstrates that ERK2 interacts with DHPS through sites unrelated to CD and FRS, revealing a new mode of ERK2 recognition by its interaction partner.

Figure 7. ERK2 interacts with DHPS through a novel interface unrelated to CD and FRS.

(A) Structure of the ERK2-DHPS complex determined by this study (PDB: 8PVU). DHPS interacts with ERK2 through an atypical interface near the FRS of ERK2, while its putative D-domain is too distal to the actual interaction surface.

(B) Structure of the ERK2-RSK1 complex (PDB: 4NIF). The D-domain of RSK1 interacts with the CD site of ERK2.

(C) Structure of the ERK2-PEA-15 complex (PDB: 4IZ5). PEA-15 interacts with both FRS and CD of ERK2.

(D) Structure of the DHPS-ERK2 complex with superimposed NAD, spermidine (SPD), spermidine and NAD, or the trapped DHPS transition state from appropriate DHPS crystal structures (PDB: 6XXI, 6XXK, 6XXJ, and 8A0G).

(E) Sites of missense variants of DHPS and ERK2 reported in ClinVar (green), and sites of potentially pathogenic mutations are marked with a darker color and bold. On the right is a closeup on N173^DHPS and G162^DHPS, which are located near the DHPS-ERK2 interface, superimposed with the crystal structure of DHPS N173S (PDB: 7A6S; gray) and with an exchanged side chain of G162^DHPS to arginine (gray). See Figure S7 for N173S and G162 A/R MST data.

The role of the SPS motif in ERK physical interactions is relatively less known.^8,9 Our data suggest that this motif is involved in the regulation of the DHPS-ERK1/2 interaction. It has been shown that phosphorylation of the SPS motif facilitates nuclear localization of ERK1/2, and the phosphomimetic mutagenesis of the motif renders this ERK localization even in the absence of the TEY site phosphorylation in the activation loop.⁴⁶ Given our data suggesting that phosphorylation of the SPS motif negatively regulates the formation of the DHPS-ERK1/2 complex, we hypothesize that the ERK1/2-DHPS interaction is also regulated by a kinase that controls the phosphorylation status of the SPS motif. For example, casein kinase II was shown to phosphorylate the SPS motif, although the molecular mechanism underlying this regulation needs to be better established.⁹ The molecular mechanisms underlying the regulation of the ERK1/2-DHPS interaction through the SPS site and how it is coordinated with the regulation through the TEY site remain a future study.

Interestingly, we observed that the DHPS cofactor NAD enhances the ERK2-DHPS interaction regardless of its redox state and that the optimal binding partner for ERK2 is the DHPS transition state present with the availability of NAD and spermidine but the absence of eIF5A. Since the binding of NAD and spermidine to DHPS or the resulting transition state does not directly influence the DHPS-ERK1/2 interface (Figure 7D), the enhanced interaction may be attributed to the stabilization of the DHPS structure, especially in the vicinity of its active site entrance. Given this, NAD depletion could lead to the disruption of the complex but should have no functional consequences, as DHPS is catalytically inactive without its cofactor.

Strikingly, our data suggest that DHPS is not a substrate for ERK1/2. Indeed, ERK1/2 can mediate different non-kinase effects, as demonstrated by ERK1/2 regulation of poly(ADP-ribose) polymerase, MAP kinase phosphatase-3, and topoisomerase IIα.^11-13 Our current study adds DHPS to this list. Of note, contrary to our characterization, a recent study reported that ERK1/2 directly mediate Ser233 phosphorylation of DHPS induced by phorbol 12-myristate 13-acetate (PMA) in HEK293 cells.⁴⁴ While consistently observing the physical interaction between ERK1/2 and DHPS, this study and our present report are contrasted by the difference in two key experimental approaches. First, although ERK1/2 substrates require the presence of Pro at the +1 position of Ser or Thr, the former study used an antibody that recognizes Phe instead of Pro at this position. Second, the in vitro ERK assay in the former study relied upon the autophosphorylation of ERK1 for the presence of active ERK1 in the reaction. Autophosphorylation is a very inefficient mechanism for activating ERK without the assistance of mutations that facilitate the hydrogen-bonding interactions between the phosphoryl acceptor and catalytic nucleophile.⁴⁷ Moreover, our structural data clearly show that ERK2 cannot phosphorylate Ser233 in DHPS because this residue is beyond the reach of the catalytic center of the kinase. Probably, PMA induces DHPS phosphorylation on the Ser233 site via a kinase other than ERK1/2. For example, PMA can activate protein kinase C,^48,49 and it was previously shown that protein kinase C inhibition blocks PMA-induced DHPS phosphorylation in HeLa and CHO cells, although the phosphorylation site(s) was not identified.³⁶

What is the interplay between ERK1/2 and DHPS? In ΔRaf-1:ER-activated cells, DHPS overexpression did not alter the phosphorylation status of ERK1/2 and its substrate RSK, making it unlikely that DHPS regulates ERK1/2. In contrast, unphosphorylated ERK1/2 can compete with eIF5A for DHPS, while phosphorylation of their activation loop hinders this ability of ERK1/2. Our data also suggest that prolonged kinase activity of ERK1/2 can upregulate cellular levels of DHPS and eIF5A proteins. Taking these results together, we speculate that the Raf/MEK/ERK pathway promotes eIF5A hypusination via at least two different mechanisms that utilize different functions of ERK1/2. It is known that DHPS and DOHH are not normally limiting and that eIF5A normally becomes hypusinated efficiently after translation, which renders the majority of eIF5A in cells to be present in the fully hypusinated form.⁵⁰ Nevertheless, our data suggest that this process is still subject to a regulation mediated by signaling pathways. Importantly, it has been reported that eIF5A hypusination is mediated via the cAMP-protein kinase A-ERK1/2 pathway to downregulate leuteinizing hormone receptor expression in the ovary⁵¹ and that expression of the oncogenic KRAS protein is regulated by a self-governing eIF5A-pseudopodium-enriched atypical kinase 1 feedforward regulatory loop.⁵² Our findings may provide insight into the molecular mechanisms by which the Ras/Raf/MEK/ERK pathway regulates eIF5A hypusination.

Recessive rare DHPS variants and some mutations in ERK2 have been recently associated with neurodevelopmental disorders.^53,54 Our analysis of missense variants reported in ClinVar located most of the ERK2 mutations within the active site and the CD site, but not in the DHPS-ERK2 interface, while locating G162 and N173S of DHPS near the interaction surface (Figure 7E). G162^DHPS does not interact directly with ERK2, and its side chain would point away from the DHPS-ERK2 interface. Therefore, the reported G162 A/R mutations, despite the bulky side chain of arginine, should not affect DHPS-ERK2 binding. N173^DHPS may interact with K231^ERK2, but the reported N173S mutation should not affect this. Moreover, K231^ERK2 is mainly coordinated by the main-chain carbonyl of W327^DHPS (Figure 7E). Indeed, our MST measurements show that these DHPS mutations do not affect the DHPS-ERK2 interaction (Figure S7).

Lastly, we previously demonstrated that active site-, but not activation loop-, disabled ERK1/2 can promote Raf/MEK-induced growth arrest responses.¹⁴ Given that DHPS is a non-kinase ERK target, it is an intriguing question whether it would have a role in these responses. This will be addressed in the next chapter of our research. In summary, we report a novel regulation of the DHPS-eIF5A pathway that requires a non-kinase activity of ERK1/2.

Limitations of this study

This study provides molecular details of a non-kinase effect of ERK1/2 on deoxyhypusination. We used crosslinking to stabilize the DHPS-ERK2 complex for cryo-EM, which could potentially limit the conformational landscape of both proteins. We also did not obtain a map of the fully saturated complex with a sufficiently high resolution; thus, we built a molecular model only for the 1:1 complex. Although we were able to resolve the ERK2 TEY motif, we lack density for the part of the ERK2 activation loop preceding this site. Hence, we lack the full picture of the interaction of the ERK2 activation loop with DHPS. Different ERK2 mutants were ectopically expressed in cells, and unnatural protein expression levels limit the evaluation of hypothesized effects and data interpretation. Precise determination of non-kinase ERK1/2 functions in cells would require depletion of endogenous ERK1/2, but complete ablation of their kinase activity is lethal. The future line of investigation will have to take into consideration these limitations.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to Jong-In Park (jipark@mcw.edu).

Materials availability

Reagents and materials produced in this study are available upon request from the lead contact pending a completed materials transfer agreement.

Data and code availability

The atomic model of the DHS-ERK2 complex and cryo-EM maps representing different complex modalities have been deposited in the Protein Data Bank and the Electron Microscopy Data Bank. Tandem affinity purification-mass spectrometry (MS) raw data (RAW files) are deposited in the MassIVE database (ftp://massive.ucsd.edu/MSV000088047/). All deposited data are publicly available as of the date of publication. Accession numbers are listed in the key resources table. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-ERK2 antibody	Santa Cruz Biotechnology	Cat#sc-154; RRID: AB_2141292
Anti-His-Tag Antibody	Santa Cruz Biotechnology	Cat#sc-8036; RRID: AB_627727
Anti-phospho-ERK 1/2 Antibody (Thr 202/Tyr 204)	Santa Cruz Biotechnology	Cat#sc-16982; RRID: AB_2139990
Anti-DHPS Antibody	Santa Cruz Biotechnology	Cat#sc-365077; RRID: AB_10846806
Anti-eIF5A Antibody	Santa Cruz Biotechnology	Cat#sc-390202
Anti-HA-Tag Antibody	Santa Cruz Biotechnology	Cat#sc-7392; RRID: AB_627809
Anti-ERK1/2 Antibody	Cell Signaling Technology	Cat#9102S; RRID:AB_330744
Anti-phospho-p90RSK Antibody (Thr359/Ser363)	Cell Signaling Technology	Cat#9344S; RRID:AB_331650
Anti-RSK1/2/3 Antibody	Cell Signaling Technology	Cat#9347S; RRID:AB_330803
Anti-Thiophosphate ester Antibody	Abcam	Cat#ab92570; RRID: AB_10562142
Anti-Phosphothreonine-Proline/Phosphoserine-Proline Antibody	Abcam	Cat#ab9344; RRID: AB_307195
Anti-β-actin Antibody	Sigma-Aldrich	Cat#A1978; RRID: AB_476692
Anti-β-tubulin Antibody	Thermo Fisher Scientific	Cat#PA5-16863; RRID: AB_10986058
Anti-eIF5A(Deoxy)Hypusine Antibody (IU-88)	gift from Raghavendra G Mirmira, MD, PhD	N/A
Anti-(Deoxy)Hypusine Antibody (clone Hpu98)	Creative Biolabs	Cat#PABL-582; RRID: AB_3111662
Bacterial and virus strains
Escherichia coli TOP10	Thermo Fisher Scientific	Cat#C404003
Escherichia coli BL21(DE3)	Thermo Fisher Scientific	Cat#EC0114
Chemicals, peptides, and recombinant proteins
Recombinant MAPK1 Human, Active (phospho-ERK2)	ProSpec	Cat#PKA-214
Recombinant MAPK1 Human, Unactive (ERK2)	MyBioSource	Cat#MBS515277
Recombinant EGF Human	Thermo Fisher Scientific	Cat#A42556
Bovine serum albumin Fraction V (pH 7.0)	AppliChem	Cat# A6588
AZD6244	Selleck Chemicals	Cat#HY-50706
PLX4032	Selleck Chemicals	Cat#HY-12057
SCH-772984	ChemieTek	Cat#CT-SHC772
LY-5214996	ChemieTek	Cat#CT-LY321
4-Hydroxytamoxifen	Sigma-Aldrich	Cat#H7904
NanoTemper Protein Labeling Kit RED-MALEIMIDE 2ND Generation	NanoTemper Technologies	Cat# MO-L014
Superscript III	Thermo Fisher Scientific	Cat#18080093
Q5 High-Fidelity 2x Master Mix	New England Biolabs	Cat#M0492S
Adenosine 5’-(γ-thio)-triphosphate	Cayman Chemical Company	Cat#14957
p-Nitrobenzyl mesylate	Cayman Chemical Company	Cat# 21456
293Expresso™ Transfection Reagent	Excellegen	Cat# EG-1089
MEM (Minimum Essential Medium)	Thermo Fisher Scientific	Cat#11095080
RPMI 1640 Medium, no phenol red	Thermo Fisher Scientific	Cat#11835030
DMEM, high glucose	Thermo Fisher Scientific	Cat#11965092
Sodium Pyruvate (100 mM)	Thermo Fisher Scientific	Cat#11360070
MEM Non-Essential Amino Acids Solution	Thermo Fisher Scientific	Cat#11140050
Fetal Bovine Serum	Thermo Fisher Scientific	Cat#16000044
HyClone Bovine Growth Serum	Cytiva	Cat#SH30541.03
Pierce™ Anti-HA Agarose	Thermo Fisher Scientific	Cat#26181
Ni Sepharose High Performance	Cytiva	Cat#17526801
Protein G PLUS-Agarose	Santa Cruz Biotechnology	Cat#sc-2002
BS3 (bis(sulfosuccinimidyl)suberate)	Thermo Fisher Scientific	Cat#21580
Deposited data
Molecular model of DHS-ERK2 complex with 1:1 stoichiometry	This study	PDB: 8PVU
Cryo-EM map of DHS-ERK2 complex with 1:1 stoichiometry refined in C1 symmetry	This study	EMDB: EMD-17972
Cryo-EM map of the third of three possible DHS-ERK2 complexes with 1:2 stoichiometry refined in C1 symmetry	This study	EMDB: EMD-17977
Cryo-EM map of the third of three possible DHS-ERK2 complexes with 1:2 stoichiometry refined in C2 symmetry	This study	EMDB: EMD-17978
Cryo-EM map of the first of three possible DHS-ERK2 complexes with 1:2 stoichiometry refined in C1 symmetry	This study	EMDB: EMD-17983
Cryo-EM map of the first of three possible DHS-ERK2 complexes with 1:2 stoichiometry refined in C2 symmetry	This study	EMDB: EMD-17984
Cryo-EM map of the second of three possible DHS-ERK2 complexes with 1:2 stoichiometry refined in C1 symmetry	This study	EMDB: EMD-17981
Cryo-EM map of the second of three possible DHS-ERK2 complexes with 1:2 stoichiometry refined in C2 symmetry	This study	EMDB: EMD-17982
Cryo-EM map of DHS-ERK2 complex with 1:3 stoichiometry refined in C1 symmetry	This study	EMDB: EMD-17985
Cryo-EM map of DHS-ERK2 complex with 1:4 stoichiometry refined in C1 symmetry	This study	EMDB: EMD-17986
Cryo-EM map of DHS-ERK2 complex with 1:4 stoichiometry refined in D2 symmetry	This study	EMDB: EMD-17987
TAP-MS raw data	This study	MassIVE: MSV000088047
Experimental models: Cell lines
HEK293	American Type Culture Collection	CRL-1573™
293T	American Type Culture Collection	CRL-3216™
LNCaP	American Type Culture Collection	CRL-1740™
A375	American Type Culture Collection	CRL-1619™
Mia-PaCa-2	American Type Culture Collection	CRL-1420™
HeLa	American Type Culture Collection	CCL-2™
HEK293-ΔRaf:ER	Hong et al., 2009	N/A
LNCaP-ΔRaf:ER	Hong et al., 2018	N/A
Oligonucleotides
RT-PCR forward primer for eIF5A: GCGGCCGCACCATGGCAGATG ACTTGGACTTCG	This study	N/A
RT-PCR reverse primer for eIF5A: CGCAAGCTTCTATTTTGCCATG GCCTTGATTGC	This study	N/A
RT-PCR forward primer for DHPS: AAGTTTGAGGACTGGCTGATG	This study	N/A
RT-PCR reverse primer for DHPS: CAGGGATGTGGTTCTTCTGG	This study	N/A
RT-PCR forward primer for β-actin: GTCCTCTCCCAAGTCCACAC	This study	N/A
RT-PCR reverse primer for β-actin: GGGAGA CCAAAAGCCTTCAT	This study	N/A
See Table S2 for list of mutagenic primers used in this study.	N/A	N/A
Recombinant DNA
Full-length human deoxyhypusine synthase with N-teminal 6xHis-tag followed by TEV cleavage site in pET-24d(+) for bacterial expression	Genescript	N/A
Human eukaryotic translation initiation factor 5A-1 (residues 15–151) with N-teminal 6xHis-tag followed by TEV cleavage site in pET28-MHL for bacterial expression	Addgene	Cat#25260
Full-length human mitogen-activated protein kinase 1 with N-teminal 6xHis-tag followed by TEV cleavage site in pNIC28-Bsa4 for bacterial expression	Addgene	Cat#73248
pHAGE-ΔRaf:ER	Hong et al.Ref. 14	N/A
Full length DHPS cDNA in pCMV3-ORF-HA	Sino Biological	Cat#HG14407-CY
pCEFL-eIF5A	Clement et al.Ref. 58	N/A
pHAGE-GFP-ERK2WT	Hong et al.Ref. 14	N/A
pHAGE-GFP-ERK2-K52R	Hong et al.Ref. 14	N/A
pHAGE-GFP-ERK2-T183A/Y185F	Hong et al.Ref. 14	N/A
pHAGE-GFP-ERK2-Y261N	Hong et al.Ref. 55	N/A
pHAGE-GFP-ERK2-D316/319A	Hong et al.Ref. 55	N/A
pHAGE-BRAF-V600E	Wu et al.Ref. 57	N/A
pHAGE-MEK1-R4F	Hong et al.Ref. 14	N/A
pLL3.7-shERK1	Hong et al.Ref. 14	N/A
pLL3.7-shERK2	Hong et al.Ref. 14	N/A
Software and algorithms
RELION	Zivanov et al.Ref. 62	https://relion.readthedocs.io
CryoSPARC	Punjani et al.Ref. 63	https://cryosparc.com
DeepEMhancer	Sanchez-Garcia et al.Ref. 64	https://github.com/rsanchezgarc/deepEMhancer
UCSF ChimeraX	Pettersen et al.Ref. 65	https://www.cgl.ucsf.edu/chimerax/
PyMol	Schrödinger	https://pymol.org/
COOT	Emsley et al.Ref. 66	http://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot
Phenix	Adams et al.Ref. 67	https://phenix-online.org
MolProbity	Williams et al.Ref. 68	http://molprobity.biochem.duke.edu
MO.Affinity Analysis 3	NanoTemper Technologies	Cat# MO-S021A
Image Lab	Bio-Rad	https://www.bio-rad.com/product/image-lab-software
ImageJ	NIH	https://imagej.nih.gov
Prism	GraphPad	https://www.graphpad.com
Other
TEM grids Quantifoil R2/1, Cu, mesh 200	Quantifoil	Cat#X-102-Cu200
Vitrobot Mark IV	Thermo Fisher Scientific	N/A
Titan Krios G3i	Thermo Fisher Scientific	N/A
NanoTemper Monolith	NanoTemper Technologies	N/A

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines

The human embryonic kidney cell line HEK293 (ATCC) was maintained in minimal essential medium (Invitrogen) supplemented with 5% bovine growth serum (Hyclone) and 5% fetal bovine serum (Gibco). The human prostate cancer cell line LNCaP (ATCC) was maintained in phenol red free RPMI (Invitrogen) supplemented with 10% fetal bovine serum. The human BRAF^V600E melanoma cell line A375 and human K-Ras^G12C pancreatic cell line Mia-PaCa-2 (ATCC) were cultured in Dulbecco’s modified Eagle’s medium (Invitrogen), and minimal essential medium (Invitrogen) supplemented with 10% fetal bovine serum, 100U/ml penicillin and 100 μg/mL streptomycin (Gibco), 1% sodium pyruvate (Gibco), and 1% non-essential amino acids (Gibco) respectively. The human epithelial adenocarcinoma cell line HeLa (ATCC) was maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum. Generation of LNCaP-ΔRaf:ER and HEK293-ΔRaf:ER cells that are stably transduced with the lentiviral pHAGE-ΔRaf:ER was previously described.^14,55 The tamoxifen-controlled ΔRaf:ER was activated in cells using 1 μM of 4-Hydroxytamoxifen (Sigma-Aldrich), as previously described.⁵⁶ All cultures were kept at 37°C and 5% CO₂ with saturating humidity.

Bacterial strains

E. coli TOP10 and BL21(DE3) strains (Thermo Fisher Scientific) were used for molecular cloning and protein expression, respectively. Bacteria were cultured in TB media at 37°C.

METHOD DETAILS

Lentiviral constructs and plasmids

Construction of the lentiviral pHAGE-ΔRaf:ER, pHAGE-BRAF^V600E, pHAGE-MEK1-R4F (ΔN3/S218E/S222D, constitutively active MEK1), pLL3.7-shERK1, and pLL3.7-shERK2 was previously described.^14,57 C-terminally HA-tagged DHPS in pHAGE-GFP was generated by ligating the full length DHPS cDNA from pCMV3-ORF-HA (Sino Biologicals) to the NheI/XhoI sites in pHAGE-GFP. Full length pCEFL-eIF5A vector was previously described.⁵⁸ Generation of pHAGE-GFP-ERK2WT, pHAGE-GFP-ERK2-K52R, and pHAGE-GFP-ERK2-T183A/Y185F as well as pHAGE-GFP-ERK2-Y261N, and pHAGE-GFP-ERK2-D316/319A was previously described.^14,55 pHAGE-GFP-ERK2-S244/246A, pHAGE-GFP-ERK2-D165A, pHAGE-GFP-ERK2-H230A, pHAGE-GFP-ERK2-D165A/H230A were generated by site-directed mutagenesis of pHAGE-GFP-ERK2WT using the primers listed in Table S2.

The DNA encoding full-length human deoxyhypusine synthase (Uniprot: P49366, residues 1–369) and optimized for bacterial expression was synthesized along with N-teminal 6xHis-tag followed by TEV cleavage site and cloned into pET-24d(+) vector using NcoI/BamHI restriction sites (Genescript). All single point mutants of DHPS were ordered from Genescript and generated in the above-described construct. The plasmid bearing full-length human mitogen-activated protein kinase 1 (Uniprot: P28482, residues 1–360) with N-teminal 6xHis-tag followed by TEV cleavage site was acquired from Addgene (Cat#73248) as well as the construct with human eukaryotic translation initiation factor 5A-1(Uniprot: P63241, residues 15–151) with N-teminal 6xHis-tag followed by TEV cleavage site (Addgene Cat#25260). All ERK2 mutants were generated in-house with the use of a Quickchange-type approach, for primers see Table S2. Q5 High-Fidelity 2x Master Mix (New England Biolabs) was used for the PCR reactions. After DpnI treatment the reaction products were transformed into TOP10 chemically competent E. coli cells (Thermo Fisher Scientific). Plasmids were isolated from single colonies and presence of the required mutations verified by sequencing (Eurofins Genomics).

Viral infection and transfection

293T (ATCC) cells were utilized for lentivirus production by transfecting packaging vectors and lentiviral expression vector pHAGE as previously described.¹⁴ Cells were transfected using 293Expresso (Excellegen) according to the manufacturer’s instructions. After 72h, viral supernatants were collected, concentrated, resuspended in fresh cell culture medium and administered to cells. Culture medium was replaced with fresh medium 24h after infection.

Immunoprecipitation

Immunoprecipitation was performed as previously described.^59,60 Briefly, cells were lysed in 50 mM Tris (pH7.4), 120 mM NaCl, 1% NP-40, 1mM EGTA and mixed with protease and phosphatase inhibitor cocktail. 500 μg of protein lysates were incubated with 20 μL slurry of anti-HA-conjugated agarose beads (Thermo Fisher Scientific) or the mixture of anti-DHPS antibody (Santa Cruz), anti-HIS antibody (Santa Cruz), anti-ERK2 antibody (Santa Cruz), or anti-eIF5A antibody (Santa Cruz) and Protein G agarose (Santa Cruz) at 4°C for 16h with rotation. Precipitates were washed three times with buffer containing 50 mM Tris (pH7.4), 120 mM NaCl, 1% NP-40, 1mM EGTA with rotation at 4°C, and proteins were eluted in the sample buffer containing 31.25 mM Tris-Cl, 5% glycerol, 1% SDS, 2.5% 2-Mercaptoethanol, and 0.0015% Bromophenol blue for Western blotting.

In vitro binding assay

0.5 μM recombinant HIS-DHPS or popB-HIS were incubated with Ni Sepharose resin (Cytiva) in binding buffer containing 20 mM Tris, pH 8.0, 200mM NaCl, 10% glycerol, 0.5% Triton X-100, and 50 mM imidazole at room temperature for 30 min. The protein-bead complexes were collected and incubated in the binding buffer supplemented with 1% BSA at 4°C for 6h with rotation. The protein-bead complexes were then washed 4 times with the buffer, and incubated with increasing concentrations of untagged recombinant ERK2 (Mybiosource) at 4°C for 1h with rotation. The protein-bead complexes were washed 3 times in the binding buffer, and proteins were eluted in the sample buffer for Western blotting.

In vitro phosphorylation assay

25 ng of recombinant pERK2 (ProSpec) was incubated with 1.5 μM recombinant DHPS or Myelin Basic Protein in buffer containing 10 mM HEPES (pH7.4), 150 mM NaCl, 10mM MgCl₂, and 20 mM ATPyS (Cayman Chemical Company) at 20°C for 30 min with gentle rocking. 50 mM of p-nitrobenzylmesylate (Cayman Chemical Company) was then added and incubation continued for another 2h with gentle rocking. The kinase reaction was quenched by adding the sample buffer for Western blotting. Immunoblot analysis was performed using an anti-thiophosphate ester-specific antibody (Abcam).

Immunoblot analysis

Cells were harvested in lysis buffer containing 62.5 mM Tris-HCl (pH 6.8)/2% SDS and protease and phosphatase inhibitor cocktails 2 and 3 (Sigma-Aldrich). Protein concentrations were measured using the bicinchoninic acid reagent (Thermo Fisher Scientific). Proteins were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane. For the in vitro deoxyhypusination assay membranes were then stained with Panceau S (Serva), imaged with ChemiDoc MP (Bio-Rad) and destained with Tris-buffered saline with 0.1% Tween 20 (TBST). After transfer, membranes were blocked at 20°C for 1 h (or at 4°C, overnight) in TBST containing 5% nonfat dry milk or bovine serum albumin. For immunodetection, membranes were incubated with designated primary antibody overnight at 4°C with rotation (or for 1h at RT in case of the in vitro deoxyhypusination assay). The following antibodies were used at the designated dilution: Santa Cruz Biotechnology: ERK2 1:2000, HIS 1:1000, phospho-ERK1/2 1:1000 (Thr202,Tyr204), DHPS 1:2000, eIF5A 1:2000, HA 1:5000; Cell Signaling Technology: ERK1/2 1:2000, pRSK 1:2000 (T359/S363), RSK1/2/3 1:1000; Abcam: Thiophosphate ester 1:1000, pS/T-P 1:500; Sigma-Aldrich: β-actin 1:10000; Thermo Fisher Scientific: β-tubulin 1:10000; Creative Biolabs: (Deoxy)hypusine 1:1000 (Hpu98) and eIF5A^Hyp 1:5000 (gift from Dr. Mirmira, IU-44). SuperSignal West Pico and Femto chemiluminescence kits (Thermo Fisher Scientific) were used for visualization of the signal. Densitometry was done with Image Lab (Bio-Rad) or ImageJ.⁶¹

RT-PCR

Total RNA was isolated from cells using TRIzol reagent (Invitrogen). Reverse transcription was performed using Superscript III (Invitrogen) and oligo-dT according to the manufacturer’s instructions. The resulting cDNA was amplified by PCR (for primers see key resource table).

Recombinant protein production

All proteins were expressed in Escherichia coli BL21(DE3) (Thermo Fisher Scientific). The cells were grown in a shaking incubator (37°C, 190 RPM) until reaching the OD of around 1. After that the cultures were cooled to 18°C, and protein expression was induced with IPTG (0.5 mM; A&A Biotechnology) and continued overnight. Cells were collected by centrifugation (5000 x g, 15 min, 4°C) and resuspended in lysis buffer (50 mM Tris-HCl pH 7.8, 300 mM NaCl, 20 mM imidazole, 10% glycerol, 5 mM β-ME). Cells were disrupted by sonication (10 min, 3 s pulse/5 s pause cycles, amplitude 70%; Vibra-Cell VCX 500, Sonics & Materials) in the presence of lysozyme and benzonase (Sigma-Aldrich). Lysates were cleared by centrifugation (50,000 x g, 20 min, 4°C) and the supernatants were applied onto an equilibrated 5mL HisTrap Excel column (Cytiva), and the column was washed with 20 CV of buffer (50 mM Tris-HCl pH 7.8, 200 mM NaCl, 30 mM imidazole, 5% glycerol, 3 mM β-ME). The protein of interest was eluted with the same buffer but containing 400 mM imidazole, dialyzed against storage buffer overnight (50 mM Tris-HCl pH 7.8, 200 mM NaCl, 3 mM β-ME) and subjected to overnight TEV protease cleavage during dialysis. The tag-free protein was isolated with reverse HisTrap Excel column chromatography, concentrated with Amicon Ultra Centrifugal Filter (10kDa MWCO; Merck) and subjected to size-exclusion chromatography on a HiLoad 16/600 Superdex 200 pg column (Cytiva) in storage buffer. Fractions corresponding to the desired protein of the highest purity were pooled, concentrated, aliquoted, flash-frozen in liquid nitrogen and stored at −80°C for further analysis.

Complex reconstitution

3.2 μM of DHPS was mixed with 12.8 μM of ERK2 in HEPES-based buffer (20 mM HEPES pH 7.6, 150 mM NaCl, 5% glycerol, 5 mM β-ME) and crosslinked with 0.63 mM bis(sulfosuccinimidyl)suberate (from 100 mM stock in DMSO; Thermo Fisher Scientific) for 30 min at RT. The reaction was quenched with 50 mM Tris-HCl pH 7.8 for 10 min at RT. The products were subjected to size-exclusion chromatography on a Superdex 200 10/300 GL column (Cytiva) in Tris-based buffer (50 mM Tris-HCl pH 7.8, 200 mM NaCl, 3 mM β-ME). Fraction corresponding to a mixture of DHPS-ERK2 complexes was used for grid preparation.

Grid preparation and cryo-EM data collection

Sample of the stabilized DHPS-ERK2 complex was concentrated to 0.35 mg/mL and 4 μl of the solution was applied on glow-discharged TEM grids (Quantifoil R2/1, Cu, mesh 200) and plunge-frozen in liquid ethane with the use of Vitrobot Mark IV (Thermo Fisher Scientific; blot time: 2 s, blot force: 0, humidity: 95%, temperature: 4°C). Frozen grids were stored in liquid nitrogen. Data was collected with Titan Krios G3i microscope (Thermo Fisher Scientific) at National Cryo-EM Center SOLARIS (Krakòw, Poland) (accelerating voltage: 300 kV, magnification: 105k, corresponding pixel size: 0.86 Å/px). Images were acquired with K3 direct electron detector (Gatan) with BioQuantum Imaging Filter (Gatan) setup with 20 eV slit enabled. The detector was operated in counting mode with physical pixel resolution. The sample was exposed to 40.16 e–/Å2 total dose (corresponding to ~16.05 e–/px/s dose rate measured on vacuum). The images were acquired at under-focus optical conditions with a defocus range of −3.3 to −0.9 μm with 0.3 μm steps.

Cryo-EM data processing

All micrographs were motion-corrected using RELION’s own, CPU-based implementation of the UCSF MOTIONCOR2.⁶² Motion-corrected micrographs were imported into CryoSPARC⁶³ where they were subjected to CTF estimation using Patch CTF. After the selection of the best micrographs suitable for further processing, about 400 particles were picked manually. Extracted particles were 2D classified and the best 2D classes were used for first template picking. Template-picked particles were subjected to reference-free 2D classification and, best 2D class averages were selected and used in the second round of template picking. Improved templates allowed picking about 1.9 million particles that underwent: 2D reference-free classification and ab initio reconstruction (using a subset of the particles selected) and finally heterogeneous refinement. After selecting the best-looking volume, the particles were further processed to clean the final dataset. The best-looking particles (after final reference free 2D classification) were classified in 3D into 6 classes by feeding artificially generated maps of all possible complex modalities as initial sorting constraints into the cryoSPARC heterogeneous refinement job. The resulting classes were subjected to the final homogeneous refinement (Figure S1).

Model building, refinement and validation

Before model building the map was post-processed with DeepEMhancer.⁶⁴ Models of ERK2 (PDB ID: 4QTA) and DHPS (PDB ID: 8A0E) were manually docked in a map with the use of UCSF ChimeraX.⁶⁵ The generated model was iteratively improved by manual model rebuilding in Coot⁶⁶ and real-space refinement in Phenix.⁶⁷ The final model was validated using MolProbity.⁶⁸

Microscale thermophoresis (MST)

All the DHPS variants were labeled on surface-exposed cysteines with fluorophore-tagged maleimide according to manufacturer protocol (NanoTemper Technologies), and the labeled protein was eluted to MST buffer (30 mM HEPES pH 8.0, 150 mM NaCl, 0.05% Tween 20). The degree-of-labelling was verified to be between 0.5 and 1 with absorbance measurements at 280 nm and 650 nm according to manufacturer protocol. ERK2 was used in a 6xHis-tagged form and was buffer exchanged to MST buffer by 3 rounds of dilution and concentration in Amicon Ultra Centrifugal Filter (10kDa MWCO; Merck). ERK2 variants, were used at a concentration achieved after the last round of ultrafiltration (>250 μM in assay), while for studies of DHPS mutants, the highest ERK2 concentration in assay was 800 μM and for studies of NAD/NADH/spermidine effect on binding was 293 μM. In all the above-described assays, a series of 16 2-fold dilutions of the unlabeled protein in MST buffer was prepared, and the labeled DHPS concentration was constant in all samples (100 nM). The measurements were done with NanoTemper Monolith in the red channel set as follows: nanoRED 15% intensity, medium MST power, 25°C, before MST 3 s, 20 s MST on time, 1 s after MST. For every assay data from four measurements was averaged and the signal after 1.5 s was analyzed in MO.Affinity Analysis 3 software (NanoTemper Technologies).

Complex stoichiometry evaluation

The labeling approach was as described above. The MST buffer was supplemented with 1uM NAD to lower the K_d and generate a more clearly resolved saturation kink. To keep the concentration of the labeled target high enough relative to K_d and not saturate the detector 1 μM of the labeled DHS and 399 μM of unlabeled DHS were used (per monomer concentrations, 100 μM of tetramer in total). A series of 24 dilutions of ERK2 (without 6xHis-tag) was designed to best cover the area surrounding the point of saturation. The measurements were performed with NanoTemper Monolith as described above but the excitation light intensity was set to 2%. Data from four measurements was averaged in MO. Affinity Analysis 3 software and linear regression was performed for linear areas surrounding the saturation kink with Prism (GraphPad).

In vitro deoxyhypusination assay

The reaction was done in 0.1 M glycine/NaOH pH 9.5 buffer supplemented with 1 mM DTT, 1 mM NAD and 1 mM spermidine at RT. For the time course experiments, 50 nM DHPS and 30 μM eIF5A1 were used and 100 μM of ERK2 or BSA (AppliChem) was added to appropriate samples. The reaction was started by the addition of eIF5A1 with a multichannel pipet. In the same manner, 10 μL samples from all the reactions were taken at appropriate timepoints and immediately mixed with 2 μL o 6x sample buffer for Western blotting. The analysis of the effect of ERK2 concentration was executed in the same way but different concentrations of ERK2 were used and the samples were taken only after 8.5 min. Samples were denatured at 95°C for 5 min, and 6 μL of each sample was separated on 12% SDS-PAGE gel for immunoblotting. Panceau S signal was used for chemiluminescent signal normalization. All experiments were done in triplicates.

Tandem affinity purification (TAP)

TAP screening of ERK1/2-interacting proteins was performed in LNCaP-ΔRaf:ER cells using ERK2-K52R tagged with calmodulin-binding protein and streptavidin-binding protein at the N-terminal in tandem (TAP-ERK2-K52R) as the bait (illustrated in Figure S8A). For lentiviral TAP-ERK2-K52R expression, a cDNA encoding ERK2-K52R was ligated into the pNTAP vector (Agilent Technologies, 240101), and the resulting TAP-ERK2-K52R was then moved into pHAGE. LNCaP-ΔRaf:ER cells infected with this lentivirus expressed TAP-ERK2-K52R to a level similar to endogenous ERK1/2 levels, which was increasingly phosphorylated in a time-dependent manner upon 1 μM 4-hydroxytamoxifen treatment (Figure S8B). This much expression of the exogenous ERK2 did not notably affect RSK phosphorylation, suggesting that it did not hinder ERK1/2 signaling in cells (Figure S8B). After 6-h ΔRaf:ER activation, total cell lysates were subject to TAP purification performed using streptavidin and calmodulin resins (Agilent Technologies, 240105 and 240106) according to the InterPlay mammalian TAP system protocol (Agilent Technologies, 240107). Purified proteins were separated by SDS-PAGE and visualized by silver staining to reveal only several notable protein bands, likely due to the high stringency of purification conditions (Figure S8C). Selected gel bands were tryptic-digested and analyzed using nanospray-liquid chromatography-mass spectrometry. Two technical replicate injections of 2 μL in succession were performed using instrument settings outlined in Table S3. MS-grade solvents and reagents were used for all these procedures. For protein identification, the mass spectrometry data were searched against the human protein database using ProteomeDiscoverer 2.4 (Thermo Fisher Scientific) according to the details in Table S3. This analysis identified DHPS as a significant component in a 40-kDa band (Table S4). DHPS was selected for further study because its peptide sequence contains two potential ERK1/2 substrate signatures and a D-domain motif (Figure S8D).

QUANTIFICATION AND STATISTICAL ANALYSIS

The two-tailed unpaired student’s t-test was performed to determine statistical significance using Prism (GraphPad), unless otherwise specified. The details of statistical analysis of each experiment are described in the figure legend.

Highlights.

• Identification of a kinase-independent function of ERK1/2

• The structure of ERK2-protein complex addresses kinase-independent ERK2 function

• The structure of DHPS complexed with a protein from outside of the hypusination pathway

• The mechanism by which the direct DHPS and ERK1/2 interaction regulates hypusination