Sugarcane bagasse (SCB) is an underutilized byproduct, rich in cellulose and hemicellulose (mainly xylan) with limited amounts of extractives and ash, and is hence suitable for upgrading. Therefore, the present work aims to develop an integrated scheme for the separation of xylan, lignin and cellulose, to upgrade the raw material. The separation is followed by enzymatic conversion of xylan into XOS, which to date has been made to limited extent from this material [49, 50]. The used methods combine physico-chemical treatments (pressure, alkali and temperature) for xylan extraction, precipitation allowing separation of xylan from lignin, and acidic cleaning of the cellulose from the solid residuals (Fig. 1).
Chemical characterization of the sugarcane bagasse
The first criterion for product valorization from lignocellulosic biomass is the suitability of the raw material, with high hemicellulose and cellulose content vis-a-vis its overall composition. SCB is previously reported to consist of approximately 70% carbohydrates [10, 11].The overall composition of the SCB-batch used here (Table 1) and the distribution of neutral monosaccharides (Table 2) shows that the results are in agreement with the composition reported by other groups [12, 51]. The total carbohydrate content of the batch was 67.3% of the dry weight (including uronic acids, Table 1), with the predominant components being glucose (36.6%, assumed to originate from cellulose) and xylose (25.7%, from the hemicellulose xylan) (Table 2). The resulting xylose:arabinose ratio was 11:1, and the xylose:glucuronic acid ratio was 14:1. This relatively low degree of substitution indicates good potential for endoxylanases to hydrolyze non-substituted parts of the xylan backbone into XOS.
Extraction of xylan and separation from lignin
Thermochemical pretreatment with alkali combined with autoclaving
The SCB was pretreated with 0.25, 0.5 and 1 M of NaOH at 121 °C (1 bar) for 30 and 60 min to investigate the effect of the alkali concentration and reaction time on the xylan and lignin solubilization. As seen in Fig. 2A–B, the concentration of alkali exhibited a significant influence on the glucose/xylose ratio in the remaining residue, with higher relative amount of glucose in the insoluble fraction after extraction with higher concentration of alkali. Solubilization of xylan from SCB thus increased with increasing concentration of alkali (Fig. 2D). The 30 min treatment with 0.25, 0.5 and 1 M NaOH released xylan fractions with a dry weight corresponding to 0.08 ± 0.07, 1.12 ± 0.28 and 7.22 ± 0.26% of the dry starting material, respectively, solubilizing 0.29 ± 0.09, 4.39 ± 0.2 and 28.4 ± 0.50% of the xylan present in the SCB (p-value < 0.0002). Extension of the treatment to 60 min, resulted in a small increase in solubilization, with xylan fractions corresponding to 0.30 ± 0.09, 1.86 ± 0.19 and 8.45 ± 0.22% of the dry starting material, which is 1.16 ± 0.37, 7.30 ± 0.74 and 33.31 ± 0.88% of the xylan content, respectively (p-value < 0.00004). After the treatment, a solid fraction enriched in cellulose was left. The increase in NaOH concentration from 0.25 to 1 M also resulted in co-extraction of lignin, from zero (using 0.25 M NaOH) to 57.7 ± 0.21% of the total lignin after 30 min and to 64.7 ± 0.52% after 60 min pretreatment with 1 M NaOH (Fig. 2C). The sugar composition of the obtained xylan was not significantly affected by the increased extraction time (Table 3).
Introduction of three consecutive extractions with 1 M NaOH, further increased the fraction of extracted xylan (monitored as dry weight) (Fig. 3A, B and D) which after repeated 30 min treatments, resulted in an accumulated release of 6.7 ± 0.10, 8.8 ± 0.21 and 10.9 ± 0.01% of the dry starting material, after the first, second, and third extraction respectively, corresponding to 26.3 ± 0.41, 35.0 ± 0.84 and 42.9 ± 0.03% of the xylan in the raw material (p-value < 0.0002). Three repeats of the 60 min treatment increased the extraction yield to an accumulated weight corresponding to 8.4 ± 0.55, 12.1 ± 0.68 and 13.0 ± 0.45% of the dry starting material (corresponding to 33.2 ± 2.2, 47.8 ± 2.7 and 51.1 ± 1.8% of the xylan in the raw material (p-value < 0.008)). The accumulated lignin release in the respective treatment was in this case 66.7 ± 3.9 and 76.7 ± 1.0% of the total lignin present in the raw material after 30 and 60 min repeated pretreatments, respectively (Fig. 3C), showing that the ratio between xylan and lignin remained the same, irrespective of the extension in time to 60 min. The xylose/arabinose ratio is increasing in the remaining solid residue after each extraction, showing that the substitution degree is reduced in the residuals after each extraction (Table 4).
Both NaOH concentration and time were thus shown to influence the xylan extraction efficiency at the selected temperature (121 °C). The increase in xylan yields with time showed that a longer extraction time increased the overall yields, but that this also led to a corresponding co-extraction of lignin. The extraction time was, however, not as significant as the concentration of alkali for releasing xylan. To investigate the effect of a further increase in NaOH concentration the consecutive experiment was repeated using 2 M NaOH, and 60 min extraction time (Fig. 4). This condition led to release of 17.4 ± 0.62, 20.1 ± 0.04, and 21.7 ± 0.28% of the dry starting material, which based on the xylan content in the material would correspond to an accumulated yield of 68.7 ± 2.46, 79.3 ± 0.17, and 85.6 ± 1.10% of the xylan in the bagasse (p-value < 0.004). This repeated pretreatment also solubilized 69.65 ± 0.21, 78.31 ± 0.35, and 84.13 ± 0.33% of the lignin.
To separate xylan and lignin, xylan was subsequently precipitated with ethanol and washed, while most of the lignin remained in the liquid phase (Fig. 1). A significant amount of the lignin is thus released at NaOH concentrations of 1 M and higher, and is kept in the liquid phase after the downstream precipitation step, allowing separate use of xylan and lignin. Calculation of the lignin content showed that 80% of the solubilized lignin was kept in the liquid phase after precipitation, also motivating capture of this stream in a biorefinery perspective. Although outside the scope of this study, capture of lignin from this step could be explored by membrane technologies , to allow further valorization of this material.
An added advantage of the 2 M NaOH extraction procedure was that this resulted in a relatively pure residual cellulose fraction (see further below). The cellulose purity in all samples of sugarcane bagasse pretreated with 2 M NaOH was twice as high as in the raw material due to removal of large amounts of both lignin and hemicellulose.
Xylan yields in the same range as in this work was reported by Jayapal et al. , who obtained 85% xylan recovery, although using even higher concentration of NaOH (12% or 3 M) combined with steam processing. Peng et al. , have reported relatively high yields using a different approach, applying sequential extractions of hemicellulose from bagasse with increasing concentration of NaOH in the sequence, starting with water, then using 1% (0.25 M) NaOH and finally 3% (0.75 M) NaOH at 50 °C for 3 h in each round of extraction. This approach released 14.3, 32.5, and 28.1% of the hemicellulose in each round of extraction (releasing 74.9% of the hemicellulose), and 2.2, 31.5, and 20.4% of the lignin (releasing 54.1% of the lignin) , which is slightly lower than achieved here, but using a lower amount of NaOH in the process.
Pressurized liquid extraction (PLE)
To explore the possibility to decrease the concentration of NaOH by introducing high pressure (100 bar), pressurized liquid extraction (PLE) was chosen as an extraction technology, previously reported as a green alternative . Another interest was to investigate if the high pressure would change the ratio between extracted xylan and lignin.
An experimental design (Box–Behnken type) was set up, varying the three parameters temperature (50, 100, and 150 °C), alkali concentration (0, 0.05 and 0.1 M NaOH) and extraction time (10, 20 and 30 min) at three levels. The model fitting plots (Additional file 1: Figure S1) show that the values of R2 are greater than 0.9 and Q2 ranges between 0.6 and 0.8 which indicates valid models and the quadratic equation well represents the system under the given experimental domain. The coefficient plots indicate that alkali concentration and temperature have direct and positive influence on solubilizing xylan while decreasing solubilization of cellulose. The contour plots (Additional file 1: Figure S2, showing remaining residues in the solids) revealed that, at the conditions chosen, high temperature and high NaOH concentration enhanced solubilization of xylan (i.e., resulted in lower residual amounts) and decreased solubilization of cellulose (i.e., higher concentration in the residuals). Long extraction time at high temperature and high NaOH concentration significantly increased lignin solubilization. An optimizer function, based on a simplex algorithm with a non-linear desirability, was applied to find conditions with minimum responses of hemicellulose and lignin, combined with a maximum response of cellulose. The best condition was found to be at 150 °C, 0.1 M NaOH, with an extraction time of 22 min.
The extraction yields of xylan at the proposed conditions were however low, with a maximum yield of 16.7% [experiment No.4 (150 °C, 0.1 M NaOH, 20 min), Table 5]. However, this yield is higher than what was obtained using 0.5 M NaOH in a single 60 min extraction at 1 bar (which yielded 7.3 ± 0.74% of the xylan, see also above), showing that the increased pressure indeed reduced the need of alkali. A higher severity factor (increased temperature or alkali concentration which was not possible in our study as that might damage the PLE equipment) is, however, probably necessary to release more xylan from SCB. A further increase in the extraction temperature (to 180–190 °C) reported by Kilpeläinen et al  showed to be useful to reach higher yields in extractions of xylan from birch sawdust. Lignin was, however, efficiently extracted at the conditions used here (experiment 4, Table 5), showing that PLE may be considered for early lignin removal, as only 13% lignin was remaining in the residual residue after the extraction.
Monosaccharide composition in extracted xylan fractions
To investigate the influence of the extraction conditions on the obtained xylan structure, eight xylan fractions from the various trials were selected for composition analysis. These included the three 60 min cycles (121 °C, 1 bar) extracted by 2 M NaOH, and by 1 M NaOH, respectively, as well as the 60 min extraction by 0.5 M NaOH (121 °C, 1 bar), and the 20 min PLE extraction by 0.1 M NaOH, at 100 bar, 150 °C. The xylan obtained after ethanol precipitation in the respective fraction was hydrolyzed and the monosaccharide composition was determined (Table 6).
The major monosaccharide in all extracted xylans was xylose, which dependent on the extraction conditions comprised 48.5 ± 1.12–76.4 ± 0.81% of the total monosaccharides in the different fractions (Table 6). Varying amount of glucose (4.0 ± 0.12–21.9 ± 0.67%) was observed, indicating some cellulose solubilization. In addition, smaller amounts of arabinose (5.3 ± 0.11–7.9 ± 0.13%), uronic acids (1.3 ± 0.04–4.05 ± 0.8%) and galactose (0.11 ± 0.04–0.6 ± 0.03%) were verified in the eight extracted xylan fractions, confirming presence of substituents. Some variation in the xylose/arabinose and xylose/uronic acid ratio is observed (Table 6), leading to slightly different predicted xylan structures (Fig. 5). The xylan should, however, according to the scheme for naming xylan developed by Ebingerova , based on the ratio of arabinose and glucuronic acid, in all cases be classified as glucuronoarabinoxylan (GAX). Variation in substituent groups of SCB xylan has previously been shown to depend on both the SCB variety  and the conditions of the extraction method . In our study, it is clear that the conditions affected the substituent content. The xylose/arabinose ratio in extracted xylans increased from 11.2 ± 0.31 to 14.1 ± 0.16 (in the 2 M NaOH consecutive process) and from 9.2 ± 0.02 to 13.6 ± 0.08 (in the 1 M NaOH consecutive process) with increasing treatment cycles. This suggested that the degree of arabinose branching decreased in every cycle. Also, the degree of arabinose branching decreased with increasing alkali strength, which in the analyzed samples show highest arabinosylation in the 60 min, 0.5 M NaOH extraction. No such clear trend was seen for the glucuronic acids, but their content clearly varied in the different fractions.
Xylans in SCB have previously been defined as acetylated glucurono-arabinoxylans (GAX) with a β-1,4 linked xylose backbone carrying substitutions at the positions C2 and C3 including arabinosyl, uronic acid and acetyl groups [35, 46, 58, 59]. The xylose content of the polymer has a reported range from 43 to 93%, with arabinose being the most common substituent linked to the xylan backbone by α-1,2 or α-1,3 linkages . This is in accordance with the xylose range and arabinose content in the different fractions of this work (Table 4, Fig. 5). Galactose was also found in minor amounts (Table 4), and has been reported to be linked by a β-1,5 linkage to the xylan backbone . Acetyl groups were, however, not found in any of the extracted xylan fractions in this work.
Molecular weight determination and FT-IR spectra of selected xylan fractions
The molecular weights of the xylan extracted and purified from three fractions (1st extraction by 2 M NaOH [2 M-C1], 1st extraction by 1 M NaOH[1 M-C1] and by 0.5 M NaOH [0.5 M-60 min], in all cases using 60 min procedures) were determined by size exclusion chromatography. The results showed one major broad peak for the two fractions, 2 M-C1 and 1 M-C1, with a retention time of 276.5 min corresponding to a molecular weight of 212 kDa (Additional file 1: Figure S3). This molecular weight is higher than previously reported from other studies of sugarcane bagasse xylan: (65–86 kDa)  or (7–45 kDa) , which either indicate xylan of a higher molecular weight, or xylan interacting (or aggregating) with other chains or with lignin. The range of molecular weights observed for xylans however varies broadly. For example, arabinoxylan from wheat bran is reported in a molecular weight range of 470–600 kDa  while xylan extracted from corn stover have a reported molecular weight of 49 kDa . The fraction from the 0.5 M NaOH extraction, also showed a broad peak, but with a maximum at a lower molecular weight, 47 kDa (Additional file 1: Figure S3). This suggests that the lower concentrations of alkali solubilized more of smaller molecular sized xylans and that increased alkali concentration led to interactions, and apparent higher molecular weight xylans.
FT-IR spectra of the same three xylan fractions, (2 M-C1, 1 M-C1 and 0.5 M-60 min) showed that the three fractions clearly illustrated the typical signal pattern for xylans (Fig. 6). A major peak at 1035 cm−1 was corresponding to vibrations of the C–O, C–C and C–OH stretching [57, 62]. The peak at 896 cm−1 indicated glycosidic linkages in the xylan backbone . The absorption at 1633 cm−1 is attributed principally to the water absorbed by xylans. The C–H stretching vibration gave a signal at 2915 cm−1. The band at 3358 cm−1 corresponded to the hydroxyl stretching vibrations of xylans, as well as the water involved in the hydrogen bonding .
Cellulose purification and FT-IR spectra
Alkali extraction is the most efficient method for separating cellulose from the various biomass sources by releasing a large amount of hemicellulose polysaccharides in alkaline solution . In this study, after three consecutive pretreatments of bagasse with 2 M NaOH at 121 °C for 60 min, the remaining solid residue had a high content of cellulose, consisting of 80% glucose. Therefore, this fraction was acid purified (Fig. 1, see also methods), resulting in cellulose of high purity, with a sugar composition of 95.8 ± 0.59% glucose, with only minor amounts of xylose (3.1 ± 0.13%) (Additional file 1: Figure S4). The remaining amount of lignin in the cellulose was determined to 1.8 ± 0.07%. This material, can either be used in polymeric form, or be hydrolyzed to glucose residues, e.g., for application as carbon source in ethanol production.
The FT-IR spectrum of the extracted cellulose is very similar to the spectra of microcrystalline cellulose and Avicel (Fig. 6). A minor difference can be seen in intensity around the peak at 1729 cm−1, which is judged to correspond to vibration of water molecules absorbed in cellulose. The broad peak at 3333 cm−1 is characteristic for stretching vibration of the hydroxyl group in cellulose. The band at 2898 cm−1 is attributed to CH stretching vibration of all hydrocarbon constituent in polysaccharides and the band at 1054 cm−1 is assigned to the C–O stretching vibration. The band at 1428 cm−1 is associated with the amount of the crystalline structure of the cellulose, while the band at 897 cm−1 is assigned to the amorphous region in cellulose [64,65,66]. Peaks around 1500 cm−1 are defined for aromatic backbone in lignin  and clearly no such peaks are present in the purified cellulose fraction.
Enzymatic production of xylo-oligosaccharides
Xylans have lately gained interest for the development of protocols for XOS production, for applications as prebiotic food ingredients, and XOS are regarded as safe for use by both FDA and EFSA . XOS with a degree of polymerization (DP) of 2–5 are often preferred in functional food production owing to the utilization of these compounds by probiotic microorganisms , and within this range XOS with an average DP of 2–3 (especially xylobiose (X2)) gain importance as they present faster fermentation kinetics in the probiotic microorganisms that can promote a favorable intestinal environment .
Thus far, only limited attention has been given to SCB xylan in the field of XOS production, despite its large availability. Zhang et al. , has used a thermo-chemical/physico-chemical study in which birch and SCB is mixed to utilize autohydrolysis where acetyl groups released from the birch xylan aids in converting xylan to oligosaccharides. Dai et al., used a treatment with furoic acid to obtain XOS from SCB . Enzymatic hydrolysis, is another mild processing alternative to gain XOS. For SCB, this has previously been explored using native crude mixes of xylanases from P stipidis, on alkaline extracted xylan (using KOH 10%) from SCB  or enzyme mixes from Thermoascus aurantiacus . The present study explored the use of three individual enzymes from glycoside hydrolase family 10 (GH10) and 11 (GH11) for XOS production in a screening procedure comparing the xylan fractions obtained using the different alkaline conditions. For this purpose the endoxylanases RmXyn10A, and BhXyn10A (from GH10) and the commercially available GH11 enzyme Pentopan (from Novozymes), were screened using four fractions of purified extracted xylans (2 M-C1 (with a Xyl/Ara ratio of 11, and Xyl/UA ratio of 23), 1 M-C1 (with a Xyl/Ara ratioof 9, and Xyl/UA ratio of 29), 0.5 M-60 min (a Xyl/Ara ratio of 6, and Xyl/UA ratio of 17) and PLE No.4 (with a Xyl/Ara ratio of 9, and Xyl/UA ratio of 16) (Table 6). Maximum enzymatic conversion yields, for all enzymes in the screen, were obtained using the low molecular weight xylan, extracted by 0.5 M NaOH during 60 min, which after 1 h incubation gave a maximum yield of 42.29 ± 1.7% XOS, using BhXyn10A (Figs. 7 and 8). The predominant XOS produced was xylobiose, followed by xylotriose (Fig. 7). The lowest total production of XOS was found in PLE extracted xylan and in the xylan extracted by 2 M NaOH (Fig. 8). These two xylan fractions represent one of the more substituted (PLE(No4), Table 6), and one less substituted xylan (2 M-C1, Table 6), showing that the type of branching is not the major bottleneck here. Both these fraction however, displayed the highest amount of co-extracted lignin, and xylan of apparent high molecular weight, which indicates that interactions with lignin occur that may pose a problem for efficient enzymatic conversion. This is in line with previous conclusions by Lin and coworkers  who proposed that adsorption of lignin on the enzyme, reduced the hydrolysis, but that this could be circumvented by introduction of surfactants in the system.