Molecular dynamics studies on the thermostability of family 11 xylanases

Molecular dynamics studies on the thermostabilityof family 11 xylanases

Mikko Purmonen1, Jarkko Valjakka2,3,Kristiina Takkinen4, Tuomo Laitinen5

and Juha Rouvinen1,6

1Department of Chemistry, University of Joensuu, PO Box 111, 80101Joensuu, Finland, 2Institute of Medical Technology, University of Tampere,33014 Tampere, Finland, 3Tampere University Hospital, Biokatu 6-8, 33520Tampere, Finland, 4VTT Technical Research Centre of Finland, PO Box1000, 02044 VTT, Finland and 5Department of Pharmaceutical Chemistry,University of Kuopio, PO Box 1627, 70211 Kuopio, Finland

6To whom correspondence should be addressed.E-mail: [email protected]

Twelve members of the family 11 xylanases, includingboth mesophilic and thermophilic proteins, were studiedusing molecular dynamics (MD). Simulations of xylanaseswere carried out in an explicit water environment at fourdifferent temperatures, 300, 400, 500 and 600 K. A differ-ence in thermotolerance between mesophilic and thermo-philic xylanases became clear: thermophilic xylanasesendured heat in higher simulation temperatures betterthan mesophilic ones. The unfolding pathways seemed tobe similar for all simulations regardless of the protein.The unfolding initiates at the N-terminal region or alter-natively from the a-helix region and proceeds to the‘finger region’. Unfolding of these regions led to denatu-rated structures within the 4.5 ns simulation at 600 K.The results are in agreement with experimental mutantstudies. The results show clearly that the stability of theprotein is not evenly distributed over the whole structure.The MD analysis suggests regions in the protein structurewhich are more unstable and thus potential targets formutation experiments to improve thermostability.Keywords: denaturation/molecular dynamics/thermostability/unfolding pathway/xylanase family 11

Introduction

The use of many enzymes in industrial applications requireshigh thermostability. Thermostable enzymes have beencreated from mesophilic enzymes by using site-directed orrandom mutagenesis. One alternative is to use natively ther-mophilic enzymes from corresponding organisms. Whereasmesophilic organisms live near room temperature,‘heat-loving’ thermophilic organisms live in considerablyhigher temperatures (optimal growth 50–808C) andhyperthermophilic organisms in even more extreme tempera-tures (optimal growth .808C) (Collins et al., 2005).Hyperthermophilic organisms are usually isolated from hotsprings and other geothermally heated sources (Rees andAdams, 1995). The structures of natively thermophilic pro-teins have shed light on these factors that contribute to thethermostability of proteins.

With molecular dynamics (MD), it is possible to studycases that are difficult to investigate experimentally, e.g.protein folding and unfolding. MD offers a detailed

description of thermal motion at different temperatures andby following the conformational changes as a function ofsmall time steps, it is possible to study, for example, unfold-ing or denaturation (irreversible unfolding) pathways of pro-teins. MD has been used a few times earlier to study proteinunfolding. Chymotrypsin inhibitor 2 (CI2) was simulated atseven temperatures, ranging from 290 to 498 K and from 94to 20 ns. It was proposed that an increase in temperature canaccelerate protein unfolding without changing the unfoldingpathway of proteins (Day et al., 2002). Para-nitrobenzyl ester-ase was simulated at 300 K until 500 ps using MD (Wintrodeet al., 2003). The data suggested that adaption for high temp-erature stability has resulted in a structure in which largerdeviations from the native state are restricted and the confor-mational flexibility of thermophilic enzymes at room temp-erature is reduced, in comparison with the mesophilic ones.MD simulations in solution have also been performed for arubredoxin from a hyperthermophilic archaeon Pyrococcusfuriosus (RdPf) and from a mesophilic organismDesulfovibrio vulgaris (RdDv) (Lazardis et al., 1997).Temperatures of 300, 375, 473 and 500 K were used andsimulations extended from 0.2 to 1.0 ns. Although theunfolding behavior of RdPf and RdDy seemed to be verysimilar, simulation pathways differed slightly from eachother. MD simulations have also been used as a tool toimprove the protein stability of haloalkane dehalogenase(DhlA), by identifying flexible regions from the structure(Pikkemaat et al., 2002). These results showed that it waspossible to use MD simulations as a tool to identify flexibleprotein domains that can be used as a target for stabilityenhancement by introducing a disulfide bond. MD simu-lations at 300 K, 400 K, 600 K, and 800 K have also beenused to study the thermostability of glucoamylase fromAspergillus awamori. An improvement in thermostabilitywas achieved by introducing disulfide bond to lock one ofthe a-helices on the surface of the catalytic domain (Liu andWang, 2003). In addition, some point mutations were testedfor the same protein by using MD.

Xylanases have applications in food, feed and technicalindustries. For example, family 11 endo-b-xylanases arewidely used in the pulp industry, e.g. in the biobleaching ofcraft pulp. Xylanases are also used in the baking industry toimprove the handling of dough and the quality of the finalbaked product (Beg et al., 2001). The addition of xylanase tothe rye-based diet of broiler chickens improved weight gain,feed intake and feed efficiency, and decreased water con-sumption (Mathlouthi et al., 2002).

We have previously determined several crystal structuresof small-molecular weight (family 11) endo-1,4-b-xylanases(EC 3.2.1.8). (Torronen et al., 1994; Torronen and Rouvinen,1995; Hakulinen et al., 2003). Endo-1,4-xylanases break upxylan, a major hemicelluloce polysaccharide in plant cellwalls, into xylo-oligosaccharides with variable lengths.Xylans are heteropolysaccharides consisting of a b-D-xylopyranosyl backbone with variable side chains. Backbone

# The Author 2007. Published by Oxford University Press. All rights reserved.

For Permissions, please e-mail: [email protected]

551

Protein Engineering, Design & Selection vol. 20 no. 11 pp. 551–559, 2007Published online October 30, 2007 doi:10.1093/protein/gzm056

by guest on June 1, 2013http://peds.oxfordjournals.org/

Dow

nloaded from

http://peds.oxfordjournals.org/

substituents depend on the source of the polymer, e.g. xylanreceived from softwood and hardwood differs significantlyfrom each other. The most common substituents are a-linked4-O-methyl glucuronic acid on C2, a-linked arabinose on C2or C3, and acetyl esters on C2 or C3 of some xylose residues(Brett and Waldron, 1996). Xylanases are created from ana-helix and two b-sheets, A and B, forming a so-calledb-sandwich structure (Fig. 1). The shape of xylanasesresembles a right hand, where b-sheet A is composed of fiveantiparallel b-strands from A1 to A6 (sometimes the A1strand is lacking among the protein family 11 xylanases andthere is only a loop region, e.g. TRX II in Fig. 1) andb-sheet B contains b-strands B1–B9. The commonly usednames for different parts of the xylanases are ‘fingers’,‘palm’, ‘thumb’ and ‘cord’. Fingers are formed from A andB b-sheets, the palm is made up of an a-helix and a twistedb-sheet B, the thumb is a loop between strands B7 and B8,partly closing the cleft, and the cord is a long loop betweenB6 and B9 partly closing the side of the cleft (Torronenet al., 1994). This family of xylanases consists of �190–200amino acids and has a diameter of 35 A (12). Xylanases’dimensions vary depending on their amount of amino acidresidues, packing and the number and size of cavities. In astudy by Hakulinen and co-workers, factors that affect the

thermostability of family 11 endo-1,4-b-xylanases were rep-resented by analyzing similarities and differences both at theamino acid sequence and three-dimensional structure level(Hakulinen et al., 2003). Several minor modificationsappeared to be responsible for increased thermostability,such as a higher Thr/Ser ratio, an increased number ofcharged residues and a higher number of residues in theb-strands and stabilization of the a-helix region.

In this study, MD have been used to study the unfoldingof the family 11 xylanases. The simulations were carried outat different temperatures, 300, 400, 500 and 600 K.High-temperature simulations eventually led to the break-down of the tertiary structure of protein. An analysis of theresults will give an estimation of the initial steps of denatura-tion in this protein family. We conducted MD simulationsfor 12 members of family 11 xylanases of which three-dimensional structures are available in the Protein Data Bank(Berman et al., 2000). The studied xylanases were fromTrichoderma reesei (I and II, TRX I and TRX II) (Torronenet al., 1994; Torronen and Rouvinen, 1995), Bacillus circu-lans (BCX) (Wakarchuk et al., 1997), Aspergillus kawachii(AKX) (Fushinobu et al., 1998), Aspergillus niger (ANX)(Krengel and Dijkstra, 1996), Bacillus agaradhaerens (BAX)(Sabini et al., 1999), Trichoderma harzianum (THX)(Campbell et al., 1993), Chaetomium thermophilum (CTX)(Hakulinen et al., 2003), Dictyoglomus thermophilum (DTX)(McCarthy et al., 2000), Thermomyces lanuginosus (TLX)(Gruber et al., 1998), Paecilomyces Varioti Bainier (PVX)(Kumar et al., 2000) and Nonomuraea flexuosa (NFX)(Hakulinen et al., 2003). A list of the studied xylanases, theirPDB-codes and optimum growth temperatures are shown inTable I. A sequence comparison of the studied family 11xylanases is shown in Fig. 2. These enzymes can be dividedinto mesophilic (TRX II, BCX, THX, TRX I, AKX, ANXand BAX) and thermophilic (NFX, CTX, DTX, TLX andPVX) enzymes. Whereas BAX, DTX, BCX and NFX werefrom bacterial sources, the other studied xylanases were froma fungal source (Hakulinen et al., 2003).

Methods

The MD simulations were carried out using the Amber 7program (Assisted Model Building with Energy Refinement)(Pearlman et al., 1995; Case et al., 2005) installed on ourDell Precision computers operating under Linux withInter(R) Xeon(TM) 2.40 GHz processors. Atom interactionswere described with Amber ff99 (Wang et al., 2000). First,before starting MD, all bad contacts need to be removed,exclusively minimizing water molecules and the wholesystem, containing protein and water, without SHAKE(Ryckaert et al., 1977). In addition to minimization (0.3 ns),the system was also heated and brought into equilibriumbefore MD simulation (15 ps). Simulations were performedusing a canonical NVT ensemble. A constant temperature forthe simulated system was implemented with a weak-couplingalgorithm, with which the system was coupled to an externalheat bath that was fixed at the desired temperature using adefault time constant (tT ¼ 1.0 ps) for the heat bath couplingfor the system (Berendsen et al., 1984). The solvent aroundthe proteins was described with explicit water moleculesconsisting of TIP3P water molecules (Jorgensen et al., 1983).

Fig. 1. (A) Topology diagram and (B) a schematic 3D structure of thefamily 11 xylanases. Family 11 xylanases are constructed with two b-sheetsand an a-helix forming a so-called b-sandwich structure. The different partsof the enzyme are named after the right hand (Torronen et al., 1994), thuscontaining regions named fingers, thumb, palm, cord and a-helix.

M.Purmonen et al.

552


Dow

nloaded from


The resultant structure was a protein surrounded with watermolecules. The range of the truncated octahedron solventbuffer was 10 A from the surface of the protein, so thetotal amount of the water molecules varied from 4278 to5494 molecules, depending on the area and shape of themolecule surface. Sodium and chloride ions were used ascounter-ions in order to uncharge the molecules. Thenumber of counter-ions varied from 0 to 14 among thexylanases. A cut-off distance of 8 A was used to carryoutlong-range electrostatics with Particle Mesh Ewald (PME)and for van der Waals forces. Simulations were conductedin periodic boundary conditions (PBC) using SHAKE, inwhich bonds involving hydrogen atoms were constrained.Simulations were carried out for 4.5 ns including 3 000 000cycles. The magnitude of a time step was 1.5 fs. Every500th step was saved to the trajectory, resulting in a set of6000 coordinates for the analysis of the unfolding pathway.The results were analyzed using an Amber module, ptraj.In addition, programs MOIL-View (Simmerling et al.,1995), VMD and Microsoft Excel were used to analyzetrajectories. Each simulation required �1 month of CPUtime, which corresponds a 4.5 ns computational model ofunfolding. Other programs used in this study wereClustalW 1.8.1, BOXSHADE 3.2.1, PyMOL 0.97 (DeLanoScientific) and HBPLUS, below described.

Density of the system was calculated to be �1.04 g/cm3 atall temperatures (300, 400, 500 and 600 K). The HBPLUSprogram uses the same minimum angles and maximumdistances as Baker and Hubbard (Baker and Hubbard, 1984).Minimum angles between D-H-A, H-A-AA and D-A-AAare 90.08, maximum angles for amino-aromatic interactionsD-A-AX and H-A-AX are 20.08 and maximum distances forD-A, H-A, and S-S are 3.9, 2.5 and 3.0 A, respectively.

Results

The simulations at different temperaturesThe denaturation of protein typically occurs on a micro-second time scale (Duan and Kollman, 1998). Unfortunately,it was not possible to perform MD simulations within thetypical time period in which proteins normally denaturate.Therefore, shorter simulation times are needed, and it isnecessary to increase the temperature to detect the denatura-tion of protein. Although the drawback is that the system ismore artificial, it is probably more realistic at the beginningof the simulation, so it may reveal details of the initialunfolding process which might be the most interestingwhen one considers the use of mutagenesis in improvingthermostability.

In total, 48 simulations were made for 12 xylanases at300, 400, 500 and 600 K. The differences between nativestructures and simulated structures were investigated withCA RMS deviations. RMS deviation is a numerical measureof the difference between two structures. An example of theeffect of higher temperatures is shown in Fig. 3, wherehigher temperatures inflict a more rapid denaturation. As canbe seen in Fig. 3, the RMS deviation values of mesophilicTRX II are considerably higher than the corresponding RMSdeviation values of DTX at 600 K. In addition, a notableobservation is that the RMS deviation values of DTX at 300and 400 K do not deviate much from each other. The sameobservation was found also for other thermophilic xylanases.

The effect of increased temperature to the protein structureA more comprehensive picture was obtained from the simu-lation when the RMS deviation values were presented as afunction of the amino acid residue. Figure 4 shows the

Table I. Studied family 11 xylanases and their properties

Organism Symbol PDB code Number of amino acid Optimum growth (8C) pH optimum References

ThermophilicDictyoglomus thermophilum DTX 1F5J 199 75 6.5 Sapag et al. (2002)

85 6.0–7.0 Xiong et al. (2004)Nonomuraea flexuosa NFX 1M4W 197 70–80 Haulinen et al. (2003)

80–85 Xiong et al. (2004)Thermomyces lanuginosus TLX 1YNA 194 60–70 6.5 Xiong et al. (2004)

65 7.0 Sapag et al. (2002)70 Damaso et al. (2003)75 Madlala et al. (2001)

Paecilomyces varioti PVX 1PVX 194 50 4.0 Beg et al. (2001)65 5.5–7.0 Sapag et al. (2002)

Chaetomium thermophilum CTX 1H1A 191 70–75 Hakulinen et al. (2003)Mesophilic

Bacillus agaradhaerens BAX 1QH7 207 4.0-6.0Trichoderma reesei II TRX II 1ENX 190 40–45 4.0–4.5 Beg et al. (2001)

4.5–5.5 Kumar and Nussinov (2001)45

Trichoderma harzianum THX 1XND 190 50 5.0 Beg et al. (2001)Bacillus circulans BCX 1XNB 185 45 6.0–7.0 Srinivasan and Rerle (1999)Aspergillus kawachii AKX 1BK1 182 50 2.0 Christov et al. (1999)

60 3.0 Sapag et al. (2002)Aspergillus niger ANX 1UKR 181 45 3.0 Beg et al. (2001)

50 4.0 De Vries and Visser (2001)Trichoderma reesei I TRX I 1XYN 178 40–45 3.0–3.5 Beg et al. (2001)

24 h at 508C 3.5–4.0 Lappalainen et al. (2003)

Thermophilic and mesophilic xylanases are divided into separate categories. Enzymes can be found from the Protein Data Bank using the PDB codesmentioned above. Optimum growth temperatures have been found from the literature according to references.

Thermostability of family 11 xylanases

553


Dow

nloaded from


average RMS deviation values for TRX II. This presentationclearly highlights some areas in which differences from thenative structure are large. These areas are the ‘thumb’,a-helix, ‘cord’ and finger/N-terminal areas. One also see thatsimulations at 300, 400 and 500 K are quite similar, but the600 K simulation shows markedly larger changes.

An even more complete view can be obtained if theRMSD values are presented as a function of time and residuenumber simultaneously. This is shown in Fig. 5, whereRMSD plots of TRX II simulations are represented at fourdifferent temperatures, 300, 400, 500 and 600 K. On thex-axis there is an amino acid sequence, on the y-axis time

and on the z-axis RMS deviation values. If RMS deviationsfluctuate substantially in certain regions, it can indicate thoseareas where there are reversible conformational changes. IfRMS deviations rise but do not descend back to the initiallevel, this can indicate an irreversible change in the structureand lead to denaturation.

At 300 K, as expected, either mesophilic or thermophilicxylanases did not show any signs of denaturation at the endpoint of the simulation (4.5 ns); the structures were stableand remained close to the crystal structure during the wholesimulation. At 400 K, there was an increase in structuralchanges, but the structures did not yet show significant

Fig. 2. Amino acid sequence comparison of studied family 11 xylanases.

M.Purmonen et al.

554


Dow

nloaded from


unfolding yet. At 500 K, local unfolding occurred mainly inthe mesophilic but to some extent also in the thermophilicxylanases and at 600 K, we observed larger unfolding whichled to comprehensive denaturation, especially in mesophilicxylanases. Consequently, on a 4 ns time scale only the use of600 K causes denaturation of the protein structure and offersa sketch of the denaturation pathway.

The influence of the initial structureThe initial three-dimensional structures of the studied proteinfamily 11 xylanases were obtained from the Protein DataBank. However, protein molecules may exist natively indifferent conformations. For example, protein crystal struc-tures often contain more than one molecule in the asym-metric unit which may have different conformations. Westudied the influence of the initial structures by performingsimulations for TRX II and DTX with the A and B mol-ecules of the asymmetric unit at 600 K. Although the struc-tures are similar (RMSD 0.1 A), there were differences in thesimulations. In TRX II, molecule B unfolded slightly fasterand ended the structure in which the average RMSD was�2 A higher. On the other hand, the unfolding pathwayswere similar regardless of the choice of the molecule. Wealso modified the structures of TRX II with the programMoleman (Kleywegt et al., 2001), which enables the randomshifting of CA atoms. The use of an average 0.1 A change inthe structure did not significantly change the unfolding beha-vior. An average 0.2 A change, however, caused seriousstructural problems in the simulations indicating that thechanges in the structure were too large. The influence ofdifferent minimization times (0.3, 0.6 and 0.9 ns) was alsostudied for TRX II and DTX at 300 K. Again, RMSD valuesbetween different simulations varied between 1 and 2 A.

Analysis of hydrogen bondsIntramolecular hydrogen bonding is known to improve thethermostability of proteins and to increase fractional polarsurfaces (Vogt et al., 1997). Intramolecular hydrogen bondswere calculated from pdb-structures with the HBPLUSprogram with default parameters at different time spots(McDonald and Thornton, 1994). Generally, thermophilicxylanases of family 11 have more hydrogen bonds andslightly more hydrogen bonds per residue in their initialstructures than mesophilic ones. With mesophile proteins thenumber of hydrogen bonds vary from 157 to 214 correspond-ing to 0.88–1.03 hydrogen bonds per residue, whereasamong thermophiles it is from 184 to 205, with a correspon-dence of 0.96–1.03. The number of intramolecular hydrogenbonds decreased during simulations at 600 K as the RMSdeviations increased. At the beginning of the simulations,there is a rapid 14–26% decrease in the number of hydrogenbonds. Later, the decrease in the number of hydrogen bondsis smaller. The differences in the number of hydrogen bondsbetween mesophilic and thermophilic xylanases during simu-lations are small and within error limits of the simulations.

Discussion

The influence of the initial structureThe influence of the initial structure on the MD simulationresults has not been extensively studied. However, some pre-vious papers have shown that the differences in initial con-formation can lead to quite different results in simulations.In the case of acetyl-CoA dehydrogenase, the key residuewas in bad conformation, which make it difficult to investi-gate the reaction mechanism theoretically. In the secondcase, it was concluded that different initial structures ofxylose isomerase from S. olivochromogenes affect the simu-lation results of the two Mg2þ cofactors, substrate and

Fig. 3. CA RMS deviation values of (A) TRX II and (B) DTX at differenttemperatures as a function of time. Black, green, violet and red colors areused to represent RMS deviation values obtained from 300, 400, 500 and600 K simulations, respectively. The duration of each simulation is 4.5 ns.

Fig. 4. Average RMS deviation values between the initial structure and theaverage of the simulated structure of TRX II as a function of amino acidsequence numbers. Values were calculated with the use of CA atoms. Black,green, violet and red lines represent the simulations at 300, 400, 500 and600 K, respectively. From the picture, a more widespread movement ofdifferent regions at higher temperatures can be seen. b-Sheets and thefingers, cord, thumb and helix are marked on the plot.


555


Dow

nloaded from


enzymatic residues during hydride transfer (Garcia-Vilocaet al., 2004). We tested the effect of similar initial structuresusing A and B molecules of TRX II and DTX alike and usedstructures distorted by the program called Moleman. The sig-nificance of the initial structure for the simulation is slight.The differences in the total energy are due to the differentnumbers of water molecules in the simulation. In any case,care should be taken when defining the initial coordinates forsimulations. In the Protein Data Bank, there are several struc-tures in which part of the protein structure, e.g. some loops,can be disordered or completely missing in the model.Therefore, it is much more challenging to perform simu-lations for such models.

Time scale and the effect of the simulation temperatureAt present, nanosecond simulations are routine but they arestill too short to observe events in protein motion that occuron the microsecond to second time scale (Daggett, 2000).Because of a lack of sufficient computer power, we wereforced to use higher simulation temperatures to accelerate theprotein unfolding. Previous studies have suggested that dena-turation, using higher temperatures, happens without

changing the unfolding pathway (Daggett et al., 2002). Dauraand co-workers have also shown that protein folding andunfolding are independent of the temperature and the mostcommon intermediates are the same at different temperatures(Daura et al., 1999). The use of high temperatures allowsreal-time MD unfolding simulations to be tested in somecases. For example, the engrailed homeodomain protein hasvery-high refolding and folding rate constants and the half-life of unfolding has been extrapolated to be �7.5 ns at1008C, which enables real-time MD simulations to be con-ducted (Mayor et al., 2000). For testing, we have also madesimulations in the longer time scale (9 ns) at 400 K for TRXII (mesophilic) and DTX (thermophilic). The RMSD plotshows that DTX protein stays in a folded state but RMSDvalues continue to increase for TRX II indicating that in thefuture with better computing power, it might be possible touse lower (more realistic) temperature and longer simulationtime to study thermostability.

MD simulations of the studied family 11 xylanases wereperformed at four different temperatures, 300, 400, 500 and600 K. By increasing the simulation temperature, it allows usthe possibility to observe many interesting events, e.g.

Fig. 5. RMSD plots of TRX II at different temperatures including (A) 300, (B) 400, (C) 500 and (D) 600 K. The x-axis represents amino acid sequencenumber, the y-axis time (trajectory number). Information used to construct this plot was saved every 500 time-steps, so 6000 trajectories correspond to 4.5 ns,and the z-axis represents the amount of movement (A). White, grey, yellow, red and black represent the amount of movement from 0 to 3, 3 to 6, 6 to 9, 9 to12 and over 12 A, respectively. Values have been calculated with the use of CA atoms.

M.Purmonen et al.

556


Dow

nloaded from


protein unfolding on a much shorter time scale. The use of asignificantly high temperature (600 K) is well foundedbecause the denaturation of some structures can be obtainedin 4.5 ns. The 300 K simulation can be used to representnative thermal fluctuation. The 400 K simulations do notgreatly differ from the native thermal fluctuation duringsimulation time. Instead, the differences for 400 and 500 Ksimulations are easier to notice simply because at 500 Ksimulations local denaturations happen. The highest simu-lation temperature (600 K) represents highly vigorous cir-cumstances in which global denaturation happens within4.5 ns. When comparing simulation results calculated atdifferent temperatures, we can see that structural changesoccur in the same regions regardless of the temperature. Theonly difference is that the structural changes have proceededthe furthest at 600 K. Therefore, the denaturation pathway offamily 11 xylanases seems to be highly similar regardless ofthe used simulation temperature and different initial struc-tures of xylanases.

Structural changes at high temperatures in different regionsStructural changes during the simulations were screened byplotting RMSD values as a function of amino acid residuesand time, calculating intramolecular hydrogen bonds, andvisualization, by using the VMD program (Humphrey et al.,1996). The unfolding of mesophilic TRX I and thermophilicNFX xylanases are shown in Fig. 6 as snapshots. In the fol-lowing, we describe in detail general structural changes indifferent regions of xylanases.

The N-terminal end is a region that is constructed fromloops of variable length and A1 (not in all structures), A2,A3, B1, B2 and B3 b-strands. Irreversible structural changeof the N-terminal can lead to a comprehensive denaturation.Denaturation can happen when b-strands are gradually lost,starting from the N-terminus (ANX and TRX I) and the

structural changes can proceed to the whole finger area(CTX). In some cases, even though the RMSD values of thefinger area are large, an increase in the movement of theN-terminal does not lead to protein denaturation (NFX, simu-lation at 500 K). As we consider the importance of theN-terminal and finger of xylanase family 11, one should notethat thermophilic xylanases have generally longer N-terminalregions than mesophilic ones. Mesophilic xylanases oftenhave a long, structurally irregular loop in the N-terminal,whereas thermophlic xylanases commonly have a b-strand inthe same position. This may explain larger RMSD values formesophilic xylanases in the N-terminal. There is also exper-imental evidence that the N-terminal affects the thermostabil-ity of protein family 11 xylanases. TRX II mutants DS1 andDS5 have been stabilized with a C110–C154 disulfidebridge and other site-directed mutations, whereas the highlythermostable DB1 mutant was a combination of the DS5mutant improved with the N-terminal C2–C28 disulfidebridge (Turunen et al., 2001; Janis et al., 2004; Xiong et al.,2004). The importance of the N-terminal has also beenshown in the study of Sun et al., where the thermostabilityand catalytic activity of A. niger xylanase A (AnxA) wasimproved by an N-terminus substitution with the correspond-ing region of Thermomonospora fusca xylanase A (TfxA)(Sun et al., 2005). Fenel et al. have also stabilized the struc-ture of TRX II by engineering a disulfide bridge into theN-terminal (Fenel et al., 2004). A combination of weakly sta-bilizing mutations with a disulfide bridge in the a-helixstructure of TRX II have been studied and produced exper-imentally; several combinations of mutations substantiallyincreased thermostability in comparison with the wild typeTRX II (Turunen et al., 2001).

The finger region is a stable region that is made up ofseveral b-strands and b-turns, including A4, A5 and A6strands, which are connected with loops to B4, B5 and B6strands. The finger region forms a central part of the hydro-phobic core of the protein. Consequently, the loss of itsstructure is important for the denaturation of protein family11 xylanases. If the b-sheet structure is lost, the wholeprotein structure is threatened by comprehensive collapse(MD simulations of CTX, BAX, BCX, ANX and TRX I at600 K). If the structure in the finger region stayed native-like,denaturation of the protein structure did not occur (4.5 nsMD simulations of DTX, NFX and TLX at 600 K). Thedenaturation of the finger region can proceed gradually fromthe N-terminal end (MD simulations of ANX and TRX I at600 K) or the whole finger region can lose the structure atonce (CTX).

An a-helix is situated at the back of the xylanases behindthe palm. Although this helix structure is very stable in thesimulations, its C-terminal end is able to fluctuate.Experimental mutations have proven the importance of thisregion to the thermostability of protein family 11 xylanases(Turunen et al., 2001; Janis et al., 2004; Xiong et al., 2004),and xylanases in general (Wakarchuk et al., 1997; Georiset al., 2000). However, the unfolding does not in most of thecases begin from here, according to the simulations. Themost well known modification in this region is the C110–C154 disulfide bridge which significantly stabilizes the struc-ture. TLX and PVX have a native C110–C154 disulfidebridge which may have a significant role in addition with astable N-terminal to in preventing the unfolding.

Fig. 6. Unfolding simulations of (A) mesophilic TRX I and (B) thermophilicNFX performed at 600 K during 4.5 ns.


557


Dow

nloaded from


The cord is a flexible loop connecting B6 and B9 strands.The cord is placed on the outer surface of the protein and isconnected to the movements of the thumb, fingers anda-helix. In the case of the CTX cord area, the ‘cord loop’,the B9 strand, the N-terminal of B8 and the loop connectingB9 and B8 areas all seem to especially affect denaturation.

The thumb is a short loop, which is the most fluctuatingregion of xylanase family 11, placed between B7 and B8strands. The movement of the loop affects the movement ofB8 and B7 strands. Mobility is probably essential for thefunction of the enzyme (Muilu et al., 1998). The fluctuationof the thumb region seems to be widespread in family 11xylanases. However, the fluctuation of the thumb does notaffect the rest of the structure, which remains stable. This isseen in the 600 K simulations of thermostable xylanases, ase.g. in DTX.

The palm is the part of the hydrophobic core and the moststable region in the protein family 11 xylanases. It consistsof a b-sheet B containing b-strands B4–B9 coiled into aform of a palm. The region consisting of the longestb-strands B6 and B7 is especially stable. In several MDsimulations, the palm region was the last region to maintainthe native structure, e.g. CTX, BCX and TRX II. In xylanasefamily 11, catalytic amino acids are situated in the palm cleft(Torronen et al., 1994).

Structural factors affecting thermostabilityUnfortunately, similar thermostability parameters are notavailable for all the studied xylanases. The only experimentalparameter that can be used to evaluate relative thermotoler-ance is the optimum growth temperature of an organism,which is available for all cases except for BAX. Correlationbetween the average RMSD values and temperatures is pre-sented in Fig. 7. The RMSD values represent average valuesduring the whole 600 K simulation. There is a correlationbetween the RMSD values and growth temperature.However, it must be kept in mind that because RMSD valuesbetween different test simulations varied between 1 and 2 A,this comparison should be considered only as an indicativeone. Most of the thermophilic enzymes have lower RMSdeviation values than the mesophilic ones, which correspondto the experimental optimum temperature data. PVX, DTX,NFX and TLX are the most thermostable structures

according to the MD simulations. AKX and molecule Afrom TRX II with the immediate presence of CTX and BCXfollow the most thermostable structures, leaving THX, ANXand TRX I to be the least thermotolerant structures. BAX isnot included in the dissection because of a lack of exper-imental data, but it would be placed on the thermophilic sideof the segment of the line rather than on mesophilic side.

Crystal analysis, sequence comparison and mutagenesisresearch shows that mesophilic and thermophilic xylanasesof family 11 are very similar (Fig. 2). Comparison of theamino acid sequence was made with programs BOXSHADE3.2.1 written by K. Hofmann and M. Baron, and ClustalW1.8.1. (Higgins et al., 1994). As a consequence of the simi-larity between mesophilic and thermophilic xylanases, it islikely that the increase in thermostability derives fromseveral small modifications. Although factors that affect ther-mostability have been studied for a long time, none of thestudies have been able to offer a prioritized list of all factors.

The number of hydrogen bonds is known factor that mayaffect the thermostability of proteins. It was found that ther-mophilic xylanases typically contain more hydrogen bondsnatively, but the differences in the amounts of hydrogenbonds during simulation is difficult to estimate because ofthe error in simulation. The simulations pointed out theimportance of the N-terminal for the initiation of unfolding.All the thermophilic enzymes (NFX, CTX, DTX, TLX andPVX) have the B1 b-strand containing five residues, whereasamong the mesophiles, only BAX, TRX II and THX havethe B1 strand containing five, five and four amino acids,respectively. In addition, thermophilic enzymes have chargedresidues at the N-terminal. NFX has acidic Asp12 and CTXbasic His11 in the B1 strand. TLX and PVX contain bothacidic (Glu7) and basic (His10) amino acids at the B1 strand,whereas the only thermophilic enzyme that does not haveacidic/basic residues in the B1 strand is DTX. BAX is theunique enzyme from the mesophilic xylanases, containingone basic (His11) amino acid. Another interesting obser-vation is that TLX and PVX have a significant conservedsequence including B1 and B2 strands; only Gly1 of PVX isreplaced with PCA1 compared with the TLX structure. BothTLX and PVX show outstanding thermotolerance and main-tain their common xylanase shape during the simulations. Inaddition, the native C110–C154 disulfide bridge stabilizesboth structures.

In this study, we used MD simulations for the first time tostudy the unfolding and denaturation of a large enzymefamily. Because a large number of proteins, five thermophilicand seven mesophilic enzymes were studied, we can try tosketch a general view of the structural factors which contrib-ute to the unfolding of family 11 xylanases. The mostinteresting finding is the importance of the N-terminal anda-helix for the initiation of the unfolding process. Thissupports previously published experimental data whichdiscussed the stabilizing effect of a disulfide bridge mutatedin the N-terminal of family 11 xylanases (Turunen et al.,2001; Janis et al., 2004; Xiong et al., 2004). Simulations alsorevealed that the palm area is the most resistant againstunfolding. Interestingly, this is also the area in which cataly-tically important residues are situated.

The MD simulations used in this study may offer analternative strategy for the study of factors that affect thethermostability of proteins. MD simulations can offer a

Fig. 7. Temperature optimums as a function of average RMS deviationvalues collected during the 4.5 ns simulation at 600 K. Temperatureoptimum values were obtained from the literature referred to Table I.

M.Purmonen et al.

558


Dow

nloaded from


general picture of the first steps of unfolding and denatura-tion. It might be very helpful to identify the parts of theprotein structure in which unfolding starts. Thus, mutagen-esis to improve thermostability could be focused on theseareas.

Funding

Finnish National Technology Agency (TEKES).

ReferencesBaker,E.N. and Hubbard,R.E. (1984) Prog. Biophys. Mol. Biol., 44, 97–179.Beg,Q.K., Kapoor,M., Mahajan,L. and Hoondal,G.S. (2001) Appl.

Microbiol. Biotechnol., 56, 326–338.Berendsen,H.J.C., Postma,J.P.M., van Gunsteren,W.F., DiNola,A. and

Haak,J.R. (1984) J. Chem. Phys., 81, 3684–3690.Berman,H.M., Westbrook,J., Feng,Z., Gilliland,G., Bhat,T.N., Weissig,H.,

Shindyalov,I.N. and Bourne,P.E. (2000) Nucleic Acids Res., 28, 235–242.Brett,C.T. and Waldron,K.W. (1996) Physiology and Biochemistry of Plant

Cell Walls, 2nd edn. Chapman & Hall, Cambridge, pp. 32–33.Campbell,R.L., Rose,D.R., Wakarchuk,W.W., To,R.J., Sung,Z. and

Yagushi,M. (1993) In Suominen,P. and Reinikainen,T. (eds), TrichodermaReesei Cellulases and other Hydrolases. Foundation for biotechnologicaland industrial fermentation research, Espoo.

Case,D.A., Cheatham,T.E., III, Darden,T., Gohlke,H., Luo,R., Merz,K.M., Jr,Onufriev,A., Simmerling,C., Wang,B. and Woods,R. (2005) J. Comp.Chem., 26, 1668–1688.

Christov,L.P., Szakacs,G. and Balakrishman,H. (1999) Process Biochem.,34, 511–517.

Collins,T., Gerday,C. and Feller,G. (2005) FEMS Microbiol. Rev., 29, 3–23.Daggett,V. (2000) Curr. Opin. Struct. Biol., 10, 160–164.Daggett,V., Day,R., Bennion,B.J. and Ham,S. (2002) J. Mol. Biol., 322,

189–203.Damaso,M.C., Almeida,M.S., Kurtenbach,E., Martins,O.B., Pereira,N., Jr,

Andrade,C.M.M.C. and Albano,R.M. (2003) Amer. Soc. Microbiol., 69,6064–6072.

Daura,X., van Gunsteren,W. and Mark,A.E. (1999) Proteins, 34, 269–280.Day,R., Bennion,B.J., Ham,S. and Daggett,V. (2002) J. Mol. Biol., 322,

189–203.De Vries,R. and Visser,J. (2001) Microbiol. Mol. Biol. Rev., 65, 497–522.Duan,Y. and Kollman,P.A. (1998) Science, 282, 740–744.Fenel,F., Leisola,M., Janis,J. and Turunen,O. (2004) J. Biotech., 108,

137–143.Fushinobu,S., Ito,K., Konno,M., Wakagi,T. and Matsuzawa,H. (1998)

Protein Eng., 11, 1121–1128.Garcia-Viloca,M., Poulsen,T.D., Truhlar,D.G. and Gao,J. (2004) Protein

Sci., 13, 2341–2354.Georis,J., de Lemos-Esteves,F., Lamotte-Brasseur,J., Bougnet,V.,

Devreese,B., Giannotta,F., Granier,B. and Frere,J.-M. (2000) Prot. Sci., 9,466–475.

Gruber,K., Klintschar,G., Hayn,M., Schlacher,A., Steiner,W. and Kratky,C.(1998) Biochemistry, 37, 13475–13485.

Hakulinen,N., Turunen,O., Janis,J., Leisola,M. and Rouvinen,J. (2003)J. Biochem., 270, 1399–1412.

Higgins,D., Thompson,J., Gibson,T., Thompson,J.D., Higgins,D.G. andGibson,T.J. (1994) Nucl. Acids Res., 22, 4673–4680.

Humphrey,W., Dalke,A. and Schulten,K. (1996) J. Mol. Graph., 14, 33–38.Janis,J., Turunen,O., Leisola,M., Derrick,P.J., Rouvinen,J. and Vainiotalo,P.

(2004) Biochemistry, 43, 9556–9566.Jorgensen,W.L., Chandrasekhar,J., Madura,J., Impey,R.W. and Klein,M.L.

(1983) J. Chem. Phys., 79, 926–935.Kleywegt,G.J., Zou,J.Y., Kjeldgaard,M. and Jones,T.A. (2001) In

Rossmann,M.G. and Arnold,E. (eds), International Tables forCrystallography. Vol. F. Crystallography of Biological Macromolecules.Kluwer Academic Publishers, Denmark, pp. 353–356, 366–367.

Krengel,U. and Dijkstra,B.W. (1996) J. Mol. Biol., 263, 70–78.Kumar,S. and Nussinov,R. (2001) Cell Mol. Life Sci., 58, 1216–1233.Kumar,P.R., Eswaramoorthy,S., Vithayathil,P.J. and Viswamitra,M.A. (2000)

J. Mol. Biol., 295, 581–593.Lappalainen,A., Siika-Aho,M., Kalkkinen,N., Fagerstrom,R. and

Tenkanen,M. (2003) Biotechnol. Appl. Biochem., 31, 61–68.Lazardis,T., Lee,I. and Karplus,M. (1997) Protein Sci., 6, 2589–2605.Liu,H.-S. and Wang,W.-C. (2003) Protein Eng., 16, 19–25.

Madlala,A.M., Bissoon,S., Singh,S. and Christov,L. (2001) Biotech. Lett.,23, 345–351.

Mathlouthi,N., Lalles,J.P., Lepercq,P., Juste,C. and Larbier,M. (2002)J. Anim. Sci., 80, 2773–2779.

Mayor,U., Johnson,C.M., Daggett,V. and Fersht,A. (2000) Proc. Natl Acad.Sci. USA, 97, 13518–13522.

McCarthy,A.A., Morris,D.D., Bergquist,P.L. and Baker,E.N. (2000) ActaCrystallogr. D, 56, 1367–1375.

McDonald,I.K. and Thornton,J.M. (1994) J. Mol. Biol., 238, 777–793.Muilu,J., Torronen,A., Perakyla,M. and Rouvinen,J. (1998) Proteins, 31,

434–444.Pearlman,D.A., Case,D.A., Caldwell,J.W., Ross,W.R., Cheatham,T.E., III,

DeBolt,S., Ferguson,D., Seibel,G. and Kollman,P. (1995) Comp. Phys.Commun., 91, 1–41.

Pikkemaat,M., Linssen,A.B.M., Berendsen,H.J.C. and Janssen,D. (2002)Protein Eng., 15, 185–192.

Rees,D.C. and Adams,M.W.W. (1995) Structure, 3, 251–254.Ryckaert,J.-P., Ciccotti,G. and Berendsen,H.J.C. (1977) J. Comp. Phys., 23,

327–341.Sabini,E., Sulzenbacher,G., Dauter,M., Dauter,Z., Jorgensen,P.L.,

Schulein,M., Dupont,C., Davies,G.L. and Wilson,K.S. (1999) Chem. Biol.,6, 448–455.

Sapag,A., Wouters,J., Lambert,C., Ioannes,P., Eyzaguirre,J. andDepiereux,E. (2002) J. Biotech., 95, 109–131.

Simmerling,C., Elber,R. and Zhang,J. (1995) In Pullman,A. (ed.), Modellingof Biomolecular Structure and Mechanisms. Kluwer Academic Publishers,Denmark, pp. 241–265.

Srinivasan,M.C. and Rele,M.V. (1999) Curr. Sci., 77, 137–142.Sun,J.-Y., Liu,M.-Q., Xu,Y.-L., Xu,Z.-R., Pan,L. and Gao,H. (2005) Protein

Expr. Purif., 42, 122–130.Torronen,A. and Rouvinen,J. (1995) Biochemistry, 34, 847–856.Torronen,A., Harkki,A. and Rouvinen,J. (1994) EMBO J., 13, 2493–2501.Turunen,O., Etuaho,K., Fenel,F., Vehmaanpera,J., Wu,X., Rouvinen,J. and

Leisola,M. (2001) J. Biotech., 88, 37–46.Vogt,G., Woell,S. and Argos,P. (1997) J. Mol. Biol., 269, 631–643.Wakarchuk,W.W., Sung,W.L., Campbell,R.L., Cunningham,A., Watson,D.C.

and Yaguchi,M. (1997) Protein Eng., 7, 1379–1386.Wang,J., Cieplak,P. and Kollman,P.A. (2000) J. Comp. Chem., 21,

1049–1074.Wintrode,P.L., Zhang,D., Vaidehi,N., Arnold,F.H. and Goddard,W.A., III

(2003) J. Mol. Biol., 327, 745–757.Xiong,H., Fenel,F., Leisola,M. and Turunen,O. (2004) Extremophiles, 8,

393–400.

Received December 15, 2006; revised September 20, 2007;accepted September 25, 2007

Edited by Jane Clarke


559


Dow

nloaded from


Molecular dynamics studies on the thermostability of family 11 xylanases

Documents