Conformational Remodeling Underlies Activity Loss in Disease-Linked Asparagine Synthetase Variant

Article activity feed

1. 

   PREreview

   May 11, 2026

   This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/20126934.

   This preprint was reviewed as part of course STB 612 in the Department of Structural Biology at University at Buffalo lead by Dr. Alex Vecchio and students on the PhD track.

   Overview

   This study investigates the structural and molecular basis of asparagine synthetase deficiency (ASNSD) caused by the R48Q missense variant in human asparagine synthetase (ASNS). ASNS is a bifunctional enzyme composed of an N-terminal glutaminase domain and a C-terminal synthetase domain connected by a ~20 Å ammonia tunnel. The authors ask: how does a single amino acid substitution at position 48 lead to near-complete loss of enzyme activity? They address this question using an integrated approach combining …

   This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/20126934.

   This preprint was reviewed as part of course STB 612 in the Department of Structural Biology at University at Buffalo lead by Dr. Alex Vecchio and students on the PhD track.

   Overview

   This study investigates the structural and molecular basis of asparagine synthetase deficiency (ASNSD) caused by the R48Q missense variant in human asparagine synthetase (ASNS). ASNS is a bifunctional enzyme composed of an N-terminal glutaminase domain and a C-terminal synthetase domain connected by a ~20 Å ammonia tunnel. The authors ask: how does a single amino acid substitution at position 48 lead to near-complete loss of enzyme activity? They address this question using an integrated approach combining steady-state enzyme kinetics, single-particle cryo-EM and 3D variability analysis, and molecular dynamics simulations. The major findings are: (1) glutamine-dependent kcat is reduced ~50-fold in R48Q while KM values remain largely comparable to WT; (2) cryo-EM structures reveal that Loop 1 is locked in an open conformation in apo-R48Q, contrasting with the flexible open/closed sampling observed in apo-WT; (3) MD simulations suggest the R48Q substitution alters correlated motions between N- and C-terminal domains. The authors propose a model in which Arg48 is essential for maintaining Loop 1 in a catalytically competent conformation, and its loss disrupts both local active-site geometry and long-range interdomain communication. They extend this finding to the broader Class II glutamine amidotransferase family, showing conservation of both Arg48 and Loop 1 residues across multiple enzymes.

   Most importantly, this study challenges our current understanding (altered thermostability) usually attributed to how individual missense mutations impair enzyme activity. The authors posit that their study moves beyond that default explanation by providing an atomic-level account of how R48Q missense mutation disrupts ASNS function through conformational remodeling rather than destabilization, a distinction with implications they claim we can use as a generalizable approach for understanding disease missense mutations in other human multidomain enzymes. In all, this study is important and the integrative approaches to address a single mechanistic question is commendable. The manuscript is well written, findings fairly organized, and some data appears well collected to high quality. However, the authors did not provide history background to assess the novelty of the study and some findings are not well presented or communicated. In our review, we find that the manuscript has major and minor weaknesses that temper the strength of its conclusions. We, therefore, raise several issues that we believe should be addressed and clarified to make it suitable for publication.

   Major Comments

   Kinetics

   • Cysteine redox modifications in enzymes can directly alter catalytic activity and conformational landscapes, and their partial oxidation can also create heterogeneous populations with depressed rates. Can the authors clarify whether the active-site fraction was measured for enzyme preparations used across kinetics, cryo-EM and MD, and acknowledge that kinetic parameters are reported per total enzyme rather than per active site?

   • Kinetics for the wt protein are imported from an older paper, this does not seem methodologically or experimentally robust. We understand the rationale, but for accurate comparative enzymology, measuring both mutant and wt kinetics side-by-side would better support the study's conclusions.

   • The manuscript reports PPi:Asn ratios that deviate from 1:1 to 15:1 under ammonia-dependent conditions (p. 5) yet conclude that the ammonia tunnel is "likely intact." If the tunnel is intact, what explains the 15-fold excess of PPi? The authors should discuss the chemical basis for this uncoupled PPi production, specifically, whether the 𝛽-aspartyl-AMP intermediate is being hydrolyzed by water rather than attacked by ammonia, and why a "rigidified" C-terminal domain (per MD results) would permit this solvent access.

   • The manuscript claims appKM values for ammonia, ATP, and L-aspartate "remain unchanged" under ammonia-dependent PPi production conditions (p.4-5). However, Table 1 shows appKM for L-aspartate is 0.65 ± 0.02 nM (R48Q) vs. 1.3 ± 0.04 mM (WT), a 2-fold difference well outside experimental error. Elsewhere in the manuscript, the authors treat 2-fold, 3-fold, 5-fold, etc changes as noteworthy (e.g., "the appKM for ammonia decreases by 2-fold, "p.5). The manuscript could be improved by more quantitative reasoning, either 2–5-fold changes are always significant or they never are.

   • Under NH3-dependent conditions, the appKM for the same substrates differs up to 6.4-fold depending on whether PPi or Asn is monitored (ATP: 0.109 vs. 0.7 mM; Asp: 0.65 vs. 2.1 mM, Table 1). For a single-pathway enzyme, shouldn't appKM be independent of the product measured? Maybe this discrepancy, combined with the substrate-concentration-dependent PPi:ASn ratio (3:1 to 15:1), indicates a branched mechanism where PPi production reports on both productive and futile pathways. Likely fitting these data to simple Michaelis-Menten equations yields composite parameters whose mechanistic interpretation is ambiguous. Please clarify.

   Cryo-EM & Stability

   • A significant and unacknowledged result in this paper is that the Cryo-EM 2D classes show both monomer and dimer populations.

   • How do the authors explain their Tm and kinetics data knowing that in solution mixed populations of enzymes are present?

   • How can allostery be assessed with mixed oligomeric species in equilibrium?

   • Moreover, the authors should justify why they chose to discard the monomer populations from their Cryo-EM dataset. One class of monomers appears to be resolvable to high resolution and may inform on the probed loop structure. We suggest processing this class fully to shed light on this.

   • We appreciate the authors testing of the stability of the enzyme. However, the authors use CD at 222 nm to conclude that R48Q is thermally stable (Tm shift of 2.5 °C, Fig. S1), several concerns limit this interpretation:

   • 222nm primarily reports on alpha-helical content, yet the R48Q mutation resides in the beta-sheet-rich N-terminal glutaminase domain. Local destabilization of beta structure would be invisible at this wavelength;

   • No unfolding control (e.g., chemical denaturation by urea or GdnHCl) is shown to confirm that 80°C achieves complete unfolding, partially unfolded protein retaining local secondary structure elements would appear stable by CD;

   • ASNS can be a dimer and monomer in solution (as confirmed by the cryo-EM images), so the measured Tm reflects dimer dissociation convolved with monomer unfolding;

   • DSF or DSC would provide a more sensitive and secondary-structure-independent measure of global and local stability. The CD spectra in Fig. S1b-c are also not well resolved. The authors should acknowledge these limitations rather than concluding thermal stability is unaffected. As the structures verify foldedness, these results could also be placed in the supplement as their impact is minor.

   • The paper proposes that Gln48 disrupts Loop 1 dynamics, impairing substrate entry and intermediate stabilization (p.7). This is the central mechanistic claim of the paper, yet it is based on a single variant (R48Q). Is the effect due to loss of positive charge, loss of hydrogen-bonding capacity, or altered side-chain size? Additional substitutions (e.g., R48K to preserve charge, R48A to remove the side chain) or Loop 1 truncations would directly test these hypotheses. Without such experiments, the proposed mechanism remains correlative. The authors should acknowledge these as testable predictions or provide additional evidence.

   • The study attributes subtle C-terminal structural differences between WT and R48Q, "Asp261 rotamer changes, Ser257 PP-loop variability, and Trp540/Ile541 conformational locking (p. 7)", to propagation of the N-terminal R48Q substitution. However, these are minor side chain rotamer differences that could arise from factors unrelated to the mutation, such as differences in protein concentration, grid preparation, particle orientation, dimerization, or position in ice. Since the paper's central allosteric model (Loop 1 disruption leads to C-terminal rigidification, and subsequently reduced synthetase activity) depends on these observations being mutation-driven, the authors should provide evidence that these C-terminal changes are specifically caused by R48Q rather than experimental variability.

   • Local resolution estimates of the wt and mutant proteins show that the mutant is more dynamic, but the authors conclude wild type is based on kinetics and structural results. How do the authors reconcile these two discrepancies?

   Figure Design & Interpretability

   • Many of the figures displaying structural changes are currently difficult to interpret in the way they are displayed.

   • The results displayed in figure 2 (a - e) would benefit from super positioning, which would increase clarity and avoid concerns about orientation differences across the panels.

   • In Fig. 3a, the RMSF plot highlights selected regions with blue shading but does not label what these regions correspond to, a reader must cross-reference with Fig. S19a/b to understand what is being highlighted. Consider labeling the shaded regions directly.

   • Fig. S11 (Loop 1 cryo-EM density with model fit for both WT and R48Q) is arguably more important than several main-text panels, as it is the direct experimental evidence for Loop 1 disorder.

   • Fig. S19a (zoomed RMSF at Loop 1, synthetase domain, and C-terminal tail) provides the detail that Fig. 3a lacks. Consider promoting these.

   MD

   • We appreciate the authors employing MD to further understand the dynamics at play in their system. We notice that the MD simulations were initiated from a ROBETTA homology model (p.8), and R48Q was generated by in-silico mutation, meaning both simulations start from the identical loop 1 backbone conformation. The authors have their own cryo-EM structures at 2.78 Å (WT) and 2.81 Å (R48Q).

   • Why were these not used as starting structures?

   • What conformation is Loop 1 in the ROBETTA model (open, closed, or something else)?

   • Does it match the experimental structures? No comparison is shown. Can the authors justify this choice and demonstrate whether the ROBETTA model agrees with their experimental structures in the regions of interest.

   • The authors claim R48Q causes specific "global motion reprogramming" based on DCCM analysis (p.9). To establish that this preprogramming is specific to the R48Q substitution and not a generic response to any destabilizing mutation, the study should show that mutations at other sites do not produce similar DCCM changes, or that alternative substitutions at the same position produce different global dynamics. Without such controls, can the authors clarify how they distinguish R48Q-specific allosteric effects from a nonspecific perturbation response?

   • In Fig. 3b, representative structures are extracted at specific time points from what appears to be a single replicate (95 and 151 ns for WT; 22 and 62 ns for R48Q).

   • No justification is given for why these time points or replicates were chosen, are these cluster centers, most populated conformations, or visually convenient snapshots

   • Additionally, the R48Q RMSD trace is visibly noisier than WT, which contradicts the claim that R48Q Loop 1 is "locked" in an open conformation.

   • No MD convergence metrics are reported. Without convergence analysis, it is impossible to know whether the observed differences between WT and R48Q reflect genuine equilibrium properties or sampling artifacts. Can the authors show convergence metrics and clarify which portion of the simulations was used for analysis? Figs. S19c and S20 show substantial inter-replicate divergence, suggesting the simulation have not converged. If so, all MD-derived analyses (RMSF, RMSD, distance trajectories, and DCCM) cannot be reliably interpreted. Averaging DCCMS across non-converged replicates mixes non-equivalent conformational states, making the ∆DCCM analysis (Fig. 3c) unreliable.

   Minor Comments

   Methods

   • The 3DVA methods state analyses used, "5 principle components" (SI p. 4).

   • First, "principle" should be "principal." CryoSPARC uses modes, are the authors referring to modes as principal components, or does this reflect additional processing or something different? Can the authors clarify?

   • Additionally, no 3DVA input parameters beyond the number are reported, what were the input parameters used? Please state explicitly. Each mode produces a different eigenvector depicting distinct motions from the same data. It is unclear which mode(s) are shown in Fig. 2f, and if only one mode is presented, the authors should justify why that mode was selected and whether other modes show consistent or contradictory results.

   • The 3DVA-derived structures were refined using Phenix.varref (SI p. 4), but no validation statistics are reported for these models. What validation metrics were output from the Phenix.varref job? Table S2 covers only the consensus refinements. Without model-to-map fit statistics for the variability-refined structures, it is difficult to assess whether the conformational differences shown in Fig. 2f are robust or refinement artifacts.

   • All R48Q kinetic measurements are reported as means of "independent duplicates" (n = 2). With n = 2, standard error is derived from a single difference value, providing no meaningful estimate of variability. The authors should clarify whether these are technical or biological replicates and consider repeating with n ≥ 3 to enable robust statistical comparison with WT if desired.

   • The cryo-EM maps (EMD-75273, EMD-75275) are listed as deposited but are currently on hold and not publicly available. Without access to the maps, it is not possible to independently evaluate the quality of the reconstructions or the modeling of individual residues. This is particularly important given that Fig. S11, which shows the key structural comparison between WT and R48Q, is placed deep in the supplementary material. Can authors provide the maps or release the maps prior to publication?

   • The SI method state PROPKA was used to assign protonation state at pH 7.4, which would predict Cys1 to be predominantly protonated. The main text argues Cys1 must be deprotonated for catalysis. Maybe the authors should address this discrepancy by saying that the simulation models ground-state enzyme dynamics rather than the reactive state and acknowledging that the thiolate form of Cys1 might exhibit different dynamics.

   Introduction & Results

   • The authors state "...the substitution primarily affects the concentration of catalytically competent enzymes during glutamine hydrolysis." (p.5). The authors should clarify what specific evidence, beyond unchanged KM values, supports the "reduced catalytically competent population" model.

   • The introduction states Arg48 is "essential" for maintaining Loop 1 in a competent conformation and that conformational perturbation "propagates" to the C-terminal domain.

   • The first claim is based on a single variant; therefore, additional substitutions might be needed to establish that arginine specifically is essential.

   • The second claim implies directionality, but the cryo-EM data show correlation, not causation, the C-terminal changes could arise independently during folding rather than propagating from Loop 1. The authors should provide evidence for these claims or temper these claims.

   • The Introduction claims this work "establishes a generalizable framework for understanding pathogenic missense mutations in other multidomain human enzymes." This overstates the scope of a single-variant study in one enzyme. Each multidomain enzyme has distinct regulatory mechanisms, catalytic chemistry, and domain-domain communication. The authors should reframe this as demonstrating the value of integrated experimental characterization of individual missense variants, rather than claiming a generalizable framework.

   • The significant figures in table 1 (kinetic results) should be consistent across the table. The authors should further acknowledge if they are truly able to get sub-one thousandth accuracy.

   • The authors observe that the WT C-terminal tail (Trp540/Ile541) samples two conformations while R48Q is locked in one (p. 7, Fig. 2e) and conclude this "supports a model of coordinated structural remodeling." The relevance of C-terminal tail dynamics to N-terminal glutaminase activity is not established. Conformational heterogeneity in a distal region ~50 Å from the mutation site does not, by itself, demonstrate coordination with Loop 1.

   • The authors highlight reduced correlated motion between Loop 1 and C-terminal residues 352-360 and 409-422, noting these regions "are in close contact with one another" (p.9). If these regions are in direct contact with the loop, altered correlated motion upon loop disruption is expected, not informative. What would be more meaningful is identifying changes in regions not in direct contact with loop 1 like residues 236-246, which are not emphasized. Can the authors clarify mechanistic insight the loss of correlated motion between contacting regions provides beyond the obvious?

   • Fig. 3a and S19a show RMSF profiles for WT and R48Q with no convincing difference at Loop 1. The figure legends do not specify whether RMSF is averaged across replicates or from a single replicate.

   • In Fig. S19a, the shaded regions around the traces are not defined, SD? SEM? Range? The authors highlight residues 49-52, 375-385, and 435-455, but other regions with apparent differences are not discussed.

   • Fig. S19c labels the y-axis as "Backbone RMSF" but the legend describes it as "Backbone RMSD", which is correct?

   • Referenced wrong figure, should be Fig 3c&d. See text: "Additionally, the R48Q variant exhibited enhanced coupling across several regions between the N-terminal and C-terminal domains relative to WT, including the pairs 32-38 and 419-424, 134-166 and 320-343, and 91-102 and 373-391. Interestingly, 91-102 includes the key residue Lys94, which is in direct contact with Loop 1 (Fig. 2c&d)."

   • Fig. (3c) label the colormap

   • The legend for SI Fig. S4b reads "…the R48Q variant shows elevated uncoupled glutaminase activity, indicating impaired interdomain communication and loss of catalytic coordination." However, the fig. shows a bar graph of % specific activity of asparagine production, with WT at ~100% and R48Q at ~2%, showing asparagine production is lower in R48Q not that glutaminase activity is elevated. On p.5 states "L-glutamate production in the R48Q variant is significantly lower than in WT", which also directly contradicts the legend's claim of elevated glutaminase activity. It is unclear what is being communicated, maybe a typo.

   • The W540A kinetics (Fig. S21) and the sequence alignment of the GATase (Fig. S22) family members are presented in the discussion section but constitute original experimental and computational results. Why are these not in the Results section?

   Discussion

   • The structural analyses conducted in this paper capture the apo, ligand-free state of ASNS. No substrate-bound or transition-state analog-bound structures of R48Q are presented. This limits conclusions of how the mutation truly affects catalytically relevant conformations, which should be addressed as a limitation in the discussion section.

   • The discussion states the reduced activity "can be directly attributed to severe dysfunction of Loop 1 at the N-terminal active site" (p. 9). Loop 1 is poorly resolved in apo-WT, and its precise function is not established, "severe dysfunction" implies a known functional role that has been disrupted. The data more accurately show that perturbation of Arg48 causes activity loss through multiple factors. The authors should tone down this language.

   • The discussion attributes the allosteric effects to "conformational reprogramming" through dynamics and bound water (p. 10). However, the structural data more directly shows altered residue contacts, hydrogen bonds, salt bridges, hydrophobic packing, between Loop 1 and other regions of the protein. Dynamics may contribute, but the loss of specific contacts is the more parsimonious explanation. The authors could frame the mechanism in terms of disrupted contacts rather than leaning heavily on dynamics as there is evidence here of the former while the latter is lacking.

   Competing interests

   The authors declare that they have no competing interests.

   Use of Artificial Intelligence (AI)

   The authors declare that they did not use generative AI to come up with new ideas for their review.

   Read the original source
2. 

   Version published to 10.64898/2026.02.01.703106 on bioRxiv

   Feb 3, 2026