Stabilising mannose using sodium dithionite at alkaline conditions

Materials

D-(+)-Mannose from wood ≥99% and D-mannitol ≥99% as well as Glycine ReagentPlus^® ≥99% and Deuterium Oxide 99.9 atom% were procured from Sigma Aldrich Co. (St. Louis, MO, USA). Sodium hydrosulphite ca. 85% technical grade was procured from Acrōs Organics (NJ, USA). Sodium hydrogen carbonate AnalaR NORMAPUR was procured from VWR International BVBA (Leuven, Belgium). Ethanol absolute AnalaR NORMAPUR was procured from VWR International (Fontenay-sous-Bois, France).

Reduction of mannose using sodium dithionite

Samples of mannose were reduced with sodium dithionite in batches using different processing parameters in a system inspired by the conditions and setup of de Vries and Kellogg (1980) for the reduction of aldehydes; samples of mannose and varying amounts of dithionite depending on the experiment were added to a round bottom flask in which deionised water at an alkalinity set by the specific experiment (using one of the buffers detailed below) had been added. The round bottom flask with buffer was placed in an oil bath and preheated to the requisite temperature before the mannose and dithionite were added. The mannose concentration was 25 mg ml⁻¹. The solution was bubbled continuously with nitrogen during the reaction to keep an inert atmosphere. Aliquots were collected after 0 min, 15 min, 30 min, 60 min and 120 min, which were acetylated and analysed using GC-FID. Later experiments used a different setup; mannose and varying amounts of dithionite and buffer were added to vials, to which solutions of 10% deuterated water and 90% deionised water were added. The mannose concentration was 60 mg ml⁻¹. The vials were sealed with Teflon tape (but could not be bubbled continuously for an inert atmosphere) and placed in an oil bath for 120 min. After the reaction, the solutions were cooled. Ethanol was added as a standard after which the solution was analysed using inverse gate ¹³C-NMR.

For either reaction system, three variables were varied: temperature, alkalinity and equivalents of dithionite to mannose. The temperatures used were 70°C, 85°C or 100°C. Alkalinity was set to pH 7 (using no buffer), pH 8 (using a sodium bicarbonate buffer, unadjusted), pH 10 (using a glycine buffer, adjusted using NaOH) or pH 14 (using 1 M NaOH). The dithionite to mannose ratio was 0.5:1, 1:1, 2:1 or 4:1, molar basis.

Rather than performing a full analysis of variance, a reduced set of experiments was performed, where each variable was varied individually from a common baseline. For clarity, the full set of experiments is detailed in Table 1. Each experiment was performed twice.

Quantification using GC-FID

Aliquots of each sample were acetylated and dissolved in ethyl acetate. The acetylated samples were then analysed on an Agilent 7890 (Agilent Technologies, Santa Clara, CA, USA) instrument with a split injector, an HP-5 column [30 m, 320 mm inner diameter (I.D.) and 0.25 mm thickness], a flame ionisation detector and the OpenLab CDS ChemStation software (Agilent Technologies, Santa Clara, CA, USA). The oven was set to 80°C and held at 1 min followed by a ramping of 30°C min⁻¹ to 200°C, followed by a ramping of 5°C min⁻¹ to 230°C, followed by a ramping of 30°C min⁻¹ to 250°C, held for 3 min. Meso-erythritol was used as internal standard for the quantification.

Select acetylated samples were also analysed on a ThermoQuest TRACE 2000 GC (ThermoQuest Corporation, Austin, TX, USA) instrument with a split injector, a DB-5MS column (30 m, 320 mm I.D. and 0.25 mm thickness), a Finnigan TRACE MS (Finnigan Corporation, San Jose, CA, USA) mass spectrometer (MS) and the Xcalibur software (Finnigan Corporation, San Jose, CA, USA). The oven was set to the same temperature ramping program as for the GC-FID analysis mentioned above.

Using the weight of the molecular ions and the fragmentation patterns observed in GC-MS as well as corroborating suggestions from the NIST library, the peak at a retention time of 9.7 min with a molecular ion of 375 m/z was identified as acetylated mannitol, while peaks at retention times of 9.0 min, 9.1 min and 9.4 min with molecular ions of 331 m/z were identified as isomers of acetylated mannose. Quantification was done by normalising the sum of the respective integrals to the meso-erythritol standard and comparing with a standard curve. See Figure 2 for an example spectrum from the Centre, bicarbonate experiment.

Figure 2:

Example spectrum for quantification of mannose and mannitol using GC-FID, using the parameters for Centre experiment, bicarbonate (85°C, pH 8, 1:1 dithionite mannose ratio, 2 h).

Spectrum is cut for brevity. Select peaks were identified from similarly acquired GC-MS spectra using their molecular ions and suggestions from the NIST library.

Quantification using ¹³C-NMR

Deuterated samples with ethanol added as the internal standard were analysed on a Bruker 400 DMX (Bruker Corporation, Billerica, MA, USA) instrument equipped with a 10 mm Bruker BBO probe (Bruker Corporation, Billerica, MA, USA) by applying the carbon inverse gate zgig pulse program, with p1=8 μs (resulting in a 50° tip angle); d1=20 s; aq=1.6 s; ns>3100; rg=32 k; pl12=10 dB. Finally, each FID was processed with 3.0 Hz exponential apodisation, manual phase adjustment and automatic baseline correction using a third order polynomial function. This processing was done using the software MestReNova (Mestrelab Research, Santiago de Compostela, Galicia, Spain).

An example spectrum from the Centre, bicarbonate experiment is shown in Figure 3. The peaks at 61.0 ppm and 63.2 ppm were identified as the C6 peak of mannose and the overlapping C1/C6 peaks of mannitol, respectively, by comparison to spectra of pure standards. The peaks at 57.4 ppm and 16.9 ppm originate from ethanol. Quantification was done by comparing the specified mannose and mannitol integrals with the 16.9 ppm ethanol integral and multiplying with the molar amount of added ethanol.

Figure 3:

Example spectrum for quantification of mannose and mannitol using NMR, using the parameters for Centre experiment, bicarbonate (85°C, pH 8, 2:1 dithionite mannose ratio, 2 h).

In order to validate the selection of relaxation delay time, a relaxation delay study was performed. Mannose, mannitol and ethanol were dissolved in high concentration in deuterated solvent and subjected to the t1irpg pulse program with p1=14 μs; d1=60 s; vdlist=[0.1; 0.3; 0.5; 0.7; 0.9; 1.5; 2.0; 3.0; 5.0; 7.0; 10; 15; 25; 35; 45; 55]s; aq=1.6s; ns=64 (at each level of vd); rg=32 k; pl12=10 dB. The resulting stack of spectra is shown in Supplementary Figure S16 together with their respective variable delays.

By fitting the relaxation delay data obtained from the t1irpg experiment to ln(I0−It)=ln(2I0−tT1), which can be derived from the Bloch equations [see (Traficante 1992)], the T₁ relaxation delay times for the respective nuclei of interest were found to be 0.4 s for the C6 mannose peak, 0.4 s for the C6 mannitol peak and 6.3 s for the aliphatic ethanol peak. Ninety-nine percent relaxation in the quantification experiment is therefore achieved after 6.3 * ln(100–99 cos(50°))=22.6 s. While this is higher than the pulse repetition time (21.6 s) used in the experiment, the difference is lower than 0.25% of relaxation.

Yields of mannitol at varying temperatures, concentrations and alkalinities

The viability of sodium dithionite as a reductant of mannose was determined by applying dithionite to mannose using a variety of processing parameters as shown in Table 1 in the Experiment section. Yields of mannitol and conversions of mannose were then determined using GC-FID as well as NMR.

The starting point for the study was to vary alkalinity, temperature and dithionite concentration independently and measure the effects on mannitol formation using GC-FID. For each experiment, aliquots were collected and analysed after 0 min, 15 min, 30 min, 60 min and 120 min.

It was found that alkalinity is crucial to the viability of the reaction. Comparing the GC-FID results obtained by varying alkalinity while keeping the other parameters constant, one can find that only small amounts (less than 1% reduction yield) of mannitol are formed at pH 7 and that no mannitol (0% reduction yield) is formed at pH 14 (see Figure 4) – at the latter pH, the mannose reactant is also seen to completely degrade (see Supplementary Figures S12–S15). This observation is in agreement with the conclusions of Regnfors (1957), who stated that sodium dithionite would be unviable at pulping conditions while further predicting degradation at lower alkalinities, with later studies finding “mildly basic” alkalinities in general and a sodium bicarbonate buffer specifically to be preferable when working with dithionite (de Vries and Kellogg 1980). At pH 8 in this study, however, a reduction is seen to take place, and non-negligible amounts of mannitol are formed with a final yield of 17%. The full GC-FID results for these experiments are presented in Figure 4 and the corresponding chromatograms after 120 min are presented in Supplementary Figures S2, S3, S12–S15. Supplementary Figure S1 is also provided for reference.

Figure 4:

Yield of mannitol formed at various time intervals by varying alkalinity, with GC-FID used for quantification.

All experiments were done using a temperature of 85°C and a 1:1 molar ratio of dithionite to mannose. Error bars depict standard deviations of double tests.

From the initial GC-FID experiments, it was also found that yield of mannitol at pH 8 is improved by increasing the temperature (although the difference between 85°C and 100°C was not statistically significant after the full 120 min) as well as by increasing the dithionite concentration. Just as for alkalinity, aliquots were procured and analysed multiple times during each reaction, where it was found that an increase in temperature had a particularly large effect at shorter reaction times, see Figure 5. This temperature asymptote could suggest that all dithionite has been consumed or degraded after 30 min of reaction at 100°C. Such an explanation is corroborated by the study by Lakshmi Veguta et al (2017), where the inorganic stability of the sodium dithionite salt was investigated, and remaining dithionite was found to decrease from 50% to 20% when the temperature was increased from 80°C to 100°C at pH 9. Differences in yield at varying dithionite concentrations, as seen in Figure 6, are less pronounced, with no significant differences in the yield between the 1:1 and 2:1 experiments until the final time point. The 0.5:1 experiment, on the other hand, shows a significantly lower yield throughout. The corresponding chromatograms for all of these experiments after 120 min are presented in Supplementary Figures S4–S11. Supplementary Figure S1 is also provided as a reference.

Figure 5:

Yield of mannitol formed at various time intervals by varying temperature, with GC-FID used for quantification.

All experiments were done at pH 8 and a 1:1 molar ratio of dithionite to mannose. Error bars depict standard deviations of double tests.

Figure 6:

Yield of mannitol formed at various time intervals by varying molar ratios of dithionite to mannose, with GC-FID used for quantification.

All experiments were done using a temperature of 85°C and a pH of 8. Error bars depict standard deviations of double tests.

Having identified alkalinity as a crucial parameter, the next step was to investigate the yield of mannitol formation at an intermediate alkalinity between pH 8 and pH 14. It has previously been shown by Lakshmi Veguta et al. (2017) that the stability of the dithionite salt increases with pH until it reaches a maximum between pH 11 and pH 13. In order to investigate if the higher stability of dithionite at these alkalinities could improve reduction efficiency, a new set of experiments was performed. These experiments used the same conditions as the GC-FID experiments, except that the dithionite to mannose ratios were doubled, the pH 7 and pH 14 experiments were obviated and inverse gate ¹³C-NMR was used for quantification. In addition, five more experiments were added to this new series, using the same variations in temperature and dithionite to mannose ratios but conducted at pH 10 rather than pH 8. The results of all these experiments are presented in Table 2 together with the previous GC-FID results. The corresponding spectra for all NMR experiments are presented in Supplementary Figures S18–S37, with Supplementary Figure S17 also provided as a reference for the chemical shifts. As previously mentioned, GC-FID chromatograms are provided in Supplementary Figures S2–S15, with Supplementary Figure S1 provided for reference.

Comparing the NMR results at pH 8 and pH 10 at a standard temperature and dithionite concentration, the yields are seen to be similar (16% and 18%, respectively), while the losses of mannose are seen to be higher at pH 10 (42% and 73%, respectively). Increasing the dithionite concentration at either alkalinity is seen to give marginal increases in yield, with the pH 8 experiments giving a 17% yield at a 4:1 ratio of dithionite to mannose while the pH 10 experiments gave a 21% yield at a 4:1 ratio of dithionite to mannose. This would suggest that a 2:1 ratio of dithionite to mannose is sufficient for the reaction. Crucially, the low temperature pH 10 experiment is seen to have the highest yield out of all the NMR experiments at a 26% yield and 55% losses. It is our hypothesis that the system in the present study can be seen as a reaction system where mannose can either be degraded by the alkali – the kinetics of which is increased with alkalinity and temperature – or reduced by the dithionite – the kinetics of which is also increased with alkali and temperature. The exact nature of the observed alkaline degradation in this study is unknown, but alkaline degradation of monosaccharides is generally accepted to proceed through the formation of multiple enediol intermediaries which are further degraded through a combination of benzilic acid rearrangement, α-dicarbonyl cleavage and retro-aldolisation (de Wit et al. 1979; de Brujin et al. 1986; Yang and Montgomery 1996). This generally accepted system is illustrated schematically in Figure 7. This scheme does not, however, account for further reaction of the degradation products with sodium dithionite or other high molecular-weight degradation products in alkali which to this day remain unknown (Liggett and Deitz 1954; de Brujin et al. 1986).

Figure 7:

Schematic illustration of how mannose could behave in the presence of dithionite as performed in this study.

Upon ring-opening of the molecule, two reaction paths are possible, one in which the saccharide is reduced to its alditol form and one in which the saccharide is fragmented and lost. As dithionite is stabilised by alkali, and degradation is induced by alkali, the kinetics of both paths can be expected to depend on a combination of alkalinity and temperature.

According to such a reaction system, the increase in pH from 8 to 10 would increase the kinetics of reduction due to the increased stability of dithionite but would also increase degradation of mannose. When a lower temperature is used, however, the reaction with lower activation energy would be promoted. If the activation energy of reduction is lower than that of the degradation, this would explain the maxima observed in the data.

Comparing analysis methods, NMR yields are seen to be lower than GC-FID yields when the same experimental parameters are used. At a 2:1 ratio of dithionite to mannose at pH 8 and 85°C, the yield is 25% for GC-FID and 16% for NMR, and at a 1:1 ratio at the same conditions, the yield is 17% for GC-FID and 16% for NMR. The deviation at the 1:1 ratio is not statistically significant. However, while part of the deviation at the 2:1 ratio could be attributed to the high standard deviation of 3.2% at that GC-FID data point (the corresponding value for the NMR data point is 1.2%), there is still a significant difference in yield at the 2:1 ratio. This might be due to the fact that the NMR experiments could not be done in an oxygen-free atmosphere. This might have caused additional degradation of dithionite, as previously detailed by de Vries and Kellogg (1980).

Relevance of results to the industrial scale

These results show that sodium dithionite could indeed be used to protect carbohydrates, although not at pulping alkalinities. Interestingly, one of the buffers used in this study – sodium bicarbonate – is also present in green liquor (in the form of sodium carbonate). One possible industrial concept would therefore be to use part of this liquor as buffer in order to perform the reduction in a pre-impregnation step. If needed, the alkalinity could be further increased by adding white liquor. Possible temperatures for the impregnation step could be 100°C using a pH of 8 or 70°C using a pH of 10 depending on whether kinetics or reduction yield is being optimised. Oxygen levels in the wood chips could be lowered by steaming, with the remainder being consumed using a slight excess of dithionite.

The preferred dithionite charge in such a pre-impregnation merits a discussion of its own. In this study, 2:1 molar amounts of dithionite to mannose were seen to provide good results. As the molecular weight of sodium dithionite is close to that of mannose, charges of dithionite between approximately 50 w% and 400 w% were used in this study. In an industrial scenario on wood chips, however, the charges would need to be adjusted with consideration to the relative amount of glucomannan as well as the average molecular weight of the glucomannan.

Normal softwood contains 19% (galacto)glucomannan by weight with an average molecular weight of 14.7 kDa (Teleman 2009). For such a material, a 2:1 dithionite charge would require: w%=2×0.19×174.114 700=0.45% (4.5 g kg−1) on wood. This does not take diffusion gradients or loss of dithionite due to reduction of other substances into consideration, of course, but could serve as a starting point for further optimisation.