The structure of galactoglucomannan impacts the degradation under alkaline conditions
- 670 Downloads
Galactoglucomannan (GGM) from spruce was studied with respect to the degradation behavior in alkaline solution. Three reference systems including galactomannan from locust bean gum, glucomannan from konjac and the linear water-soluble carboxymethyl cellulose were studied with focus on molecular weight, sugar composition, degradation products, as well as formed oligomers, to identify relative structural changes in GGM. Initially all mannan polysaccharides showed a fast decrease in the molecular weight, which became stable in the later stage. The degradation of the mannan polysaccharides could be described by a function corresponding to the sum of two first order reactions; one slow that was ascribed to peeling, and one fast that was connected with hydrolysis. The galactose side group was stable under conditions used in this study (150 min, 90 °C, 0.5 M NaOH). This could suggest that, apart from the covalent connection to C6 in mannose, the galactose substitutions also interact non-covalently with the backbone to stabilize the structure against degradation. Additionally, the combination of different backbone sugars seems to affect the stability of the polysaccharides. For carboxymethyl cellulose the degradation was linear over time which further suggests that the structure and sugar composition play an important role for the alkaline degradation. Molecular dynamics simulations gave details about the conformational behavior of GGM oligomers in water solution, as well as interaction between the oligomers and hydroxide ions.
KeywordsHemicellulose Spruce Alkaline hydrolysis Peeling Structural analysis Molecular dynamics simulations
Today large quantities of softwoods are being processed under alkaline conditions in order to produce cellulose-rich wood fibers for a variety of pulp and paper products. Apart from cellulose and lignin, hemicelluloses are a fundamental component in lignocellulosic biomass, with galactoglucomannan (GGM) followed by arabinoglucuronoxylan being the most abundant hemicelluloses in softwood. Hemicelluloses are the second most abundant class of polysaccharides in nature, and an improved exploitation of hemicelluloses is fundamental for the development of sustainable wood-based biorefineries. Moreover, many fiber-based products would benefit from retaining the hemicellulose fraction, resulting in an increased yield and improved properties in mainly cellulose based products (Tavast et al. 2015; Salmén and Lindström 2015). Hemicelluloses have also shown good properties as oxygen barriers (Mikkonen et al. 2011; Kisonen et al. 2014), in hydrogels (Edlund and Albertsson 2008) and as emulsifiers (Mikkonen et al. 2016), gaining great attention of the pulping industry in the context of an integrated biorefinery. However, the number of hemicellulose-based products is still limited. One main reason is the degradation of hemicelluloses during traditionally important alkaline processes, such as kraft and soda pulping, for separation of the wood polymers and production of cellulose-based products.
The known mechanisms involved in degradation of polysaccharides are peeling and hydrolysis caused by ionic reactions. In kraft cooking radicals originating from formed polysulfides can be present and reactions with lignin has been observed (Gellerstedt et al. 2004) but, to the knowledge of the authors, no reactions with carbohydrates has been reported. During peeling, single sugar units are, in modified form, cleaved off from the reducing end. This mechanism is a larger problem in the case of GGM since xylan is stabilized from peeling due to reactions involving the side chains, where the linkage position of the side chain on the backbone plays a central role (Sjöström 1993). Alkaline hydrolysis of the glycosidic linkages can occur anywhere within the polymeric chain and is not hindered by the stopping reaction observed for peeling-type reactions. Hydrolysis results in additional reducing end groups and secondary peeling can take place subsequently. The peeling reaction removes one sugar unit at a time resulting in a slow decrease in molecular weight while hydrolysis can cut the chain anywhere within the backbone leading to a faster decrease in molecular weight (Sjöström 1993). Thus peeling has a low impact on molecular weight, but the yield will still decrease and the yield losses of GGM are especially high during the initial part of the kraft cook (Sjöström 1977). Anthraquinone has industrially been used as a way to decrease carbohydrate degradation during alkaline treatment (Löwendahl and Samuelson 1978). However, health concerns connected to this compound is now debated and alternative carbohydrate retaining methods are being evaluated. These alternative methods are based on the oxidation agent polysulfide or the reducing agents sodium borohydride (Wang et al. 2015) and sodium dithionite (Veguta et al. 2017). The concentration of alkali is also of importance. The degradation of GGM is increasing with the increased alkali concentration. However, at high alkali concentrations, the GGM yield is higher indicating a stabilization reaction (Brännvall and Lindström 2007). Applying a harsher pre-treatment has also showed an increased yield of GGM (Annergren et al. 1961; Paananen et al. 2010). The processes can therefore be optimized, and in order to develop sufficient methods to retain these polysaccharides it is important to increase the understanding about the degradation mechanisms and how it is related to the structure.
The structure of hemicelluloses is complex and varies widely among different types of plants. GGM consists of a backbone built up by d-mannose (Man) and d-glucose (Glc) units linked through β-(1 → 4) glycosidic linkages. The mannose units can be branched with α-(1 → 6) liked d-galactose (Gal), and also O-acetylated at the C2 and C3 positions. In softwoods, GGM make up about 20% of the total mass, and xylan, the second most abundant hemicellulose, correspond to around 5–10% (Sjöström 1993). The ratio between Gal:Glc:Man in softwood GGM has been reported to 0.1–1:1:3–4 (Timell 1967; Lundqvist et al. 2002) and the degree of acetylation to 0.2–0.3 (Lundqvist et al. 2002). In our study, the degradation of GGM and two other mannan polymers: locust bean galactomannan (LBG) and konjac glucomannan (KGM) has been studied. LBG and KGM both consist of structures that in different ways are similar to GGM and therefore are good models to evaluate structural effects in pure systems. LBG has a backbone built up by β-(1 → 4) d-Man and is substituted with α(1 → 6) liked d-Gal, while KGM consist of a backbone of β-(1 → 4) linked d-Glc and d-Man units, without any Gal side groups, but with a low degree of β-(1 → 6) d-Glc and d-Man units branched to the Glc backbone units (Katsuraya et al. 2003). Another reference system is carboxymethylated cellulose (CMC), specifically chosen since it is a water soluble linear β-(1 → 4) linked homopolysaccharide of d-Glc containing CH2CO2H substituted hydroxyl groups at a DS of around 0.8. The alkaline degradation of the polysaccharides was evaluated by analysis of molecular weight, sugar composition, oligomer and organic acid formation. Furthermore, molecular dynamics (MD) simulations were used to evaluate details in the structure of GGM and study conformations in water solution as well as coordination with sodium hydroxide (NaOH). Hence, the main objective with this study was to reveal more information about how the structure impacts the degradation of GGM as well as understanding the various mechanisms involved.
Theory of alkaline hydrolysis
The immediate reaction products of the attack are an epoxide and a deprotonated hydroxyl group (Fig. 1a). These components can either (1) react with water, completing the hydrolysis as earlier described (Gellerstedt 2009) or, as we propose, (2) react with each other, i.e., a relegation. Our view is that the probability for reaction (2) (the relegation) is higher if the newly formed chain ends are kept close to each other. One factor that potentially could affect this is the inherent stiffness of the glycosidic linkage, which also varies depending on the involved sugars. Another hypothesis is that the structure can be stabilized by a galactose side chain positioned over the backbone glycosidic linkage, forming a “sandwich” structure (Fig. 1b). This would imply that main chain glycosidic linkages close to galactose residues should be more resistant to alkaline hydrolysis. Finally, the accessibility of water is important for completing the alkaline hydrolysis. The above described hypotheses have been evaluated further in this work.
We have earlier classified glycosidic linkages in the glucomannan backbone as belonging to either of two types: the M-type which is a β-(1 → 4) linkage between two Man units or Man-β-(1 → 4)-Glc, or the C-type which corresponds to Glc-β-(1 → 4)-Man or a β-(1 → 4) linkage between two Glc units (Berglund et al. 2016). These classifications are based on the configuration of the C2 hydroxyl group (equatorial for C-type and axial for M-type) adjunct to the glycosidic linkage, as well as the hydrogen bonding possibilities around the glycosidic linkage. In the previous computational study in water solution, it was observed that these glycosidic linkages experience small differences in their conformation and hydrogen-bonding pattern around the glycosidic linkage. Here, we study GGM with the suggested types of glycosidic linkages (Fig. 1b) in mind, aiming for an increased understanding about the mechanisms involved in the degradation during alkaline conditions.
Materials and methods
Spruce GGM was extracted according to (Willför et al. 2003). In short, thermo-mechanical spruce pulp was extracted with water at 60 °C and the polysaccharides were thereafter precipitated with ethanol (method 2.3 in Willför et al.). LGM (low viscosity) and KGM (low viscosity) were purchased from Megazyme (Ireland), and CMC (low viscosity) from Sigma-Aldrich (Germany).
Alkaline treatment of polysaccharides
First, polysaccharides were dissolved in MilliQ-water at a concentration of 4 mg/ml by mixing and increasing the temperature to 60 °C for the extracted GGM, CMC, KGM and LBG in a thermomixer (Eppendorf, Germany). The dissolved samples were then allowed to cool to room temperature and 4 ml were pipetted to 12 ml vials. Thereafter, 4 ml of 1 M NaOH were added to each vial obtaining a final volume of 8 ml, a concentration of 2 mg/ml polysaccharide and 0.5 M NaOH. The glass vials were then sealed with PTFE caps and incubated in a water bath at 90 °C for different times. The incubation lasted for 0, 5, 10, 20, 30, 40, 60, 90, 120 or 150 min; the reaction was then stopped by cooling of the glass vial in an ice bath. Thereafter 1 ml of each sample was saved for analysis of organic acids, 1 ml was saved for oligomer analysis and 6 ml was dialyzed in Float-A-Lyzer G2 tubes (Spectrum labs, Netherlands) with a 100–500 Da cutoff. After dialysis, 1 ml was used for analysis in the 10 mM NaOH SEC system and the remaining sample was freeze dried and used for sugar analysis. The same procedure was performed for all samples except that sugar composition and oligomers were excluded for CMC.
Size exclusion chromatography (SEC)
After dialysis the samples were filtered with 0.2 µm nylon syringe filters and analyzed on an Ultimate-3000 HPLC system (Dionex, Sunnyvale, CA, USA) equipped with both RI (Waters, Milford, MA, USA) and UV/Vis (Dionex, Sunnyvale, CA, USA) detectors. The separation was performed using a column set-up consisting of a Suprema guard column (50 × 8 mm, 10 µm particle size), a Suprema 30 Å column and two Suprema 1000 Å columns (300 × 8 mm, 10 µm particle size) connected in series (Polymer Standard Services, Mainz, Germany). Standard calibration with 10 pullulan standards (Polymer Standard Services, Mainz, Germany) ranging between 342 and 708,000 Da was used for determining the apparent molecular weight, and in the results we mainly consider the weight average molecular weight, Mw. The mobile phase was 10 mM NaOH and the oven was set at 40 °C and the cooler at 30 °C, the flow rate was 1 ml/min.
The spruce GGM was also analyzed on a SECurity 1260 system (Polymer Standard Services, Mainz, Germany) coupled to both RI and MALLS (BIC-MwA7000, Brookhaven instrument Corp., US) detectors. For separation GRAM columns (Polymer Standard Services, Mainz, Germany) were used, a PreColumn (50 × 8 mm, 10 µm particle size), a 100 Å, and a 10,000 Å (300 × 8 mm, 10 µm particle size) were connected in series. The eluent was DMSO 0.5% LiBr, the flow rate 0.5 ml/min, the column oven was set at 60 °C and the RI detector at 40 °C. For calibration the same pullulan standards described above were used. The dn/dc value used for MALLS estimation was 0.068, as described for pullulan (Vilaplana and Gilbert 2010).
A second setup with the same SEC-RI-MALLS system was also applied for analysis of the raw materials. Here the eluent 0.1 M NaNO3 and 5 mM NaN3 together with the earlier described PSS Suprema columns and pullulan standards were used. The flow rate was 1 ml/min, the oven was set at 40 °C and a dn/dc value of 0.146 was used for the MALLS estimation, which corresponds to pullulan (Podzimek 2010).
After dialysis, the samples were freeze dried and hydrolyzed by 2 M triflouracetic acid at 120 °C for 3 h (Albertsheim et al. 1967). The concentration of sugars was measured by high-pH anion exchange chromatography with pulsed amperometric detection (ICS-3000, Dionex, Sunnyvale, CA, USA) on a Dionex CarboPac PA1 column, using the eluent gradients previously reported by McKee et al. (2016). For quantification, eight standards containing all detected sugars with concentrations ranging between 2 and 400 mg/l was used. The samples were filtrated with 0.2 µm nylon syringe filters prior to injection.
After quenching the alkaline treated samples on ice one fraction was saved for the oligomer analysis. The samples were either stored at 4 °C if analyzed within 24 h or stored at − 20 °C and moved to 4 °C prior to oligosaccharide analysis. After filtration with a 0.2 µm nylon syringe filter, the oligosaccharides were analyzed by HPAEC-PAD (ICS-3000, Dionex, Sunnyvale, CA, USA) using a Dionex CarboPac PA200 column with an isocratic flow of 0.5 ml/min of 45 mM NaOH at 30 °C during 15 min (Wang et al. 2014). Linear β-(1 → 4)-linked d-mannan oligomers ranging from DP 1–6 were used as standards. Additionally, a sample containing a mannotriose oligomer substituted with one α-(1 → 6) linked d-Gal unit at the reducing end Man unit (M3Gal), and a similar oligomer with five Man units substituted with two Gal units at the Man which are 3 and 4 units from the reducing end (M5Gal) were used as substituted standards (Megazyme, Ireland). The linear mannose oligomers were used for quantification with 8 concentrations ranging between 0.1 and 100 mg/l.
Organic acid analysis
Directly after the alkaline treatment 1 ml was saved for HPLC analysis. It was neutralized with 37% HCl and stored at 4 °C overnight prior to filtration with a 0.2 µm nylon syringe filter and injected to an Ultimate-3000 HPLC system (Dionex, Sunnyvale, CA, USA). A Phenomenex Rezex ROA organic acid column was used and the detection was performed by a UV-detector (Dionex, Sunnyvale, CA, USA) set at 210 nm. The oven was set to 50 °C, the mobile phase was 2.5 mM H2SO4, and the calibration was performed by making a standard curve with 6 concentrations ranging between 0.1 and 10 mM acetic acid and formic acid. Possible glycolic acid formation could be interesting for CMC and was therefore also tested.
The degree of acetylation of the raw materials was determined in the same way as in Bi et al. (2016). The concentration of acetic acid was, after saponification, analyzed on the same HPLC system described above.
2D heteronuclear single quantum correlation (HSQC) NMR spectroscopy
For 2D-NMR, 50 mg of GGM was dissolved in 1 ml of deuterated water (D2O) or DMSO-d 6 by mixing at 60 °C for 24 h. The sample was centrifuged and the supernatant was transferred to an NMR tube before analysis. NMR spectra were recorded at room temperature on a Bruker Avance III HD 400 MHz instrument with a BBFO probe equipped with a Z-gradient coil. Sensitivity improved with gradient (e/a TAPPI). 2D HSQC NMR experiments were carried out with the Bruker pulse program ‘hsqcetgpsi’ where the spectra were acquired with the following parameters: size of FID 1024, pulse 9.2 μm, number of dummy scan 16, spectral width 13 ppm and the relaxation delay of 1.5 s. The number of scans was set to 56 which lead to a run time of 6.5 h. Data were processed with MestreNova (Mestrelab Research) using 90° shifted square sine-bell apodization window; baseline and phase correction was applied in both directions.
Molecular dynamics simulations
The molecular dynamics simulations were performed according to the methodology described in Berglund et al. (2016), where the models also were verified by NMR analysis. The software used was GROMACS 2016.3 (Hess et al. 2008; Abraham et al. 2015), the force field was GLYCAM06 (Kirschner et al. 2008), the water model was TIP3P (Jorgensen et al. 1983), and replica exchange molecular dynamics (REMD) (Sugita and Okamoto 1999) was used for improved sampling. The temperatures were controlled by the Nosé-Hoover algorithm (Nosé 1984; Hoover 1985) and were ranging between 300 and 366 K and divided into 12 evenly spaced replicas. The simulations lasted for 50 ns and were using a leap-frog integrator with a 2 fs basic time step. The simulation box contained one saccharide and the number of water molecules were just over 1500, the volume around 50 nm3, and the pressure was 1 atm controlled by the Parrinello–Rahman barostat (Parrinello and Rahman 1981). A straight cutoff at 1.2 nm was applied for the Lennard–Jones interactions, and for the electrostatic forces the real space cutoff was 1.2 and particle mesh Ewald summation (PME) was used for the long-range part (Darden et al. 1993; Essmann et al. 1995). To keep bonds at their equilibrium lengths, P-LINCS (Hess 2008) and SETTLE (Miyamoto and Kollman 1992) were used for saccharides and water, respectively.
For the simulations in 0.5 M NaOH 13 hydroxide ions (OH−) were added and the OH− model described in Hub et al. (2014) and Wolf et al. (2014) was adapted to our system and used together with the Na+ model from Yu et al. (2010). A more detailed description of the NaOH model is available in supplementary materials. The NaOH-simulations were executed in the same way as the simulations in pure water.
Free energy maps of (φ, ψ) and (ψGal, ωGal) dihedral spaces were constructed by converting the probabilities to free energies by the Boltzmann inversion. Contour plots were constructed with a grid spacing of 5°. Molecular graphics were made by using Visual Molecular Dynamics (Humphrey et al. 1996). To study hydrogen bonding the “gmx hbond” tool in GROMACS were used with default cutoffs.
Composition and properties of extracted GGM and mannan model systems
Considering lignin contamination, the GGM seems to have a high purity. 2D HSQC NMR spectra (Fig. 3c) showed no aromatic signals corresponding to lignin but low levels of lignin methoxyls were detected, the signals of which were enhanced due to the presence of three protons (see full spectrum in Fig. S2). Volume integrals of the methoxyl signal relative to those of the GGM fraction approximate the lignin content in the sample at less than 1%. The lignin-sensitive UV detector in SEC also showed clear responses, which might indicate that traces of lignin could be connected to the carbohydrates.
The degree of acetylation (DSac) in GGM was estimated to 0.15 by integrating the volume integrals of the acetylated C2 and C3 regions and the GGM sugars in the anomeric region. This was in agreement with the value of 0.2 obtained from the HPLC analysis (Fig. 3b). Furthermore, of the acetyls determined by NMR, about 65% are O-acetylated at the C2 of Man and the rest at the C3 position. Integration of the peaks in NMR also show that 61% of the mannose units were acetylated which is consistent with the 65% obtained by Willför et al. (2003).
The molecular weight analysis showed that the spruce GGM consist of two fractions with different apparent molecular weight. To evaluate if this really is the case or if aggregations could be the explanation for the peak at large molecular weights SEC–MALLS analysis with both aqueous eluent and DMSO with 0.5% LiBr were compared. Both solvent systems showed similar bimodal distributions for the molecular weight. Additionally, the intensities from the refractive index (RI) and light scattering (MALLS) detectors in both the DMSO-LiBr eluent and the aqueous NaNO3–NaN3 eluent made aggregation seem improbable (Figs. S3, S4). Furthermore, the same phenomenon with two molecular weight fractions has been observed earlier both with GGM from spruce (Willför et al. 2003) and with glucomannan from birch (Teleman et al. 2003). Therefore, the molecular weight distribution observed from SEC is assumed to be correct (Fig. 3a). The full fraction is referred as GGM, but the two fractions were also integrated separately and called as GGM-P1 for the high Mw fraction and GGM-P2 for the low Mw fraction.
The sugar composition of the model compounds KGM and LBG is also presented in Fig. 3b. The molecular weight distribution of these mannans were unimodal from the RI-analysis (Fig. S5). From SEC-RI-MALLS analysis a small content of aggregates was observed for both KGM, LBG and CMC as showed by the high MALLS signal in relation to very low RI signal at high molar masses in Figs. S6–S8.
SEC analysis with standard calibration was used to study the degraded samples, however standard calibration can result in an overestimation of Mw (Berggren et al. 2003). Thus a comparison between molar mass distributions from both light scattering and standard calibration for the GGM material is presented in Fig. S9, and Table S2 compares the Mw for all raw materials analyzed by aqueous NaNO3–NaN3 SEC-RI-MALLS.
Kinetic evaluation of the molar mass during alkaline degradation
Constants for the models of the degradation behavior according to Eq. (1)
The curves in Fig. 4 shows that GGM experiences a fast drop in Mw during the first 10 min, for KGM it took around 20 min, while this first phase was as long as 40 min for LBG. The difference is also reflected by the time constant τ1 in Table 1, which is the largest for LBG resulting in the slowest degradation rate. Thereafter the decrease of Mw continued at a slower pace for all mannans, which is also revealed by the time constant where τ2 is significantly larger than τ1. However, for CMC the degradation was single exponential throughout the treatment with only the slow term present (Fig. 4).
The yield of the mannans decreased with time and a gravimetrical analysis of LBG, GGM and KGM showed that compared to 0 min the yield decreased with 16, 29, and 39% respectively after 60 min, and 27, 44, 45% respectively after 150 min. Also the concentration sensitive RI-signal in SEC give a similar impression as presented in Fig. S11.
Sugar composition and low molar mass products during alkaline treatment
The analysis of released organic acids showed that the concentration of formic acid, a degradation product from the peeling reaction, initially increased for all mannans, but stabilized for GGM after 60 min while it continued to increase for both KGM and LBG (Fig. 5b). Acetic acid mainly originates from acetyl substitutions, which are cleaved off, but might also be formed during other degradation reactions (Sjöström 1993). The concentration of acetic acid was stable for all compounds and the highest concentration was observed for the mannan with the highest degree of acetylation, GGM (Fig. 5c). Thus, most acetyl groups were cleaved off early during the alkaline treatment. For CMC traces of formic, acetic and also glycolic acids were detected, however the signals were out of detection limit and no significant variations were observed.
Oligosaccharide profiles obtained when the mannan polymers were treated for different times in 0.5 M NaOH and 90 °C are presented in Fig. 6. The data was analyzed with principal component analysis (PCA) by the software SIMCA-P 12.0.1 and the plots are presented in Fig. S12. If the retention times matched the standards (M1-M6, M3Gal or M5Gal) this name was used. Otherwise the RX indicates unknown oligomer of a degree of polymerization (DP) estimated to X based on retention time. The second part of the name indicates which mannose the oligomer originates from by the letters G for GGM, L for LBG, and K for KGM.
Generally, the analysis showed that larger oligomers (DP 4–6) became more common with longer treatment times. The concentration of monosaccharides was low, probably due to further degradation partially shown by the concentration of organic acids. For GGM, the R5.G3, R4.G3 and M3 oligomers were significant for longer times, while R6.G2 was more pronounced at average times. KGM showed a similar behavior as GGM and the oligomers R4.K2, R3.K1 and M6 were the most important for long times, and also here an oligomer of DP 6, R6.K, showed an increased concentration up to average times and then started to decrease. LBG, on the other hand, experienced a somewhat different behavior mainly due to that one type of oligomer, R5.L2, highly dominated the spectrum after long times in alkali. Otherwise mainly oligomers of DP 4 showed increased concentration with time.
The load scattering plot from the PCA analysis (Fig. S12) showed a similar behavior for all mannans with raw materials (0 min) present in the first quadrant, thereafter the samples moved in an elliptical movement to the second, and finally, the third quadrant with longer treatment times. The samples between 5 and around 30 min were in the second quadrant while longer treatment times made the samples more concentrated in the third quadrant which indicates that their spectrums are more alike. The score scatter plots confirm as described above, that R6.G1 and R6.G2 are the most important for GGM treated for 5–30 min, and R4.G3, R5.G3 and M3 are significant for longer times. The corresponding peaks are R6.K for average times, and R4.K2, R3.K1 and M6 for longer treatment of KGM. GGM and KGM were more similar with regards to the oligomer profiles than LBG. Indeed, alkaline degradation of LBG generates one oligomer (R5.L2) that significantly stood out from KGM and GGM, probably due to the fact that LBG contains only Man units in the backbone.
Oligosaccharide conformation in water
The conformation of the backbone glycosidic linkages is the most important factor for the macromolecular conformation of polysaccharides. Therefore, manno-oligosaccharides were simulated and the effect of Gal substitution was evaluated. For the analysis, the backbone glycosidic linkages in glucomannans are classified into M-type and C-type as explained earlier.
The probability for hydrogen bonding between the backbone and the Gal side group was 11% for the Glc and Gal units in MGLM, but only 3% for the Gal and corresponding Man-unit in MMLM. This can be attributed to that the C2 hydroxyl in Man is axial instead of equatorial as in Glc, thus pointing away from the Gal side group in the most common conformations. Nevertheless, the probability for hydrogen bonding between backbone and side group is still relatively low. Additionally, the effect on backbone flexibility from having a α-(1 → 6)-Gal substitution is small, and this suggests that other effects are more important for the degradability of GGM.
The conformation of the ωGal dihedral can be classified into three groups; gg, tg, and gt, where the dihedral angles O5-C5-C6-O6 and C4-C5-C6-O6, respectively, are in gauche for g and trans for t. For all oligomers the conformation at ψGal ≈ 180° and ωGal ≈ 290° (gg) was preferred, but with varying probability. In MGLM it was as high as 81%, and if the backbone is purely Man the probability was 70% indicating that the Glc unit promotes this conformation. The exception here is the MLLM (with substitution at the second Man from the non-reducing end) which showed a significantly lower chance of attaining gg. This is probably explained by steric effects when two side group substitutions are present close to each other, and the structure is illustrated in Fig. S14. To quantify the distribution between tg, gt, and gg the ωGal curves were integrated (Table S4). This was also compared to the ω distribution that was obtained from the simulations of the unsubstituted oligomers. Previous results on the conformation of the hydroxymethyl rotamer in a single mannose suggest a gg:gt:tg relation of about 50:50:0 (Kräutler et al. 2007). Here, a similar probability for gg but a relatively even distribution between tg and gt was observed, nevertheless, a difference is expected between a single hexose or end group, and a hexose unit in the middle of an oligomer (Angles d’Ortoli et al. 2015). Furthermore, for a pure mannose backbone, Gal substitution on O6 increased the probability for gg from 55 to 75%. The gg conformation is generally more common for MGMM and here the probability increased from 60 to 85% with Gal substitution.
Simulations at increased temperatures affected the dihedral distributions (Fig. 8) and, specifically, resulted in a slightly lowered probability for gg (Table S4). However, previous work where the conformation of Glc ω was simulated using the same water model (TIP3P) and compared to NMR spectroscopy (Angles d’Ortoli et al. 2015) show that this most likely is an artifact, possibly explained by the fact that the parameters are developed for room temperature simulations.
MD simulations in 0.5 M NaOH
The four oligomers MMMM, MGMM, MMLM and MGLM were also simulated in 0.5 M NaOH. The purpose was to investigate if the Gal substitution affected the interaction between OH− and the backbone hydroxyl groups. Previous work on cellobiose has also showed that deprotonation of the hydroxyl group is likely to occur on C2 (Bialik et al. 2016). The C2 hydroxyl group is of largest interest since alkaline hydrolysis of the glycosidic linkage has been suggested to start by deprotonation of this group (see mechanism described in Fig. 1).
The OH− coordination number of the hydroxyl groups in the middle sugar units are of large interest since these are involved in the alkaline hydrolysis reaction. Figure 9 shows that two adjacent sugar hydroxyl groups simultaneously can coordinate one OH−. For the two central sugars, OH2 and OH3 both had a coordination number of around 0.20–0.35 at room temperature, which then decreased at higher temperatures due to more dynamic hydrogen bonds (Fig. 9c, d). A higher coordination number gives a larger probability for deprotonation and therefore further reactions, such as alkaline hydrolysis. However, this requires that the temperature is high enough for upcoming reaction steps, which means that the probability for a complete hydrolysis can still be larger at 366 K. Furthermore, the simulations showed that Glc experienced an increased coordination in comparison to Man at OH2 and OH3, and a further increase from having a Gal side group substitution. For visualization spatial distribution functions (SDFs), showing a three-dimensional density distribution of hydroxide ions in the close area around the manno-oligomers are presented in Fig. 9e.
To correlate the structure of GGM and the mechanisms involved in alkaline degradation, a number of different analytical methods have been integrated. We have shown that the molecular weight of the studied mannans (GGM, LBG, and KGM) decreased in two phases during alkaline treatment at elevated temperature. The curve describing the decrease in molecular weight for the mannans can be described by the sum of two first order reactions, showing that there are two dominating time scales: one fast at short times and one slow at longer times.
Alkaline degradation of softwood glucomannan has been studied before (Meier 1962; Hansson and Hartler 1970; Sjöström 1977), also observing a decreased degradation rate with longer times. The stabilization observed in previous works is attributed to the stopping reaction, where the end group is rearranged to a metasaccharinic acid type configuration. This reaction is specific to the peeling reaction. However, it is hardly possible for peeling to result in the fast initial degradation phase seen in our data. Therefore, the mechanism of alkaline hydrolysis of the backbone glycosidic linkages seems more likely. The two time scales are here proposed to correspond to the two degradation reactions: alkaline hydrolysis and peeling. Alkaline hydrolysis, resulting in a fast decrease in molecular weight, dominated during the first phase, whereas the slower peeling dominated in the second phase. Indeed, the initial rapid decrease in molar mass coupled with the increasing concentration of oligomers is evidential of ongoing hydrolysis; while detection of degradation products such as formic acid is typically a result of peeling reactions. An additional explanation, correlating well to our results, is that there are more and less stable structural regions, and that the weaker regions hydrolyze in the beginning and the more stable structures remain. Also an early study on kraft cooking of pine wood concluded that there must be other mechanisms rather than stopping reactions involved in the stabilization of GGM, and proposed that the remaining GGM might have a more orderly structure (Aurell and Hartler 1965). Thus, the sugar composition and structure of the mannans must be of importance for the degradation.
The use of model systems allows for systematic variation of key structural variables and decreases systematic errors arising from polysaccharide impurities that are present in native GGM. Here, the mannan polysaccharides KGM and LBG were studied; where the KGM chain is built up by Glc and Man units, with no Gal decorations, and LBG consist of only Man in the backbone and Gal substitutions. Also CMC was chosen because of the single sugar composition, the β-(1 → 4) glycosidic linkage, and the water solubility. However, the system is not ideal due to the carboxymethyl substitutions, but other water soluble polysaccharides with a pure glucan backbone, such as starch and methylcellulose, have drawbacks that are even more critical (different glycosidic linkages in starch, and thermoresponsive behavior of methylcellulose). An interesting observation was that CMC showed a first order degradation behaviour throughout the treatment. This could suggest that a possible stopping reaction is not the only explanation, however, the exact mechanisms for the CMC behavior requires more focused studies.
Temperature is of importance for the degradation behavior. A study of CMC in water showed that higher temperatures give a faster decrease in chain length, and the side groups was relatively temperature stable (Hiltunen et al. 2018). Also previous work on GGM degradation in alkali showed that a higher temperature speeds up the degradation but the shape of the curves was the same at 90, 100, and 110 °C (Azhar 2015). In the current work the main purpose was to investigate details between structure and degradation in alkali and a temperature of 90 °C was therefore considered sufficient.
A comparison between the mannans LBG and KGM is especially illuminating since these are pure polysaccharides with a similar starting molecular weight. Despite this, the shape of the curves describing the decrease in Mw differs and the initial degradation phase is longer for LBG compared to KGM. This is also reflected by the time constant τ1, which is about 5 times larger for LBG compared to KGM (and more than 10 times larger than for GGM). A larger value of τ1 correponds to a slower degradation rate meaning that LGB experience the slowest decrease of molar mass during the first phase. Comparing LBG and KGM, there are two aspects in the structure that can explain this, either that the pure Man backbone is more resistant than the Glc-Man backbone in KGM, or that the Gal side groups result in a stabilizing effect that slows down the first degradation phase.
Considering the Gal side group, it is here observed for both LBG and GGM that the amount of Gal in relation to Man is constant throughout the alkaline treatment which indicates that Gal substitution is significant for the stability of the structure. The Gal content in GGM is however affected by to the content of arabinogalactan, but since a similar behavior was observed for LBG this observation is believed to also resemble the GGM fraction in the spruce sample. Previous studies have also indicated a stable Gal side group in softwood GGM treated by kraft cooking (Croon and Barbro 1962; Wang et al. 2011). In Wang et al. (2011) they saw that during the initial impregnation of spruce wood the ratio of Gal in relation to Man actually increased, and for 2 h of harsher kraft cooking conditions the ratio was similar as in the starting material, but after 4 h the amount of Gal decreased. Nevertheless, this confirms that also in technical conditions the Gal sidegroup shows a high stability and is an important part of the structure. In this context, the effect of the interactions of Gal in GGM with other non-polysaccharide polymers on stabilzation cannot be neglected. The presence of native lignin carbohydrate bonds is one such example. Gal has been shown to have elevated presence in milled wood lignins (Lawoko et al. 2005) and coupling through C6 hydroxyl to lignin has been hypothesized. A bulky lignin presence connected to Gal could have its effects on backbone stability.
The other main structural aspect is the composition of the backbone sugars in the studied mannans. For GGM, the concentration of Glc in relation to Man increased during the alkaline treatment, whereas such increase was not clearly observed for the model system KGM. Furthermore, a previous study of softwood hemicelluloses during kraft cooking showed a stable composition and only a slight increase of Glc after a 40 min treatment (Meier 1962). Thus the effect from backbone is not obvious and if the stability of Glc and Man is the same the backbone composition is expected to be rather constant. The difference observed for GGM and KGM can either be explained by that GGM also have the Gal substitution which impact stability, that the distribution of sugars in the backbone differs significantly and that this affects the degradation, or that the extracted spruce sample contain another Glc rich polysaccharide that is more stable than GGM and significantly influence the Glc/Man ratio. Furthermore, in an early study on methyl glycosides is was observed that the glycosidic linkage in β-Glc is more reactive in alkaline conditions compared to β-Man and α-Gal (Janson and Lindberg 1960). This alone would speak for that Glc is the sugar unit in GGM that is the most sensitive to alkaline hydrolysis, however, their system is limited and does not contain neighbour sugars making the proposed stabilisation suggested in Fig. 1b impossible.
Chemical substitution on O2 and O3 is and additional factor possibly affecting the alkaline hydrolysis of backbone glycosidic linkages. Substitutions such as the carboxymethyl groups in CMC, or even acetyl groups in mannans, would inhibit hydrogen bonding of OH− at the hydroxyl groups. Nevertheless, acetyl groups are easily cleaved off in alkali and the analysis of organic acids showed that the major part of acetyl group were instantly released, which should result in minor effects on alkaline hydrolysis.
From the present experimens it is difficult to determine which structural effect is dominating, and therefore MD simulations were used to evaluate the properites of the specific mannan structures. Since MD simulations include no chemical reactions, we use computer modeling to evaluate selected stages along the proposed rection path (Fig. 1), namely (1) the possibility for OH− to be coordinated by the polysaccharide hydroxyl groups, a necessary step for deprotonation and subsequent nucleophilic attack; (2) the factors that could affect the stability of the reactive intermediate, thus promoting relegation; and (3) the accessibility of water to the oxygen in the glycosidic linkage, which is necessary to complete the hydrolysis.
The simulations showed that OH− was strongly coordinated by C2 and C3 hydroxyl groups in the backbone sugars and that Gal substitution was not a hinder for hydrogen bonding. Moreover, a more pronounced effect was that backbone Glc showed an increased coordination of OH− by the OH2 and OH3 hydroxyl groups compared to Man. This can be explained by that the OH2 hydroxyl group is equatorial in Glc, but axial in Man. Similar coordination has been seen previously in simulations of cellobiose, and has been suggested to induce deprotonation of cellulose in alkali solvents systems (Bialik et al. 2016; Lindman et al. 2017). Furthermore, Glc experienced an even higher coordination of OH− in the Gal substituted oligomer (MGLM), possibly a consequence of that the Glc and Gal groups can coordinate a OH− together as showed in Fig. S15. Since higher coordination number leads to higher probability for deprotonation, larger amounts of Glc in the backbone would lead to a faster degradation rate. This is supported by the observation that KGM degrades faster than LBG, which has a pure mannan backbone and by the reactivity observed for methylated glycosides in Janson and Lindberg (1960). However, it is not in agreement with the observation that the Glc/Man ratio in GGM increased during the alkaline treatment, however, that could suggest that Man rich regions is more rapidly cleaved off than Glc by subsequent peeling reactions. Thus, it seems likely that there are several mechanisms involved in stabilization and alkaline degradation of the backbone.
Alternatives for non-covalent stabilization of the reaction intermediate were presented in Fig. 1, and here the importance of intrinsic stiffness of the glycosidic linkage and the effect of Gal substitution were highlighted. With respect to the glycosidic linkages in the backbone, the simulations give at hand that while C-type linkages are less flexible than M-type linkages, and are further stiffened by the addition of a Gal side group substitution, the differences are too small to have any appreciable effect on degradability. In this work we have seen that rotation of the C5–C6 bond where there is a Gal substitution favors the gg conformation, which corresponds to a conformation where the side group is covering the mid glycosidic linkage in the backbone chain. This supports the hypothesis of a more stable region due to the formation of a “sandwich” structure. If this would significantly affect the stability of the intermediate structure, it would give an explanation why the Gal/Man ratio is constant throughout the alkaline treatment. Additionally, an important factor for the hydrolysis reaction is the accessibility of water, therefore this was evaluated by the MD simulations. However, no difference was observed between unsubstituted and substituted structures, or between Glc and Man units. This points to that this will not be a contributing factor for the observed differences in degradation. Finally, one important factor in the mechanism in Fig. 1a not considered here is the time scale of the different reaction steps. To completely verify the proposed explanations and reaction mechanisms further work with additional MD simulations and quantum chemical calculations are proposed.
The discussion has so far assumed that the pH is uniform and that the chains are totally separated from each other. This is a simplified situation and during chemical pulping the system is more complex, especially inside the wood chips where polysaccharide chains are located close to one another. In such system hydrogen bonds and other interactions are created between chains, and alkalinity gradients can be formed due to deacetylation and influence of charged groups. These conditions could be stabilizing on hemicelluloses such as glucomannans, since interactions to other chains might stimulate relegation of a reactive intermediate in a similar way as we have suggested for the galactose side chain and as described in Fig. 1. A lower local concentration of OH− will also be a stabilizing factor in a wood chip during alkaline pulping processes and to fully understand all reactions involved in a technical process requires more studies. Still, we think that this work helps us in the endeavors to understand hemicellulose reactions also in more complex alkaline systems.
From this study it has been showed that galactoglucomannan, glucomannan and galactomannan degrade according to different kinetics than carboxymethylated cellulose. The molecular weight of the mannans quickly decreases in a first phase, then the rate slows down. Here we suggest that this can be explained by more or less stable structural regions in mannans. One factor is the effect from composition and distribution of different backbone sugars, and hence also different types of glycosidic linkages, on the stability. Furthermore, substitution by galactose seem to result in a stabilizing effect, possibly by forming a “sandwich” structure with the backbone where a reversible state of an activated intermediate is stabilized. The MD simulations support this hypothesis since the most common conformation of the galactose side group is when it faces the middle glycosidic linkage in studied manno-oligosaccharides. Also the interaction with OH− was investigated by MD simulations, which showed that OH− was strongly coordinated by the hydroxyl groups on C2 and C3, a necessary initial step for deprotonation of OH2 and subsequent hydrolysis. It was observed that the interaction between manno-oligomers and OH− was structure dependent. Glc experienced a higher coordination of OH− compared to Man, and this further increased when also a Gal substitution was present, possibly increasing the probability for hydrolysis at these sites. However, this picture does not explain every detail pertaining to the degradation behavior, such as why the ratio Glc/Man increases significantly during degradation of GGM, but not in KGM. Evidently, there are other mechanisms at play simultaneously, which highlights the complexity of the degradation process.
In order to develop the technical processing of softwood in a way where we can extract/retain polymeric hemicelluloses, it is important to understand the mechanisms involved in degradation and stabilization of these polysaccharides, and in this work we have advanced this knowledge by combining experimental and theoretical methods to obtain a detailed view of the atomic-scale degradation mechanisms. An increased understanding of the degradation behavior is key for being able to utilize our resources in the best possible way and replace oil-based products for a more sustainable society.
The Knut and Alice Wallenberg’s research foundation is acknowledged for funding within the Wallenberg Wood Science Centre. The authors thank Dr. Pan Chen for helpful discussions concerning the NaOH simulation parameters. Computational resources were provided by the Swedish National Infrastructure for Computing (SNIC) at the PDC Center for High Performance Computing, KTH.
- Annergren GE, Croon I, Engström BF, Rydholm S (1961) Stabilization of spruce glucomannan in wood and holo-cellulose. Sven Papperstidn 64:386–393Google Scholar
- Aurell R, Hartler N (1965) Kraft pulping of pine. I. Changes in the composition of the wood residue during the cooking process. Sven Papperstidn 68:59–68Google Scholar
- Azhar S (2015) Extraction of polymeric hemicelluloses from spruce wood. Doctoral thesis, KTH Royal Institute of Technology, School of Chemical Science and Engineering, StockholmGoogle Scholar
- Ballou CE (1954) Alkali-Sensitive Glycosides. Adv Carbohydr Chem 9:59–95Google Scholar
- Berglund J, Angles d’Ortoli T, Vilaplana F, Widmalm G, Bergenstråhle-Wohlert M, Lawoko M, Henriksson G, Lindström M, Wohlert J (2016) A molecular dynamics study of the effect of glycosidic linkage type in the hemicellulose backbone on the molecular chain flexibility. Plant J 88:56–70CrossRefGoogle Scholar
- Bialik E, Stenqvist B, Fang Y, Östlund Å, Furó I, Lindman B, Lund M, Bernin D (2016) Ionization of cellobiose in aqueous alkali and the mechanism of cellulose dissolution. J Phys Lett 7:5044–5048Google Scholar
- Brännvall E, Lindström M (2007) The hemicellulose composition of pulp fibers and their ability to endure mechanical treatment. Tappi J 6:19–24Google Scholar
- Croon I, Barbro E (1962) The hemicelluloses in sulphate pulps from scots pine. Sven Papperstidn 65:595–599Google Scholar
- Gellerstedt G (2009) Chemistry of chemical pulping. In: Ek M, Gellerstedt G, Henriksson G (eds) Pulp and paper chemistry and technology. Pulping chemistry and technology, vol 2. Walter de Gruyter Gmbh and Co., Berlin, pp 91–120Google Scholar
- Meier H (1962) On the behaviour of wood hemicelluloses under different pulping conditions. Part 2. Spruce hemicellulose. Sven Papperstidn 65:589–594Google Scholar
- Podzimek S (2010) Light scattering, size exclusion chromatography and asymmetric flow field flow fractionation. Powerful tools for the characterization of polymers, proteins and nanoparticles. Wiley, HobokenGoogle Scholar
- Sjöström E (1977) The behavior of wood polysaccharides during alkaline pulping processes. Tappi J 60:151–154Google Scholar
- Sjöström E (1993) Wood chemistry: fundamentals and applications, 2nd edn. Academic, LondonGoogle Scholar
- Veguta VL, Stevanic JS, Lindström M, Salmén L (2017) Thermal and alkali stability of sodium dithionite studied using ATR–FTIR spectroscopy. Bioresources 12:2496–2506Google Scholar
- Wang Y, Lindström ME, Henriksson G (2011) Mild alkaline treatment activates spruce wood for enzymatic processing: a possible stage in biorefinery processes. Bioresources 6:2425–2434Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.