Xylan adsorption on cellulose: Preferred alignment and local surface immobilizing effect

Interaction between xylan and cellulose microfibrils is required to maintain the integrity of secondary cell walls. However, the mechanisms governing their assembly and the effects on cellulose surface polymers are not fully clear. Here, molecular dynamics simulations are used to study xylan adsorption onto hydrated cellulose fibrils.

Available online 17 February 2022

0144-8617/© 2022 The Authors Published by Elsevier Ltd This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)

Xylan adsorption on cellulose: Preferred alignment and local surface

immobilizing effect

Emilia Heinonena,b, Gunnar Henrikssona,c, Mikael E Lindstr¨omc, Francisco Vilaplanaa,b,

Jakob Wohlerta,c,*

aWallenberg Wood Science Center, KTH Royal Institute of Technology, Teknikringen 56-58, Stockholm 10044, Sweden

bDivision of Glycoscience, Department of Chemistry, KTH Royal Institute of Technology, AlbaNova University Centre, Roslagstullsbacken 21, Stcokholm 10691, Sweden

cDepartment of Fibre and Polymer Technology, KTH Royal Institute of Technology, Teknikringen 56-58, Stockholm 10044, Sweden

A R T I C L E I N F O

Keywords:

Xylan

Cellulose

Cell wall,

Molecular dynamics

Adsorption

Conformation

A B S T R A C T Interaction between xylan and cellulose microfibrils is required to maintain the integrity of secondary cell walls However, the mechanisms governing their assembly and the effects on cellulose surface polymers are not fully clear Here, molecular dynamics simulations are used to study xylan adsorption onto hydrated cellulose fibrils Based on multiple spontaneous adsorption simulations it is shown that an antiparallel orientation is thermo-dynamically preferred over a parallel one, and that adsorption is accompanied by the formation of regular but orientation-dependent hydrogen bond patterns Furthermore, xylan adsorption restricts the local dynamics of the adjacent glucose residues in the surface layer to a level of the crystalline core, which is manifested as a three-fold increase in their 13C NMR T1 relaxation time These results suggest that xylan forms a rigid and ordered layer around the cellulose fibril that functions as a transition phase to more flexible and disordered polysaccharide and lignin domains

1 Introduction

The interaction between xylan and cellulose is one of the most

important molecular scale phenomena in plants, both from a biological

perspective and in an industrial context Together with glucomannan

and lignin, they constitute the major structural components of the wood

secondary cell wall (Donev, Gandla, J¨onsson, & Mellerowicz, 2018) It is

known that a certain amount of xylan is necessary to maintain the

integrity of the cell wall and to withstand the large gravitational forces

and turgor pressures associated with supporting the growing tree while

securing water and nutrient transport Glucuronoxylan deficient

Arabi-dopsis mutants result in plants with dwarfed stems and roots and reduced

cellulose content (Persson et al., 2007), highlighting the importance of

xylan-cellulose interactions from the viewpoint of the living organism

From an industrial point of view, on the other hand, the tight

asso-ciation between xylan and cellulose can become an obstacle for

suc-cessful fractionation and modification of the individual lignocellulose

components, and for biotechnological transformation through

sacchar-ification and fermentation processes Adsorbed xylan has lower

accessibility for xylanolytic enzymes (Teleman, Larsson, & Iversen,

2001; Viikari, Kantelinen, Buchert, & Puls, 1994) and may also impede the action of cellulolytic enzymes on cellulose surfaces (Meng & Ragauskas, 2014; Zhang, Tang, & Viikari, 2012) It also interferes with the conversion of dissolving pulp into further products (Sixta, 2006) Nevertheless, the tendency of xylan and cellulose to interact can also be beneficial as it results in reduced hornification (Yang, Berthold, & Ber-glund, 2018), higher strength of paper and higher pulping yield (Ribe, Lindblad, Dahlman, & Theliander, 2010) An increased understanding of cellulose-xylan interactions can thus be of importance for a better un-derstanding of the structure-function relationships of the plant cell walls, for the development of more efficient processes for selective extraction and purification of cell wall polysaccharides, and for the design of high-performance materials

On the molecular level, the association of xylan to cellulose is accompanied by a change in conformation; from a twisted 3-fold screw found in solution to a pseudo-flat 21-fold conformation (Berglund et al.,

2016; Busse-Wicher et al., 2014; Mazeau, Moine, Krausz, & Gloaguen,

2005), similar to that of the native cellulose polymers found in the

* Corresponding author at: Wallenberg Wood Science Center, KTH Royal Institute of Technology, Teknikringen 56-58, Stockholm 10044, Sweden

E-mail addresses: sehei@kth.se (E Heinonen), ghenrik@kth.se (G Henriksson), mil@kth.se (M.E Lindstr¨om), franvila@kth.se (F Vilaplana), jacke@kth.se (J Wohlert)

Contents lists available at ScienceDirect Carbohydrate Polymers

journal homepage: www.elsevier.com/locate/carbpol

https://doi.org/10.1016/j.carbpol.2022.119221

Received 8 December 2021; Received in revised form 31 January 2022; Accepted 1 February 2022

elementary fibrils Such conformational change is believed to extend the

cellulose crystal lattice Thereby, it is also affecting the physical

prop-erties of the cellulose, such as increasing the apparent crystallinity of the

surface cellulose determined from xylan-cellulose cross-peaks (Simmons

et al., 2016) On a related note, 13C NMR spin-lattice-relaxation times in

cellulose was observed to increase upon adsorption of xyloglucan,

another hemicellulose (Terenzi, Prakobna, Berglund, & Fur´o, 2015),

which indicates that the segmental dynamics of surface cellulose chains

may be affected as well

Xylan is composed of a β-(1→4)-D-xylopyranosyl (Xylp) backbone,

which can be either acetylated or carry α-L-arabinopyranose

sub-stitutions, and further substituted by 4-O-methylated

β-(1→2)-D-glu-curonic acid units, depending on source (Ebringerov´a, Hrom´adkov´a, &

Heinze, 2005) Both the amount of substitutions and the specific

sub-stitution pattern is believed to affect xylan-cellulose association (

Bos-mans et al., 2014; Busse-Wicher et al., 2016; Kabel, van den Borne,

Vincken, Voragen, & Schols, 2007; Martnez-Abad, Giummarella,

Law-oko, & Vilaplana, 2018), as well as xylan degree of polymerization (DP)

(Crowe et al., 2021) Indeed, adsorption of xylan to cellulose has been

shown to increase with the removal of both O-acetyl (Kabel et al., 2007)

and arabinosyl (Andrewartha, Phillips, & Stone, 1979; Bosmans et al.,

2014; Kabel et al., 2007) side groups, as well as with decreasing amounts

of 4-O-Me-GlcA (Chimphango, G¨orgens, & van Zyl, 2016; Linder,

Bergman, Bodin, & Gatenholm, 2003) However, low substituted xylan

has a strong tendency to self-associate already at low concentrations,

hence adsorbing to cellulose surfaces as aggregates rather than a

monolayer (Bosmans et al., 2014) Due to this partial insolubility, it is

challenging to experimentally determine the mode of adsorption of

single unsubstituted xylan chains, which is why molecular dynamics

(MD) simulations has become a convenient alternative to study xylan-

cellulose interactions in detail (Busse-Wicher et al., 2014, 2016;

Fal-coz-Vigne et al., 2017; Martnez-Abad et al., 2017; Mazeau et al., 2005)

MD simulations confirm both the strong interaction between xylan

and cellulose, and the benefits of a pseudo-flat conformation (Busse-

Wicher et al., 2016; Falcoz-Vigne et al., 2017; Martnez-Abad et al., 2017;

Mazeau & Charlier, 2012; Pereira, Silveira, Dupree, & Skaf, 2017)

However, how xylan adsorption affected the underlying cellulose

sub-strate has not been analyzed Moreover, most simulations were started

by placing the xylan directly on top of the cellulose surface, parallel to

the glucan chains (Busse-Wicher et al., 2016; Martnez-Abad et al., 2017;

Pereira et al., 2017) One recent exception is the study by Gupta, Rawal,

Dupree, Smith, and Petridis (2021) where the xylan chains were placed a

small distance away from the surface and let to adsorb spontaneously,

although only the parallel orientation was examined Thus, the extent to

which simulation results are affected by the initial conditions has been

poorly investigated In the present work, cellulose-xylan complexes

were modeled without pre-adsorbing the xylans on the cellulose, and

using different relative starting orientations From several microsecond-

long MD simulations the spontaneous adsorption of unsubstituted xylan

to cellulose in water was analyzed with respect to xylan structure, its

relative orientation to cellulose, and also changes in both structure and

dynamics of the cellulose surface chains Such detailed molecular level

understanding is helpful for developing more accurate cell wall models

at the supramolecular level and interpreting results from experimental

studies on lignocellulose

2 Methods

The cellulose polymer consists of β-(1→4)-linked D-Glcp residues In

nature the chains crystallize into fibrils of co-existing crystalline forms:

Iα and Iβ, where the latter dominates in higher plants (Nishiyama,

2009) Current understanding of the dimensions and the organisation of

the cellulose synthesis complex in wood gives an 18-chain fibril as the

most probable one (Cosgrove, 2014), which is also proposed by

exper-imental measurements of wood fibril dimensions (Newman, Hill, &

Harris, 2013) However, fibril models containing 24 or 36 chains have

also been proposed (Fernandes et al., 2011; Thomas, Martel, Grillo, & Jarvis, 2020) The specific arrangement of chains into a fibril cross section is still a matter of debate However, a recent computational study comparing different cross sections found that the hexagonal arrange-ment proposed by Newman (Newman et al., 2013) was the energetically most favourable one (Yang & Kubicki, 2020), and is therefor used in the present study The number and arrangement of the cellulose chains determines the ratio of hydrophilic to hydrophobic surface (Fig 1), which affects the possible interactions with hemicelluloses (Cosgrove,

2014)

The cellulose model used herein was generated using Cellulose Builder (Gomes & Skaf, 2012) and consisted of 18 fully periodic chains arranged as fibrils with a hexagonal cross-section in the cellulose I form (Nishiyama, Langan, & Chanzy, 2002) Both cellulose and xylan topol-ogies were generated using the tLeap module in Amber tools (Case et al.,

2019) with the most recent release of GLYCAM06 force field (Kirschner

et al., 2008) Amber topologies were converted to GROMACS (Abraham

et al., 2015; Hess, Kutzner, van der Spoel, & Lindahl, 2008) using ACPYPE (Bernardi, Faller, Reith, & Kirschner, 2019; da Silva & Vranken,

2012) Furthermore, the 1–4 interaction scale factors were set to 1.0 For water, the TIP3P model was used (Jorgensen, Chandrasekhar, Madura, Impey, & Klein, 1983)

Four different configurations were prepared (Fig 1) A cellulose fibril of DP 12 was placed parallel to the z-axis Next, a number of xylan chains of DP 6 were placed in the vicinity of the fibril, on average 6.1 Å` away from the cellulose surface The first system (denoted S1) has six xylan chains, where the top three ones are aligned parallel to the cel-lulose fibril and the bottom three are antiparallel In the second system (S2) every other chain is parallel The third system (S3) has three chains initially placed perpendicular to the fibril, and in the fourth system (S4) has six perpendicular chains An additional fifth system (S5) used a similar arrangement as in S1, except that both cellulose and xylan were elongated to DP 16 and DP 12, respectively All systems were fully sol-vated using explicit water

After energy minimization, all systems were subject to 1 μs of MD at ambient temperature and pressure allowing the xylans to adsorb freely

to the cellulose surface The time step was 0.002 ps and energies and coordinates were saved every 5000 steps Non-bonded interactions used

a 1.2 nm cutoff and long-range electrostatic interactions were included using PME (Darden, York, & Pedersen, 1993; Essmann et al., 1995) Temperature was controlled by the velocity-rescale thermostat (Bussi, Donadio, & Parrinello, 2007), and pressure by semi-isotropic Parrinello- Rahman pressure coupling (Parrinello & Rahman, 1981) with xy- and z- axis compressibilities set to 4.5 × 10− 5 and 4.5 × 10− 6 Pa− 1, respec-tively Constraints were applied on all bonds using P-LINCS (Hess,

2008)

Simulations of S1 and S5 were extended by 100 ns, saving the co-ordinates every 5 ps to obtain sufficient data for calculating 13C NMR

spin-lattice (T1) relaxation times from 1

T1

=n H

10χ2(j(ω H− ω C) +3j( ω C) +6j( ω H+ω C) ) (1)

where j( ω ) is the spectral density of the P2 rotational autocorrelation function of the C–H bond vectors at the frequency ω This is the appropriate expression for an isotropic system, i.e., where all C–H orientations are equally probable Here, a carbon Larmor frequency of

100 MHz was used for comparing with previously published data (Chen, Terenzi, Fur´o, Berglund, & Wohlert, 2019) The constant in front

de-pends on the number of protons bonded to the carbon (n H, which is equal

to one in the case of C4) and on the dipole-dipole coupling constant In

the present case the numerical value becomes n H χ 2/10=2.3 × 109 s− 2 Calculations were performed separately for each C4-H4 bond in the cellulose after which the values were grouped based on their location: inner surface (meaning that the C6 hydroxymethyl is pointing inwards,

to the crystalline core) or outer surface (meaning that C6 points

outwards) Reported values are logarithmic averages for each group

3 Results

MD simulations were run on systems with varying initial

configu-ration, xylan concentconfigu-ration, and degree of polymerization The resulting

trajectories were analyzed with respect to xylan orientation and

conformation, cellulose-xylan hydrogen bonding patterns, and finally

the effects on segmental mobility of surface cellulose chains

3.1 Xylan aligns spontaneously on the cellulose fibril's surface

The systems were characterized with respect to the orientation and

alignment of the xylan polymers, and the effects of initial configuration,

chain length and concentration, respectively Xylan orientation

(paral-lel/antiparallel or perpendicular) and location (hydrophobic or

hydro-philic surface) at the beginning and at the end of the simulations are

presented in Table 1 Snapshots of systems S1, S4 and S5 (Fig 2), as well

as S2 and S3 (see Fig A.1) are used to visualise the migration of xylans

once adsorbed

The alignment of xylan oligomers was assessed by calculating the

xylan end-to-end distance and projecting it onto the fibril axis Fig 3

shows the alignment per xylan chain in each system as a function of time, where a value of +1 indicates perfect alignment in a parallel orientation, while − 1 means antiparallel Generally, most xylan chains are well aligned most of the time, which is shown by the histogram in

Fig 3 In S5 fluctuations are small compared to those observed in S1-4 This is an effect of the chains being twice as long, which reduces the mobility of the xylans on the cellulose surface and the relative contri-bution from the more mobile chain ends In S1 and S2 relatively large fluctuations on the scale of several tens of nanoseconds occur Xylan X4

in S2 rotated such that it interacted with both the fibril and its periodic image and is therefore excluded from further analysis However, it is curious that this xylan-mediated fibril-fibril connection via xylan was stable for more than 350 ns In S3, xylans that were initially oriented perpendicular to the fibril quickly aligned with the fibril axis whereas in the more concentrated system (S4) it took longer time before they settled in a partially aligned state, approximately 80–90% of full alignment An explanation for this is the crowding imposed by neigh-boring xylan chains, which is supported by the snapshots shown in

Fig 2

hydrophilic hydrophobic

reducing end

Cellulose Iβ Xylan DP12

200

1-10

Z 110

reducing end

1

X1

X2 X3

X4

X5

X6

X1 X2

X3

X4

X5

X6

X1 X2

X3 X1

X2

X3

X4

X5

X6

2

Systems

Φ Ψ

Fig 1 A) β-(1→4)-xylan, 2-fold helix and 3-fold screw conformations B) Cellulose I intrasheet hydrogen bonding pattern and dihedral angles of the glycosidic bond,

ϕ (O5’-C1’-O4-C4) and ψ (C1’-O4-C4-C3) Possible conformations of the hydroxymethyl group are trans-gauche (tg), gauche-gauche (gg) and gauche-trans (gt) Cross-

section of 18 chain cellulose fibril used in this study, exposing the 110 and 110 (hydrophilic), and 200 (hydrophobic) crystallographic planes C) Starting points of the simulations: placement of the xylan chains around the cellulose either pre-aligned (S1 and S2) or perpendicular (S3 and S4) System S5 is similar to S1 except that xylan chains are extended from DP 6 to DP 12, and the cellulose fibril from DP 12 to DP 16

Table 1

Alignment and location of the xylan chains with respect to the cellulose fibril

3.2 Xylan prefers antiparallel orientation to cellulose

Curiously, all xylan chains in system S3 settled in an antiparallel

orientation To investigate this behavior further, the simulation was

repeated 30 times (using different random seeds and for 150 ns) to

ac-quire more statistics Of the total 93 xylan chains (31 simulations using 3

chains each), the final orientation was antiparallel in 61 cases, parallel

in 30 cases, and unaligned in 2 cases Here, as our xylan is unsubstituted and thus differs from cellulose by just lacking the hydroxymethyl group,

a comparison can be made with the difference between the two cellulose crystal allomorphs I and II In cellulose I (native cellulose) all chains are parallel, but upon alkali treatment it can be irreversibly converted to cellulose II (Okano & Sarko, 1985; Simon, Glasser, Scheraga, & Manley,

1988), which has been shown to be thermodynamically the most stable

1000 ns

1000 ns

10 ns

10 ns

500 ns

1000 ns

1000 ns

500 ns

System 4 (S4) System 1 (S1)

System 5 (S5)

Fig 2 Snapshots of the systems S1, S4 and S5 at 10, 500 and 1000 ns The fibril surface is visualised with the VMD drawing method QuickSurf and the water

molecules are excluded for clarity Blue-red end of xylan marks the non-reducing end (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

allomorph (Goldberg et al., 2015) In this structure, every other chain is

antiparallel Thus, xylan seems to be no different from cellulose in that

alternating chain polarity is slightly favoured However, the native

orientation of xylan with respect to the cellulose microfibrils may still be

a consequence of biosynthetic mechanisms during cell wall formation,

rather than equilibrium thermodynamics

The aspect of preferred orientation of has only seldom been

addressed in the simulation literature Hanus and Mazeau (2006) found

that the differences in interaction energies in vacuum between parallel, antiparallel and perpendicular xyloglucan on cellulose surface were within the calculated error and hence all of them were well tolerated Later work on xylan adsorption to cellulose in vacuum indicated that the parallel orientation would be preferred (Mazeau & Charlier, 2012) In that work, a single simulation using three xylans in random starting orientations was performed, and it was found that two out of three finished parallel and one antiparallel More recently Falcoz-Vigne et al

System 2 (S2) System 1 (S1)

System 4 (S4) System 3 (S3)

Fig 3 Time evolution of xylan end-to-end-vector projections on the z-axis (cellulose fibril axis), where +1 means fully parallel and -1 means fully antiparallel

Systems S1, S2 and S5 initially have xylans pre-aligned, while S3 and S4 initially have xylans perpendicular to the cellulose fibril axis The xylan numbers are same as

in the Fig 1 The bottom right panel displays a histogram of normalised absolute projections of all xylans

(2017) studied the preferred orientation by placing a xylo-oligomer at

either 30, 60 or 90 degree angle to the cellulose surface Their 10 ns

simulation results show that xylan (DP 10) at 30 and 60 degrees tends to

rotate towards parallel orientation during the 10 ns simulation time,

while the xylan at 90 degrees did not, even after 30 ns In the present

simulations, using xylans of DP 6, 30 ns is typically enough for achieving

alignment of the xylan chains It could be that longer oligomers need

longer times to align Moreover, since the present simulations show that

the statistical preference for antiparallel orientation after 60 trials

re-mains fluctuating between 0.4 and 0.5:1 (parallel:antiparallel), any

outcome is probable if predictions are based based on too few

obser-vations Recent studies have concentrated on other aspects of xylan-

cellulose interactions and usually assumed a parallel orientation

(Busse-Wicher et al., 2014; Martnez-Abad et al., 2017; Pereira et al.,

2017)

3.3 Xylan adopts a 2-fold conformation when adsorbed/aligned onto

cellulose surfaces

One of the most important structural features of polysaccharides is

the variation in how consecutive sugar residues are connected by the

glycosidic linkages This largely determines the conformational space of

the polysaccharides (Varki, 2017) The β-(1→4)-linked backbone in

xylan in combination with the lack of an hydroxymethyl group on C5

makes an extended, twisted 3-fold screw the lowest energy

conforma-tion in soluconforma-tion However, a pseudo-flat 2-fold conformaconforma-tion is

ener-getically allowed, and the barrier between the two conformations is

sufficiently low to allow inter-conversion between the two in

equilib-rium (Berglund et al., 2016) In simulations, when xylan adsorbs to a flat

surface such as cellulose, the 2-fold conformation becomes the dominant

one, since it will maximize the specific interactions between the xylan

and the substrate (Busse-Wicher et al., 2016; Martnez-Abad et al., 2017)

This conformational change was also experimentally observed in solid

state CP/MAS 13C NMR (Simmons et al., 2016; Teleman et al., 2001)

It has been shown that the sum of the two dihedral angles ϕ and ψ is a

good indicator of the polysaccharide chain conformation (French &

Johnson, 2009) Thus, the conformational space can be reduced to just

one dimension Here, ϕ and ψ were defined by the sequences O5’-C1’-

O4-C4 and C1’-O4-C4-C3, respectively In this representation, a sum of

300◦corresponds to a right-handed 3-fold helix, 360◦to a 2-fold screw,

and 420◦to a left-handed 3-fold helix (see Fig 1 for the xylan

confor-mations and Fig 5 for examples of conformational plots, remaining

conformational plots are shown in Fig A.2-A.6)

Histograms of ϕ + ψ collected from the simulations compared to a

histogram from a simulation of a free xylan in solution (Fig 4) confirms

that unsubstituted xylan prefers the 2-fold conformation This is in

agreement with previous MD simulations (Busse-Wicher et al., 2016;

Martnez-Abad et al., 2017; Pereira et al., 2017), although recently Gupta

et al (2021) observed a different kind of behavior In their study an unsubstituted xylan oligomer remained in 3-fold conformation and even desorbed from the cellulose surface This is possibly related to the choice

of interaction potentials, which will be addressed further down

Fig 4 Left: Histograms of the sum of and from all simulations combined Adsorbed xylan (yellow) is compared to a reference in water (magenta) Middle and right:

Xylan chains in S4 split in two groups based on the equilibrium projection length (see text Fig 3) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

X4

A)

B)

X6

Fig 5 Example conformational plots of 3rd the glycosidic linkage of the xylan

xylan X6 and X4 (S4), where the diagonals indicate the sum of φ and ψ 300◦ corresponds to a right-handed 3-fold helix, 360◦to a 2-fold screw, and 420◦to a left-handed 3-fold helix

Also histograms pertaining to adsorbed xylans displays a minor peak

around 440◦that corresponds to a left-handed 3-fold conformation To

investigate the origin of this peak, separate histograms of fully and

partially aligned xylan oligomers were constructed based on their

projections For S4, full alignment correlates well with the 2-fold conformation, and the 3-fold peak is suppressed when partially aligned xylans are removed from the analysis Instead, the distribution

of partially aligned xylans from the S4 overlaps perfectly with the

A)

C)

D)

B)

Fig 6 Short-range xylan-cellulose interactions in S3 Xylan chains are grouped based on (A) orientation (parallel/antiparallel) only, or (B–D) both orientation and

the closest cellulose surface Xylan chains that could not unambiguously be assigned to either the hydrophilic or the hydrophopic surface were excluded Panels (A) and (B) show the total non-bonded energy while (C) displays Coulomb interactions, and (D) the Lennard-Jones (LJ) interactions

reference This indicates that xylans may indeed form stable complexes

with cellulose while retaining the 3-fold conformation, at least within

the time scale of these simulations

In S5 the xylans were quite immobile and the 3-fold conformation

arises mainly from individual linkages that makes one of the chains (X4)

to bend on the cellulose surface (Fig 2), which results in alternating sections of 2- and 3-fold conformation This is an curious observation considering that recently Simmons et al (2016) detected a minor

0 1 2 3 4 5

X2-G2 X2-G3 X2-G6 X3-G2 X3-G3 X3-G6 X2-G2 X2-G3 X2-G6 X3-G2 X3-G3 X3-G6

acceptor (O) donor (OH)

0 2 4 6 8 10

A)

B)

Xylan antiparallel

Xyl

Glc

Xyl

Glc

Xyl

C)

Xylan parallel

Fig 7 A) Average number of hydrogen bonds in each system per 10 xylose residues (Note: chain end residues excluded) Hydroxyl groups in xylose and glucose

residues are denoted with X and G, respectively, followed by the carbon number to which they are attached B) Average number of hydrogen bonds in system S5 per

10 xylose residues located on hydrophilic surfaces, for parallel and antiparallel orientation separately (Note: chain end residues excluded) C) Snapshots of system S5 illustrating the hydrogen bonding patterns of parallel and antiparallel xylan

fraction of 3-fold xylan in a wild type Arabidopsis, but only as a relatively

immobile component in CP-INADEQUATE spectra, not as a mobile

component in DP-INADEQUATE Furthermore, their study showed that

the 3-fold xylan in the wild-type had higher dipolar order parameter

compared to the major 3-fold fraction in the cellulose deficient mutant

Similarly, the 3-fold xylan of the wild type had 13C NMR T1-relaxation

time in same range with 2-fold xylan and cellulose These observations

suggest that even 3-fold xylan in the secondary cell wall can be relatively

rigid, and perhaps close to cellulose Thus, the presence of a minor 3-fold

fraction could indicate occasional bends in otherwise tightly bound

xylan, either in the middle of the polymer or at the chain ends

3.4 Hydrogen bonding of xylan is orientation dependent

When xylan adsorbs to cellulose, hydrogen bonds (H-bonds) are

formed between the polymer and the substrate However, H-bonding is

not required neither to drive the adsorption, nor for the xylan to adopt a

2-fold conformation, since both adsorption and a shift from 3-fold to 2-

fold occurs also on the hydrophobic surface On the other hand, H-

bonding does to some extent define the crystal lattice of cellulose, and

since it has been suggested that adsorbed xylan could extend the

crys-talline structure, it is of interest to investigate the H-bonding

charac-teristics of the xylan-cellulose systems To that end, H-bonds between

the xylan and cellulose, for those chains adsorbed to hydrophilic surfaces,

were analyzed using standard geometric criteria: a cutoff angle of 30◦

and a cutoff distance of 3.5 Å The first 100 ns of each simulation are

excluded from the analysis Results are presented in Fig 7 In the

following, specific hydroxyl groups are denoted by either X or G

indi-cating whether they belong to xylan or cellulose residues, followed by

their respective carbon number

The average number of interchain H-bonds varied between 7.5 and

10.4 per 10 xylosyl residues (Fig 7A) This makes sense, since the xylans

adsorb on the cellulose surface primarily in 2-fold conformation, hence

every other residue has X2 and X3 directed towards the cellulose and

apparently at least one of them is participating in H-bonding X2 is

somewhat more likely to participate in intermolecular H-bonding

compared to X3 in systems S1–3 and S5, while in S4, X2 and X3 both

participate in, on average, 4.6 and 4.2H-bonds, respectively This

dif-ference can be explained by the tendency of X3 to form an intrachain H-

bond to O5 of the preceding xylose residue

Next, the effect of xylan orientation on H-bonding pattern was

analyzed using data from S5 where chains are well aligned but not

forming aggregates, thus only forming H-bonds to cellulose and not to

neighboring xylans There is a remarkable difference in the number of H-

bonds depending on orientation: parallel xylan forms on average five H-

bond to cellulose per 10 xylosyl residues while antiparallel forms nine

This agrees with the visual observation that parallel xylans generally

form one bond per every other residue, whereas antiparallel xylans gives

a more complex pattern that often involves both of the free hydroxyls,

X2 and X3 (Fig 7C)

Parallel xylan preferably forms a regular network of X2-G6 bonds

with a minor fraction of X3-G6 (Fig 6B), which is in perfect agreement

with Busse-Wicher et al (2014) In this pattern, X2 may be either donor

or acceptor, while X3 almost solely functions as an acceptor This pattern

is reminiscent of that found in native cellulose (Nishiyama et al., 2002),

but not identical since xylan lacks the hydroxymethyl group

Antiparallel xylan exhibit a more complex pattern Of all possible H-

bond combinations, only X3-G3 is rare H-bonds are formed equally

likely through X2 and X3, often simultaneously The most common H-

bonds are X2-G3 and X3-G2 followed by X2-G6 and X3-G6 Again, X3

usually participates as an acceptor, which means that X2 in X2-G3, and

G2 in X3-G2, typically act as donors On the other hand, X2-G2 and X2-

G6 may be formed either way Visual observation reveals two kinds of H-

bonding networks, here denoted A and B Network A is characterized by

hydrogen bond pairs of X2-G3 and X3-G2, with G6 hardly participating

at all but rather pointing away in a gg conformation On the contrary,

network B reminds more that of the parallel xylan with G6 participating

in the H-bonding patterns: the two most common bonds being X2-G2 and X3-G6 These results show that a regular hydrogen bonding pattern is observed between each xylan-cellulose pair, although the type

of the pattern may vary depending on the xylan orientation It is inter-esting to note that the statistically more likely antiparallel orientation also leads to the more extensive hydrogen bond pattern, possibly indi-cating a physical mechanism for this preference However, the anti-parallel orientation is preferred also on the hydrophobic surfaces of the cellulose microfibril where no hydrogen bonds are formed, pointing to a more complex selection process

To investigate this further, the short-range non-bonded energy be-tween xylan and cellulose in the multiple repeats of S3 were extracted from the final 50 ns of the simulations and displayed as histograms based

on their respective orientation (Fig 6) This quantity is a measure of the proximity between xylan chains and the cellulose surface and should not

be interpreted as a binding energy The distributions for antiparallel chains are shifted towards lower energies compared to the parallel ones This indicates a more effective packing between the xylan and the cel-lulose in the former case In both cases the energy is more negative for chains adsorbed to hydrophilic surfaces compared to hydrophobic ones However, the fact that the binding has been shown to be consistently stronger to the hydrophobic surfaces (Martnez-Abad et al., 2017) shows the importance of other interaction terms, especially those mediated by water When the energy is further decomposed into contributions from Coulomb and dispersion (Lennard-Jones) interactions, one finds, not surprisingly, that the Coulomb energy dominates the short-range in-teractions at the hydrophilic surfaces, presumably a consequence of forming hydrogen bonds, while dispersion dominates at the hydropho-bic surfaces where no hydrogen bonds are formed The present analysis shows that an antiparallel orientation leads to tighter interactions be-tween xylan and cellulose, which tentatively could contribute to the statistical preference of that orientation over a parallel one However, since the difference arises in the already adsorbed state and no inter-conversion between orientations occurs (at least on MD timescales), the actual selection process is likely governed by more long-ranged in-teractions, such as the molecular dipole moments of the individual polymers

3.5 Adsorption of xylan as aggregates decreases the molecular mobility of adjacent glucose residues

Previously Terenzi et al (2015) have shown using 13C CP/MAS NMR

T1 relaxation times that the mobility of surface carbons in a hydrated CNF/xyloglucan (XG) composite were significantly lower compared to what would be expected from the weight averages of the individual relaxation times of the same carbons in pure CNF and XG They assigned this to a decrease in XG dynamics due to the restrictions imposed by the CNF interface, but also speculated whether the adsorbed XG could affect the dynamics of cellulose surface chains Here, such effects are

investi-gated for the case of adsorbed xylan T1 relaxation times can be calcu-lated from MD-simulations from the Fourier transform of the rotational autocorrelation functions of specific C–H bond vectors, yielding com-parable results with experimental ones (Chen et al., 2019) The present investigation concentrates on the dynamics of C4-H4 since the splitting

of the C4 peak in 13C NMR is commonly interpreted to arise from the rigid core and more mobile surface/amorphous regions, respectively, and to correlate with the hydroxymethyl conformation (Bardet, Emsley,

& Vincendon, 1997) To this end, systems S1 and S5 were simulated for

an additional 100 ns starting from the end of the 1 μs simulations From

these simulations T1 relaxation times were calculated as described above The results were compared to reference simulations performed of the same CNF models, but without xylan present For the analysis, the cellulose chains were divided into core chains, as well as hydrophobic and hydrophilic surfaces Individual residues of the hydrophilic surface

chains were further divided into hydrophilic (inner), having the

hydroxymethyl group pointing into the crystal, and hydrophilic (outer),

having the hydroxymethyl group exposed to the xylan and/or water

Average T1 values are presented in Table 2, where longer relaxation time

indicates more restricted molecular dynamics The upper part of Table 2

shows averages of all glucose residues of both the xylan-cellulose

ag-gregates and their respective references In general, reference values

agrees with those previously reported using the same force field (Chen

et al., 2019): C4 relaxation times in hydrophilic (outer) are about half of

that in hydrophilic (inner), and within 35–42 s Residues in hydrophobic

surfaces are more rigid compared to those in hydrophilic surfaces, and

overall relaxation times are 10–40 s higher in the xylan-cellulose

sys-tems compared to the references

In order to isolate possible effects from adsorbed xylan oligomers,

relaxation times in residues directly adjacent to adsorbed xylans (and

the corresponding residues of the reference) were calculated This shows

that in S1 the T1 values of hydrophilic (outer) in contact with xylans

in-crease from 30 to 121 s, compared to the same residues in the reference

system, clearly showing that adsorbed xylan has a direct effect on

cel-lulose mobility The corresponding increase in S5 is smaller; from 42 to

55 s This was actually quite surprising since one would think that the

longer xylan chains in S5 would bind more stably to the cellulose and

therefor would confine the motions of adjacent glucoses to a larger

extent However, while the longer xylans were indeed relatively

immobile on the cellulose surface, the shorter ones migrated to form

aggregates This suggests that a certain amount of continuous surface

coverage is required to convert the mobile surface glucose residues into

more core-like

Recently, it was proposed based on solid state NMR studies on native

Arabidopsis that xylan may induce a conformational change of the

cel-lulose hydroxymethyl groups from the gt and gg conformations

domi-nating in surfaces to tg, which is the dominant conformation in the

crystalline core, thus effectively make surface chains more crystal-like

(Dupree et al., 2015; Terrett et al., 2019) To investigate whether the

xylan adsorption affects the C6OH conformation, probability

distribu-tions of the torsion angle ω of surface residues in S1 and S5 (and

cor-responding references) were determined (Fig B.1-B.4) As expected,

hydrophilic (inner) groups display very narrow distributions of a single

conformation, mainly tg, while distributions for hydrophilic (outer)

groups exposed to water are bi- and trimodal, meaning that they rotate

between states more often For hydrophilic (outer) residues close to xylan,

there is no visible increase in tg, but rather a shift away from it to either

gt or gg On the other hand, when compared to the reference, the peak

pertaining to the most probable conformation is amplified, indicating

that the time spent in that conformation increases, i.e., the motion

around the C5-C6 bond slows down This is consistent with the

forma-tion of regular H-bonding network, although that network is different

from that in native cellulose Iβ However, from the perspective of

cel-lulose mobility this appears to be of no consequence since the response

in T1 relaxation times clearly shows that these chains become crystal-

like regardless of H-bonding pattern However, one cannot rule out

the possibility of the hydroxymethyls eventually converting to tg based

on these simulations alone It is possible that significantly longer

simulations are required to observe it

4 Discussion

Cellulose and hemicelluloses are the major structural components of the plant cell wall and their molecular level association affects the properties of the cell wall as whole Cellulose has a straight and flat structure due to the β-(1 → 4)-glycosidic bonds in two-fold conformation and equatorially oriented hydroxyls at positions one and four in the glucose β-pyranoside residue (Nishiyama, 2009) Xylans and gluco-mannans share these molecular features and can therefore adapt to the cellulose structure, but also to more helical conformations (Berglund

et al., 2016; Busse-Wicher et al., 2014; Mazeau et al., 2005) This is hardly a coincidence; one biological role of the non-cellulosic cell wall polysaccharides is to bind the cellulose fibrils together through a less crystalline and a more flexible phase, forming a composite material with distinct biomechanical properties (Berglund et al., 2020) An ability to adapt to different conformations seems to fulfill this function Indeed, the data in this work is in line with a view that xylans can form an extension of the cellulose crystal structure that can function as a tran-sition phase both between rigid crystalline microfibrils, and to more flexible polysaccharide phases and amorphous lignin While MD simu-lations can indicate what type of molecular organizations are possible,

to this day, it is not known with certainty how the cellulose and hemi-celluloses are deposited in the cell wall of a growing tree

Possible interactions between cellulose- and hemicellulose chains include hydrogen bonding, van der Waals dispersion forces and hydro-phobic forces, which partially can be understood as the increase in solvent entropy from excluding water molecules from the cellulose surface (Chandler, 2005) The role of hydrogen bonding has often been exaggerated in the field of cellulosics (Wohlert et al., 2021), but a recent MD-study by Kishani, Benselfelt, Wågberg, and Wohlert (2021) on the hemicellulose xyloglucan, suggested that the hydrophobic forces were central for adsorption on cellulose In addition, Nishiyama (2018)

showed that the cohesive energy of crystalline cellulose is dominated by dispersion interactions Yet, hydrogen bonds may still play an important role in the formation of xylan-cellulose complexes: after the adsorption, alignment and aggregation, a regular hydrogen bonding network can be formed, and once formed, the interactions may be strong enough to stabilize the chain conformation of xylan

Based on polarized FTIR, Stevanic and Salm´en (2009) and Olsson, Bjurhager, Gerber, Sundberg, and Salm´en (2011) have suggested that xylan is oriented parallel to the cellulose fibrils, while results presented here show a preference for the antiparallel orientation It is interesting if both cellulose and xylan exhibit the same tendency for parallel organi-sation despite antiparallel being energetically more stable The biosyn-thetic process seems to be the reason for the parallel oriented β-glucans within a cellulose crystal: the individual chains are built in the cellulose synthesizing complex (CSC) by sequentially adding new glucose units to the reducing end of the chains, which then immediately coalesce forming cellulose I (Cosgrove, 2014) However, antiparallel packing of polysaccharides does occur in nature as the currently accepted crystal unit cell of α-chitin consists of two antiparallel chains in a 2-fold conformation (Ogawa, Lee, Nishiyama, & Kim, 2016; Sikorski, Hori, & Wada, 2009) For xylan, more studies are needed on its biological as-sembly in the secondary cell wall to elucidate if there is a preferred orientation, how it is controlled and whether it is similar throughout the cell wall layers and the cell types

One thing that seems increasingly certain, based on solid state NMR- studies (Dupree et al., 2015; Terrett et al., 2019) and the simulations presented herein, is that the close association of 2-fold xylan to cellulose microfibrils restricts the mobility of the surface polymers when quan-tified by 13C CP/MAS NMR T1 relaxation times One possible reason for this could be the exclusion of water from the cellulose surface upon adsorption of a xylan aggregates, which causes a local decrease in the mobility of cellulose, but the crowded environment imposed by the

Table 2

13C NMR T1-relaxation times, logarithmic means in seconds

S1 Reference CNF S5 Reference CNF All Glc

Glc next to Xylan

Hydrophilic (inner) 135 132 116 93