The authors compared the wild-type apo lectin domain of FimH (PDB 4XO8) with three single-point substitutions: R60P, Q133K and Y137A. They reported three independent molecular-dynamics (MD) replicates of 500 ns for each system; the manuscript treats each replicate as the unit of analysis (n = 3). The work combined dynamics with three folding-stability prediction methods: FoldX, a Cartesian-ddG-type Rosetta protocol and ERIS-2. The preprint states these tools were used to generate comparative ΔΔG-like stability estimates, but the source did not provide additional experimental folding measurements to validate the predictions.
The simulation series focused on the isolated apo lectin domain rather than the full adhesin or its ligand-bound states. The authors reported aggregate simulation time and emphasized replication as a control for stochastic MD variability. No global unfolding events were described for any system across the reported trajectories.
Prediction outputs did not all agree. FoldX and ERIS-2 produced a consistent ranking: R60P emerged as the most destabilising substitution, while Q133K was ranked near-neutral by those two tools. The Rosetta Cartesian-ddG-style protocol yielded an outlying, anomalously high value for Q133K relative to the other predictors, as the authors noted. The preprint does not report experimental thermodynamic ΔΔG values; the comparison is therefore confined to in silico predictors and the MD observations.
This mixed result underlines that different algorithms can give divergent quantitative estimates for single-residue effects on folding stability. The authors used the predictors together with dynamics to build a consistent ordinal ranking of the three variants, rather than to assert precise thermodynamic numbers.
Among the three substitutions, only R60P produced a locally reproducible structural perturbation in the simulated apo lectin domain. The study team observed loss of beta-strand structure at residues 60–61 across replicates, accompanied by the loss of three local hydrogen bonds. These changes were reproduced between independent MD runs, which the authors present as evidence of a genuine local destabilisation induced by the proline substitution.
All systems remained folded overall; no replicate showed wholesale domain unfolding. The authors therefore interpret the R60P outcome as a clear case where a substitution disrupts secondary structure locally (here, breaking a beta strand), while preserving global fold integrity in the isolated lectin domain trajectories they ran.
The MD results for Q133K and Y137A revealed variant-specific rearrangements of polar contacts rather than the strand-breaking observed for R60P. For Q133K, the authors reported that an existing polar contact between Gln133 and Asp54 became a Lys133–Asp54 salt bridge during dynamics. That change in side-chain interaction was documented in the trajectories and is presented as the first dynamic characterisation of Q133K in the lectin domain, according to the preprint.
For Y137A, the authors specifically report that the alanine substitution did not cause Thr158 to enter the binding pocket in the apo-lectin simulations they ran. The preprint does not provide ligand-bound simulations or direct affinity assays. Therefore, the absence of Thr158 movement is an observation limited to the isolated, apo-domain MD conditions described.
The authors applied principal-component analysis to the MD ensembles to define a conformational landscape and to compare variant versus wild-type sampling. Departures from the WT region of the principal-component space were not uniform across all replicates for a given variant. Instead, the reported excursions were driven by individual replicates: the preprint cites Y137A replicate 1 and Q133K replicates 2–3 as showing displacement from the WT basin. The authors emphasised that these were replicate-specific events rather than a consistent variant-wide shift across all three MD runs.
This pattern led the study team to treat replicate-level variability as a meaningful component of interpretation: some conformational changes arise stochastically in individual runs and should not be automatically generalised across every replicate or taken as systematic for the entire variant without replication.
The preprint states limitations explicitly in scope. The simulations involve the isolated apo lectin domain only; no ligand-bound dynamics or experimental binding/affinity data are provided. The authors caution that their results are intended to characterise apo-domain structural and dynamic behaviour and to provide a reproducible basis for interpreting the consequences of single-residue substitutions in FimH, not to infer changes in adhesive affinity or in vivo function.
Prediction tools produced differing numerical outputs; FoldX and ERIS-2 agreed on an ordinal ranking but Rosetta reported an anomalous value for Q133K. The source did not report experimental folding free energies or biophysical validation experiments, so the stability-predictor results remain in silico comparisons paired with MD observations. The study is a preprint and has not undergone peer review. The authors made their code and supplementary material available via the repository link provided in the manuscript footnotes.
Open questions that remain from the reported work include how these local dynamic effects would translate in ligand-bound states, whether the observed salt-bridge formation in Q133K affects catch-bond behaviour under force, and how predictions and MD would compare with experimental thermodynamic or kinetic folding measurements. These specific follow-up items are not addressed in the source text.