In this work, we study interactions of five different hemicellulose models, i.e. Galactoglucomannan, O-AcetylGalactoglucomannan, Fuco-Galacto-Xyloglucan, 4-O-Methylglucuronoxylan, and 4-O-Methylglucuronoarabinoxylan, and their respective binding strength to cellulose nanocrystals by molecular dynamics simulations.
Carbohydrate Polymers 270 (2021) 118364 Contents lists available at ScienceDirect Carbohydrate Polymers journal homepage: www.elsevier.com/locate/carbpol Cellulose-hemicellulose interactions - A nanoscale view Ali Khodayari a, *, Wim Thielemans b, Ulrich Hirn c, Aart W Van Vuure a, David Seveno a a Department of Materials Engineering, KU Leuven, Leuven, Belgium Sustainable Materials Lab, Department of Chemical Engineering, KU Leuven, campus Kulak Kortrijk, Etienne Sabbelaan 53, 8500 Kortrijk, Belgium c Institute of Bioproducts and Paper Technology, TU Graz, Graz, Austria b A R T I C L E I N F O A B S T R A C T Keywords: Cellulose Hemicellulose Free energy of adsorption Shear Hemicellulose folding on cellulose Hydration effect Molecular dynamics simulations In this work, we study interactions of five different hemicellulose models, i.e Galactoglucomannan, O-AcetylGalactoglucomannan, Fuco-Galacto-Xyloglucan, 4-O-Methylglucuronoxylan, and 4-O-Methylglucuronoar abinoxylan, and their respective binding strength to cellulose nanocrystals by molecular dynamics simulations Glucuronoarabinoxylan showed the highest free energy of binding, whereas Xyloglucan had the lowest inter action energies amongst the five models We further performed simulated shear tests and concluded that failure mostly happens at the inter-molecular interaction level within the hemicellulose fraction, rather than at the interface with cellulose The presence of water molecules seems to have a weakening effect on the interactions of hemicellulose and cellulose, taking up the available hydroxyl groups on the surface of the cellulose for hydrogen bonding We believe that these studies can shed light on better understanding of plant cell walls, as well as providing evidence on variability of the structures of different plant sources for extractions, purification, and operation of biorefineries Introduction Cellulose and its derivatives have been the target of numerous experimental and numerical studies over the past decades because of their excellent mechanical performance (Moon et al., 2011) Mechanical properties of plant cell walls and hence elementary fibres are controlled by several parameters including the structure of the secondary wall S2, the microfibrillar angle (MFA), crystalline cellulose content, and the ratio of other constituents such as hemicellulose, lignin, and pectin (Eichhorn et al., 2010; Habibi et al., 2010) For instance, it is shown that MFA inversely regulates the tensile behaviour of elementary fibres (Bourmaud et al., 2013) In other words, fibres with lower MFA display higher tensile moduli when stretched along the main axis Moreover, relative humidity (RH) has also proven to impose significant drifts on the elastic modulus of plant fibres While studies show that RH can lead to swelling of fibres and according loss in mechanical strength, it is also observed that certain percentages of RH can enhance the strength and strain to failure of cellulosic fibres (Baley et al., 2005; Placet et al., 2012) Structural analysis of the plant cell wall proposes that the cell wall assembly is mostly responsible for the mechanical properties of the fi bres (Zhong et al., 2019) As an example, the secondary cell wall consists of fibrils made up from cellulose nanofibrils, often referred to as cellu lose microfibrils (CMF), which can be up to thousands of nanometers in length, connected longitudinally through disordered regions (Khodayari et al., 2021; Khodayari et al., 2020; Kontturi et al., 2016; Nishiyama et al., 2003) and laterally through amorphous media including mostly hemicellulose (Cosgrove, 2005; Gibson, 2012), as depicted in Fig The content and ratio of the constituents can affect the moisture uptake, as well as the mechanical interlocking of these constituents together Due to the alignment of the microfibrils along a director, the MFA, it is proposed that when fibres are exposed to a tensile load, a shear force is induced between the hemicellulose and the surface of the cellulose fi brils (Placet et al., 2014) This shear force between the cellulose and hemicellulose can be the reason for the non-linear shape of the stressstrain curves of elementary fibres One important matter of concern when studying interactions of cell wall constituents, and resulting mechanical behaviour is water (Moore et al., 2008; Tang et al., 1999) It has been shown that water-binding capacity of cell walls can be modified by degradation of certain pectic polysaccharide side-chains (Klaassen & Trindade, 2020) Water can alter molecular conformations and mobility of pectic contents in the plant cell wall As minor changes of the rheological properties of the matrix, being the stress transmitters between cellulose microfibrils, can lead to * Corresponding author E-mail address: ali.khodayari@kuleuven.be (A Khodayari) https://doi.org/10.1016/j.carbpol.2021.118364 Received April 2021; Received in revised form 14 June 2021; Accepted 17 June 2021 Available online 23 June 2021 0144-8617/© 2021 The Author(s) Published by Elsevier Ltd This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/) A Khodayari et al Carbohydrate Polymers 270 (2021) 118364 modulations of the overall behaviour of plant cell walls (Ulvskov et al., 2005), it is critical to account for the role of water on the structure of the walls (Evered et al., 2007) From a broader point of view, the type of hemicellulose could also play a role in the strength of the cellulosic fibres In particular, side groups of the hemicellulose can make different types of bonds with the free hydroxyl groups on the surface of the cellulose fibrils, provoking specific shear behaviours In terms of extraction of hemicellulose, this leads to different yields and molecular weights depending on the plant source (Gallina et al., 2018) The chemical composition of each hemi cellulose and the ratio of its constituents depend on the plant species and the stage of tissue development (Barbieri et al., 2017) The most abun dant hemicellulose types are Xyloglucans (XyG), Glucuronoarabinox ylans (GAX), Glucuronoxylans (GX), and Galactoglucomannan (GGM), mostly present respectively in the primary cell walls of hardwood, softwood, and grasses, in the secondary cell wall of grasses, in the sec ondary cell wall of hardwood, and in the secondary cell wall of softwood (Sorieul et al., 2016) The acetylated form of GGM (ACE-GGM) is believed to be more abundant in softwoods (Hannuksela & Du Penhoat, 2004) The diverse structures of XyG that have already been characterized (Hsieh & Harris, 2009; Picard et al., 2000) are formed of β-D-glucopyr anose (Glc) backbones, mostly branched with an α-D-xylopyranose (Xyl) on the C6 The branched xylose units themselves also usually contain galactosyl, fucosyls, or arabinosyl residues (Lerouxel et al., 2006; Pauly et al., 2001) GAXs mainly consist of a β-D-xylopyranose backbone, where α-L-arabinofuranosyl (Ara) residues are occasionally substituted on the O3, and sometimes are branched with α-D-glucopyranosyl acid (GlcA) on the O2, in a non-repeating manner Ferulic acid is also observed to reside on the arabinose groups (arabinofuranose) in a random fashion, and the backbone is also acetylated to a minor extent (Harris, 2006; Kozlova et al., 2012; Vogel, 2008) The Ara:Xyl ratio can vary significantly as a function of the elongation phase or the plant family For instance the molar ratio of Xyl:Ara:GlcA was found to be 45:12:1 in wheat (Zeng et al., 2010), 100:28:8 in Guadua chacoensis ´ndez et al., 2019), or 100:67:8 in woody bamboo (Zelaya et al., (Ferna 2017) GX on the other hand has the same backbone, with more frequent acetylations on either O2/O3 or both (Heinen et al., 2019) Minor 4-Omethyl-α-D-glucopyranosyl acids are substituted on the backbone as well (de Carvalho et al., 2019) GGM consists of β-D-mannopyranose (Man) and β-D-glucopyranose backbone, with occasional β-D-galactopyranose (Gal) residues substituted on the O6 of the mannose units (Eichinger et al., 2019) Acetylations in ACE-GGM occur on the mannose units, and happen on either C2 or C3 (Berglund et al., 2019) The reported ratios for Man:Glc:Gal also vary in different plant cells with Man being the majority and Gal the minority in GGM structures (Barbieri et al., 2017; Lundqvist et al., 2002; Sims et al., 1997; Yu et al., 2018) Note that, despite concrete findings on the type and position of linkages and sub stitutions on the backbone of each mentioned hemicellulose, the ratio of the constituents in hemicelluloses differs from one plant to the other, and also within one specific plant itself For instance, it has been shown that GAX substitution type and frequency of side chains differ within the stem and leaf of grasses (Tryfona et al., 2019) In other words, compo sition of the hemicelluloses has shown to be tissue specific (Pauly et al., 2001) Degree of substitution and its pattern, playing an important role on the functionality of hemicellulose models is also shown to differ in plants, as well as their development stages (Scheller & Ulvskov, 2010) In this study, we model the most abundant type of each hemicellulose from the substitution point of view The Iβ cellulose crystal structure has been a source of debate for years Several structures, including different number of chains have been already proposed Proposed models for Iβ cellulose mainly include 18-chain (Kubicki et al., 2018; Nixon et al., 2016; Ros´ en et al., 2020; Vandavasi et al., 2016; Zhong et al., 2019), 24-chain (Fernandes et al., 2011; Willhammar et al., 2021), or 36-chain structures (Elazzouzihafraoui et al., 2008; Endler & Persson, 2011; Mutwil et al., 2008; C Zhang et al., 2021) Despite all, there are still discussions in the field, whether one model should be favored over the other (Cosgrove, 2014; Hill et al., 2014) Hill et al (2014) did come up with a convincing argument that from the possibilities of 12, 18, 24 and 36 chains, the 18 chain CMF was the most likely, while 36 would also be possible but less probable, and that 12 and 24 would be the least likely to exist, also supported by Cosgrove and Jarvis (2012) There is however also con tradicting work For instance, Wang and Hong (2016) provides NMR Fig Structural hierarchy of plant fibres Technical fibres are composed of elementary fibres Elementary fibres consist of primary cell wall, secondary cell walls, and a lumen at the core Cellulose microfibrils connected through hemicellulose and lignin, mostly make up the secondary cell wall A Khodayari et al Carbohydrate Polymers 270 (2021) 118364 evidence that the CMF consists of at least 24 chains in various wild type and mutant Arabidopsis primary cell walls, which given the 18 enzyme complex and the arguments of Hill et al (2014) that 24 chains are not possible might indicate 36 glucan chains crystals exist To try to avoid the complexity, we have chosen the hydrophilic surface of the × model to interact with hemicellulose, as the actual state of the surfaces at the outer side of the crystals will be very similar to those of an 18chain model Hence, the results of this study should remain intact, whether a 36-chain or an 18-chain model is used Conformational analyses of crystalline cellulose and hemicellulose have been already studied in detail by molecular dynamics simulations in many studies (Berglund et al., 2020; Nishiyama et al., 2008; Nish iyama et al., 2012) Flexibility of hemicelluloses has been a matter of concern, as the molecular conformation, and consequently, function ality of hemicelluloses would be directly affected Berglund et al (2016) inspected the flexibility of different combinations of D-xylopyranosyl, Dmannopyranosyl, and D-glucopyranosyl units as di- and tetra saccharides The observation was that there are differences in the flex ibility of different linkage types in their models Particularly, the backbone of the xylan seemed to show more flexibility that those seen in cellulose, glucomannan, and xyloglucan The higher flexibility is then related to weaker interactions of xylan and cellulose, compared to glu comannan and cellulose (Åkerholm & Salm´en, 2001) From a modelling point of view, amongst the available force fields parametrized to model carbohydrates, GLYCAM06 is shown to be a proper choice to model cellulose and hemicellulose structures (Foley et al., 2012) In a study performed by Matthews et al (2012), authors simulated Iβ crystalline cellulose structures with three different force fields, all parametrized to model carbohydrates for near microseconds GLYCAM06 was shown to be capable of modelling cellulose, consistent with experimentally observed structures It must be noted that, the complexes proposed in this study not aim to model the plant cell wall, as other important substances such as pectin (McCann & Roberts, 1996) and lignin, which are inherently determinant in defining the behaviour of the cell wall are not present (Kang et al., 2019) These components, despite being minor in some plant walls, can be crucial in determining the architecture, and even ´nchez et al., tually the mechanical behaviour of the cell walls (Moneo-Sa 2020) Moreover, we aim not to favor any particular plant cell wall model, as it has been argued for quite a while that load-bearing segments in the plant cell walls could be possibly formed of contacts made by CMFs, bridged by hemicellulose matrices (Park & Cosgrove, 2012) Therefore, our focus in this work is to model the shear behaviour of cellulose and hemicellulose, and governing mechanisms controlling their interactions We start by calculating and comparing the binding free energy be tween five hemicellulose models and a cellulose nanocrystal (CNC) to provide enough information on the way these hemicelluloses interact with the surface of cellulosic fibrils In particular, results of this study tend to rationalize why different plant cell walls show different me chanical properties, from the point of view of hemicellulose type con tent We further inspect the folding behaviour of hemicellulose models onto cellulose hydrophilic surfaces and compare this with the folding behaviour of the models in solvated states Further, we investigate the shear behaviour between CNCs and hemicellulose employing molecular dynamics simulations In particular, we model two types of shear: the first model mimics the shear between two surfaces, one made of cellu lose and the other made of hemicellulose, and the other model mimics the situation where a CNC is pulled out of a hemicellulose bath, mimicking the shear in elementary fibres The relationship between the observed stick-slip behaviour and the (de)formation of hydrogen bonds is demonstrated and explained These simulations are provided to shed light on the internal mechanisms of natural fibres when mechanically loaded We believe this study could assist in better characterization and modelling of the plant cell wall constituents and mechanisms of deformation Materials and methods 2.1 Molecular dynamics simulations Molecular dynamics simulations have been performed by GROnin gen MAchine for Chemical Simulations (GROMACS) 2019.1 version (van der Spoel et al., 2005) GLYCAM06 parameter sets (Kirschner et al., 2008) from Amber (Case et al., 2018) are used, and conversions of the Amber topology files to GROMACS format is done through Acpype py thon code (Sousa Da Silva & Vranken, 2012) Leap-frog algorithm is used to solve the Newton's equations of motion Bonded hydrogens are constrained throughout the ensemble and production runs with the LINear Constraint Solver (LINCS) algorithm, which is performed to speed up the calculations (Hess et al., 1997) All bonds were constrained for the free energy calculations, allowing a time step of fs Verlet scheme is used for neighbour searching with a cut-off of 1.4 nm for both Coulombic and van der Waals interactions Long range electrostatics are treated by Particle Mesh Ewald (PME) method (Darden et al., 1993) Nos´e-Hoover thermostat is used to keep the temperatures at 300 K with a time constant of 1.0 (Hoover, 1985; Nos´ e, 1984), and Pressure coupling is performed using Parrinello-Rahman barostat, applying isotropic pressure at 1.0 bar with a time constant of 2.0, and compressibility of 4.5 × 10− (Parrinello & Rahman, 1981) Periodic boundary conditions are applied in all directions Two box sizes were used for two sets of simulations: a triclinic box of 17 × 17 × 30 nm3 for the first configu rations, and a rhombic dodecahedron box of 25 nm, 25 nm, 17 nm being v1(x), v2(y), and v3(z), respectively, and with v3(x) and v3(y) equal to 12.5 nm TIP3P water molecules are used to solvate the simulation boxes (Mark & Nilsson, 2001) Steepest descent minimization algorithm is used to minimize the configurations Molecular images in this work are rendered by PyMOL (DeLano, 2009) 2.2 Cellulose model × Iβ crystalline cellulose models have been built according to the crystallography data of (Nishiyama et al., 2002) Two finite degree of polymerizations (DP) are used in these simulations, i.e DP 10 for the free energy calculations, and DP 30 for the second set of simulations (Khodayari, Van Vuure, et al., 2020) Fig 4a shows the CNC (DP 30) structure 2.3 Hemicellulose models Five hemicellulose types are considered in this work, namely Gal actoglucomannan (GGM), O-Acetyl-galactoglucomannan (ACE-GGM), 4-O-Methylglucuronoxylan (GX), 4-O-Methylglucuronoarabinoxylan (GAX), and Fucogalactoxyloglucan (XyG) The hemicellulose models for the free energy calculations all have units of backbone (except for the GGM model) as shown in Fig This is to eliminate length effects when potential of mean forces are computed for each hemicellulose case Hence, comparison between the GGM free energy, and the rest of the models might be biased in this work 2.4 Free energy simulations Free energy calculations are performed to measure the free energy of (un)binding between each hemicellulose model and a CNC A combi nation of center-of-mass (COM) pulling and umbrella sampling is done for this analysis (Lemkul & Bevan, 2010; Torrie & Valleau, 1977) The CNC is considered to be short (10 DP), to decrease the computational demand Larger CNCs would have required bigger simulation box and would not be cost-efficient to perform extended molecular dynamics Each hemicellulose model was positioned in the vicinity of the CNC, on the (110) hydrophilic face with exposed hydroxyl groups (Besombes & Mazeau, 2005), within the non-bonded cut-off range (