Few signals from pectin and xyloglucan were detected in these experiments, due to both their relatively high mobility and low abundance in the secondary cell wall. Since the glucan chains in cellulose fibrils are polymorphic in structure, the two domains of cellulose were detected in at least three identified environments each. The three distinguished environments for both amorphous and crystalline cellulose may arise from differences in hydrogen bonding patterns between glucan chains, slight changes in glucan chain conformation, variations in bond geometries, and changes and inconsistencies in neighboring chain environments within the microfibrils different environments as follows: superscript [C] or [A] represents the. Here, for each carbon in a glucosyl unit from cellulose we represent these crystalline or amorphous cellulose domain respectively, and superscript represents the three distinguished environments of cellulose in each domain. The major chemical shift differences between the two domains of cellulose are present in the cellulose carbon-4 at ~89 ppm and ~84 ppm , and C6 at ~65 ppm and ~62 ppm . The signals from xylosyl units of three-fold screw xylan backbone were identified , but there were a lack of signals from xylosyl units of two-fold screw xylan . In addition, high intensity signals from arabinosyl units decorating the xylan backbone were identified, which is consistent with our monosaccharide analysis on the non-cellulosic components from stem internodes . Lower intensity signals from carbons 1 and 2 of glucuronic acid were also detected due to the lower abundance of glucuronic acid substitutions on the xylan backbone, as compared to arabinose. A DP-INADEQUATE experiment with relatively short recycle delay, 2 s,ebb and flow bench was also performed on the sorghum stem internode samples to enhance the detection of the relatively mobile components in the secondary cell walls, and spectrum was labeled and shown in Supplementary Fig. 3.

Two-fold screw xylan with a flat-ribbon shape has been reported as the dominant conformation detected by CP in the secondary cell walls of dicot plants and softwoods, which facilitates the binding with the cellulose fibrils on their hydrophilic surface via hydrogen bonding. However, our CP-INADEQUATE result indicated a significantly higher fraction of xylan was present in the three-fold screw conformation than in the two-fold screw conformation in the sorghum secondary cell wall. According to the intensities of cross peaks from the Xn4 and Xn5 of the xylosyl units from both two- and three-fold screw xylan backbone, the immobile three-fold screw xylan was the dominant conformation. This could be due to the large quantity of arabinosyl substitutions on the sorghum xylan backbone. Dupree et al. previously demonstrated that the spacing between the substitutions is critical for the formation of two-fold screw xylan due to the steric hindrance effect. For instance, xylan in dicots has little or no arabinosyl substitution, and acetylations and glucuronic acid substitutions are evenly spaced on a major fraction of the xylan backbone, which allows the formation of two-fold screw conformation. Similarly, although softwood xylan is substituted by both glucuronic acid and arabinose, substitutions are evenly spaced on every six and two xylosyl units respectively on the xylan backbone. Hence, we speculate that a large number of closely spaced substitutions on sorghum xylan would likely disrupt any pattern of regular spacing on a major fraction of the xylan. The result of this would be more xylan in the three-fold screw conformation in sorghum secondary cell walls. This spacing pattern has yet to be confirmed, due to a lack of a glycosyl hydrolase with the required specificity for cleaving at an arabinosyl substitution, akin to the glucuronic aciddependent xylanase XynC used to determine the spacing pattern in Arabidopsis. However, since CP experiments emphasize the signals from relatively immobile components of the cell wall, the large amount of three-fold screw xylan detected suggests that this confirmation could be important for xylan-cellulose interactions in sorghum.

In contrast, in recent work by Kang et al., two-fold screw xylan was detected by CP-INADEQUATE as the dominant conformation present in maize, a grass closely related to sorghum. However, their CP-INADEQUATE data did not detect arabinose signals, which is surprising given the reported structure of maize xylan. We hypothesized that this could be due to the additional lyophilization step they performed on the plant samples prior to analysis, which may alter the native wall structure. To test this, we lyophilized the sorghum stem tissue and rehydrated the sample according to the procedure that Kang et al. described in. A CP-INADEQUATE experiment was conducted on the rehydrated sample and compared with our previously collected CP-INADEQUATE spectrum on the untreated sample . We found that the previously strong signals from arabinosyl units were lost in the spectrum of lyophilized-rehydrated sample . The intensities of 3fXn4 and 3fXn5 also decreased in the lyophilizedrehydrated sample, suggesting that lyophilization may disrupt the immobile three-fold screw xylan interactions with other wall components. Rehydration of the sample did not restore these interactions, but instead, this free three-fold xylan became more mobile in the rehydrated water, and therefore no longer detectable by the CP. In addition, we also collected DP-INADEQUATE spectrum on the lyophilized-rehydrated sample and compared it with the previously collected DP-INADEQUATE spectrum from the untreated sample . The spectra show that there is a significant enhancement of signals from arabinosyl and xylosyl units of the three-fold screw xylan in the lyophilizedrehydrated sample, which indicates a significant amount of arabinosyl and xylosyl units from three-fold screw xylan have become more mobile after lyophilization and rehydration. On the other hand, the cross peak intensity from crystalline cellulose and two-fold screw xylan was enhanced in the lyophilized-rehydrated sample. Two-fold screw xylan has a flat-ribbon shape that can form a crystalline structure that is similar to cellulose.

Lyophilization of the sample increases the crystallinity of these structures. Although rehydration of the sample could reduce crystallinity to a certain extent, the cell wall architecture will remain altered permanently as compared to native cell walls. Hence, the increased rigidity of twofold screw xylan and cellulose led to enhanced CP signals for the lyophilized-rehydrated sample. For comparison, we also performed additional CPINADEQUATE experiments on never-dried sorghum leaf and root samples of sorghum, which are richer in primary cell walls, for comparison with the stem internode samples. The leaf samples show that the majority of xylan detected by CPINADEQUATE are in a three-fold screw conformation, but to a lower extent compared to the stem . However, analysis of the root material indicates that there is almost no immobile xylan in either conformation, as detected by CP-INADEQUATE .Xylan in the intact plant cell wall can populate both mobile and immobile states. The states can be distinguished based on the extent of their interactions with the immobile cellulose, which are mediated via either hydrogen bonding or Van der Waals forces: the cellulose-bound fraction of xylan is immobile, whereas the fraction of xylan that fills the inter-microfibril space is highly mobile. In contrast to reported data for dicots, we found that the three-fold screw xylan, not two-fold screw xylan, was the dominant conformation of xylan in the sorghum secondary cell wall,4x8ft rolling benches in both immobile and mobile forms . To investigate how this impacts xylan-cellulose interactions, 13C-13C proton-driven spin diffusion experiments were performed with CP with three different mixing times . Due to the CP transfer used in the experiment, these measurements report on the immobile fraction of xylan in the sample characterized by limited larger-scale molecular motions about an average structure with motional timescales on the order of µs to ms. This is in contrast to the mobile fraction of xylan, which is characterized by much faster molecular reorientations with motional timescales on the order of ns to µs. PDSD experiments provide information on carbons in close spatial proximity. The longer the mixing time, the longer the distance observed on the spectra. The short-mixing time CP-PDSD experiment is dominated by intramolecular peaks, such as carbons from glucan chains with six identified allomorphs in cellulose microfibrils, xylosyl units in the xylan backbone in both two- and three-fold screw conformations, and arabinosyl units from xylan .In addition, some intermolecular cross peaks between cellulose and xylan are observed. There are no cross peaks representing the interaction between the two-fold screw xylan and either the crystalline or amorphous domain of cellulose, likely due to the limited amount of two-fold screw xylan in the sorghum secondary cell wall. Instead, cross peaks, such as 3fXn4-2AC4 , 3fXn3-2AC4 , 3fXn3-1AC6 , 3fXn3-3AC6 , 3f,AXn3-2AC4 , and 3f,AXn3-3AC6 , all indicate that the xylosyl units from xylan in a three-fold screw conformation is closely interacting with amorphous cellulose. Additionally, carbons from the xylosyl units of three-fold screw xylan, 3fXn2 to 3fXn5, show cross peaks with the C1 from cellulose , which derives from cellulose in multiple environments. However, since we observed no cross peaks between carbons from three-fold screw xylan and CC2 to CC6, we conclude that the 3fXn2- to 3fXn5-C1 cross peaks were primarily contributed by the C1 from amorphous cellulose. Furthermore, arabinosyl units from xylan show close interactions with the cellulose C1, as indicated by two cross peaks, A2-C1 and A4-C1 . Since a low abundance of two-fold screw xylan was observed , the arabinose signals are likely from three-fold screw xylan. The close interactions between the arabinosyl units and the cellulose imply that the less ordered amorphous cellulose is able to bind with helical three-fold screw xylan. In the CP-PDSD experiment with 100 ms mixing time, the spectrum showed similar intermolecular cross peaks to the 30 ms mixing time, but with enhanced intensities . Together, these data suggest the xylan-cellulose interaction is dominated by an immobile xylan with three-fold screw conformation and amorphous cellulose across short distances.

It remains unclear to us what type of forces are facilitating such interactions, but we speculate it involves both Van der Waals contacts and some hydrogen bonds. The less ordered amorphous cellulose may have a distorted flat-ribbon shape and therefore create more surface space to occasionally enable the formation of hydrogen bonds with the three-fold screw xylan on the hydrophilic side. Interactions with Van der Waals forces are mainly from the hydrophobic surface of cellulose fibrils. Hence, the xylan-cellulose interactions in sorghum secondary cell walls are significantly weaker than those in dicot plants and softwoods which are dominated by hydrogen bonds between two-fold screw xylan and cellulose fibrils on the hydrophilic surface. To further explore the interaction between three-fold screw xylan and amorphous cellulose, we measured the spin-lattice relaxation times at various chemical shifts representing different components . The higher T1 indicates slower molecular dynamics of the cell wall component. The results show that carbons from crystalline cellulose, such as C1, 1CC6, and 2CC6, have the highest T1 values, ~9 s, and the carbons from amorphous cellulose, such as 1AC3/5, 1AC6, and 2AC6, have similar T1 values as the carbons from arabinosyl and xylosyl units in relatively immobile three-fold screw xylan, ~5 s. T1 measurements of the cell wall components further demonstrate that amorphous cellulose shares similar molecular dynamics with the relatively immobile fraction of the three-fold screw xylan, while crystalline cellulose has significantly reduced molecular motion. This supports our interpretation that amorphous cellulose and threefold screw xylan are closely interacting with each other. One-dimensional spectra were extracted at seven chemical shifts from the F1 plane of the CP-PDSD spectra with both 30 and 1500 ms mixing times and compared in Supplementary Fig. 8. No interaction between the three-fold screw xylan and the crystalline cellulose was detected in the short-mixing time . This is consistent with previous work reported by Dupree et al. using Arabidopsis, which showed that the flat-ribbon shape of two-fold screw xylan with even pattern of substitutions is required for binding on the highly ordered crystalline cellulose hydrophilic surface. In addition to the intramolecular interactions, many more intermolecular interactions were also detected with long mixing time . Although cellulose domains in different environments were expected to be close to each other, cross peaks between crystalline and amorphous cellulose were only detected in the long-mixing time CP-PDSD experiment .We interpret this as being due to the relatively low ratio of crystalline to amorphous cellulose in the sorghum cell wall, as described in the following experiments, and that crystalline cellulose is spatially further away from the amorphous cellulose than the three-fold screw xylan.