Selective transport of plasma-derived reactive species through the plant aquaporin channels: a molecular dynamics study Authors: Davronjon Abduvokhidov, Parthiban Marimuthu, Akbar Kodirov, Mukhammadali Niyozaliev, Chen Zhou, Dingxin Liu, Jamoliddin Razzokov Published: 2026 ABSTRACT AbstractThe selective permeability of reactive oxygen and nitrogen species (RONS), generated by cold atmospheric plasma (CAP), through plant aquaporins was investigated to identify plasma-derived species capable of intracellular delivery. Using atomistic molecular dynamics and enhanced sampling methods, we quantified the free energy profiles of eight RONS (HNO3, HO2, cis-HNO2, trans-HNO2, N2O4, NO, NO2, and O3) across the PIP2;1 aquaporin channel embedded in a lipid bilayer. Permeation was found to be controlled by two complementary features: the free-energy barrier opposing entry at the aromatic/arginine (ar/R) constriction, and the depth of the free-energy minimum within the pore interior. The strongly polar, hydrogen-bonding species HNO3 exhibited the highest entry barrier (≈ 16 kJ·mol− 1) and is thus the least permeable, whereas weakly polar species such as NO crossed the channel with minimal resistance. Notably, several species, most strikingly N2O4, combined facile entry with deep internal minima, indicating that they are readily admitted into the pore but strongly retained within it rather than freely transported. These results show that RONS selectivity in PIP2;1 is governed by molecular size, polarity, and hydrogen-bonding capacity, and provide a mechanistic basis for the rational selection of CAP-generated species in sustainable plasma-based agricultural technologies.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-61939-z.Keywords: Cold atmospheric plasma, Plant aquaporin, AQP channel, Reactive oxygen species, Molecular dynamics, Umbrella sampling, Free energy barrier, Selective permeationSubject terms: Biophysics, Plant sciences Introduction Recent advances in sustainable agriculture have pointed to cold atmospheric plasma (CAP) technology as a promising eco-friendly technology to enhance seed germination, stimulate plant growth, and improve crop yield1–3. CAP produces a mixture of short-lived and long-lived reactive oxygen and nitrogen species (RONS), including H2O2, HO2,·OH, NO, NO2, O3, OH, N2O4, HNO3, s- cis -HONO, and s- trans-HONO which are believed to influence plant physiology when applied externally to seeds or tissues1,4. Experimental studies have demonstrated that plasma treatment can accelerate germination rate, early plant development, shorten the ripening period, and reduce irrigation demands - yielding significant agronomic and environmental benefits5–8. However, the fundamental molecular mechanisms underlying these effects remain unclear. One hypothesized pathway involves the interaction of plasma-generated RONS with cellular transport systems. In this context, aquaporins (AQPs) have emerged as potential mediators of RONS entry into plant cells, linking plasma exposure to downstream physiological effects. AQPs are integral membrane proteins that facilitate the selective and passive transport of water and small solutes across cellular membranes in all forms of life, including plants9–11. In higher plants, the plasma membrane intrinsic protein (PIP) subfamily, particularly isoforms like PIP2;1, plays a crucial role in regulating water homeostasis, nutrient uptake and adaptation to environmental stresses12,13. Each AQP monomer forms a distinct water-conducting channel composed of six transmembrane helices and conserved structural elements such as the Asn-Pro-Ala (NPA) motifs and the aromatic/arginine (ar/R) selectivity filter14,15. In plants, additional regulation of AQP activity occurs via phosphorylation, pH gating and oxidation-sensitive features such as disulfide bonds between extracellular cysteine residues16,17. While AQPs are well established as water channels, several isoforms have also been shown to facilitate the transport of small neutral solutes and, in some cases, hydrogen peroxide18,19. Nonetheless, comprehensive computational studies on the ability of plant AQPs to conduct plasma-generated RONS are still lacking. Earlier computational work by R Cordeiro20 used molecular dynamics (MD) simulations to investigate the permeability of reactive oxygen species such as H2O, H2O2, HO2 and·OH through both mammalian and plant AQPs, revealing that small radicals may transiently access the pore. Yusupov et al.21 extended this line of inquiry by modeling RONS interactions with AQP1 channels and lipid bilayers containing plasma-induced oxidative modifications, highlighting that membrane composition and oxidation state strongly influence permeation. However, these studies primarily focused mainly on mammalian AQP1 which is one of the important transmembrane protein in biomedical contexts and did not specifically address plant growth-related RONS transport. In this work, we perform atomistic MD simulations22,23 to explore the permeability of plant aquaporin PIP2;1 to selected RONS (e.g., H2O2, HO2, NO, O3, OH, NO2, N2O4, HNO3, s- cis-HONO, and s-trans-HONO that are commonly generated by CAP. Our objective is to quantify the free energy barriers and structural interactions that govern RONS entry through AQPs, with the broader goal of identifying species most likely to penetrate plant cells and mediate growth-promoting effects. This mechanistic study represents a fundamental step within a larger agricultural project aimed at optimizing CAP-based technologies for enhanced crop performance. Understanding the molecular basis of RONS permeation through AQPs will inform the design of plasma devices optimized for enhanced nutrient uptake and stress tolerance in crops. Methods We conducted atomistic MD simulations to investigate the transmembrane transport behavior of selected RONS through the plant AQP in a model cell membrane. Particularly, the investigated RONS encompassed both hydrophilic species (HNO3, HO2, cis-HNO2, trans-HNO2) and hydrophobic species (NO, NO2, N2O4, and O3), covering a diverse range of chemical reactivity, polarity, and molecular size. These RONS were selected based on their relevance in environmental and plasma-treated agricultural systems, where they are known to interact with biological membranes. In addition, the molecular system comprised a tetrameric plant aquaporin (PIP2;1) obtained from the Protein Data Bank (PDB ID: 1Z98), embedded in a 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) lipid bilayer (see Fig. 1). DOPC was chosen as a single-component model membrane because phosphatidylcholine is one of the most abundant glycerophospholipids in the plant plasma membrane24, and because its fluid-phase, well-characterized bilayer provides a reproducible matrix that allows direct comparison with previous RONS-permeation studies; since the permeation barriers of interest arise predominantly at the ar/R constriction within the channel, a single-component bilayer is appropriate for isolating the role of the pore. The bilayer was solvated with explicit water layers on both the extracellular and cytoplasmic sides, mimicking the native plant cellular environment. Moreover, chloride ions () were added to neutralize the system’s net charge. Additionally, the structural renderings of the systems were prepared using Visual Molecular Dynamics25. Structural configuration of the PIP2;1 aquaporin system: (A) side view showing the protein embedded in a DOPC lipid bilayer, (B) chemical structure of the DOPC lipid molecule, showing its amphipathic character, (C) side view of the tetrameric AQP arrangement, and (D) top view of the AQP. The protein and lipids were modeled using the GROMOS 53A6 united-atom force field26, with water represented by the SPC/E model27. Parameters for the reactive species were adopted from the GROMOS-compatible sets developed by Cordeiro et al.28–30, which were derived specifically to reproduce the experimental water-to-alkane partition coefficients of RONS – a property essential for the reliable description of their transmembrane permeation. The NO parameters were taken from the same source, parameterized by an equivalent procedure28,29,31. All MD simulations were performed using the GPU-accelerated version of GROMACS 202232–34. The system was energy-minimized using the steepest descent algorithm for 50,000 steps, with a convergence criterion of 100 kJ·mol− 1·nm− 1. Long-range electrostatics were treated using the Particle Mesh Ewald (PME) method, and a 1.0 nm cutoff was applied for both Coulomb and van der Waals interactions. The system was equilibrated in the NPT ensemble, and the temperature was maintained at 303.15 K via the Nose-Hoover thermostat, and pressure was controlled semi-isotropically at 1.01325 bar using the Parrinello-Rahman barostat. Separate thermostat coupling groups were defined for the protein-lipid complex and the water-ion phase. Bonds were constrained using the LINCS algorithm, allowing a 2 fs integration time step. Following equilibration, a 1 µs production MD simulation was conducted using a 2 fs timestep (total of 500 million steps). All thermostat and barostat parameters remained unchanged. The structural stability of the system was confirmed using root mean square deviation (RMSD) analysis of the protein backbone (see Fig. 2). A well-equilibrated frame from the final frame of this trajectory was selected as the starting configuration for umbrella sampling (US). Global structural stability. (A) RMSD of PIP2;1 during the 1 µs MD simulation. (B) Box dimension along the Z-axis demonstrating stable bilayer thickness. The free-energy profiles (FEPs) for RONS permeation were computed using US35,36. Each RONS was restrained along the channel (z) axis by a harmonic biasing potential (force constant 2000 kJ·mol− 1·nm− 2), while lateral diffusion was limited by a flat-bottomed restraint in the xy-plane (radius 0.5 nm, force constant 500 kJ·mol− 1·nm− 2). To obtain a fine resolution of the permeation coordinate while making efficient use of the four pores of the AQP tetramer, a staggered US protocol was employed: within each setup, eight windows were placed at 1.0 nm intervals along the z-axis in each of the four channels, and eleven independent setups were generated whose window centers were mutually offset by 0.1 nm (see Fig. 3). Superimposing these eleven staggered setups yielded a combined set of 81 umbrella windows per channel, spanning z = − 4.0 to + 4.0 nm at an effective window spacing of 0.1 nm (1 Å). Within a given pore the eight simultaneously restrained solutes were separated by at least 1.0 nm, the non-bonded cutoff distance, so that no direct non-bonded interactions occurred between them. Each window was simulated for 5 ns in the NPT ensemble; the first 1 ns was discarded as equilibration and the final 4 ns were used for analysis. In total, 8 windows × 4 pores × 11 staggered setups × 8 RONS species = 2816 umbrella sampling simulations were performed to obtain the FEPs, amounting to ≈ 14 µs of cumulative umbrella sampling. Preparation of umbrella sampling configurations: (A) side view of the equilibrated AQP-lipid system (water omitted for clarity), (B) side view after insertion of RONS along the z-axis, (C) top-down view of the tetrameric AQP system, and (D) top view showing RONS distribution within individual channels. For each RONS, the biased histograms from all four pores and all eleven staggered setups were unbiased and combined together in a single weighted histogram analysis method (WHAM) reconstruction37 as implemented in the gmx wham tool. Because the permeation coordinate spans a periodic simulation cell, the profiles were reconstructed using the cyclic (periodic) option of gmx wham, which enforces consistency of the two bulk-water regions on either side of the membrane; this yielded free-energy profiles whose bulk reference values on both sides agree to within ≤ 0.05 kJ·mol− 1, confirming convergence. Statistical uncertainties were estimated by bootstrap analysis (50 bootstrap samples of complete histograms), with the integrated autocorrelation time of each window taken into account; the reported error bars correspond to ± 1 standard deviation across the bootstrap estimates. The permeation free-energy barrier (ΔGmax) for each species was defined as the maximum of the FEP relative to the bulk-water reference. Results and discussion To ensure reliable sampling of RONS permeation through the PIP2;1 aquaporin, a 1 µs atomistic MD simulation was conducted following standard minimization and equilibration protocols (see Methods). Structural stability of the channel was monitored via the RMSD of the protein backbone atoms relative to the initial structure (see Fig. 2A). The RMSD increased during the initial relaxation phase and reached a stable plateau of approximately ~ 0.32–0.34 nm after ~ 80 ns, remaining steady for the remainder of the trajectory. This stable behavior confirms that the aquaporin maintained its native fold and achieved dynamic equilibrium prior to umbrella sampling calculations. In addition to RMSD, the stability of the membrane environment was monitored by tracking the box length along the Z-axis, which reflects the bilayer thickness (see Fig. 2B). The Z-dimension remained highly stable around ~ 10 nm over the entire 1-µs trajectory, with only small thermal fluctuations. This consistency confirms that the lipid bilayer preserved its structural integrity and that no artificial compression or expansion occurred during the production simulation. Local flexibility of the protein was assessed using residue-wise RMSF analysis of the four PIP2;1 monomers (see Fig. 4). The transmembrane helices remained highly stable, with fluctuations generally below 0.20 nm, consistent with a rigid pore architecture. In contrast, increased mobility was observed in the N-terminal region, in extracellular loops (around residues 60–75 and 140–160), and in the C-terminal tail (residues 270–280), where RMSF values reached up to 0.5–0.6 nm. These flexible regions correspond to non-helical segments typical of plant AQPs and do not affect the structural stability of the conduction pathway. Residue-wise RMSF of PIP2;1, highlighting stable helices and flexible loop/terminal regions. Figure 5 shows the free energy profiles (FEPs) for the relatively hydrophilic species - HO2, cis-HNO2, trans-HNO2, and HNO3 - calculated along the z-axis of the PIP2;1 channel. All four species entered the extracellular vestibule of the channel with minimal energetic barriers but exhibited significantly different free energy barrier profile near the channel’s narrowest region, corresponding to the NPA motifs and the ar/R selectivity filter. Among these, HNO3 exhibited by far the highest permeation barrier, with ΔG_max ≈ 16 kJ·mol− 1 located at the selective region. This pronounced barrier is consistent with its strong polarity and hydrogen-bonding capacity, which favour retention in the aqueous phase and produce unfavourable interactions when the molecule is forced through the constriction. FEP of hydrophilic RONS - HO2, cis-HNO2, trans-HNO2, and HNO3 - along the z-axis of the PIP2;1 channel obtained from US simulations. In contrast, HO2 and cis-HNO2 encountered essentially no ascending barrier relative to bulk water (ΔG_max ≈ 0.1 and 1.0 kJ·mol− 1, respectively); instead, their profiles descended into deep free-energy minima of approximately − 11 and − 10 kJ·mol− 1 within the pore interior. These deep wells indicate that both species are thermodynamically stabilized inside the channel through favourable interactions with the polar pore-lining residues, rather than being excluded by it. trans-HNO₂ displayed an intermediate profile, with a modest barrier of ΔG_max ≈ 4.6 kJ·mol− 1 superimposed on a shallow internal minimum, reflecting a more even partitioning between the aqueous and pore environments. The presence of multiple local minima along the permeation axis for HO2 further points to transient binding at successive polar sites lining the pore, in line with earlier reports that small radicals can be accommodated within narrow AQP pores20,21. The permeation FEPs for more hydrophobic or weakly polar species, NO, NO2, N2O4, and O3, are presented in Fig. 6. Notably, NO exhibited a remarkably flat FEP with only asmall barrier (ΔGmax ≈ 3.5 kJ·mol− 1), suggesting near-free diffusion across the pore. This result is consistent with its small size, neutral charge, and minimal interaction with water or channel residues. O3 moderate barrier (ΔG_max ≈ 6.7 kJ·mol− 1) together with a pronounced internal minimum of about − 9.6 kJ·mol− 1, indicating that it crosses the constriction at a modest energetic cost but is markedly stabilized within the pore. NO2 displayed a somewhat higher barrier (ΔG_max ≈ 8.3 kJ·mol− 1) concentrated nearthe channel center. FEP of hydrophobic or weakly polar RONS - NO, NO2, N2O4, and O3 - across the PIP2;1 channel. N2O4 exhibited the most distinctive behaviour. Rather than facing a large ascending barrier, it encountered only a small entry barrier (ΔG_max ≈ 4 kJ·mol⁻¹) and instead fell into a very deep free-energy minimum of approximately − 17 kJ·mol⁻¹ at the centre of the pore - the deepest well observed among all studied species. This profile indicates that N2O4 is not excluded by the channel but is strongly trapped within it: while entry is facile, the depth of the central well implies that escape toward the cytoplasmic side requires surmounting a substantial free-energy difference. This behaviour is attributable to its bulky, quadrupolar geometry, which interacts favourably with the pore interior while remaining poorly hydrated. These findings highlight the dependence of RONS permeation on molecular size, polarity, and flexibility. Species with smaller van der Waals radii and weaker hydrogen bonding potential permeate more efficiently. Importantly, the energetics reported here correspond to passive, non-interacting transport; reactive events such as proton transfer, oxidation, or radical reactions are not captured within this framework but may further influence the biological impact of these species. The maximum permeation barriers (ΔGmax), defined as the highest point of each FEP relative to the bulk-water reference, are compared in Fig. 7, together with the corresponding internal free-energy minima (ΔGmin). The resulting order of barriers, HNO3 > NO2 > O3 > trans-HNO2 > N2O4 ≈ NO > cis-HNO2 > HO2, shows that the most strongly polar, hydrogen-bonding species, HNO₃ and, to a lesser extent, NO2, face the greatest resistance to entry, whereas weakly polar or radical species such as HO₂, cis-HNO2, NO, and N2O4 encounter little or no ascending barrier. Importantly, a low ΔGmax does not by itself imply rapid net translocation: several of the readily entering species (most strikingly N2O4, but also HO2, O3, and cis-HNO₂) display deep internal minima, down to ≈ − 17 kJ·mol− 1 for N2O4, indicating that they are readily admitted but then strongly retained within the pore rather than freely transported through it. The channel therefore discriminates among RONS through two complementary features. Strongly polar species such as HNO3 are excluded by a barrier at the ar/R constriction, whereas weakly polar species such as N2O4 are admitted but retained within a deep internal well. Both effects are governed by molecular size, polarity, and hydrogen-bonding capacity. Maximum free-energy barriers (ΔGₘₐₓ, red) and free-energy minima within the pore (ΔGmin, blue) for plasma-derived RONS permeating through the plant aquaporin PIP2;1, obtained from US MD simulations. ΔGmax is referenced to bulk water (zero); error bars denote ± 1 standard deviation from bootstrap analysis. Species are ordered by decreasing ΔGmax. To further contextualize these results, the obtained ΔGₘₐₓ values were compared with those previously reported for RONS permeation across lipid membranes without AQP31. While the absolute barriers differ between the two systems, a similar trend is observed for polarity-dependent transport efficiency (Table 1). Notably, species such as NO and trans-HNO₂ remain the most permeable in both systems, although the energy barriers for the polar species are generally higher in the PIP2;1 aquaporin, reflecting the tighter ar/R constriction. This suggests that the AQP pore imposes additional steric and electrostatic selectivity on polar species relative to a lipid membrane. Table 1Comparison of calculated maximum free-energy barriers (ΔGₘₐₓ) for selected plasma-derived RONS across nitro-oxidized lipid membranes31 and the plant aquaporin PIP2;1 (this study). Higher barriers observed for polar species in the AQP highlight the selective nature of the ar/R filter.SpeciesΔGₘₐₓ (kJ·mol⁻¹),Membrane (2023)31ΔGₘₐₓ (kJ·mol⁻¹),Plant AQP (this work)HNO37.5516.3 ± 2.3NO20.98.3 ± 1.4O31.66.7 ± 1.3trans-HNO25.64.6 ± 0.8N2O40.14.0 ± 0.9NO0.53.5 ± 0.6cis-HNO2−0.11.0 ± 0.2HO26.40.1 ± 0.1 Comparison of calculated maximum free-energy barriers (ΔGₘₐₓ) for selected plasma-derived RONS across nitro-oxidized lipid membranes31 and the plant aquaporin PIP2;1 (this study). Higher barriers observed for polar species in the AQP highlight the selective nature of the ar/R filter. The differential permeability profiles observed among RONS provide important mechanistic insight into how CAP-generated species might interact with plant tissues at the molecular level. Hydrophobic and weakly polar species like NO and O3 can efficiently cross the cell membrane through AQP such as PIP2;1, potentially reaching intracellular compartments and modulating physiological processes such as growth or stress resistance. In contrast, the entry of highly polar or reactive species (e.g., HNO3) may be kinetically or sterically hindered, limiting their biological activity unless facilitated by alternative mechanisms (e.g., transient pore expansion or lipid-mediated diffusion). These results suggest that the composition of plasma effluents could be tuned to favor the generation of species with high membrane permeability, thereby maximizing agricultural benefits such as early ripening or stress adaptation. Conclusion This study provides a molecular-level investigation into the passive permeation of plasma-derived RONS through the plant aquaporin PIP2;1, using extensive atomistic MD and US simulations. By computing the free-energy profile of eight RONS along the channel, we found that permeation is governed by two complementary energetic features: the barrier opposing entry at the ar/R constriction, and the depth of the free-energy minimum within the pore interior. Strongly polar, hydrogen-bonding species, most notably HNO3, encounter the highest entry barriers and are therefore the least likely to permeate passively through the channel. In contrast, weakly polar species such as NO cross the pore with little energetic penalty. Several species, including N2O4, HO2, and O3, combine facile entry with deep internal minima, indicating that they are readily admitted into the pore but become thermodynamically retained within it rather than freely transported. This distinction between barrier-limited exclusion and well-limited retention, together with the dependence of both on molecular size, polarity, and hydrogen-bonding capacity, constitutes the central energetic signature of RONS selectivity in PIP2;1. These findings should be interpreted within the scope of the present model. The energetics describe passive, non-reactive transport and do not capture chemical reactions of RONS (e.g., proton transfer, oxidation, or radical recombination), nor the kinetic factors that ultimately determine net flux; a deep internal minimum, for instance, may either favour accumulation or slow throughput depending on the escape kinetics. Likewise, the use of a single-component DOPC bilayer and a single AQP isoform represents a simplified model of the plant plasma membrane. The system was neutralized with counterions but did not include an explicit physiological salt concentration; because all permeants studied here are neutral, this is expected to have only a minor effect on the computed barriers. Within these limits, the permeation profiles obtained here provide a predictive framework for identifying which plasma-generated species are most likely to enter plant cells through aquaporin channels. This contributes to the broader effort of optimizing CAP-based technologies for sustainable agriculture, and motivates future studies incorporating reactive transport, mixed-lipid membranes, and additional plant AQP isoforms. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements J.R gratefully acknowledges the financial support received from the Agency for Innovative Development of the Republic of Uzbekistan, Grant Number FZ-2020092817. P.M. gratefully acknowledge the use of the bioinformatics infrastructure facility supported by Biocenter Finland, the Sigrid Juselius Foundation and the CSC-IT Center for Science (Project: 2000461) for the computational support, Dr. Jukka Lehtonen (Structural Bioinformatics Laboratory) for IT support, Prof. O. Salo-Ahen (SBL - Pharmacy) Åbo Akademi University for providing laboratory support.The authors acknowledge the use of ChatGPT (version 5.2) for editorial support, specifically to enhance the clarity, structure, and readability of the manuscript. The tool was employed only for language refinement and text improvement. All scientific ideas, experimental design, computational work, data collection, analysis, and interpretation were carried out independently by the authors and were fully verified by them. Author contributions All authors contributed to the conception and design of the study. D.A. developed the research methodology and performed the molecular dynamics simulations. D.A. and P.M. conducted data analysis and visualization of the simulation results. D.A., P.M., A.K., C.Z., and D.L. interpreted the results and contributed to the preparation of figures and tables. J.R. supervised the research and critically revised the manuscript for important intellectual content. All authors reviewed and approved the final manuscript. Funding This work was supported by the Agency for Innovative Development of the Republic of Uzbekistan (Grant No. FZ-2020092817) and by the bioinformatics infrastructure facility supported by Biocenter Finland, the Sigrid Juselius Foundation, and the CSC – IT Center for Science (Project No. 2000461). Data availability The datasets generated and analysed during the current study are available from the corresponding author, (D. Abduvokhidov, davronjonabduvokhidov@gmail.com), upon reasonable request. Declarations Competing interests The authors declare no competing interests. Ethical approval This research does not involve any studies with human participants or animals conducted by the authors. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Davronjon Abduvokhidov, Email: davronjonabduvokhidov@gmail.com. Parthiban Marimuthu, Email: Parthiban.Marimuthu@abo.fi. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets generated and analysed during the current study are available from the corresponding author, (D. Abduvokhidov, davronjonabduvokhidov@gmail.com), upon reasonable request.