2.1. Differences in particle reactivity during saccharification assays of maize shoot powder
First a saccharification assay of the whole maize shoot powder (M) was carried out using the torus reactor Cinetore to analyze the saccharification yield coupled with real time monitoring of changes in the number and size of the particles through image analysis. Before degradation, the maize shoot powder was made of particles of different sizes and shapes (Figure 1a), including long rod-shaped particles. The longest particles were as long as 1,500 µm, (measured by hand).
In the control experiment, the stirring into the reactor caused the disappearance of 19% of the particles and the release of 2.1% of total sugars. This result suggests that some particles were broken down by the mechanical action of the screw and/or the water. After seven hours of saccharification, the number of particles observed was lower (Figure 1b). Changes in the total sum of gray levels relative to time T0 were computed (Figure 1c) to account for the number of particles in the reactor during saccharification. The number of particles rapidly decreased to about 75% after one hour. After seven hours, 26-28% of particles disappeared. In parallel, the amount of sugars released was evaluated from the hydrolysates (Figure 1d). The percentage of disappearance rapidly increased at first and then stabilized at 36-37% after two hours. This result is consistent with results obtained in non-pretreated maize . Clearly, some particles were recalcitrant to enzymatic hydrolysis while others disappeared or were modified.
Changes in particle size during the reaction were also studied. Figure 1e shows the granulometric curves at times T0 and 7 h. The granulometric curves normalized relative to time T0 showed variations both in the number of particles and in the decrease in size. A mode was observed at 90 µm for time T0 and 75 µm for time 7 h. Three size classes were built from these granulometric curves: [0-45 µm], [45-225 µm], [>225 µm]. The three classes represented respectively, 6%, 44% and 50% of the original sample at T0, and 8%, 47% and 45% of the sample after 7 h of hydrolysis. Changes in the number of particles in each class were evaluated after normalization relative to their original number (Figure 1f). A strong relative increase in small particles from 1 to 1.40 and a slight relative increase in medium particles from 1 to 1.06 were observed. In parallel, the number of large particles decreased slightly from 1 to 0.90. Several interpretations are possible. A certain proportion of large, medium, and small particles were fully degraded while others were recalcitrant. Some medium and large particles were degraded and broke into many small particles thereby increasing the relative proportion of small particles. Medium and large particles may be partly eroded during degradation or broken down into medium size particles. Whatever the size, some particles were recalcitrant as small, medium, and long particles were still present after seven hours of degradation. To better understand this heterogeneous behavior, maize shoot powder was sub-fractioned to isolate fractions that were expected to be more homogeneous according to the saccharification process.
1.2. Exploring variability of maize shoot powder
The enzymatic hydrolysis of each particle might be linked to the origin of the tissue in the maize shoot. To explore this hypothesis, dry fractionation was carried out to try to obtain fractions with specific tissue enrichment from leaves, pith, or rind. The particle size obtained in the fractions is not only the result of the efficiency of the grinder but is also due to interactions between plant structures and mechanical loading modes developed in the grinder . The particle size could thus reflect differences in grinding ability in relation to the mechanical properties of the tissues. On this basis, we chose a separation process mainly based on particle size characteristics [44, 45]. Air classification enabled separation of the original maize shoot powder into three fractions (fine, medium, and coarse). A subsequent step was undertaken to take advantage of differences in particle composition . Electrostatic separation, based on tribocharging, which is an interface phenomenon that depends on surface composition ,was then used to separate each fraction into two more fractions, giving a final total of six fractions (Figure 2). The coarse, medium, and fine fractions accounted for respectively 43%, 22% and 35% of the weight of the original sample (Table 1). Their median particle sizes (D50) were 310-315 µm, 159-201 µm, and 58-63 µm, for the coarse, medium, and fine fractions, respectively (Table 1). Size dispersion was still observed within each fraction but was less than in the original maize shoot sample, as shown by the narrower range of span values. Electrostatic separation led to no further substantial separation according to particle size, as shown by optical observations (Figure 2) and by laser granulometry (Table 1). The median particle size of samples recovered on the negative electrode side (coded “-”) was slightly higher than on the positive ones (coded “+”), in the coarse, medium, and fine fractions. As expected, the reduction in particle size was consistent with an increase in specific surface area measured by physisorption (Table 1). Moreover, the hue of the particles separated by electrostatic separation also differed from the negative fractions with more brownish hues than in the positive fractions.
The chemical composition of each fraction was assessed by focusing on the analysis of the main cell wall components (polysaccharides, lignins, ester-linked p-coumaric acid and ferulic acid) that are known to be involved in enzymatic sensitivity. If the total amount of sugars remained relatively stable between the fractions (about 66.7% ± 1.2%), the relative proportion of neutral sugar associated with hemicellulose (arabinose, xylose) was much more variable (relative standard deviation ranged from 8-15%). No major differences in glucose concentrations were observed. Cellulose crystallinity, determined in CP-MAS solid state NMR experiments, was around 25% and no significant difference between fractions was observed. Lignin content, measured using the Klason method, varied significantly from 14.52 to 19.56% in the fractions of cell wall material. Lignin structure was studied by thioacidolysis that estimates the monomer amount of guaiacyl (G), syringyl (S) and p-hydroxyphenyl (H) involved in β-O-4 bonds. The thioacidolysis yield ranged from 720 to 999 µmol.g-1 KL, suggesting different degrees of lignin condensation in the maize fractions. The S/G ratio of thioacidolysis monomers ranged from 0.9 to 1.42.
Major differences in composition were observed depending on the particle size of the fractions and opposed fine and coarse fractions. The coarse fractions, which could be seen as the most difficult to grind, showed higher concentrations of xylose and Klason lignin and also differences in lignin structure as evidenced by the lower S/G ratio. Lower concentrations of protein, galactose and uronic acid were also observed in the coarse fractions compared to the fine ones. Differences according to the electrostatic separation were also observed: the positively deviated samples (‘+’) had more acetyl groups, ester-linked p-coumaric and ferulic acid, and higher thioacidolysis yields, but lower ash content. Smaller differences were observed between the two coarse fractions compared to between the fine and medium fractions, in particular ash and phenolic acid contents. Electrostatic separation, which is based on surface properties, might be less efficient for bigger particles whose specific surface area is lower .
Likewise, particle size fractions isolated from the maize shoot powder differed in chemical composition, supporting the hypothesis that they are enriched in specific tissues. The higher concentrations of galactose and uronic acids in the finest fractions, associated with lower lignin content, suggests enrichment in pith from the stem . However the higher amount of ash, also expected in this case , was not observed in this sample set. Leaves were also expected to be more present in the finest fractions . In Poaceae, compared to the stem, leaves are characterized by less xylose and more protein, like in sorghum , more ash [51, 52], a lower S/G ratio, like in sugarcane, and less ester-linked p-coumaric . More leaves in the fine fraction can be assumed from lower xylose or higher protein amounts. The fine fractions were found to be the richest in lignin syringyl units (%S > 55%), despite their lower lignin content. This observation suggests that these fractions are enriched in leaf blade sclerenchyma, a tissue repeatedly shown to contain lignins rich in S units [34, 53]. In negatively deviated fractions (‘-‘), an enrichment in ash was observed combined with a brownish hue (Figure 2). This could reflect more leaves, even if protein enrichment was not that high. Nevertheless, based on our compositional analysis, we propose the following interpretation: coarse fractions could originate from the stem rind, while the fine fractions could be enriched in pith and leaves. The enrichment in leaves may also differ according to the electrostatic separation, with more leaves in the Mf- fraction. Medium fractions could contain tissues from both origins, again with more leaves in the Mm- fraction.
1.3. Degradation kinetics by imaging and chemical analyses: Cinetore experiments
The parallel release of sugar and changes in the number and size of the particles during saccharification was analyzed in Cinetore experiments. The coarse fraction considered corresponded to the whole Mc fraction (obtained before electrostatic separation) as the compositions of the two fractions Mc- and Mc+ were close.
1.3.1. Fractions at time T0
Examples of images at time T0 are given in Figure 3 for the five fractions. Although the same mass of particles was loaded in the reactor, the images of particles differed considerably. At T0, particles in Mc were mainly elongated with variable thickness and length. As expected, particles in the medium fractions (Mm+ or Mm-) were clearly more numerous and differed in shape: thin and elongated to almost isotropic and cuboid shapes. In the raw images (see Additional file 2), particles in the fine fractions (Mf+ or Mf-) were so numerous that they could hardly be distinguished. Considering the D50 measured by laser granulometry (60 µm) and the pixel size (8.2 µm), only a few pixels are required to observe one particle and their overlapping led to a low contrast between background and particles. The bright background attested to the high density of particles and to particles smaller than the pixel size. After subtracting the background (Figure 3), many isotropic particles were observed together with small pointed particles and a few elongated particles. For these fine fractions, the estimation of the relative number of particles from the total gray levels would be underestimated and should be considered relative to the number of particles larger than 8.2 µm.
For the fine fraction, the mode was found at 55-75 µm while it was 90-125 µm and 155-205 µm for the medium and coarse fractions, respectively. The mode corresponded to the D50 (Table 1) in the case of the fine fractions and was lower for the other fractions. It should be noted that particle size evaluated by image analysis is more sensitive to the smaller dimensions of the particles , i.e. the width rather than the length, compared to laser granulometry. In addition, the variations in gray level observed for large sizes in the case of fine fractions were caused by the overlapping of the numerous particles. To a lesser extent, variations in gray level were observed for small sizes in the medium and coarse particles, which actually corresponded to irregularities in their contour.
1.3.2. Changes in the number and size of particles during saccharification
After seven hours of saccharification (Figure 3), particles could still be seen in the images, regardless of the fraction concerned. Fewer particles were observed in the coarse fraction Mc, and the contrast was lower. In the fine fractions, the contrast was lower after seven hours compared to T0 and fewer particles could be distinguished. A decrease in the number of particles was also visible in the medium fractions. Considering the gray level granulometric curves normalized with respect to time T0, the lower intensity of the curves clearly confirmed the decrease in the number of particles in all the fractions, with the biggest decrease in the fine (Mf-, Mf+) and Mm- fractions. In the fine fractions, a decrease in particle size was evidenced with an absolute increase in the number of particles smaller than 25 µm, whereas a moderate decrease in particle size was observed in the coarse and medium fractions.
To investigate physical changes in the particles depending on the period of saccharification in more detail, the number of particles was deduced from the total amount of gray levels, and mean particle size was estimated from the gray level mean sizes (Figure 4a and 4b, respectively). In all the fractions except Mc, the decrease in the number of particles was always greater than the decrease in particle size. The kinetics were analyzed through the time needed to obtain a 50% reduction in the number of particles or in their size (t1/2): in most cases the reduction in the number of particles took place more quickly than the reduction in particle size.
The least change in the number and size of particles was observed in the coarse fraction Mc. The total decrease in gray levels between images after seven hours of degradation was 12% and the gray level mean size decreased by 8%. The size and number of particles decreased with quite similar t1/2 (t1/2 = 1.08 h and 0.82 h respectively).
Conversely, the fine fractions (Mf+ and Mf-) were the most impacted during saccharification: after 7 h, the total decrease of gray levels was 36% and 32% in Mf+ and Mf, respectively, and the gray level mean size also decreased by 20% and 17% in Mf+ and Mf-, respectively. For the sake of comparison, the decrease in the total gray level in the control experiments was 28% and 18% in Mf+ and Mf-, respectively. Particle disappearance was clearly enhanced by enzymatic hydrolysis. The size and number of particles varied consistently with a rapid and important decrease occurring in the first 15 min. These two phenomena were observed following similar kinetics in the Mf+ fraction, whereas in the Mf- fraction, the decrease in particle size was slightly slower (t1/2=19 min) than the decrease in the number of particles (t1/2=10 min).
Different patterns were observed in the medium fractions. In Mm- samples, after seven hours of saccharification, almost no significant change in particle size (9%) was observed, while the number of particles decreased by about 34%. Mechanical stirring alone could not explain such a discrepancy as the gray level decrease in control experiments was only 4%. This result suggests that saccharified particles were fully dissolved in the aqueous medium with no noticeable particle erosion or breaking. The reduction in the total number of particles was similar to that obtained in the fine fractions but was less rapid (t1/2 = 49 min for Mm- compared to t1/2 = 10 min and t1/2 = 7 min for Mf- and Mf+, respectively). This could suggest similar tissue composition but differences in surface accessibility (in relation to particle size) between Mm- and fine particles. In the Mm+ sample, the particle size reduction was also very small (10%) but the reduction in the number of particles was also smaller (20-24%). Particle changes in Mm+ were the slowest observed in these fractions, as indicated by the higher values of t1/2 (around 2.5 h). The reduction in both the number and size of the particles were moderate, even less than in the coarse fractions during the first 90 min.
Differences in the physical change pattern were next observed according to particle size. Fine particles were the most impacted, regardless of the criteria used (number of particles or size), the coarse ones the least. Concerning the disappearance of particles, the medium fractions were between the two but concerning the decrease in particle size, their behavior was closer to that of the coarse fraction.
1.3.3. Sugar release during saccharification
After seven hours of hydrolysis, sugar release was 22%, 34%, 38%, 52% and 55% in Mc, Mm-, Mm+, Mf-and Mf+, respectively (Figure 4c). The main differences were observed according to the overall particle size of the fraction, with coarse, medium, and fine fractions ranked according to increasing hydrolysis yield. Concerning electrostatic separation, the medium and fine fractions obtained at the positive electrode were slightly more degraded than those recovered at the negative electrode.
Concerning kinetics, the saccharification of the fine fractions was the most rapid as shown by the lowest t1/2 (around 6 min), while the coarse fraction had the highest t1/2 value (10.4 min). As suggested by Mansfield et al. , the smallest size fractions were hydrolyzed preferentially during the first stage of the hydrolysis reaction. Contrasted kinetic values were observed for medium fractions: saccharification of Mm+ was as slow as in the coarse fraction (t1/2 =10.1 min) whereas Mm- behaved like the fine fraction Mf-.
1.3.4. Coupled physical and chemical changes during saccharification
Saccharification was described by combining the analysis of saccharification yield, the reduction in the number and size of particles (Figure 5). Whatever the fraction, the maximum relative change in the number of particles (dotted lines) was still lower than the chemical saccharification yields obtained at the plateau. The times to reach 50% of the modification were always shorter for chemical changes than for physical changes. This result suggests that at least some particles were fully degraded, as evidenced by the decrease in the total number of particles, while other particles were only partially degraded. Sugar release could be mainly due to the hydrolysis of few particles potentially originating from highly hydrolysable tissues, such as pith parenchyma  or leaves .
1.4. Relationships between physicochemical characteristics, water mobility distribution and saccharification patterns
High sugar release prior to the reduction in the number and size of the particles, observed particularly in the Mm+ fraction, could be explained by the open porosity of the particles. The specific surface (Ssp) area of all the fractions measured by physisorption was systematically higher than the Ssp area calculated from granulometric curves (Figure 6), also supporting the presence of open porosity at mesoscale (2-50 nm pore size). A high correlation (R2 = 0.915) was observed between the values obtained with the two methods. Mm+ did not deviate from this relationship (Figure 6). However, the specific surface area measured by physisorption corresponded to the surface area available to krypton molecules and but not to the surface accessible to cellulase, whose mean size is about 5.9 nm . Moreover, values were determined in dry state at -195.8 °C, quite far from hydrolysis conditions implying swollen substrates.
Swelling properties and water retention capacity (WRC) whose values include both water at the surface and between the particles and water within the biomass were considered (Table 3). The coarse fractions had the lowest values for both swelling and WRC, and the fine fractions the highest. A positive correlation was found between these properties (in particular swelling capacity) and sugar release (R2=0.915) and the reduction in particle size (R2=0.968) after seven hours of saccharification. However, here again, Mm+ did not differ from Mm- and their values were intermediate between those of the coarse and the fine fractions (Table 3). Differences in swelling and WRC did not reflect the specific saccharification pattern of the Mm+ fraction.
Low-field nuclear magnetic resonance (LF-NMR) was used to obtain information on water-biomass interactions at the molecular-scale. relaxation times of water proton distribution and their relative proportions were determined. To take advantage of the high sensitivity of this approach to any changes in sample supramolecular structure associated with water distribution and diffusion specificities, water-fraction interactions were investigated at five different moisture contents (MC) ranging from 15% to 67%. It should also be noted that at low water content (15% MC), the NMR signal is expected to be dominated by water arising from the hydration shell of macro-molecules and from water in interaction within the small pores of matrices, reporting on their microstructural specificities.
As illustrated in Figure 7, the analysis of relaxation curves led to profiles with multiple distinct peaks. Because no simple direct relationship exists between components and the morphological compartments in biological tissues , each peak can be preferentially assigned to a pool of water at a given range of mobility corresponding to specific molecular environment/interactions. Both the number of components, the relative surface area, and the individual mean relaxation time values associated with these peaks changed with increasing water concentration and differed between the five fractions Mc, Mm-, Mm+, Mf- and Mf+. Overall, increasing moisture content led to an increase in the number of components (from 1-2 at 15% MC to 5-6 at 67% MC) with, in most cases, a shift towards higher water mobility modes, which could partially result from swelling. However, it should be noted that a short mode, centered around 2 ms, was present from 40% to 67% MC with only small changes in value/mobility, indicating that the physical-chemical environment associated with the highly constraint water molecules can remain relatively unaffected in this range of moisture contents. In any case, differences between profiles of maize fractions were observed (Figure 7), regardless of the particle size, the electrostatic deviation, and the water content.
The water content associated with the [5. 15[ pore size range (in nm) that typically represents the average diameter of enzymes, was investigated in more detail (Figure 7, shaded area). The water content associated with this pore size range at 67% MC was positively correlated with the saccharification yield (R2 = 0.881) (Figure 8). It was also correlated with the water content associated with pore sizes below 4.3 nm at 15% MC (R2=0.992). At 15% MC, the profiles of the five fractions showed two peaks with short relaxation times (≤ 4.3 ms). The population , associated with the component, was positively correlated with specific surface area (Figure 9-a, R2=0.969), suggesting that this component is mainly influenced by water molecules located at the surface of particles. The of Mm- fraction was of the same order as that of the fine fractions, whereas the value measured for Mm+ was between the fine and coarse ones. Therefore, Mm- differed from Mm+ by more constrained water, which is hypothesized to be essentially located at the surface of particles. Despite the very small change in the shortest relaxation time value, it proved to be positively correlated with lignin content (Figure 9-b, R2 = 0.983). This could mean that the hydrophobic character of lignin may induce an increase in water mobility, in turn resulting in an increase in the relaxation time centered from around 1.10 ms to 1.35 ms. The Mm+ fraction, with its higher lignin content, could keep the same macroscopic structure while still allowing accessibility to sugar hydrolysis.
Even at low moisture content, water/fraction interactions could be related to sugar release. The decrease in the number of particles was negatively correlated with the shortest centered around 1.2 ms observed at 15% MC (Figure 9b; R2 = 0.990) and the lignin content (R2 = 0.983). Moreover, the population , associated with the component, was positively correlated with the reduction in mean particle size (Figure 9a; R2 = 0.982). Thus, these two relaxometric parameters ( and ) could be early (at low MC) indicators of the saccharification potential, particularly when no significant swelling of the biomass matrix has yet occurred.