Specific characterization of substrate and inhibitor binding sites of a glycosyl hydrolase family 11 xylanase from Aspergillus niger †

The importance of aromatic and charged residues at the surface of the active site of a family 11 xylanase from Aspergillus niger was evaluated using site-directed mutagenesis. Ten mutant proteins were heterologously produced in Pichia pastoris, and their biochemical properties and kinetic parameters were determined. The specific activity of the Y6A, Y10A, Y89A, Y164A, and W172A mutant enzymes was drastically reduced. The low specific activities of Y6A and Y89A were entirely accounted for by a change in k(cat) and K(m), respectively, whereas the lower values of Y10A, Y164A, and W172A were due to a combination of increased K(m) and decreased k(cat). Tyr(6), Tyr(10), Tyr(89), Tyr(164), and Trp(172) are proposed as substrate-binding residues, a finding consistent with structural sequence alignments of family 11 xylanases and with the three-dimensional structure of the A. niger xylanase in complex with the modeled xylobiose. All other variants, D113A, D113N, N117A, E118A, and E118Q, retained full wild-type activity. Only N117A lost its sensitivity to xylanase inhibitor protein I (XIP-I), a protein inhibitor isolated from wheat, and this mutation did not affect the fold of the xylanase as revealed by circular dichroism. The N117A variant showed kinetics, pH stability, hydrolysis products pattern, substrate specificity, and structural properties identical to that of the wild-type xylanase. The loss of inhibition, as measured in activity assays, was due to abolition of the interaction between XIP-I and the mutant enzyme, as demonstrated by surface plasmon resonance and electrophoretic titration. A close inspection of the three-dimensional structure of A. niger xylanase suggests that the binding site of XIP-I is located at the conserved "thumb" hairpin loop of family 11 xylanases.

The three-dimensional structures of 12 family 11 endoxylanases are available, from both bacteria (13)(14)(15)(16)(17)(18)(19) and fungi (20-27). All reported structures of family 11 endo-xylanases present one single (catalytic) domain with an all b-strand "sandwich-like" fold containing two b-sheets forming a large cleft that can accommodate the xylan polymers and the overall structure has the shape of a "righthand" as described by Törrönen et al. (21). The active site contains two conserved glutamate residues located on either side of the extended open cleft, which have been identified as the nucleophilic and acid/base catalysts (28). Only a few ligand-enzyme complexes (13,16,17,29,30) have been crystallized. The subsites that bind the glycone or aglycone regions of the substrate are 4 X-ray crystal structure is available (24). The active site is located within a deep and long cleft, which is lined with many aromatic amino acid residues and is large enough to accommodate at least four xylose residues (24). The two conserved catalytic residues, Glu79 and Glu170, face each other on opposite sides (24). This fungal enzyme has been studied for its role as a bread improver (32,(34)(35)(36), in wheat processing (37) and as a supplement in animal feed (38,39).
The presence of protein inhibitors of endo-1,4-b-D-xylanase in cereals was first reported in wheat flour protein extracts (40)(41)(42). To date, two types of endoxylanase inhibitors with different structures and specificities have been described. The first type are xylanase inhibitor protein (XIP)like inhibitors and have been isolated from wheat (43,44) and rye (45). They are monomeric glycosylated proteins with M r 's of approx. 29 kDa and pI-values of 8.7-8. 9. The second type are the Triticum aestivum L. xylanase inhibitor (TAXI)-like inhibitors (36). They are high pI, nonglycosylated proteins with M r 's of approx. 40 kDa. At least two inhibitors of this type (TAXI I and TAXI II) with different pI values (8.8 and ~9.3, respectively) and varying specificities towards different endoxylanases have been identified in wheat (46). The N-terminal amino acid sequences of TAXI-I and TAXI-II showed a high degree of identity, but there was no similarity to XIP-I. The TAXI-like inhibitors are believed to be active against bacterial and fungal family 11 endoxylanases but not against family 10 endoxylanases (46,47). XIP-I inhibited both family-10 and 11 fungal xylanases apart from the family 10 Aspergillus aculeatus xylanase with K i values ranging from 3.4 to 610 nM, but bacterial family 10 and 11 xylanases were not inhibited (48).
We have previously reported the production and characterization of the A. niger xylanase in 5 inhibition mechanism of XIP-I against family 11 fungal A. niger xylanase has been studied in detail (48). The inhibition is pH-dependent in the range 4-7, as determined by activity assays and titration curves, illustrating the importance of electrostatic interactions in the strength of the interaction.
Moreover, isothermal titration calorimetry of the XIP-I/A. niger xylanase complex showed the formation of a complex with a stoichiometry of (1:1) and a heat capacity change of -1.38 kJ/mol, suggesting that the interaction was enthalpy-driven (48). Aromatic and charged amino acids are believed to play a pivotal role in protein-protein interactions, by forming hydrophobic stacking and electrostatic interactions, respectively with the target ligand. Since the inhibition is competitive, residues were selected near the active site. The three-dimensional structure of the A. niger xylanase revealed several aromatic and charged residues (Tyr6, Tyr10, Tyr89, Asp113, Asn117, Asp118, Tyr164 and Trp172) on the surface around the binding cleft. The structural analysis of the enzyme showed that these target residues are not part of the hydrogen bond network in the vicinity of the two catalytic residues (Glu79 and Glu170) (24). To investigate the importance of these residues in binding to XIP-I, appropriate mutations in the xylanase were constructed, and the biochemical properties of the mutated enzymes were evaluated in terms of kinetic properties and ability to interact with XIP-I.
Our results indicate that Asn117 is critical for the XIP-I-binding capacity of the enzyme. The mutational analysis also allowed a better understanding of the role of individual residues involved in a number of subsites in the active-site cleft of a family 11 glycosyl hydrolase. Primers which hybridize to the extremities of the A. niger xylanase cDNA were as follows: pHILD-2/XylAF 5' TTT TTT GAA TTC ATG CTT TTG CAA GCC TTC C 3' (forward primer); pHILD-2/XylAR 5' TTT TTT GAA TTC TTA AGA GGA GAT CGT GAC ACT GGC 3' (reverse primer). For the first round of PCR, 50 pmoles of the forward pHILD-2/XylAF or reverse pHILD-2/XylAR primer were used along with equimolar amounts of reverse or forward mutagenic primer, respectively. DNA amplification was carried out using 10 ng of template DNA (pHILD-2/XylA), 1.25 U of Pfu polymerase and 50 mM dNTP through 25 cycles of denaturation (1 min at 94 ºC), annealing (2 min at 35-50 ºC [depending on the mutant]), and extension (2.5 min at 72 ºC). The resulting PCR products were gel-purified using the Qiaquick PCR purification kit (Qiagen Inc., Chatsworth, CA). The final PCR was performed using 5 ng of each purified PCR products, along with 1.25 U Pfu polymerase, 50 mM dNTP. After 5 cycles of denaturation (1 min at 94 ºC), annealing (2 min at 42 ºC) and extension (3 min at 72 ºC) 50 pmoles of the forward (pHILD-2/XylAF) and reverse (pHILD-2/XylAR) primers were added and then subjected to 25 cycles of denaturation (1 min at 94 ºC), annealing (2 min at 46 ºC) and extension (3 min at 72 ºC). After completion of the amplification, 1 U of Taq polymerase was added to the reaction, and incubated at 72 ºC for 10 min in order to add adenines at the 3' end of the PCR products, for subsequent TA cloning. The PCR products were gel-purified and subcloned into a pCR ® 4-TOPO TA cloning vector according to the manufacturers instructions. The insert in the TOPO vector was subjected to DNA sequencing using the ABI prism Big Dye TM Terminator Cycle Sequencing kit to confirm the presence of the mutation and that no errors were generated during the PCR. The EcoRI-cDNA insert was gel-purified and cloned into pHILD-2 and the correct orientation checked by restriction mapping. The transformation of the P. pastoris strain GS115 (his4) (50) was achieved using the spheroplast method (51) as previously described (52) and the transformants screened for the best expression performances using routine xylanase activity assay (33). A representative His + Mut s by guest on March 22, 2020 http://www.jbc.org/ Downloaded from transformant for each mutant was selected for production of xylanase in shake-flask cultures (33).
Large-scale expression of A. niger xylanases in P. pastoris was achieved in buffered minimal glycerol-complex medium (BMGY) as previously described (52). Cells grown in BMGY at 30 °C to A 600 of 20-25 were harvested, resuspended in 200 mL buffered minimal methanol-complex medium (BMMY) pH 6.0 and incubated with shaking (200 rpm) in 250 mL flasks at 30 °C for 3 days.
Protein purification -For purification of A. niger xylanases (wild-type and variants), the culture supernatant was subjected to ammonium sulfate precipitation up to 70%, centrifuged at 10000 g for The xylanase inhibitor was purified from 2 kg wheat flour (Triticum aestivum var. Soisson) as described (43,48).

9
Protein assays and protein sequencing -Total protein in supernatants and throughout the purification procedure was estimated using the method of Bradford (53). For purified xylanase mutants, total protein was calculated using an extinction coefficient at 280 nm determined from the amino acid composition that was derived from the primary structure (50 210 M -1 cm -1 for wild-type xylanase, D113A, D113N, N117A, E118A and E118Q variants; 48 930 M -1 cm -1 for Y6A, Y10A, Y89A and Y164A; 44 520 M -1 cm -1 for W172A). Protein sequencing was performed at the Protein Sequencing and Peptide Synthesis Facility, John Innes Centre, Norwich, using an ABI 491 Procise sequencer.
Gel electrophoresis and immunoblotting -SDS-polyacrylamide gel electrophoresis (PAGE) was carried out on 10% Bis-Tris pre-cast NuPAGE gels (Invitrogen) with biotinylated marker proteins (New England Biolabs). Proteins were transferred to nitrocellulose membranes by semi-dry blotting (BioRad). The blots were probed with a 5000-fold dilution of polyclonal antiserum raised in rabbits against A. niger xylanase (a custom preparation from Unilever Laboratorium Research, The Netherlands). Immunoreactive proteins were visualized using a horseradish peroxidase anti-rabbit secondary antibody (Sigma, 1:2000) together with the chemiluminescent detection reagents (ECL Plus Detection Kit, Amersham Pharmacia Biotech, Uppsala, Sweden). Isoelectric focusing gels were run using the Phast system (Amersham Pharmacia Biotech).
Xylanase activity assays -Routine assays during screening for the highest xylanase producer was performed using a colorimetric assay from Megazyme as previously described (33). Purified xylanase activity was measured using the dinitrosalicylic acid assay (54)  where possible, at the K M (9 mg mL -1 ) of the mutants (Y6A, D113A, D113N, N117A, E118A and E118Q) or at 18 mg mL -1 for Y6A, Y10A, Y89A, Y164A, and W172A. Optimal pH for xylanase activity was estimated using the xylanase assay described above with low viscosity arabinoxylan (10 mg mL -1 ) in McIlvaine's buffer, in a range of 2.6 to 7.8. The difference in free energy (DDG) for the mutated enzymes was calculated from the equation

RESULTS
Production and structural properties of the xylanase variants -All ten mutants were efficiently produced in P. pastoris with secretion yields ranging from 30 to 270 mg L -1 , as judged by SDS-PAGE. Purified recombinant variants were obtained in yields ranging from 7 to 30 mg L -1 using a single chromatography step. The N-terminus, S-A-G-I-N, of the mutant xylanases was identical to that of the recombinant wild-type xylanase, indicating correct processing of the A. niger signal sequence. None of the mutations significantly modified the CD spectrum of the enzymes, indicating that the secondary structure content remained the same after mutation (data not shown). Therefore, loss of function and differences with the wild-type protein characteristics described hereafter for mutants Y6A, Y10A, Y89A, Y164A D113A, D113N, N117A, E118A, E118Q and W172A may only be due to minor local structural changes undetectable by UV CD.
Enzymatic activity and inhibition of the xylanase variants -To evaluate the consequences of the mutations on the enzyme, the specific activity of the xylanase variants was measured against wheat arabinoxylan (Table I). Five variants Y6A, Y10A, Y89A, Y164A, and W172A showed a significant decrease in activity. Particularly, the mutation at position Y10 dramatically reduced the specific activity by 98.3%. However these mutant proteins retained enough activity for accurate evaluation of their kinetic parameters (Table I). All variants exhibited normal Michaelis-Menten kinetics but the kinetic parameters varied with the nature of the mutation. The catalytic constant (k cat ) of the Y6A, Y10A, and Y164A variants was reduced by values similar to the specific activity (Table I).
The K M of the xylanase variants was increased for most of the mutations, indicating that these residues play a role in substrate binding. Only the Y6A mutation did not alter this parameter. The low specific activity of Y6A and Y89A was entirely accounted for by a change in k cat and K M , respectively, while the lower value of Y10A, Y164A and W172A was due to a combination of increased K M and decreased k cat . The results show that mutation of these aromatic residues caused a decrease of 92 -75% in the specificity of the xylanase for xylan, which is reflected by losses in 14 apparent binding energy [D(DG)] (58) ranging from -0.2 to 3.5 kcal mol -1 (  (Table I) comparable to those of 9.9 ± 1.7 mg mL -1 and 129 ± 11 s -1 , respectively, obtained for the wild-type enzyme. The k cat /K M value of the wild-type enzyme is 13 ± 1.1 mL mg -1 s -1 , also in good agreement with that obtained with the above variants.
XIP-I inhibited all active variants tested but a drastic loss of inhibition was observed for N117A (Table I); no inhibition was observed for the mutant up to an inhibitor:enzyme molar ratio of 5:1, as compared to 60% inhibition for the wild-type enzyme (not shown). An IC 50 value could not be measured despite a 30-fold molar excess of XIP-I which corresponds to 9 mM (Table I). IC 50 values for the D113A, D113N, E118A, and E118Q variant xylanases were measured at substrate concentrations corresponding to their K M (9 mg mL -1 ) and were comparable to that of 1 mM obtained with the wild-type enzyme. Due to the low activity of Y6A, Y89A, Y164A, and W172A variants, IC 50 values were determined at higher substrate concentration (18 mg mL -1 ) and ranged from 0.65 mM to 1.5 mM (Table I) The purified N117A xylanase migrated as a single band on SDS-PAGE identical to that of the recombinant wild-type enzyme (Fig. 1A), and no contaminating band was seen after silver-staining; this 20 kDa protein reacted with polyclonal antibodies raised against A. niger xylanase (Fig. 1B), and IEF revealed that N117A xylanase consisted of one single molecular form of approx. pI 3.5 which is similar to that of the wild-type (Fig. 1C) Da for the wild-type and variant enzyme, respectively (spectra not shown). The CD spectrum of N117A was similar to that of the wild-type (Fig. 2), indicating that this amino acid substitution did not significantly alter the secondary structure of the mutated enzyme.
The specific activities of wild-type and N117A xylanases on wheat low viscosity arabinoxylan were 172 ± 10 U mg -1 and 162 ± 5 U mg -1 and kinetic parameters were similar (Table I). Highperformance liquid chromatographic analysis of the hydrolysis products of wheat arabinoxylan showed that mutation of this residue did not alter the profile of products (data not shown). The xylanase activity of the recombinant N117A mutant as well as the wild-type enzyme was further determined on a range of different substrates (Table II). Both xylanases displayed comparable kinetic parameters on Birchwood xylan and oat spelt xylan substrates (Table II).
The effect of pH on N117A xylanase mutant was compared with the wild-type xylanase. The pH activity curves of both enzymes were identical, with the optimum pH at 3.5. Substantial amount of activity was found at pH 2.0 while a drastic decline in enzyme activity was detected at pH values above the pH optimum (not shown).
Interaction of wild-type and N117A xylanases with XIP-I -In order to determine if the loss of inhibition, as observed using activity assays, was correlated with a loss of interaction between XIP-I and N117A variant, the two protein partners were tested in binding assays in the absence of substrate.
The relative affinities and pH dependencies of the interaction of XIP-I with A. niger xylanases were studied using titration curves (Fig. 3). The pI value of the wild-type and mutant A. niger xylanase is 3.5 (Fig. 1C), whereas that of XIP-I is 8.7-8.9 (43), in agreement with the titration curves of the individual xylanases (Fig. 3A,B) and inhibitor (Fig. 3C). A. niger wild-type xylanase by guest on March 22, 2020 http://www.jbc.org/ Downloaded from formed a complex on the IEF titration gel across a pH range of approximately 4-7 (Fig. 3D) while no complex could be detected in the case of the N117A variant (Fig. 3E), in agreement with data from the activity assays.
The interaction between XIP-I and wild-type or N117A A. niger xylanases was studied by using a biosensor based on surface plasmon resonance (SPR) (59,60). XIP-I was immobilized as a ligand on the sensor surface, while the A. niger xylanase was passed in solution as an analyte over the surface. The results of the analyses of the interaction of XIP-I with increasing amounts of the wildtype or N117A xylanases are reported in Fig. 4. In Fig. 4A, the increase in RU (resonance units) from the initial baseline represents the binding of the wild-type xylanase to the surface-bound XIP-I. The plateau line represents the steady-state phase of the xylanase-XIP-I interaction while the decrease in RU from the plateau represents the dissociation phase. The SPR data showed that mutation N117A strongly affects the interaction (Fig. 4B). The kinetic constants could not be calculated due to the undetectable interaction between the mutated protein and the enzyme. The K D is therefore higher than 1 mM, which is the limit of detection of the BIAcore, as compared to 5.68 mM calculated for the wild-type xylanase (48).

DISCUSSION
The active site of glycosyl hydrolases often contain aromatic residues, such as tyrosine and tryptophan, which hydrophobically stack against sugar rings, as well as side chains which hydrogen bind to hydroxyl groups of the substrate. The replacement of the aromatic amino acid residues, Tyr6, Tyr10, Tyr89, Tyr164, and Trp172 with Ala significantly reduced the enzyme activity of A. niger glycosyl hydrolase family 11 xylanase. These mutations did not affect the fold of the xylanase as revealed by circular dichroism spectra of these variants. Although these residues are not essential for the hydrolytic reaction per se (24), the present findings indicate that they play an important role in ligand binding and catalysis. Recently the structure of an inactive mutant of the Xyn11 from Bacillus agaradhaerens in complex with xylotriose has been obtained (17). The interactions are very similar to that described at subsites -2 and -1 for the covalent intermediate complex (16), with the addition of the interactions for the -3 subsite.

18
The active site residues in the A. niger xylanase correspond remarkably closely in both position and orientation with the residues that contact the sugar rings in these complexes. In Fig. 5, structural alignment of the above xylanases with that of A. niger enzyme showed that four of the A. niger xylanase variants (Y10A, Y89A, Y164, and W172A) have been mutated at positions corresponding to residues which have been involved in discrete subsites of other family 11 xylanases. Tyr10 corresponds to Trp9 BCX , Tyr9 XYNI , Trp18 XYNII , Trp18 Xyna , and Trp19 Xyn11 , all involved in stacking interaction with the xylose ring in substite -2 of these xylanases (13,16,17,22,26). Tyr89 corresponds to Tyr96 XYNII thought to determine subsite +3 of the T. reseei enzyme (22). Tyr164 is equivalent to Tyr166 BCX and Tyr172 Xyna , both shown to form hydrogen bonds with the substrate in subsite -2 of the xylanases (13,26). Trp172 corresponds to Trp166 XYNI and Tyr179 XYNII which are thought to determine subsites +1 and +2 of XYNI and XYNII enzyme, respectively (22). Moreover the oxygen atom OH of residue Tyr6 is in the same position as atom Glu17 Xyn11 OE1 (not shown) which, in B. agaradhaerens, is involved in one of the solvent-mediated hydrogen bonds associating the -3 substite sugar (17). These findings are consistent with the substantial decrease in the activity of the Y6A, Y10A, Y89A, Y164, and W172A mutants against xylan, although these residues do not necessarily play equivalent roles in different family 11 enzymes. In Fig. 6, the two xylose rings found at substites -1 and -2 of the BCX mutant complexed with a substrate (13) are superimposed to the A. niger xylanase three-dimensional structure. In this view Tyr6, Tyr10 and Tyr164 are in close contact with the ligand, also suggesting their involvement in substites of the A. niger enzyme.
Mutational analyses of active site residues have been carried on other family 11 xylanases but, to our knowledge, none of the present mutations were reported apart for Tyr166 BCX (corresponding to Tyr164 in A. niger xylanase), which, replaced by Phe in B. circulans xylanase, led to a small decrease of enzyme activity (13). The substrate binding cleft of A. niger xylanase contains at least four xylose-binding subsites (24). Our data clearly show that Tyr10, Tyr164 are likely to play an important role in ligand binding in the glycone region of the substrate binding cleft (probably at subsite -2) while Trp172 and Tyr89 are involved in the aglycone subsites (probably at subsites +1 19 and +2, respectively) of A. niger xylanase. Interestingly, the finding that Tyr6 played a role in enzyme activity together with its position in the 3D structure might indicate the presence of another subsite (-3) in the substrate binding cleft of A. niger xylanase.
Interestingly, all the fungal xylanases tested so far are inhibited by XIP-I apart from the family 10 A. aculeatus xylanase, whereas none of the bacterial enzymes are (48). The specific inhibition/recognition of fungal xylanases was not due to the binding of XIP-I to glycosylation on the fungal enzymes, since both the native and E. coli-expressed recombinant forms of P. funiculosum xylanase were inhibited to the same extent (48). In the same way, the loss of XIP-I inhibition towards the N117A xylanase mutant was not due to a potential requirement for protein glycosylation on Asn117 residue since the wild-type and mutant enzymes showed the same glycosylation content (approx. 150 Da), probably corresponding to a single hexose or a hexosamine.
Alignment of the A. niger xylanase with other family 11 xylanases from fungal and bacterial origin showed that the Asn117 residue was conserved in most sequences (8), indicating that the environment of this amino acid in the three-dimensional structure is more likely to be responsible for the difference in binding. The A. niger xylanase three-dimensional structure comprises a single domain containing one a-helix and 13 b-strands, which are arranged in two mostly antiparallel bsheets A and B (Fig. 7A). The data obtained from titration curves and surface plasmon resonance clearly showed that the Asn«Ala mutation at the solvent-exposed position 117 abolished the capacity of the molecule to interact with XIP-I. This residue is present on the b-strand 8 of the larger, eight-stranded b-sheet (sheet B) which twists around the catalytic cleft (Fig. 7A). The A. niger xylanase has been compared to the shape of a right hand with the "fingers" at the top, the "palm" at the bottom and the "thumb" at the right hand side of the molecule as represented in Fig.   7A. Asn117 is located at the end of the b-bend positioned at the tip of the "thumb" (Fig.7A/B). In all other family 11 xylanases with known structure, this loop is significantly longer and includes parts of b-strands B7 and B8 of the A. niger xylanase (Fig. 5). However, these differences in secondary structures do not influence the general shape of the "thumb", which points back towards 20 the bottom of the cleft (Fig.7A). It is stabilized in this position by hydrophobic interactions as well as by several hydrogen bonds (8) but also able to move and thus regulate the width of the active site cleft (29). The interaction of XIP-I with the thumb region could thus be responsible for the inhibition, preventing access by the substrate to the catalytic cleft (Fig. 7A). In the thumb region of A. niger xylanase a number of residues, Thr114, Thr116, Asn117, Thr124, Thr126 and Thr128, have their side-chains highly solvent exposed, and could interact with XIP-I (Fig.7B). The bacterial xylanase from Bacillus agaradhaerens is not inhibited by XIP-I (48) and the superimposition of the "thumb" to that of the A. niger xylanase shows that, although Asn117 is conserved both in position and orientation, four threonines present in the vicinity of Asn117 in the A. niger enzyme are replaced by hydrophobic or basic residues in B. agaradhaerens xylanase (Fig. 7B). Interestingly, among the xylanases from A. niger, one xylanase has a threonine instead of Asn117 at this position (61). Taken together these findings suggest that the "thumb" region, rather than a single residue, is involved in XIP-I binding to A. niger xylanase. This region is overall conserved among the family 11 fungal enzymes and is proposed as the binding site for the interaction of XIP-I with family 11 xylanases.
In summary, this is the first report investigating the molecular interactions of the proteinaceous inhibitor XIP-I with a target xylanase. Characterization of the xylanase N117A variant of A. niger xylanase and the analysis of the three-dimensional structure of the wild-type xylanase converge towards an essential role of the "thumb" hairpin loop in binding to the inhibitor. This mutational analysis of A. niger xylanase also allowed identification of key residues that play an important role in ligand binding at the various subsites.  Panel D, a mixture containing A. niger wild-type xylanase (5 mg) and XIP-I (4 mg); Panel E, a mixture containing A. niger xylanase N117A mutant (5 mg) and XIP-I (4 mg). The pH-mobility curves of the proteins, free or in complex, are indicated with an arrow.    Wheat arabinoxylan 9.9 ± 1.7 129 ± 11 13 ± 1.1 9.9 ± 1.8 104 ± 9 10 ± 0.9