Effects of species differences in molar proportions of cell wall monosaccharides, concentrations of ester-linked phenolic acids, and concentrations of xylose and individual oligosaccharides following endoxylanase digestion were determined by analysis of variance (ANOVA) using a 1-way model. Data were tested for normality via the Kolmogorov-Smirnov test. If significant differences (p<.05) were seen, means were compared using a Dunn-Sidak (cell wall monosaccharides and ester-linked phenolic acids) or Bonferroni (oligosaccharides) test. All statistical analyses were completed using the OriginPro 2017 software program from OriginLab Corporation.

The monosaccharide composition of the water-insoluble plant cell wall material was quantified via HPAEC-PAD following Saeman hydrolysis (see Table 1 ). Glucose, xylose, arabinose, galactose and rhamnose were quantified in all pasture grass samples (with the exception of rhamnose in perennial rye). The concentration of glucose in the cell wall was significantly lower in perennial ryegrass (45% of the monosaccharides released during hydrolysis) than for the other three species (52-57% of the monosaccharides released during hydrolysis), indicating smaller amounts of cellulose in the cell walls of perennial ryegrass. The proportion of glucose in the sugars released during hydrolysis of our perennial ryegrass cell wall material (45%) was also lower than the range (50-60%) reported by Gordon et al. (1985) for epidermis, mesophyll, and secondary cell walls of perennial ryegrass. Galactose levels, which could arise from complex, galactose-bearing side-chains in the AX (Schendel et al., 2016), were low overall (2-3% of the released monosaccharides), but were significantly higher in tall fescue and bluegrass compared to perennial rye and timothy. Galactose could also partially originate from rhamnogalacturonan I pectic polysaccharides (galactan side chains), which are observed at low levels in the primary cell wall of grasses (Chesson et al., 1995; Xue et al., 2013; Couture et al., 2021; Avci, 2023). Trace amounts of rhamnose, which also arise from rhamnogalacturonans, were seen in all samples. The arabinose/xylose (A/X) ratios of all samples were quite low (see Table 1 ), ranging from 0.22 to 0.29. Perennial rye had a significantly lower A/X ratio compared to bluegrass and timothy, but not when compared to tall fescue. These values indicate a low level of xylan backbone substitution and are comparable to those previously reported in the literature. Gordon et al. (1985) described A/X ratios for perennial ryegrass leaves of 0.66, 0.40, and 0.14 for mesophyll cell walls, epidermis, and the residual fiber fraction, respectively. Sijtsma and Tan (1993) and Meak (2002) reported A/X ratios of 0.36 and 0.26, respectively, for perennial ryegrass cell walls isolated from whole plants, and Xu et al. (2007) found A/X ratios ranging from 0.25 to 0.45 for several water-insoluble AX fractions from perennial ryegrass leaves. Collings and Yokoyama (1979), Åman and Lindgren (1983), Meak (2002), and Kasuya et al. (2008) reported A/X ratios of 0.24, 0.21, 0.23, and 0.25 respectively, for tall fescue. Lindgren et al. (1980) reported A/X ratios for timothy grass throughout the growing season ranging from 0.21 to 0.32. Meak (2002) found a similar ratio of 0.25 for timothy grass. A/X ratios of 0.25 and 0.29 were respectively described for Kentucky bluegrass by Collings and Yokoyama (1979) and Meak (2002).

The insoluble cell wall material of four grass samples contained ester-linked p-coumaric and ferulic acid in both their trans- and cis- stereoisomers (see Supplementary Figure 1 for sample chromatogram). The total levels of monomeric ester-linked phenolic acids were over 10,000 µg g^-1 water-insoluble cell wall material ( Table 2 ), representing a full one percent of the insoluble material. The cis isomers of ferulic and coumaric acid arise from UV light-induced isomerization (Kahnt, 1967), so although plants synthesize these compounds exclusively in their trans- forms, formation of the cis-isomers is to be expected in forage tissues exposed to sunlight (Yamamoto and Towers, 1985; Turner et al., 1993; Schäfer et al., 2019). Timothy had significantly lower concentrations of p-coumaric acid than the other species. Trans-ferulic acid concentrations were not significantly different between the pasture grasses and were roughly comparable to the trans-p-coumaric acid concentrations. Approximately equivalent amounts of ester-linked ferulates and coumarates were also previously observed in aerial parts of timothy (Theander et al., 1981), ensiled tall fescue (Sousa et al., 2021), and aerial parts of rice (Oryza sativa) plants (Dhakarey et al., 2017). The roughly equivalent proportions of ester-linked ferulates and coumarates in vegetative tissue contrast with the ester-linked phenolic acid profiles of monocotyledonous grain tissue, which are dominated by ferulic acid (Andersson et al., 2008; Li et al., 2008; Nyström et al., 2008; Shewry et al., 2008; Schendel et al., 2015). In their ester-linked form, the two compounds are directly bonded to separate cell wall polymers: ferulic acid to AX and p-coumaric acid primarily to lignin (Smith and Hartley, 1983; Lu and Ralph, 1999; Hatfield et al., 2008), although small amounts of p-coumaric acid esters acylate AX (Mueller-Harvey et al., 1986). The clear difference in ester-linked phenolics between grain and vegetative tissue springs from their polymer composition: whereas vegetative tissue contains substantial amounts of lignin (Buxton and Russell, 1988), the lignin content of grain tissue is quite low (Bunzel et al., 2011).

We did not quantify dehydrodimers and higher oligomers of ferulates, representing AX-AX cross-links (Ralph et al., 1994; Bunzel, 2012), or ether-linked ferulates, representing AX-lignin cross-links (Quideau and Ralph, 1997). Ferulate-driven crosslinking significantly reduces AX degradability by endoxylanases (Grabber et al., 1998; Beaugrand et al., 2004; de Souza et al., 2018). Supplementation of xylanase-based enzyme treatments with feruloyl esterases increases the amount of released carbohydrate from AX in some cases (Li et al., 2011; Várnai et al., 2014), and some ruminant bacteria even express bifunctional endo-1,4-β-xylanase/esterases (Pavarina et al., 2021), thus directly attacking the ferulate-driven recalcitrance to xylanase digestion. However, although ester-linked ferulate monomers are relevant to cell wall architecture as potential building blocks for future cell wall polymer crosslinking, ferulate monomers themselves do not reduce xylanase-mediated hydrolysis of AX beyond the level caused by simple arabinosylation alone of the AX backbone (Grabber et al., 1998; Beaugrand et al., 2004; Várnai et al., 2014).

Generation of oligosaccharide profiles by endoxylanase digestion followed by separation, detection, and (sometimes) quantification via HPAEC-PAD has been previously used to compare the fine AX structures of cereal grains. This approach generates and compares unique AXOS fingerprints for different grain varieties and grain tissues (Viëtor et al., 1994; Ordaz-Ortiz et al., 2005; Pastell et al., 2009; Saulnier and Quemener, 2009; Makaravicius et al., 2012; Snelders et al., 2013; Falck et al., 2014; Immerzeel et al., 2014). We have adapted this technique to fingerprint the AX contained in the vegetative tissue of cool-season forages. This method, which targets the water-insoluble hemicellulosic fraction of cool-season forages, complements the use of HPAEC-PAD to profile the water-soluble and ethanol-soluble carbohydrate fractions of various cool-season forages (Kagan et al., 2011; Kagan et al., 2018).

Hydrolysis of the β-1→4-glycosidic linkages between the xylopyranose units of the AX backbone by endoxylanases is the cornerstone of our approach, and which bonds are accessible by the enzymes is influenced by nearby side-chain residues and xylanase type (Pollet et al., 2010). Xylanases from the glycoside hydrolase (GH) family 10 require only two unsubstituted xylose units between branched units, whereas GH11 xylanases require three unsubstituted backbone units. The two enzyme types would thus produce different oligosaccharide fingerprints. GH10 xylanases are more effective in digesting more densely substituted AX, but the low A/X ratio of our vegetative grass tissue led us to expect that GH11 xylanases would still be able to access the xylan backbone. We compared a GH10 xylanase derived from Cellvibrio japonicus, a gram-negative soil bacterium (DeBoy et al., 2008), and a GH11 xylanase from Neocallimastix patriciarum, an anaerobic fungus isolated from ruminants (Orpin and Munn, 1986). Pure water was used as the incubation medium instead of buffer to avoid potential chromatography problems during HPAEC analysis. These conditions resulted in very limited oligosaccharide production by the GH11 Neocallimastix patriciarum enzyme (see Supplementary Figure 2 ), and therefore, only the Cellvibrio japonicus endoxylanase was used for further enzymatic work.

Enzyme incubation time was compared at 4, 12, 24, and 48 h. The 4 h incubation time resulted in only minimal hydrolysis, whereas by 12 h, release of various XOS and AXOS was clearly evident by examination of the HPAEC chromatogram and comparison with a standard chromatogram (data not shown). At 12 h, all of the DP 2-6 XOS were present, but at the later incubation times (24 and 48 h), the larger XOS had been mostly degraded into xylose and 2X. This was expected, given the hydrolytic patterns of GH10 endoxylanases. As reviewed by Pollet et al. (2010), 2X is the smallest XOS released by GH10 xylanases. We selected the 12 h incubation time as the standard condition for our forage AX fingerprinting method so that differences in the amount of unbranched sections along the AX backbone between materials would be more visible, and the AX fingerprint generated by the method would contain more structural information. Higher released quantities of longer XOS directly demonstrate the existence of many unsubstituted sections on the original AX polymer. In contrast, higher released quantities of 2X are less meaningful, since 2X could arise from degradation of a long stretch of unsubstituted xylan backbone, but it could also have been released from a region with a moderate degree of substitution.

We selected 10 XOS and AXOS standard compounds (purchased commercially) and used quantitative ¹H-NMR to weigh them in accurately. We chose lactose as an internal standard because it was well-separated from our standard compounds and not native to plant materials. Our optimized gradient and column temperature achieved baseline separation of the standard compounds and internal standard, with the exception of 6X and A²XX, which remained slightly shouldered (see Figure 2 ). As also reported by Alyassin et al. (2020), implementation of a gradient elution step for sodium hydroxide at the beginning of each run as opposed to an isocratic level of sodium hydroxide substantially improved oligosaccharide resolution. We also found that a relatively high concentration of sodium acetate (1 M) during the column clean-out step followed by a 20-minute rinse with 0.1 M sodium hydroxide improved separation reproducibility. Table 3 presents retention time (RT) values from a randomly selected standard chromatogram prepared using the optimized gradient. Intraday RT reproducibility was excellent (± 0.05 min); small (maximum of 0.3 min) interday variations in RT were observed between eluent batches, but because a fresh standard sample was run with each new batch of eluent, peak ID was unhindered.

The LOD and LOQ for xylose and the oligosaccharide standards ranged from 5 to 8 nM and 16 to 23 nM, respectively, underscoring the outstanding sensitivity of the amperometric detector ( Table 3 ). Due to limited quantities of standard compounds, we did not attempt to identify a precise upper concentration limit of the compounds where detector saturation would hamper quantitative work, but instead determined a reliable working concentration range, that is, one above the LOQ and well below detector saturation ( Table 1 ). Individual standard curves were prepared for each compound within the CR ( Table 1 ). We compared fitting data to both linear and polynomial (quadratic) equations and found that, although the R² values were high for the linear fit (≥0.986 for all standards), the R²-values for a quadratic fit were slightly higher for each standard ( Table 1 ). Visual comparison of the residual plots from the linear vs. quadratic fit further supported the use of a quadratic equation: whereas the linear residual plots had a distinctive parabola pattern, the quadratic residual plots were randomly dispersed ( Figure 3 residual plots). Other researchers have also utilized quadratic calibration curves for HPAEC-PAD-based oligosaccharide quantification, for instance, for human milk oligosaccharides (Kunz et al., 2017), β-glucan-derived oligosaccharides (Radva et al., 2012), and fructans (Haselberger and Jacobs, 2016; Ispiryan et al., 2019). Alyassin et al. (2020) utilized linear calibration curves for HPAEC-PAD-based quantification of XOS and AXOS, but they did not compare quadratic fitting.

We also calculated RRF values for each compound ( Table 3 ), but given the reported analyte-specific PAD response drop due to recession of the gold working electrode for XOS and AXOS (Mechelke et al., 2017), we chose to perform all oligosaccharide quantifications using quadratic calibration curves measured anew with each batch of eluent. The molar detector responses for xylose and 2X were similar to each other and higher than that of 3X, but the RRFs for the DP 4-6 XOS were higher than the lower-DP compounds. Increasing molar response values with increasing DP up to a DP 6-7, followed by a leveling of detector response, has been reported for some oligosaccharide series, including dextrins (White et al., 2003) and fructooligosaccharides (Fernández et al., 2013). In contrast, Mechelke et al. (2017) reported a decreased molar detector response for DP 4-5 XOS compared to DP 1-3, no clear trend for cellooligosaccharides, and decreasing molar detector responses for DP 2-5 laminarioligosaccharides.

The validated HPAEC-PAD method enabled quantitative comparisons of endoxylanase hydrolysates of cool-season forage samples (see Figure 4 ). Many previous studies of vegetative AX structures released by xylanase were based on alkali-pretreated material (Mazumder and York, 2010; Bowman et al., 2014; Tryfona et al., 2019), but because these treatments cleave ester linkages, AX-AX crosslinks are undone and AX is unzipped from lignin by releasing the ester bonds of ester/ether-linked ferulates, which dramatically alters the xylanase accessibility compared to the native cell wall structure. We chose to work with the native, non-pretreated structure to strengthen the practical application of our results to forage screening for livestock feeding. Only a small amount of AX was released from the native structure as AXOS during the 12 h of water-based endoxylanase incubation (0.09, 0.14, 0.19, and 0.39% of the total arabinose and xylose for Kentucky bluegrass, tall fescue, timothy, and perennial rye, respectively). Incubation in buffer would likely increase xylanase activity and AX solubilization, but the recommended buffer for most commercially available endoxylanases, sodium acetate, could influence HPAEC separation, so we chose to incubate in pure water instead. However, the intent of the AX profiling method was not full degradation of the cell wall AX of forage material into AXOS, but rather, the rapid generation of thumbnail sketches of the AX structure in forage materials. Differences in these thumbnail sketches, produced by digesting various forage materials under the same incubation conditions, serve as unique fingerprints of the AX contained in those forages and can be used to compare the AX structures contained in those forages.

Perennial rye generated significantly greater amounts of released xylose and all quantifiable oligosaccharides compared to bluegrass, and significantly greater amounts of 5X, A²XX, and A³X compared to timothy and tall fescue. Several competing AX structural factors influence endoxylanase hydrolysis efficiency in intact cell walls. Firstly, densely substituted AX backbone regions block endoxylanase access to the xylan backbone (Pollet et al., 2010) and thus require the deployment of accessory enzymes for complete depolymerization (de Vries et al., 2000). Secondly, denser AX backbone substitution patterns increase AX water solubility and limit hydrogen-bonding interactions with cellulose microfibrils (Kabel et al., 2007; Köhnke et al., 2008; Saulnier et al., 2012; de Souza et al., 2013; Selig et al., 2015). As discussed in Section 3.1, both the glucose content, corresponding to cellulose, and the A/X ratio, corresponding to AX backbone substitution, of perennial ryegrass cell wall material was lower than that of the other grass species. The higher levels of released XOS and AXOS from the perennial rye cell wall are thus partly due to the lower quantities of cellulose present in the perennial rye cell wall, which would reduce hydrogen bonding opportunities between the cellulose microfibrils and the AX, and the lower backbone substitution rate of the perennial rye AX polymer compared to the other forages.

Of the oligosaccharides and monosaccharides released, 3X was the most abundant across all grasses, but 4X, 5X, and 6X were also present in the hydrolysates from all species at the 12-h incubation point. This means that these species share a structural similarity of long stretches of unsubstituted regions on the AX backbone.

A²XX and A³X were the only quantifiable branched AXOS released under the current conditions. The smallest decorated oligosaccharide produced by a GH10 xylanase is A³X or A²X (Pell et al., 2004). Several small peaks that may have corresponded to XA²XX, XA³XX, or A²⁺³XX were seen in the chromatograms, but their identity and concentration could not be unequivocally determined because of partial coelution with other, unidentified peaks. Expansion of the method to include additional branched AXOS standards would improve the resolution of the generated AX fingerprints. Additional investigation with a mass spectrometer coupled to the HPAEC-PAD, analogous to the profiling method recently validated for enzymatic hydrolysates of xyloglucans (Steck et al., 2021), would strengthen our method by expanding our peak identification toolset beyond retention time comparison alone.

A peak that co-eluted with arabinose was seen in the monosaccharide region of the chromatogram for most, but not all, of the samples. However, xylanase-free blank incubations confirmed that the compound was released from the forage samples independently of the xylanase. As a result, we did not attempt to quantify any arabinose released during incubation, since our aim was quantification of the xylanase-generated AX profiles.