Hydroxyl radical footprinting modification reveals an intradomain communication pathway in EFL1 disrupted by a Shwachman‐Diamond syndrome‐associated mutation

Abstract

Shwachman‐Diamond syndrome (SDS) is a rare genetic disorder characterized by pancreatic insufficiency and neutropenia. While most cases are linked to mutations in the SBDS gene, some involve mutations in the GTPase EFL1. This protein works with SBDS to release the anti‐association factor eIF6 from the 60S ribosomal subunit during ribosome biogenesis. The pathogenic EFL1 R1095Q mutant (R1086Q in yeast) exhibits altered guanine nucleotide recognition and impaired eIF6 release, prompting an investigation into its structural consequences. Using the yeast Efl1 orthologue in X‐ray hydroxyl radical footprinting experiments, we tracked changes in solvent accessibility caused by the mutation. Although the mutation is situated in domain IV, widespread conformational changes were observed across the protein, particularly in domain I, suggesting a long‐range intramolecular communication. Strikingly, the growth defect caused by the pathogenic mutation was rescued by a second mutation located in a allosteric pathway that spans from the nucleotide‐binding pocket to domain IV. This compensatory mutation restored proper nuclear localization of eIF6 (Tif6 in yeast). These findings reveal that the R1086Q mutation disrupts a structural communication network within Efl1, impairing the conformational dynamics required for its activity. The loss of this coordination likely underlies the ribosome maturation defects observed in SDS cases linked to EFL1, offering new insights into the molecular basis of the disease.

1. INTRODUCTION

Ribosome biogenesis is one of the most complex and energy‐demanding processes in a dividing cell (Warner 1999). Cells have evolved quality control mechanisms and checkpoints to ensure the efficient and accurate production of mature ribosomes. This process requires the precise temporal and spatial coordination of nearly 300 assembly factors and 70 small nucleolar ribonucleoproteins (snoRNPs) (Dorner et al. 2023). Defects in ribosomal components or biogenesis factors can impair cell viability or lead to disease (D'Andrea et al. ²⁰²⁴; Farley‐Barnes et al. ²⁰¹⁹; Kampen et al. ²⁰²⁰; Narla and Ebert ²⁰¹⁰). Ribosome biogenesis begins in the nucleolus, where the RNA polymerase I transcribes a polycistronic ribosomal RNA (rRNA) that associates with structural proteins and biogenesis factors to form the pre‐90S subunit. This precursor undergoes extensive processing—including folding, chemical modification, and nucleolytic editing—to generate the mature rRNAs. A key endonucleolytic event splits the pre‐90S complex into pre‐60S and pre‐40S subunits, which mature independently through sequential steps completed in the cytoplasm (Klinge and Woolford 2019; Nerurkar et al. ²⁰¹⁵; Vanden Broeck and Klinge ²⁰²⁴; Woolford Jr. and Baserga ²⁰¹³). The final cytoplasmic maturation step of the 60S subunit requires the eviction of the anti‐association factor eIF6, facilitated by the GTPase EFL1 and its effector protein SBDS (Menne et al. 2007). eIF6 binds to the B6 inter‐subunit bridge of the 60S subunit, physically blocking the association with the 40S subunit. SBDS is recruited to the pre‐60S·eIF6 particle and assesses the integrity of structures such as the peptidyl transferase center (PTC) and the exit tunnel through its domain I, while domain III contacts the sarcin‐ricin loop (SRL) and the P‐stalk. When GTP‐bound EFL1 interacts with the pre‐60S·SBDS·eIF6 complex, it undergoes a conformational change that allows it to contact the SRL and displace eIF6 by competing for an overlapping binding site. EFL1 accommodation triggers GTP hydrolysis, prompting another conformational change that leads to the release of both EFL1 and SBDS from the 60S particle (Weis et al. 2015). Additionally, SBDS also functions as a guanine exchange factor (GEF) for EFL1, facilitating the exchange of GDP for GTP and promoting the active, GTP‐bound form of EFL1 (García‐Márquez et al. ²⁰¹⁵; Gijsbers et al. ²⁰¹³). Failure to release eIF6 disrupts the balance of mature ribosomal subunits, reducing global protein synthesis. This defect is present in patients with Shwachman‐Diamond syndrome (SDS), a condition caused by mutations in SBDS or EFL1. SDS is an autosomal recessive disease characterized by exocrine pancreatic insufficiency, skeletal abnormalities, and bone marrow dysfunction, with an increased risk of developing leukemia (Dror et al. 2011). Approximately 90% of SDS cases are linked to biallelic mutations in SBDS (Boocock et al. 2003), most of which lead to reduced protein expression. In contrast, few homozygous EFL1 missense mutations have been reported, and their effects on protein structure and function remain poorly understood. One such pathogenic mutation, EFL1 R1095Q (R1086Q in yeast), presents similar catalytic activity to the wild‐type protein but shows altered guanine nucleotides and SBDS recognition (Luviano et al. 2019; Perez‐Juarez et al. ²⁰²²). The binding of this mutant to GTP exhibits a thermodynamic signature opposite to that of the wild type: endothermic instead of exothermic, significantly impacting the binding heat capacity change. Unlike the wild type, interaction with SBDS does not induce the mutant to adopt the conformation required for GTP binding, as evidenced by a heat capacity change reduced by half. These results suggest that the wild‐type and mutant Efl1 adopt different conformations upon interaction with their effector molecules. To study the structural basis of these differences, we used synchrotron X‐ray footprinting (XFP) to map conformational changes in both the wild‐type and R1086Q yeast Efl1. These studies revealed significant differences in solvent accessibility in dissimilar regions, indicating that the two proteins adopt distinct conformational ensembles. GTP binding in the mutant induces a conformational change that does not resemble the apo or nucleotide‐bound forms of the wild‐type protein. Further analysis highlighted a region functioning as an internal communication network between the guanine nucleotide‐binding pocket and distal effector sites. Disruption of this communication network likely underlies the defects of the mutant EFL1 in ribosome maturation, offering insight into the molecular basis of SDS in EFL1‐related cases. Introduction of a second‐site mutation P151L—located within this proposed communication backbone—restored the yeast growth defect and corrected the nuclear localization of Tif6 (yeast eIF6) in the R1086Q background. This suggests the possibility of reconditioning mutant Efl1 through conformational modulation, opening avenues for therapeutic strategies that mimic molecular chaperone functions.

2. RESULTS

2.1. Conformational analysis of the Efl1 R1086Q mutant

SBDS and EFL1 facilitate the final maturation of the 60S ribosomal subunit (Finch et al. 2011; Menne et al. ²⁰⁰⁷). We previously described two independent mutations in EFL1 in SDS patients (Stepensky et al. 2017), which, in the context of yeast cells (R1086Q and L910K), recapitulate the inability to effectively release the anti‐association factor eIF6 (Tif6 in yeast) from the ribosomal subunit. Only a few EFL1 mutations have been reported to date (Stepensky et al. 2017; Tan et al. ²⁰¹⁸; Tan et al. ²⁰¹⁹), and their pathogenic mechanisms remain poorly understood. Molecular dynamics (MD) simulations (Delre et al. 2020) and biophysical studies (García‐Márquez et al. ²⁰¹⁵; Luviano et al. ²⁰¹⁹; Perez‐Juarez et al. ²⁰²²) indicate that altered conformational dynamics may underlie their pathogenic effects. To expand on this, we assessed the structural content, size, shape, and flexibility of the R1086Q mutant and contrasted them with the wild‐type protein.

Secondary structure analysis by circular dichroism (CD) revealed subtle differences between wild‐type and mutant Efl1, particularly in the β‐strand region. Both proteins exhibited spectra consistent with a mixed α‐helical and β‐sheet content, characterized by two negative bands at 222 and 208 nm and a plateau at 217 nm. However, the R1086Q mutant displayed a less defined minimum at 220 nm and a slight reduction in the signal at 198 nm (Figure 1a). Deconvolution of the spectra indicated a small increase in β‐sheet content in the mutant (23%) compared to the wild‐type (19%) (Table S1). While far‐UV CD is primarily used to assess secondary structure content, it is also sensitive to tertiary packing of aromatic groups. Thus, it can reflect changes in both backbone and side‐chain conformations. Alterations in side‐chain orientations or tertiary packing—without substantial changes in secondary structure—can influence CD band intensities and positions, such as shifting the 208/222 nm ratio (Sreerama et al. 1999; Woody ¹⁹⁹⁴). Given that Efl1 contains 6.6% of aromatic residues, the observed spectral differences may also reflect subtle changes in the conformation of the mutant protein and not only in secondary structure content.

FIGURE 1.

Structural and conformational comparison of wild‐type Efl1 and mutant R1086Q. (a) Far‐UV circular dichroism (CD) spectra. (b) Thermal stability assessed by differential scanning calorimetry (DSC). Small angle X‐ray scattering (SAXS) analysis showing the (c) SAXS intensity plot, (d) dimensionless Kratky plot; the intersection of the black dotted trace corresponds to the value for bovine serum albumin, representing a compact globular protein, (e) pair‐distance distribution function [P(r)]. Color coding: wild‐type Efl1 (black) and R1086Q mutant (red). The solid lines in (c) represent the theoretical I(q) from the derived P(r) function.

We further examined differences in tertiary structure and stability using Differential Scanning Calorimetry (DSC) and Small‐Angle X‐ray Scattering (SAXS). DSC provides information on the energetics of protein unfolding, with the shape of the thermogram offering insights into domain interactions and conformational cooperativity. Both wild‐type and mutant proteins exhibited irreversible denaturation (data not shown) with two cooperative transitions and three folding states, as indicated by the narrow intervals of 12–15°C for each transition. The first transition occurred at an identical melting temperature (Tm) of 47.5°C for both proteins. The second transition was destabilized in the mutant by 2.5°C (Tm = 62.5°C), without changes in the denaturation enthalpy (Figure 1b), suggesting only minor differences in overall stability. Nevertheless, it has been reported that mutations can produce similar DSC profiles while substantially altering residue‐level dynamics at sites distant from the mutation, even in small, rigid proteins such as Eglin C (Clarkson et al. 2006). Conformational changes without an alteration in thermal stability of the mutant Efl1 suggest a complex interplay between internal interactions and solvent dynamics. Reorganization of hydrogen bonds, where newly formed bonds compensate for the loss of existing ones, may underlie this behavior.

To further probe structural and dynamic differences between wild‐type and mutant Efl1, we performed SAXS experiments (Figure 1c–e and Table S2, Supporting Information). Both proteins were monomeric in solution, with no significant structural differences evident from their similar Kratky plots and Pair Distribution functions. Both exhibited compact, asymmetric shapes, as evidenced by the shift of q*R _g maxima to values >1.7 in the dimensionless Kratky plots. Nonetheless, subtle differences were apparent, such as an increase in compactness for the mutant, as supported by a Porod exponent of 3.1 compared to 3.6 for the wild type. Inspection of the P(r) curves revealed similar elongated shapes and comparable maximal dimensions, although D _max was slightly reduced in the mutant (152 Å vs. 156.5 Å for the wild type).

In summary, these results reveal subtle structural differences between wild‐type and mutant Efl1 that are insufficient to explain the pronounced defects previously reported for the weakened interaction with Sdo1 (SBDS in humans) and the altered thermodynamic signatures for guanine nucleotide recognition, particularly with GTP (Perez‐Juarez et al. ²⁰²²). The R1086 residue is distant from the nucleotide‐binding pocket and the Sdo1 interaction site (Asano et al. 2014; Gijsbers et al. ²⁰¹⁸; Holding ²⁰¹⁵), suggesting that the mutation perturbs Efl1 function from long‐distance effects. We propose that the R1086Q mutation disrupts the intramolecular communication within Efl1, modulating the dynamics of distal residues without grossly altering its mesoscopic structure and thermal stability. To explore this hypothesis, we turned to X‐ray footprinting (XFP) to investigate how the different domains within Efl1 interact with each other and how this communication network is altered by the mutation R1086Q or the interaction with GTP.

2.2. Structural architecture of Efl1 in solution

Samples of Efl1 wild‐type and mutant, either free or bound to GTP, were exposed to synchrotron X‐ray radiation to covalently label the solvent‐accessible amino acid sidechains of Efl1 with hydroxyl radicals generated by X‐ray radiolysis. Oxidized peptides were detected by mass spectrometry (MS), and the extent of modification was quantified as a function of exposure time (X‐ray dose). Radiation dose controls showed consistent oxidation rates of Alexa Fluor 488 across all conditions, ensuring data comparability despite minor variations in buffer composition, for example, salt concentration, reducing agents, and GTP concentration (Figure S1). Mass spectrometry analysis for all experimental conditions yielded peptide coverages ranging from 63% to 67% (Table S3), encompassing peptides from all Efl1 domains (Figure S2). The modification rate (k _fp ) of each peptide was determined by fitting the extent of modification data versus exposure time to a first‐order exponential decay function (Table S4). To correct for intrinsic differences in residue reactivity towards hydroxyl radicals, modification rates were normalized by the intrinsic reactivity (Ri) of the residues comprising each peptide. The Ri of a peptide was calculated as the sum of the reactivities of its residues (Huang et al. 2015), yielding a normalized modification rate referred to as the peptide protection factor (PF = k _Ri /k _fp ; Table S4). This protection factor reflects the peptide solvent accessibility: high PF values indicate protection from solvent, while low PF values suggest exposed regions located on a solvent accessible surface.

To date, no high‐resolution structure of yeast Efl1 is available. The closest structural model is that of the human orthologue, solved by cryo‐EM at a local resolution of 8–9 Å bound to the 60S subunit (Weis et al. 2015). Both yeast and human EFL1 proteins have a large insertion in domain II (Mendez‐Godoy et al. ²⁰²¹), which distinguishes them from Elongation Factor 2 (EF‐2) (Senger et al. 2001). To investigate domain rearrangements under the different experimental conditions studied here, and to map the hydroxyl radical footprinting (XFP) data onto a structural framework, we generated 25 AI‐based Efl1 models using the AlphaFold Server. All models, including the AlphaFold database entry P53893, have coinciding regions of low confidence scores (pLDDT <70 or <50), most notably that within the domain II insertion (residues 425–579). This region, modeled as an extended segment, has been reported to be important for Efl1 interaction with Sdo1 (SBDS in humans) (Asano et al. 2014; Gijsbers et al. ²⁰¹⁸; Holding ²⁰¹⁵). Two additional low‐confidence regions correspond to the loop comprising residues 814–832 and the C‐terminal end (1094–1100) in domain IV. In the AlphaFold predicted models, the insertion exhibits high Cα‐atom RMS fluctuations (5–40 Å), whereas the rest of the protein averaged 3 Å (Figure S3a), suggesting that the insertion may be disordered in solution, with its spatial placement less certain. Although AlphaFold excels at modeling structured regions, it tends to misrepresent disordered regions. To address this, we employed a dual approach to generate a realistic conformational ensemble of Efl1. On one hand, we submitted one AlphaFold model to the programme AFflecto (Pajkos et al. 2025) to generate >10,000 conformers that include both the structured domains and the disordered regions using stochastic sampling algorithms. AFflecto has been shown to complement the limitations of AlphaFold in modeling disordered regions and has been successfully applied to generate conformational ensembles of unstructured regions in proteins such as the measles virus nucleoprotein and human Tau40. Second, we ran a 500 ns molecular dynamics (MD) simulation on the AlphaFold server model (P53893) (Figure S4a,b). MD simulations are a powerful tool for capturing the conformational flexibility of both folded and disordered regions over time (see Data S1 for more details).

To choose models best representing Efl1 behavior in solution, we scored all candidates against experimental SAXS and XFP data using two metrics: χ ² and φ ² (Figure S3b; see section 4.4). SAXS and XFP provide complementary structural insights—SAXS for global shape and domain arrangement, and XFP for local solvent accessibility—making them valuable for modeling complex, multidomain proteins (Hsieh et al. 2017; Huang et al. ²⁰¹⁵; Huang et al. ²⁰¹⁶; Huang et al. ²⁰¹⁸). The χ ² value, calculated using the program Crysol 4.1 (Franke et al. 2017), measures the difference between the experimental and theoretical SAXS profiles. The φ ² value quantifies the average difference in solvent accessibility between the modeled conformation and the experimental XFP data. Lower χ ² and φ ² values indicate better agreement of the model candidate with the experimental data and guide the correct model selection (Figure S3b). All models showed acceptable χ ² and φ ² scores, suggesting reasonable structural representations. However, six AFflecto models appear as top candidates based on their combined scores (red dots in Figure S3b). These models mainly differ in the conformation of their insertion, with an average RMSD of 20 Å when comparing all Cα atoms, while the remaining backbone was nearly identical (RMSD 0.12 Å); this trend is consistent with the initial AlphaFold models (Figure S3a). The goodness of fit for the Efl1 six model ensemble was supported by strong agreement between the experimental and calculated SAXS and footprinting data. The calculated ensemble SAXS profile had a χ ² = 1.7 (Figure S3c), closely matching individual model scores, and an ensemble R _g of 43.6 Å, approaching the experimental R _g of 46 Å (Table S2). In contrast, the AlphaFold model subjected to MD simulation yielded the poorest agreement, with the highest χ ² and φ ² values. During the equilibration phase, the model's insertion (residues 425–579) became less extended and remained largely unchanged over the course of the simulation with an average shell R _g of 41.5 Å, lower than the experimental value. This is evidenced by the increase in the χ ² values across all simulation frames compared to most AFflecto models (Figures S3b and S4b,c). AFflecto preserves the integrity of the predefined structural regions; therefore, χ ² fluctuations arise solely from changes in the extension of the insertion, whereas φ ² variation indicates peptide‐level rearrangements (Figure S3b). In contrast, MD simulations perturb all atoms in the structure. Consequently, although the χ ² values remain relatively stable due to the compact insertion, the φ ² values span a broader range than those observed for the AFflecto models, likely reflecting global structural fluctuations. Indeed, the insertion consists of 46% disorder‐promoting residues (Theillet et al. 2013) and 48% charged amino acids (R, K, D, and E), supporting the idea that it likely adopts a disordered conformation in solution, sampling a wide range of spatial conformations.

To further evaluate the quality of the selected ensemble, we calculated the solvent‐accessible surface area of each peptide, weighted by the sum of its constituent residues' intrinsic reactivity, ⟨SASA⟩, as described in He et al. (2024) and Huang et al. (2015). Analysis of the inverse relationship between ⟨SASA⟩ and the experimentally determined protection factor (PF), plotted on a logarithmic scale, revealed a moderate correlation, with a Pearson correlation coefficient (ρ) of −0.51 (p < 0.0001) (Figure 2a). Residual analysis identified a subset of peptides that deviated from the expected trend, having either higher or lower theoretical ⟨SASA⟩ values than predicted by their corresponding PF. These outliers represent fewer than 15% of the peptide dataset and were predominantly located in domain IV, the insertion region, and the G2 motif within domain I. Excluding these outlier peptides improved the correlation significantly, yielding a Pearson correlation coefficient of −0.72 (p < 0.0001) (Figure 2a). Mapping the PF values for wild‐type Efl1 (in the absence of ligand) onto one of the selected structural models (red spots in Figure S3b) revealed a highly protected core that includes the GTPase active site, with more solvent‐exposed regions located at the protein periphery (Figure 2b). Most outlier peptides were located in loop regions. For instance, the C‐terminal peptide (residues 1090–1096) and those within the insertion (residues 441–450 and 499–511) exhibited higher ⟨SASA⟩ values than expected from their protection factors, suggesting that these regions may be less solvent‐exposed than the model predicts. Interestingly, peptide 441–450 lies within a segment that adopts a long α‐helix in some of the AlphaFold models, implying potential secondary structure transitions. On the other hand, some peptides exhibit smaller ⟨SASA⟩ values than expected from their protection factors, suggesting that the model predicts them as more solvent‐protected than they really are. For example, peptide 67–80 contains the conserved G2 motif R(G/A)ITM, also known as switch 1. This loop behaves as a loaded spring that is conformationally relaxed when bound to GDP but locked as it interacts with the γ‐phosphate of GTP and the Mg(II) ion. These discrepancies, along with the inherently dynamic nature of disordered regions, likely result from the limitations of static structural models in capturing conformational heterogeneity. In contrast, experimental techniques such as X‐ray footprinting and SAXS report ensemble‐averaged signals that reflect all conformations sampled in solution. This may underlie the observed differences between the models predicted and experimental data and further supports the notion that Efl1 exhibits considerable conformational plasticity.

FIGURE 2.

Efl1 local structural conformation in solution. (a) Correlation between experimentally measured hydroxyl radical footprinting protection factors (PF) and the peptide‐weighted solvent‐accessible surface areas (⟨SASA⟩) calculated from the six best‐fit ensemble structures. The Pearson correlation coefficient (ρ) was calculated after removing the outlier peptides (red). Representative outliers have been labeled in red. Error bars indicate the standard deviation in ⟨SASA⟩ across the six ensemble models. (b) Protection factor values mapped onto a representative Efl1 structural model. Colors indicate relative solvent accessibility, with red representing high protection (low accessibility) and blue indicating low protection (high accessibility). Transparent regions correspond to peptides not detected in the hydroxyl radical footprinting experiment. (c) 3D structural model of Efl1 with domains labeled in Roman numerals and highlighted in different colors. Close‐up of the G‐motifs in the active site.

2.3. A mutation in domain IV alters the global structure of Efl1

While the previous sections provided global insights into the stability and dynamics of wild‐type Efl1, they did not identify specific structural regions affected by the R1086Q mutation. To address this, we extended the hydroxyl radical footprinting experiments to mutant Efl1 in its nucleotide‐free state. Comparative analysis of the footprinting data revealed distinct solvent accessibility patterns between Efl1 R1086Q and wild type, affecting all domains rather than being confined to the mutation site (Figure 3). The most pronounced changes were observed in the β‐strands within domains IV and V, near the mutation, where the mutant showed substantial protection with more than threefold decrease in modification rates (peptides 588–594, 860–867, 877–884, 907–927, 971–978). Additional changes in solvent accessibility occurred in domain III (peptides 681–693 and 725–730) and segments in domain II (peptides 365–375 and 579–595), both of which are positioned near the GTPase active site in domain I (Table S5).

FIGURE 3.

The R1086Q mutation induces widespread changes in Efl1 solvent accessibility. (a) Histogram showing the distribution of normalized reactivity ratios across all detected peptides. (b) Structural mapping of solvent accessibility changes between wild‐type and R1086Q Efl1. The color gradient represents the ratio of modification rates: dark blue indicates up to 10‐fold increased exposure (ratio = 0.1), and dark red indicates up to 3‐fold increased protection (ratio = 3.0). Peptides undetected by mass spectrometry are rendered transparent. Roman numerals relate to the protein domains.

The GTPase active site of Efl1 contains the five conserved motifs essential for guanine nucleotide binding: G1 (AHVDHGKT), which coordinates the α‐ and β‐phosphates; G2 (R(G/A)ITM) and G3 (DSPGH), which interact with Mg(II) and the β‐ and γ‐phosphates (also known as switch regions 1 and 2 due to their conformational plasticity upon GTP/GDP binding), and G4 (NK(I/M)D) and G5 (SAID), which confer guanine base specificity (Vetter and Wittinghofer 2001). Our mass spectrometry data covered four of these five motifs (G1, G2, G4, and G5), providing detailed insights into the active site dynamics. Even in the absence of nucleotides, peptides containing the G1 (21–32), G2 (67–80), and G5 (228–237) motifs displayed increased protection in the R1086Q mutant, whereas the G4‐containing peptide (149–157) showed enhanced solvent exposure. Although peptide‐level analysis does not resolve residue‐specific solvent accessibility, such changes in the active site could alter nucleotide binding affinity. Explaining the calorimetric studies showing that the R1086Q mutant has a 10‐fold increase in GTP affinity (Perez‐Juarez et al. ²⁰²²), likely resulting from the conformational shifts revealed by XFP.

Detected peptides within the characteristic Efl1 insertion showed minimal changes in modification rates, suggesting this region remains unstructured in both wild‐type and mutant Efl1. A notable exception is peptide 499–511, which showed a threefold increase in modification rate in the mutant (Figure 3b and Table S5). This observation presents an apparent paradox, as the same peptide in the wild‐type protein displayed a high protection factor (Figure 2b), despite its location in a seemingly solvent‐exposed region as proposed for the insertion. Correlation analysis flagged this peptide as an outlier (Figure 2a), indicating a possible misrepresentation in the structural ensemble. We suggest that in wild‐type Efl1, peptide 499–511 adopts a partially folded conformation or engages in transient interactions with the protein core, reducing its solvent accessibility. The alternative explanation that this peptide adopts local structure but does not interact with the core of the protein seems less likely since XFP primarily probes the side chains of solvent‐accessible residues rather than backbone amide hydrogens involved in secondary structure stabilization. Therefore, it is more plausible that this peptide, even if disordered or partially folded, interacts with the core of the protein in the wild type, thus protecting it from modification. In the R1086Q mutant, this interaction is likely disrupted, resulting in increased exposure (Figure 3 and Table S5). Together, these findings indicate that the R1086Q mutation propagates long‐range conformational changes across Efl1, affecting domains spanning from the GTPase active site to the unique insertion domain. This underscores the presence of extensive allosteric communication networks essential for maintaining the protein's structural integrity and function.

2.4. Efl1 R1086Q shows different dynamics compared to the wild type upon GTP binding

Efl1 is a molecular motor whose function is mediated by a set of conformational changes promoted by the interaction with different molecules, such as the Sdo1 protein and guanine nucleotides. As it has been described for translation GTPases, guanine nucleotides drive the protein through states known as the apo‐form, the T‐form (bound to GTP), or D‐form (bound to GDP) (Vetter and Wittinghofer 2001; Wittinghofer and Vetter ²⁰¹¹). To understand the extent of these changes, we initially compared the modification rates of Efl1 in the absence of ligand and bound to GTP. Efl1 bound to GTP showed peptides with 5.0‐fold exposure (peptides 198–203, 365–373, 432–437) and a 2.7‐fold protection (peptides 33–49, 595–605, 705–716) compared to the apo‐form (Figure 4a). As expected, mapping these changes onto a structural model shows that peptides in the active site get protected upon GTP binding (Figure 4a and Table S5). This is true for peptides 149–157 and 228–237, containing residues belonging to the G4 and G5 motifs, respectively. For peptides 21–32 and 67–80, containing residues from G1 and G2 motifs that anchor the γ‐phosphate, there seems to be the opposite effect. However, the observed modification rate of a peptide is the sum of the individual residues' modification rates, and the intrinsic reactivity of each amino acid is modulated by its local sequence context, impacting the calculated Protection Factor and its quantitative relationship with ⟨SASA⟩ values (Xie et al. 2017). Results at the residue level of peptide 21–32 showed an increase in the modification rates of H27 (1.4‐fold) and V28 (2.6‐fold), that is, enhancing their solvent accessibility. Meanwhile, K32 got protected 4.2‐fold upon GTP binding (data not shown). The sequence of the G1 motif in Efl1 is 26‐AHVDHGKT‐33, complying with the consensus sequence across the GTPase family (G/A)XXXXGK(S/T) (Vetter and Wittinghofer 2001). The lysine within the G1 motif has direct contact with γ‐ and β‐phosphate of the nucleotide; hence, it is expected to become protected, as proven by the XFP results. The fact that other residues get exposed is only a reflection of the conformational change that occurs in that area. Interestingly, upon GTP binding, peptides located at different regions of the protein, including the most distant domains, showed exposure up to a 5‐fold rate change (Figure 4a). Six peptides have the most significant changes of all: 432–437 and 479–496 from domain II, 681–693 from domain III, 762–773 and 907–927 from domain IV, and 1001–1007 from domain V (Table S5). Peptides from domain II are located within the insertion. Interestingly, peptide 432–437 is adjacent to S430, a residue known to be phosphorylated (Lanz et al. 2021) and contains residue M435 that has been reported to directly interact with Sdo1 (Holding 2015), suggesting that this region may play a role in Efl1 function and may undergo conformational transition upon binding to its target. Peptide 681–693 is located in domain III, next to a loop connecting domain III‐IV, known to act as a structural hinge between domain III and the C‐terminal “super domain” of GTPases such as EF‐2 (Jorgensen et al. 2004). Lastly, peptides in domains IV and V group with a series of peptides that increase their solvent accessibility upon GTP binding (762–773, 907–927, 958–978, 1001–1007, 1034–1051), forming a β‐strand trail towards the active site. Overall, these observations agree with earlier reports (García‐Márquez et al. ²⁰¹⁵; Luviano et al. ²⁰¹⁹) and indicate that GTP binding triggers a conformational change in the protein. XFP analysis reveals that this change is global in nature, supporting the existence of a long‐range communication mechanism through which Efl1 senses its nucleotide‐binding state. The GTP hydrolysis rate of Efl1 is very low, such that it can be considered more of a GTP binding protein (k _cat = 0.5 min⁻¹) (Gijsbers et al. 2013). It is only when Efl1 interacts with the 60S ribosomal subunit that hydrolysis is triggered, so initial conformational changes result from the interaction with GTP rather than hydrolysis (Finch et al. 2011).

FIGURE 4.

Hydroxyl radical footprinting analysis reveals differences between wild‐type Efl1 and R1086Q mutant in response to GTP binding. (a) Wild‐type Efl1 and (b) R1086Q mutant. Left panels: histograms showing the distribution of normalized reactivity ratios across all detected peptides. Right panels: ratios of hydroxyl radical reactivity mapped onto a predicted model of Efl1. The color gradient indicates changes in solvent accessibility relative to the GTP‐bound condition, with red representing increased protection and blue representing increased exposure, as shown in the color scale bar. A ratio of 0.2 corresponds to a 5.0‐fold increase in exposure (dark blue), while a ratio of 2.7 indicates a 2.7‐fold increase in protection (dark red). Transparent regions indicate undetected peptides. Roman numerals relate to the protein domains. (c) Close‐up views of regions exhibiting significant differences in solvent‐accessibility between conditions. Region positions are indicated by yellow rectangles in panels (a) and (b), numbered with Arabic numerals.

Previous studies have shown that the R1086Q mutant retains the same catalytic properties as the wild‐type Efl1. However, it fails to undergo the conformational change observed in the wild‐type protein upon GTP binding, as indicated by differences in their binding heat capacity change (Perez‐Juarez et al. ²⁰²²). This thermodynamic parameter reflects global conformational rearrangements, but small changes in heat capacity do not necessarily imply the absence of structural changes—they can result from the compensation of opposing effects, such as simultaneous exposure or burial of polar residues and nonpolar ones. Our data suggest that apo wild‐type Efl1 and the R1086Q mutant adopt distinct conformations (Figure 3). Nevertheless, it remains unclear whether the mutant structure remains unaltered upon GTP binding or whether it undergoes compensatory changes. To address this, we compared hydroxyl radical modification rates of the mutant in the free and GTP‐bound states. The results reveal that structural changes do occur in the mutant upon GTP binding, but to a lesser extent than in the wild type. This is reflected in the lighter coloring in the structural representation and in the histogram distribution, where most peptides cluster around a normalized ratio of one (Figure 4b). Similar to the wild‐type, the most protected peptide in the mutant comprises residues 33–49 (Table S5) containing T33, which is part of the G1 motif and contacts the β‐phosphate and the Mg(II) ion. On the other hand, the most exposed peptide is 907–927 in domain IV, with a dramatic increase in modification rate upon GTP binding (more than 5‐fold): from a nearly zero rate in the apo form to 0.9 s⁻¹. Although this represents a significant change, it remains modest compared to the wild‐type protein, which shifts from 3.9 to 9.5 s⁻¹. Notably, this peptide lies close to residue 1086, suggesting that the mutation may disrupt local interactions, altering the conformation and solvent accessibility of this region. Striking differences emerge when comparing wild‐type and mutant responses to GTP binding (Figure 4c). Despite overall attenuated structural changes in the mutant, certain regions stand out, including peptides in domain III (Figure 4c, inset 3), the previously described trail extending from the active site to domain IV (Figure 4c, inset 4), and peptides in the insertion (Figure 4a,b). These findings suggest that GTP binding initiates long‐range conformational changes that propagate through a defined pathway along Efl1. The R1086Q mutation appears to impair this communication. We propose that this structural pathway transmits information from the GTP‐binding site to domain IV, which we hypothesize functions as an effector domain. If we were to anthropomorphise the Efl1 structure, domains I and II would represent the body, domains III and V the neck, and domain IV the head, with the conformational relay—the spinal cord—connecting them (Figure 5). In this analogy, the R1086Q mutation renders Efl1 paraplegic: the spinal cord pathway is disrupted, the signal is not transmitted, and proper conformational change fails to occur. This disruption likely explains why, despite showing a tenfold increase in GTP affinity (Perez‐Juarez et al. ²⁰²²), the R1086Q mutant is functionally defective. Supporting this model, six disease‐associated mutations (in their yeast equivalents) are located within this spinal cord region, underscoring the critical role of this structural communication path (Figure 5a), with R1086Q located nearby (Figure 5b). This observation is consistent with our previous studies on human EFL1 mutants M882K (domain IV), T127A (domain I), and R1095Q (domain IV; equivalent to R1086Q in yeast), in which we analyzed inter‐domain motions using molecular dynamics simulations. Despite the distinct locations of these mutations, all simulations revealed pronounced conformational changes in domain IV, as reflected by significant RMSD variations (Delre et al. 2020). Taken together, this orthogonal methodological approach supports the existence of a conserved connectivity pathway between the active site (domain I) and the effector domain (IV), shared by the human and yeast orthologues, enabling the sensing of nucleotide‐binding status and ensuring proper protein function.

FIGURE 5.

“Spinal cord” conformational change trajectory in Efl1. Highlighted peptides in wild‐type Efl1 (a) and the R1086Q mutant (b) form a continuous path extending from the GTP‐binding site to domain IV. This pathway likely serves as a structural conduit for sensing GTP binding. Coloring and intensity follow the scheme in Figure 4; transparent residues lie outside this pathway. Roman numerals relate to the protein domains. Solid‐colored segments correspond to peptides 21–36, 122–157, 228–242, 753–790, and 860–1051. Residues not detected by mass spectrometry are included in the model if spatially adjacent to the detected peptides with a white color. Residues labeled in yellow indicate yeast homologs of pathogenic variants reported in Shwachman‐Diamond syndrome patients, while those in green represent the suppressor variants tested in this work. Left‐hand models match the orientation used throughout the manuscript; the long‐disordered insertion, unrelated to this analysis, has been cropped for clarity.

Efl1 binds GTP and undergoes a conformational change into the T‐form, a transition that does not occur in the R1086Q mutant. To explore the possibility that the mutant already adopts a conformation resembling the T‐form—thus exhibiting only minimal structural rearrangement—we compared hydroxyl radical modification rates of GTP‐bound Efl1 with those of the mutant in both nucleotide‐free and GTP‐bound states. The analysis revealed widespread changes throughout the mutant protein, which were more pronounced than those observed when comparing the wild‐type protein in its nucleotide‐free form (Figure S5). Strikingly, when comparing structural differences between wild‐type and mutant proteins (Figure 3), as well as between GTP‐bound wild‐type and GTP‐bound or nucleotide‐free mutant proteins (Figure S4), we observed an uncannily similar pattern of changes. The same peptides showed solvent protection or exposure across these comparisons, with the only distinction being the magnitude of the effect (Table S5). Suppose the overall conformational change could be represented on a scale with the wild type apo form as a reference point. In this case, it appears that the wild‐type T‐form and apo mutant conformation move in opposite directions, supporting the notion that the mutant is not in the T‐form either. Another noteworthy point is that calorimetric results from previous studies indicated that the R1086Q mutation does not affect the conformational change triggered by GDP (as reflected by similar heat capacity changes with the wild type), nor does it alter the dissociation constant (Perez‐Juarez et al. ²⁰²²). This observation suggests that the pathway by which nucleotide‐binding information is transmitted differs between the substrate‐ and product‐bound states. Future XFP experiments with Efl1 in the presence of GDP may shed light on this behavior.

2.5. Second‐site variants restore structural communication disrupted by the R1086Q mutation

The C‐terminal region of the Elongation Factor 2 and EFL1 protein families comprises domains III–V, collectively known as the C‐terminal “super‐domain,” helps anchor these factors to the 60S ribosomal subunit (Jia et al. 2024; Sengupta et al. ²⁰⁰⁸). Previous studies demonstrated that this region is essential for Efl1 function. Specifically, deletion of the last 22 residues of Efl1 leads to growth defects in yeast, which can be rescued by intragenic suppressor mutations such as P151L and T657M (Becam et al. 2001). These final 22 residues include part of a conserved motif within the EFL1 family, R(R/K)RKGL (Mendez‐Godoy et al. ²⁰²¹), whose function remains unknown. Interestingly, the arginine at position 1086 (R1086) is the first residue of this motif. Based on the hypothesis that suppressor mutations counteract the growth defects caused by disruption of this motif, we tested whether the P151L or T657M variants could also restore aberrant intramolecular communication introduced by the pathogenic R1086Q mutation. As previously discussed, far‐UV circular dichroism provides information on the protein's overall secondary structure. Although the R1086Q mutation induces only subtle changes in the protein CD spectrum, these differences nonetheless reflect the significant conformational alterations revealed by the X‐ray footprinting data (Figure 3). We therefore examined whether the double mutants (R1086Q/P151L and R1086Q/T657M) would display CD spectra more similar to that of wild‐type Efl1. As predicted, both double mutants exhibited CD spectra nearly identical to the wild‐type protein. Notably, the R1086Q/P151L double mutant fully restored both the signal at 195 nm and that within the β‐strand region to wild‐type levels (Figure 6a). The R1086Q/T657M mutant also improved the β‐strand signal but still displayed a reduced amplitude at 195 nm, like the single R1086Q mutant. To determine whether these structural corrections were accompanied by functional recovery, we evaluated growth under translational stress and Tif6 localization in haploid efl1Δ/tif6‐GFP yeast cells complemented with centromeric plasmids encoding either the single R1086Q mutant or the two double mutants.

FIGURE 6.

Second‐site variants rescue structural and functional defects caused by the R1086Q mutation in Efl1. (a) Far‐UV circular dichroism spectra of Efl1 double mutants compared to wild‐type and the single R1086Q mutant. The inset displays the zoomed region of the CD minima. (b) Growth kinetics under translational stress in YEPD medium supplemented with 40 μg/mL hygromycin. The inset table shows calculated specific growth rates (μ); asterisks denote statistical significance difference compared to wild type (WT) (one‐way ANOVA with Dunnett's post hoc test). (c) Fluorescence microscopy showing Tif6‐GFP localization. Scale bar = 5 μm. (d) Quantification of nuclear Tif6‐GFP fluorescence. Asterisks denote statistically significant differences (n = 215 cells, Kruskal–Wallis with Bonferroni post hoc test). The dashed line indicates the median, and the dotted lines indicate the first and third quartiles. Color code: WT‐black, R1086Q‐red, R1086Q/T657M‐orange, R1086Q/P151L‐blue.

Aminoglycosides like hygromycin impair translation by binding the ribosomal decoding center and interfering with tRNA–codon pairing. Hence, sensitivity to hygromycin indicates translational dysfunction due to insufficient ribosome biogenesis. In rich YEPD medium, all strains exhibited comparable growth (Figure S6). However, under hygromycin stress, cells expressing the R1086Q single mutant showed significantly delayed growth, whereas the double mutants grew similarly to WT (Figure 6b). This was confirmed by specific growth rate (μ) analysis, where the R1086Q strain had the slowest growth, and the double mutants showed intermediate μ values. All cultures eventually reached the same maximum biomass, but the R1086Q mutant took 24 h, compared to 18 h for the wild type and 20 h for the double mutants.

To further test functional recovery, we analyzed the subcellular localization of Tif6‐GFP (eIF6 in human). Under normal conditions, Tif6 shuttles between the nucleus and cytoplasm but is predominantly nuclear at steady state (Basu et al. 2001). When either Efl1 or its cofactor Sdo1 is non‐functional, Tif6 cannot be released from pre‐60S particles in the cytoplasm and becomes redistributed throughout the cell (Menne et al. 2007; Stepensky et al. ²⁰¹⁷). Consistent with this, fluorescence microscopy revealed that Tif6‐GFP was evenly distributed in both the nucleus and cytoplasm of cells expressing the R1086Q mutant. In contrast, the double mutant strains displayed nuclear localization of Tif6‐GFP, similar to the wild type (Figure 6c). Quantification of fluorescence confirmed these observations: nuclear Tif6‐GFP localization averaged 90% in wild‐type and both double mutant strains, compared to 75% in cells expressing the R1086Q single mutant (Figure 6d). These results indicate that the second‐site mutations partially restore the functional impairment caused by the R1086Q mutation, although the extent of recovery varies slightly between variants. Doubling time analysis further supported this trend: both double mutants exhibited significantly faster growth than the R1086Q mutant alone but still grew more slowly than the wild type. The R1086Q/P151L mutant showed a statistically significant difference from wild type (p < 0.01), while the R1086Q/T657M variant differed more markedly (p < 0.0001). Additionally, Tif6‐GFP localization in the R1086Q/P151L strain was more tightly confined to the nucleus (less dispersed signal), resembling wild‐type localization more closely than that observed in the R1086Q/T657M mutant. These observations suggest that both suppressor mutations mitigate the structural and functional defects introduced by R1086Q; however, the P151L variant appears to restore Efl1 function more effectively. Notably, residue P151 is located within the spinal cord region that undergoes differential conformational changes in the mutant compared to the wild type and seems to connect the active site with the effector region of Efl1. In contrast, residue T657 is located in domain II, far away from this path (Figure 5b). The fact that it also restores Ef11 function, albeit less effectively, suggests the presence of additional compensatory communication networks.

3. DISCUSSION

The use of X‐ray hydroxyl radical footprinting (XFP), in combination with SAXS and in silico modeling, allowed us to generate a coherent structural representation of Efl1 in solution that reconciles its global architecture with local dynamics. A conspicuous feature of this model is the large insertion encompassing residues 425–579, which XFP and SAXS identifies as highly flexible and largely disordered. Such a long (~150‐residue) intrinsically disordered segment is unusual in otherwise well‐folded proteins and is therefore suggestive of a specialized regulatory role rather than a simple structural appendage. Within this insertion, several peptides—most notably 432–437 and 479–496—sample alternative conformations and become more solvent‐exposed upon GTP binding, indicating that the insertion participates in nucleotide‐dependent structural rearrangements. Additionally, peptide 499–511 becomes markedly more exposed in the pathogenic R1086Q background, a change that points to an unexpected long‐range coupling between the C‐terminal region and this insertion.

Biochemical and thermodynamic studies support a functional role for the insertion as the primary docking site for Sdo1 (SBDS in human), the guanine exchange factor (GEF) of Efl1. Asano and colleagues (Asano et al. 2014) showed that the isolated insertion retains nanomolar affinity for Sdo1, comparable to full‐length Efl1, underscoring the insertion's centrality to Efl1–Sdo1 recognition. Complementary evidence from crosslinking experiments identified M435 as directly contacting Sdo1 (Holding 2015); a residue contained in peptide 432–437 that greatly changes its solvent accessibility and gets buried in the mutant upon GTP binding, an opposite behavior to the wild‐type protein. Luviano and co‐workers (Luviano et al. 2019) further reported that Sdo1 binding profoundly alters the thermodynamics of nucleotide recognition: the presence of Sdo1 dramatically reduces the magnitude of the heat capacity change associated with GTP binding, consistent with Sdo1 pre‐organizing Efl1 into a conformation that resembles the active, GTP‐bound (T) state. Taken together with our SAXS and XFP observations, we propose that, following release form the pre‐60S particle, Efl1 remains GDP‐bound and is primed for GDP → GTP exchange through its interaction with Sdo1. Whether the two factors remain associated or transiently dissociate in the cytoplasm remains to be determined.

Our XFP data provide insights into how the R1086Q mutation undermines this functional cycle. Switch I (the conserved G2 motif) is highly responsive to nucleotide binding state in Efl1: GTP binding in the wild‐type protein causes a 2.5‐fold increase in solvent exposure of the peptide containing the G2 motif, consistent with its “unlocking” that accompanies the transition into the active conformation. By contrast, in the nucleotide‐free state, the same peptide is 1.4‐fold more protected in the R1086Q mutant relative to wild type, indicating that the mutation biases switch I towards a more closed or pre‐organized conformation even in the absence of nucleotide. Importantly, R1086Q does not abolish GTP binding per se; instead, it alters the initial conformation and attenuates the magnitude of switch I rearrangements that usually follow nucleotide engagement. This change effectively decouples nucleotide recognition from the long‐range conformational relays needed to reposition domain IV, which in turn enables eIF6 displacement. This allosteric perturbation of the switch I conceptually parallels regulatory mechanisms described for other translation GTPases. Post‐translational modifications that perturb the β‐hairpin geometry of switch I shift the conformational equilibrium towards non‐productive states (Pereira et al. 2015; Redpath et al. ¹⁹⁹³). For example, phosphorylation of EF‐Tu's switch I or of a nearby residue traps the factor in an open or GDP‐like conformation, uncoupling nucleotide binding from downstream conformational dynamics and inhibiting translation in vivo (Talavera et al. 2018). In Efl1, the DELMER motif lies immediately adjacent to the RIASSDELMER sequence (residues 427–437), whose serine residues have been identified as phosphorylation sites in proteomic datasets (Lanz et al. 2021) and therefore represent plausible regulatory nodes. Phosphorylation in this local context could modulate the insertion's conformational ensemble, alter DELMER exposure, and thereby influence Sdo1 binding. Given the long‐distance connection between the insertion and switch I revealed by XFP, such phosphorylation events could conceivably alter Efl1 ensembles much as phosphorylation regulates EF‐Tu activity by stabilizing off‐like conformations.

Functionally, decoupling nucleotide binding from conformational cycling produces a deleterious outcome: the mutant Efl1 becomes trapped in a unique conformation that is neither the native apo state nor the productive T‐form, impairing Sdo1 interaction and likely weakening engagement with the 60S particle. The net result is a failure to transmit the nucleotide‐binding signal across the intramolecular “spinal cord” that links the GTPase active site to the effector domain (IV), preventing the coordinated domain rearrangements required to displace eIF6 and complete 60S maturation. This model rationalizes how disrupted conformational coupling prevents eIF6 eviction and thereby compromises 60S maturation in EFL1‐linked Shwachman‐Diamond syndrome and suggests a mechanistic basis that could account for other pathogenic EFL1 variants. Encouragingly, our study also shows that this broken communication can be at least partially restored by intragenic suppressors. The P151L substitution, located within the proposed communication backbone, most likely restores the wild‐type pattern of conformational change as it rescues growth under translational stress and corrects Tif6 (eIF6) localization. In contrast, a more distal suppressor, T657M, produces a milder rescue. These findings demonstrate that distinct structural nodes can re‐establish dynamic connectivity, highlighting multiple leverage points for modulating EFL1 behavior.

From a translational perspective, these results raise an appealing therapeutic concept: small molecules or designed peptides that stabilize productive intramolecular communication might mimic the effect of suppressor mutations, re‐enable eIF6 eviction, and ameliorate defects in ribosome maturation in EFL1‐linked Shwachman‐Diamond syndrome. While our in vitro ensemble does not capture the full complexity of the ribosomal assembly landscape, integrating XFP, SAXS, and functional genetics offers a rich framework for targeting the conformational landscape of EFL1. More broadly, our work underscores that the functional output of translation‐associated (TRAFAC) GTPases depends as much on the integrity of their conformational relays as on their nucleotide chemistry, and that restoring dynamic coupling—rather than merely modulating nucleotide binding—may represent a promising route to correct ribosomopathies that arise from disrupted allostery.

4. MATERIALS AND METHODS

4.1. Protein expression and purification

Saccharomyces cerevisiae Efl1 wild type (WT) and mutants R1086Q, R1086Q/L151P, and R1086Q/T657M were expressed in S. cerevisiae BCY123 and purified as described before (Gijsbers et al. 2016). Finally, size exclusion chromatography was used to buffer exchange the protein. The nature of the buffer varied depending on the experiment and is described in the corresponding sections. Proteins were used fresh when possible or flash frozen in liquid nitrogen and stored at −80°C until used.

4.2. Hydroxyl labelling with synchrotron X‐rays

Four different Efl1 samples were used for the X‐ray footprinting experiments (XFP): wild type (WT), mutant R1086Q (MUT), alone and in the presence of GTP (WT + GTP and MUT + GTP). Irradiated samples at four exposure times (0, 12, 20, and 30 ms) contained 2 μM protein in a buffer consisting of 50 mM phosphate, pH 7.5, 300 mM NaCl, and 5 mM MgCl₂. The buffer for the mutant protein was identical except for the addition of 0.5 mM TCEP. To ensure more than 90% complex, the experiments done in the presence of nucleotides contained 650 μM GTP for the wild‐type protein and 200 μM GTP for Efl1 R1086Q (Luviano et al. 2019; Perez‐Juarez et al. ²⁰²²). Eight replicates for each sample and exposure conditions were used. The fluorescence decay of Alexa Fluor 488 (4 μM) was used as a proxy for radiation dose in the samples. Rates of 80 s⁻¹ were obtained in the presence of aluminum attenuators of 254 μm (WT), 51 μm (WT + GTP), 152 μm (MUT), and 76 μm (MUT + GTP) (Figure S1). A 96‐well high‐throughput apparatus was used to expose the samples with a 2.5 mm diameter X‐ray beam to ensure complete coverage of the 5 μL sample droplet (Asuru et al. 2019; Jain et al. ²⁰²¹). The hydroxyl labelling reaction was quenched with 10 mM methionine immediately after exposure, and the samples were frozen in liquid nitrogen until subsequent use for the mass spectrometry analysis.

4.3. Mass spectrometry sample preparation and analysis

Sample replicates for each condition were pooled together and prepared for protease digestion in a buffer consisting of 20 mM Tris, pH 8.5, 1.5M urea, 10 mM DTT, and incubated for 45 min at 37°C. The samples were alkylated for 60 min at 37°C with 25 mM iodoacetamide, and initially digested with 6.5 ng/mL of LysC for 2 h at 37°C. Followed by a second digestion with 6.5 ng/mL of trypsin overnight at 37°C. Trypsin was deactivated by heating the samples at 95°C for 4 min. Subsequently, each sample was divided equally into two aliquots; one aliquot was further digested with 4 ng/mL of rAspN (Promega) at 37°C for 6 h, and the reaction was stopped by acidifying the samples to pH <3.0 with 2% formic acid.

Digested Efl1 samples from both proteolysis protocols (LysC + Trypsin and LysC + Trypsin + rAspN) were mixed and diluted with 0.1% formic acid for mass spectrometry analysis. Samples were analyzed with a Waters nanoACQUITY UPLC C₁₈ trap column (100 Å, 5 μm, 180 μm × 20 mm) and eluted from a Waters ACQUITY UPLC BEH C₁₈ analytical column (300 Å, 1.7 μm, 75 μm × 250 mm) coupled to a Thermo OrbitrapTM Eclipse Tribrid mass spectrometer as previously described (Jain et al. 2024). The two mobile phases consisted of: phase A—0.1% formic acid in water, and phase B—0.1% formic acid in acetonitrile at a 300 nL/min flow rate. The UPLC gradient consisted of 1–12% B for 5 min, 12–50% B for 5–95 min, 50–70% B for 95–100 min to allow the complete elution of Efl1 peptides. The temperature of the analytical column was maintained at 40°C and 2.1 kV capillary voltage. Mass spectra were acquired in positive polarity. The Orbitrap resolution was set to 120,000, and the RF lens setting was 30 with an automatic gain control (AGC) target of 4 × 10⁵ ions, and the maximum injection time was 50 ms. Precursor ions were chosen in a Data‐Dependent acquisition (DDA) mode, followed by MS/MS fragmentation by Collision‐Induced Dissociation (CID) with energy set at 35% with an AGC target of 10,000 ions and IonTrap detection of fragment ions.

MassMatrix Xtreme (v3.0.10.25) was used to search for unmodified and hydroxyl‐labeled peptides (variable hydroxyl modifications, with iodoacetamide modification +57 Da as a fixed change). Peptides were searched with a maximum of two missed cleavages ±10.00 ppm peptide mass tolerance and ±0.80 Da (CID) fragment mass tolerance. Xcalibur software was used to manually extract chromatograms and analyze the modification of the Efl1 peptides based on MassMatrix Xtreme output. Modification rates were calculated by fitting the normalized unmodified fraction of each peptide as a function of exposure time to a single exponential decay equation using the program Kaleidagraph (Synergy Software). For comparison between protein samples, rate constants were expressed as normalized ratios. We obtained the average of each set of rates and then divided each ratio by this average. This normalization assumes that most peptide modification rates remain unchanged, effectively centering most values around one and highlighting peptides with significant deviations.

4.4. Structural model of Efl1

To choose an Efl1 structural model that closely resembles that in solution, we initially generated 25 AlphaFold models (https://alphafoldserver.com/) in addition to the existing model in the AlphaFold structure database (AF‐P53893). These models were assessed against experimental SAXS and footprinting data via two scoring functions, χ ² and φ ², respectively. The best model was used to generate a larger collection that explored the conformational diversity of Efl1 flexible regions using the programme AFflecto (Pajkos et al. 2025). We used the recommended parameters and the default structural region classification identified by the programme and ran the simulations using the local LAAS‐CNR server. The >10,000 conformations generated together with the AlphaFold models were scored again to select the best conformational ensemble. Side chains for the AFflecto models were added with the program Pulchra (Rotkiewicz and Skolnick 2008). The goodness of fit of each structural model against the experimental SAXS data was evaluated via the function χ ² calculated using the programme Crysol 4.1 (Franke et al. 2017). Similarly, φ ² represents the goodness of fit between experimental and theoretical footprinting data calculated as follows (Hsieh et al. 2017; Huang et al. ²⁰¹⁶):

$${\phi }^2=\frac{1}{N_{\mathit{fp}}}\sum _i\frac{{\left({\text{logPF}}_{\text{calc}\left(p\right)}-{\text{logPF}}_{\exp \left(p\right)}\right)}^2}{{{\delta }_i}^2},$$

where N _fp is the number of footprinted peptides used in the analysis, which in this case equals 54, and logPF_calc(p) is the predicted logPF value for each peptide calculated using the linear regression from the dependence between the experimental logPF_exp of each peptide (p) and their corresponding ⟨SASA⟩ obtained from the model in question, as follows:

$$\log {\mathrm{PF}}_{\text{calc}\left(p\right)}=m⟨\text{SASA}⟩+b$$

4.5. Small‐angle X‐ray scattering

Small‐angle X‐ray scattering (SAXS) data were collected on the bioSAXS beamline B21 operated by Diamond Light Source (Didcot, Oxfordshire, UK) (Cowieson et al. 2020) coupled to online size exclusion chromatography (SEC‐SAXS). Samples consisting of 50 μL at a concentration of 6 mg/mL were injected onto a Shodex KW‐403 column equilibrated with a buffer composed of 50 mM HEPES, pH 8, 300 mM NaCl, 5 mM MgCl₂, and 5% glycerol, with a flow rate of 75 μL/min. Initial SEC data were analyzed with the ScÅtter program. The values of I(0), molecular weight (MW), maximum dimension particle (D _max), and radius of gyration (R _g ) were calculated using PRIMUS in the ATSAS version 4.1.2 package (Franke et al. 2017). The tool CRYSOL in ATSAS was used to compute the theoretical SAXS curves from the Efl1 models and compare them with the solution scattering data. Detailed data can be found in Table S2. SAXS data were deposited in the Small Angle Scattering Biological Data Bank (www.sasbdb.org) (Kikhney et al. 2020) with the following accession codes: SASDYG4 (wild‐type yeast Efl1) and SASDYF4 (mutant R1086Q yeast Efl1).

4.6. Differential scanning calorimetry

Differential scanning calorimetry (DSC) measurements were performed using a VP‐DSC Capillary Cell Microcalorimeter‐Autosampler (MicroCal, Inc.). Protein concentration ranged between 1 and 2 mg/mL in a buffer containing 50 mM HEPES, pH 8.0, 300 mM NaCl, 5 mM MgCl₂, 10% (v/v) glycerol. Samples and buffer were degassed for 10 min before loading into the calorimetry instrument. Buffer baselines were collected under identical conditions and were subtracted from the corresponding protein samples' data. A scan rate of 1°C/min was used. Data were processed using Origin7 SR4 (v7.0552) software.

4.7. Circular Dichroism spectroscopy

Analysis of the secondary structure content was performed using a JASCO J‐1500 spectropolarimeter (Malvern Panalytical) equipped with a Peltier temperature controller. Experiments were carried out in triplicate in 20 mM phosphate buffer, pH 8, 300 mM NaF at a protein concentration of 1 μM and a 1 mm cuvette. Scan wavelength was recorded from 190 to 260 nm, with an increase of 1 nm per step, an averaging time of 5 s, and a bandwidth of 1 nm. The signal of the buffer alone was subtracted from the protein. Data were analyzed using the CDSSTR programme training set 7 in the DichroWeb server (Miles et al. 2022).

4.8. Growth assays

Plasmids to complement haploid efl1Δ yeast cells were constructed using the pRS316 vector and included a cassette comprising the Efl1 coding sequence and 500 bp upstream and downstream of it. Mutations were introduced by inverse PCR reactions using the plasmid containing the wild‐type sequence as a template. The yeast strain AM2 transformed with the corresponding plasmids was used for the complementation assays and selection of the deleted haploid progeny as previously described (Baryshnikova et al. 2010; Menne et al. ²⁰⁰⁷). The resulting complemented strain (JAZD2H; Table S6) was used to test growth rates. Cells were grown in rich liquid media supplemented with and without hygromycin (40 μg/mL) for 1 day at 30°C with continuous shaking using 100‐well plates in a Bioscreen C MBR instrument. Growth curves were generated by adjusting the initial OD_600nm to 0.05 in a volume of 200 μL of media. Curves correspond to the average of 10 biological replicates and three technical replicates each. The specific growth rate (μ) was calculated as the slope of the exponential growth phase expressed as the logarithm of the population size versus time. Data were analyzed with a one‐way ANOVA and a Dunnett's multiple comparison.

4.9. Tif6‐GFP localization by fluorescence microscopy

Subsequently, each efl1Δ haploid strain complemented with the corresponding plasmid was genomically tagged with GFP in‐frame of the Tif6 gene (strain JZDH6L; Table S6). To do so, each strain was transformed with a cassette consisting of the last 65 bp of Tif6 fused in frame with the coding sequence of GFP and the ADH1 terminator, followed by a LEU2 cassette as a selection marker and 50 bp of the sequence downstream of the stop codon after the Tif6 gene to allow recombination. Correct insertion of the GFP was verified by sequencing of the full Tif6‐GFP coding sequence.

Cells were grown to the exponential phase in SD‐Ura media supplemented with 2% glucose before fixing the cells for 1 h with 10% formaldehyde and staining with 10 μg/mL Hoescht 33258 (Sigma). Samples were visualized under a thin slab of 1% agarose using a Nikon Eclipse E600 Fluorescence microscope (Nikon Corporation, Japan) equipped with a Nikon DXM1200F camera (Nikon Corporation) and 0.02 s exposure time for the blue channel and 0.04 s for the green signal. Fluorescence images were quantified using the Fiji/ImageJ software. Statistical data analysis was performed using a Kruskal–Wallis H test with a Dunn's test post hoc test with Bonferroni‐adjusted p‐values and n = 215 cells.

AUTHOR CONTRIBUTIONS

Jonathan A. Zúñiga‐Domínguez: Data curation; methodology; formal analysis. Rohit Jain: Data curation; formal analysis; methodology; investigation; writing – review and editing. Martín González‐Andrade: Methodology; investigation; writing – review and editing. Erik R. Farquhar: Methodology; writing – review and editing. Mark R. Chance: Supervision; writing – review and editing. Abril Gijsbers: Conceptualization; data curation; formal analysis; funding acquisition; investigation; writing – original draft; project administration; supervision; visualization. Nuria Sánchez‐Puig: Conceptualization; investigation; funding acquisition; writing – original draft; visualization; formal analysis; project administration; data curation; supervision.

FUNDING INFORMATION

XFP experiments were carried out at beamline 17‐BM at the National Synchrotron Light Source II (NSLS‐II), Brookhaven National Laboratory (USA) under proposal 306055. The NSLS‐II is a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Brookhaven National Laboratory under Contract No. DE‐SC0012704. The SAXS experiments were carried out in the B21 beamline at Diamond Light Source under proposal MX36130‐3. N.S.P. acknowledges funding from UNAM‐DGAPA‐PAPIIT IG200125 and IN215122, while A.G. received funding from UNAM‐DGAPA‐PAPIIT IA200725.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Data S1. Supporting Information.

Figure S1. Fluorescence decay of Alexa Fluor 488 dye as a proxy for hydroxyl radical dose during radiolysis. Quenching of Alexa Fluor 488 fluorescence due to hydroxyl radical‐mediated modification was monitored as a function of X‐ray exposure. Similar decay rates across experimental conditions indicate comparable hydroxyl radical concentrations were delivered.

Figure S2. Sequence features of the Efl1 GTPase. Modified peptides are indicated in bold within the primary sequence of Efl1. Structural domains (I–V) are color‐coded. G‐motifs (G1–G5) are outlined with light blue boxes.

Figure S3. Structural architecture of Efl1 in solution revealed by integration of SAXS and hydroxyl radical Footprinting (XFP) data. (a) Cα‐root mean square fluctuation (RMSF) across the 26 AlphaFold‐generated models of Efl1. The largest fluctuations are observed within the domain II insertion unique to the yeast protein (residues 425–579), suggesting intrinsic flexibility. (b) Goodness of fit for all modeled Efl1 conformations based on hydroxyl radical footprinting (φ ²) and small‐angle X‐ray scattering (χ ²) scores. Lower values correspond to better agreement with experimental data. Red dots represent the six best‐fit ensemble models. AFflecto conformations (light gray) were generated using one AlphaFold model as the starting structure. Dark‐gray dots represent the AlphaFold models. Representative models (1‐3) are displayed to highlight differences. (c) Goodness of fit between experimental SAXS data (gray) and the theoretical scattering profiles from the best‐fit model ensemble (red). The standard deviation of the average SAXS data for the ensemble was below 1% and is omitted for clarity.

Figure S4. Conformational sampling of Efl1 across three independent molecular dynamics simulations. (a) Time evolution of the radius of gyration, including the hydration shell contribution, for the three independent 500‐ns simulations. (b) Time dependence of the SAXS goodness of fit (χ ²) calculated for each frame along the three trajectories. (c) Conformational landscape of each repliacte plotted as φ ² (hydroxyl radical footprinting score) versus χ ² (SAXS score). Each point corresponds to a trajectory frame and is colored according to simultaion times (frame number). Arrows in Replicate 1 highlight representative low‐, intermediate‐, and high‐φ ² states selected to illustrate domain rearragements sampled during the simulations (rigth panel).

Figure S5. Comparison of hydroxyl radical footprinting analysis between wild‐type Efl1 bound to GTP and the R1086Q mutant in the presence and absence of GTP. (a) Changes in solvent accessibility of wild‐type Efl1 bound to GTP compared to the R1086Q mutant. (b) Changes in solvent‐accessibility of wild‐type Efl1 bound to GTP compared to the R1086Q mutant also bound to GTP. Left panels: histogram showing the distribution of reactivity ratios across all detected peptides. Right panels: Ratios of hydroxyl radical reactivity mapped onto a predicted model of Efl1. The color gradient indicates changes in solvent accessibility, with red representing increased protection and blue representing increased exposure, as shown in the color scale bar. A ratio of 0.02 corresponds to a 50‐fold increase in exposure (dark blue), while a ratio of 5 indicates a 5‐fold increase in protection (dark red). Transparent regions indicate undetected peptides. Roman numerals relate to the protein domains.

Figure S6. Growth control of efl1Δ yeast cells. Cells complemented with centromeric plasmids encoding EFL1 wild type or variants grown in rich YEPD medium. Color code: WT‐black, R1086Q‐red, R1086Q/T657M‐yellow, R1086Q/P151L‐blue.

Data S2. Supporting Information.

Zúñiga‐Domínguez JA, Jain R, González‐Andrade M, Farquhar ER, Chance MR, Gijsbers A, et al. Hydroxyl radical footprinting modification reveals an intradomain communication pathway in EFL1 disrupted by a Shwachman‐Diamond syndrome‐associated mutation. Protein Science. 2026;35(4):e70504. 10.1002/pro.70504

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.