Antimicrobial peptides are promising agents for combating resistant infections. They exhibit bactericidal activity against a wide range of microbes, primarily by disrupting the permeability of the bacterial membrane and ultimately causing cell death. Effective bacterial killing requires a high number of membrane-bound peptide molecules. Therefore, it is conceivable that peptide accumulation on the membrane could also interfere with essential cellular processes by altering bilayer dynamics, a hypothesis referred to as the “sand in the gearbox” model. In this work, we systematically investigated how membrane dynamics is affected by a set of well-characterized yet highly diverse peptides: the natural AMP magainin 2, the toxin melittin, the synthetic peptides LAH4 and Killer-FLIP, and small membrane-active peptidomimetics with bactericidal activity. These effects were examined using fluorescence spectroscopy techniques, by measuring anisotropy, generalized polarization, and excimer formation of specific probes inserted at different depths within the lipid bilayer. Our results show that the activity of all compounds extends beyond membrane permeabilization, and that perturbation of membrane dynamics is a common feature among all systems analyzed. The membrane-active compounds induced a stiffening of the phospholipid bilayer by reducing lipid lateral mobility and decreasing water penetration, at least on the nanosecond timescale accessible to fluorescence measurements. Interestingly, when accounting for differences in the resulting membrane surface area coverage, the concentration range in which this behavior occurred was the same for all compounds studied. This threshold is, generally, higher than that required for membrane permeabilization and reflects near-complete coverage of the bilayer surface.

Background/Objectives: Antimicrobial peptides are promising agents for combating resistant infections. They exhibit bactericidal activity against a wide range of microbes, primarily by disrupting the permeability of the bacterial membrane and ultimately causing cell death. Effective bacterial killing requires a high number of membrane-bound peptide molecules. Therefore, it is conceivable that peptide accumulation on the membrane could also interfere with essential cellular processes by altering bilayer dynamics, a hypothesis referred to as the “sand-in-a-gearbox” model. Methods: We systematically investigated how membrane dynamics is affected by a set of well-characterized yet highly diverse peptides: the natural AMP magainin 2, the toxin melittin, the synthetic peptides LAH4 and Killer-FLIP, and small membrane-active peptidomimetics with bactericidal activity. These effects were examined using fluorescence spectroscopy techniques, by measuring anisotropy, generalized polarization, and excimer formation of specific probes inserted at different depths within the lipid bilayer. Results: The activity of all compounds extends beyond membrane permeabilization, and the perturbation of membrane dynamics, often associated with the so-called “sand-in-a-gearbox” mechanism, is a common feature among all systems analyzed. The membrane-active compounds induced a stiffening of the phospholipid bilayer by reducing lipid lateral mobility and decreasing water penetration, at least on the nanosecond timescale accessible to fluorescence measurements. Conclusions: The concentration range in which this behavior occurred was the same for all compounds studied. This threshold is, generally, higher than that required for membrane permeabilization and reflects near-complete coverage of the bilayer surface.

The prediction of RNA secondary structures is essential for understanding its underlying principles and applications in diverse fields, including molecular diagnostics and RNA-based therapeutic strategies. However, the complexity of the search space presents a challenge. This work proposes a Graph Convolutional Network (GCNfold) for predicting the RNA secondary structure. GCNfold considers an RNA sequence as graph-structured data and predicts posterior base-pairing probabilities given the prior base-pairing probabilities, calculated using McCaskill’s partition function. The performance of GCNfold surpasses that of the state-of-the-art folding algorithms, as we have incorpo-rated minimum free energy information into the richly parameterized network, enhancing its robustness in predicting non-homologous RNA secondary structures. A Symmetric Argmax Post-processing algorithm ensures that GCNfold formulates valid structures. To validate our algo-rithm, we applied it to the SARS-CoV-2 E gene and determined the secondary structure of the E-gene across the Betacoronavirus subgenera.