Sensitivity-enhanced solid-state NMR detection of expansin’s target in plant cell walls

Significance

The protein expansin loosens the cell walls of plants for cell growth, but its carbohydrate target of binding has been elusive because of the difficulty of studying the noncrystalline plant cell wall by most structural biology techniques and the trace amount of expansin needed for wall loosening. We have now combined dynamic nuclear polarization sensitivity-enhanced solid-state NMR spectroscopy with ¹³C labeling of Arabidopsis thaliana plants and ¹³C, ¹⁵N labeling of expansin, to determine that expansin binds cellulose microfibrils to loosen the plant cell wall. The expansin binding site is enriched in the hemicellulose xyloglucan and has a different cellulose structure from bulk cellulose, shedding light on the mechanism of wall loosening.

Abstract

Structure determination of protein binding to noncrystalline macromolecular assemblies such as plant cell walls (CWs) poses a significant structural biology challenge. CWs are loosened during growth by expansin proteins, which weaken the noncovalent network formed by cellulose, hemicellulose, and pectins, but the CW target of expansins has remained elusive because of the minute amount of the protein required for activity and the complex nature of the CW. Using solid-state NMR spectroscopy, combined with sensitivity-enhancing dynamic nuclear polarization (DNP) and differential isotopic labeling of expansin and polysaccharides, we have now determined the functional binding target of expansin in the Arabidopsis thaliana CW. By transferring the electron polarization of a biradical dopant to the nuclei, DNP allowed selective detection of ¹³C spin diffusion from trace concentrations of ¹³C, ¹⁵N-labeled expansin in the CW to nearby polysaccharides. From the spin diffusion data of wild-type and mutant expansins, we conclude that to loosen the CW, expansin binds highly specific cellulose domains enriched in xyloglucan, whereas more abundant binding to pectins is unrelated to activity. Molecular dynamics simulations indicate short ¹³C-¹³C distances of 4–6 Å between a hydrophobic surface of the cellulose microfibril and an aromatic motif on the expansin surface, consistent with the observed NMR signals. DNP-enhanced 2D ¹³C correlation spectra further reveal that the expansin-bound cellulose has altered conformation and is enriched in xyloglucan, thus providing unique insight into the mechanism of CW loosening. DNP-enhanced NMR provides a powerful, generalizable approach for investigating protein binding to complex macromolecular targets.

As part of the cell growth process, plants use expansins to induce wall stress relaxation, which creates the driving force for cell water uptake and consequent enlargement (1). Expansins were first discovered in studies of acid-stimulated growth of plant cells (2). Auxin, the classical plant growth hormone, rapidly stimulates growth in part by activating plasma membrane H⁺-ATPases, lowering wall pH, thereby activating expansins, which have a low pH optimum. Expansins mediate wall loosening not by lysis of the major polysaccharides of the growing cell wall (CW), but by weakening the noncovalent polysaccharide network that constitutes the load-bearing structure of the CW (3).

Structural studies of the mechanism of expansin-mediated wall loosening have been hampered by the fact that active plant expansins are difficult to produce in recombinant expression systems. This obstacle was recently circumvented with the discovery of microbial expansins, which are readily expressed in Escherichia coli and enabled mutagenesis studies of the residues required for wall loosening and X-ray analysis of protein-oligosaccharide structures (4–6). These studies showed that expansins consist of two domains, D1 and D2, which present a nearly flat surface for binding to cellohexaose and related oligosaccharides. Mutagenesis and functional assays indicate that two distinct regions of the D2 domain have carbohydrate-binding properties: three conserved and linearly arranged aromatic residues on a hydrophobic surface of D2—resembling a type-A carbohydrate-binding module—are required for binding to pure microcrystalline cellulose and for wall loosening, whereas several nonconserved basic residues on the opposite side of D2 increase binding to whole CWs but do not promote wall loosening (5, 6).

Although we now know expansin’s structure in atomic detail, its exact site of action in the native plant CW has not been structurally characterized because of the disordered and insoluble nature of CW polysaccharides and the fact that the only assays for expansin activity are biophysical, not biochemical. An attractive approach to this problem is solid-state NMR (SSNMR) spectroscopy, which can probe the structure of insoluble biomolecular complexes through intermolecular magnetization transfer, or spin diffusion (7–9). Two challenges in applying SSNMR to plant CWs are the overlapping ¹³C signals of multiple polysaccharides (10, 11) and the trace amount (∼0.1 wt% of the CW) of expansin at which it is operative (12). Using excess expansin to increase the detection sensitivity causes nonspecific binding, which complicates structural analysis. We recently overcame the first challenge by ¹³C labeling of entire Arabidopsis thaliana plants (13), which enabled the use of 2D and 3D magic-angle-spinning (MAS) NMR techniques to resolve and assign the ¹³C signals of various wall polysaccharides (14). To overcome the second challenge, we now use dynamic nuclear polarization (DNP), which enhances the NMR sensitivity by as much as two orders of magnitude by transferring the electron polarization of paramagnetic dopants to nuclear spins under microwave (MW) irradiation (15, 16). We show that DNP sufficiently enhanced the NMR sensitivity to allow determination of the functional binding target of EXLX1, a bacterial expansin from Bacillus subtilis (5). Furthermore, distinct structural features of expansin-bound polysaccharides are observed, shedding light on CW sites important for wall loosening and wall mechanics.

Results and Discussion

DNP Enhancement of CW Spectra.

We produced ¹³C-labeled Arabidopsis CW by growing whole plants in media containing ¹³C-labeled glucose (13). Non-CW molecules such as lipids and intracellular proteins were removed by detergent and protease extraction, followed by partial extraction of matrix polysaccharides to reduce nonspecific expansin binding (SI Appendix, Table S1). The ¹³C, ¹⁵N-labeled expansin was expressed by E. coli grown in minimal media, mixed with never-dried CW at a mass ratio of ∼1:60–100, and hydrated with partially deuterated glycerol/water solution containing TOTAPOL [1-(TEMPO-4-oxy)-3(TEMPO-4-amino)propan-2-ol] (17), a biradical that exhibits high DNP efficiency. Solvent deuteration directs the electron polarization to ¹H spins in the proteins and polysaccharides, whereas glycerol protects the CW structure and uniformly distributes the biradical at the cryogenic temperature (∼100 K) of the DNP experiments. A 600-MHz NMR spectrometer equipped with a 395-GHz gyrotron for high-power MW irradiation was used for the experiments. Fig. 1A shows 1D ¹³C cross-polarization (CP) MAS spectra of CW that contains wild-type (WT) expansin. The spectra are simpler than reported Arabidopsis CW spectra (13) because of the reduced amount of pectins in the wall (SI Appendix, Table S1).

Fig. 1.

The ¹³C NMR spectra of Arabidopsis CW with WT expansin. (A) The ¹³C CP-MAS spectra with MW on (black) and off (blue) at 102 K, showing 27-fold sensitivity enhancement. The DNP-enhanced spectrum has a similar pattern as the 248-K spectrum of a radical-free CW sample (red), indicating uniform sensitivity enhancement. ssb, MAS sidebands. (B) The ¹⁵N-¹³C filtered ¹³C spectra after 1-s spin diffusion. Without DNP at 248 K (blue), little intensity was observed after 22,528 scans. With DNP at 102 K (black), strong protein and polysaccharide intensities were detected after 2,560 scans.

Turning on the MW enhanced the spectral sensitivities 27-fold (Fig. 1A). Low temperature and the presence of the biradical did not cause significant line broadening, based on comparisons with the ¹³C linewidths of radical-free samples at high temperature (SI Appendix, Fig. S1) (18). Thus, the structural order of cellulose microfibrils in the plant CW is preserved at ∼100 K, similar to pure microcrystalline cellulose (19). Moreover, the DNP-enhanced spectral pattern is similar to the high-temperature spectra (Fig. 1A), and the ¹H T₁ relaxation times are uniform for all carbons (SI Appendix, Table S2). This result demonstrates uniform transfer of the electron polarization to wall polysaccharides. These favorable properties, also seen for two other expansin-containing CW samples (SI Appendix, Fig. S2), indicate that plant CWs are ideal targets for structural studies by DNP NMR (20–22).

Identifying CW Polysaccharides near Expansin.

To identify the target site of expansin in the CW, we exploited the fact that ¹⁵N labels are present only in expansin and not in any CW biopolymers, by selecting the ¹³C magnetization of expansin carbons that are directly bonded to nitrogens using a ¹⁵N-¹³C dipolar filter (SI Appendix, Fig. S3). Magnetization of the nitrogen-proximal ¹³C spins was then transferred to the nearby polysaccharides through distance-dependent ¹³C spin diffusion. Such protein-transferred ¹³C spectra could not be observed without DNP because of the low amount of expansins (<0.23 mg) in these samples (Fig. 1B) and the unavoidable dilution of the protein ¹³C magnetization among the more abundant CW polysaccharides. However, with DNP, strong protein ¹³C signals were observed in the aliphatic (10–60 ppm), aromatic (120–160 ppm), and carbonyl (170–185 ppm) regions (Fig. 1B). Moreover, characteristic polysaccharide signals at 70–110 ppm became evident after >0.5 s of mixing, i.e., as expansin ¹³C magnetization diffused to nearby polysaccharides. A 2D ¹³C spectrum that directly correlates protein Cα and polysaccharide signals (SI Appendix, Fig. S4) confirmed that the polysaccharide intensities in the 1D spectra result from spin diffusion from expansin.

To compare the proximities of different polysaccharides to expansin and, thereby, determine the makeup of the functional binding site, we measured the protein-transferred ¹³C spectra of the polysaccharides as a function of spin diffusion mixing time (Fig. 2) and compared CWs that contain WT expansin, a WWY mutant, or an RKK mutant (4). The WWY mutant (W125A/W126A/Y157A) replaces the three aromatic residues on the cellohexaose-binding D2 surface with Ala (5). The RKK mutant (R173Q/K180Q/K183Q) replaces three basic residues on the opposite surface of D2 with the uncharged Gln. The RKK mutant has stronger wall loosening activities than WT expansin, an effect that was postulated to result from reduction of nonspecific interactions of the basic protein with acidic pectins. All three samples show increasing polysaccharide intensities with mixing time (Fig. 2A and SI Appendix, Fig. S5), as expected for spin diffusion measurements. However, even after 3 s, the WT and the WWY spectra still show much lower interior cellulose C4 intensity at 89 ppm than in the equilibrium CP spectrum, indicating that neither the WT nor the WWY protein has equilibrated its magnetization with the cellulose microfibril. In contrast, the RKK mutant displayed a significantly higher 89-ppm peak (Fig. 2B), and the polysaccharide intensity envelope is closer to that of the CP spectrum, indicating closer average proximity to cellulose. Relative to the 52-ppm protein Cα signal, both the interior cellulose C4 peak at 89 ppm and the dominant polysaccharide 72-ppm peak increase in intensity in the order of WWY ≤ WT << RKK (Fig. 2C). Because the WWY mutant lacks wall-loosening activity while the RKK mutant is hyperactive, this trend in intensity strongly suggests that cellulose is the functional binding target of expansins and that the hyperactive RKK mutant binds cellulose with the shortest average distances. The short average distances may be due to short nearest-neighbor distances of RKK expansin to the cellulose microfibril, more RKK expansin bound specifically to the microfibril, or a combination of both.

Fig. 2.

DNP-enhanced ¹³C spin diffusion spectra of expansin-bound CWs. (A) Spectra of CW that contains WT expansin after 1-ms to 3-s spin diffusion. (B) Spectra of CW that contains RKK-expansin after 1-ms to 3-s spin diffusion. (C) Three-second spin diffusion spectra of CWs with RKK, WT, and WWY expansins. Normalized to the protein Cα intensity, the RKK mutant shows the highest cellulose intensities. (D) Peak height ratios between the 3-s spin diffusion spectra and CP spectra, normalized to the 72-ppm intensity ratio. The RKK expansin has the highest iC4 intensity (89 ppm), whereas the WWY mutant has the highest pectin intensities (69 and 100 ppm). Polysaccharide abbreviations are indicated. (E) Decay of the total protein intensities (filled squares) and buildup of the total polysaccharide intensities (open squares) with time. RKK expansin (red) shows overall more spin diffusion transfer than WT expansin (black). (F) Intensity buildup of the 89-ppm peak of interior cellulose C4. The y axis plots the peak height ratio to the total intensity, further normalized by the ratio of the 72-ppm peak of the RKK sample at 3-s spin diffusion.

Quantitative intensity analysis of the 3-s spin diffusion spectra relative to the equilibrium CP spectra (Fig. 2D) and spin diffusion buildup with time (Fig. 2 E and F and SI Appendix, Fig. S6) confirm that the RKK mutant brings interior cellulose closest to its equilibrium magnetization, whereas the WWY mutant transfers the least magnetization to cellulose. Moreover, among the three expansins, the WWY mutant has the highest magnetization transfer to pectins, indicating that weakening the aromatic interaction while preserving the basic residues favor expansin binding to pectins. This result is consistent with WWY’s ability to bind whole CWs, which, however, does not cause wall loosening because cellulose binding is weakened (5).

Comparison with Computational Results.

Molecular dynamics simulations of the expansin protein docked onto a cellulose microfibril (Fig. 3) verify the presence of atomic contacts between expansin and cellulose indicated by the spin diffusion NMR data. These simulations modeled cellulose as the I_β crystalline allomorph, which is the dominant form in plant CWs (23, 24) and used the crystalline structure of expansin complexed with cellohexaose (5) to guide the placement of the protein on the complex microfibril surface. The protein lies with its aromatic-triplet surface parallel to the edge of the hydrophobic (100) surface of the cellulose microfibril. The axis through the aromatic triplet is roughly parallel to the glucan axis, maximizing hydrophobic interactions. The shortest ¹³C-¹³C distances between the three aromatic rings and the pyranose rings are 4–6 Å, suggesting CH–π interactions. Similarly, the surface glucans are separated from the first layer of interior crystalline glucans by 4.2–5.4 Å. These distances are fully within ¹³C spin diffusion reach, consistent with the observed 89-ppm peak of interior cellulose in the RKK spectra.

Fig. 3.

MD simulation of expansin docking onto a cellulose I_β microfibril. (A) Side view showing the three aromatic residues (magenta) that bind to the (100) surface of the microfibril. Surface and interior cellulose carbons are colored in yellow and green, respectively. (B) Top view showing the shortest ¹³C-¹³C distances (4–6 Å) between expansin and the microfibril, and between surface and interior glucan chains. These short distances explain the protein-polysaccharide spin diffusion seen in the 1D and 2D ¹³C spectra.

Distinctive Characteristics of the Expansin Binding Site.

More detailed structure information about the expansin-bound region of the cellulose microfibril is obtained from a protein-transferred 2D ¹³C correlation spectrum of the RKK-expansin sample, whereby only cross-peaks of polysaccharides in close contact to the protein are detected (Fig. 4A). Several changes from the spin–diffusion-free 2D spectrum are observed. In the 99-ppm cross-section of xylose, galacturonic acid, and rhamnose C1, the cross-peak at 68 ppm (from galacturonic acid C1–C2 and rhamnose C1–C5) is weakened relative to the xylose C1-C2 cross-peak at 72 ppm (Fig. 4 B–D and SI Appendix, Fig. S7). This preferential enhancement of xylose intensities is also observed at several other cross-sections (SI Appendix, Fig. S7) and indicates that the expansin-bound site has higher XyG content and lower pectin content than the average CW. These results are reinforced by the spin diffusion 2D spectrum showing a cross-peak between the C1 of xylose and a Trp or Tyr aromatic carbon (SI Appendix, Fig. S8), which is consistent with atomic contact between the aromatic triplet (W125/W126/Y157) in the expansin protein and XyG-enriched regions of the cellulose microfibril.

Fig. 4.

Two-dimensional ¹³C-¹³C correlation spectra of RKK-bound CW polysaccharides. (A) DNP-enhanced 2D spectrum after 3-s intermolecular spin diffusion (black). Only polysaccharides in close contact with RKK-expansin are detected. The equilibrium 2D spectrum without spin diffusion is superimposed in red. (B–D) Comparison of the C1 region of several 2D spectra. (B) DNP-enhanced 3-s spin diffusion spectrum of the RKK CW sample at 102 K. (C) DNP-enhanced equilibrium 2D spectrum of the RKK CW sample at 102 K. (D) Equilibrium 2D spectrum of WT CW sample at 253 K without DNP (12). The spectrum in B has a higher xylose cross peak and a different iC4 chemical shift (highlighted in yellow) than the equilibrium spectra in C and D.

Along with XyG enrichment, there is evidence that the expansin-bound interior cellulose has a C4 chemical shift that is 1-ppm lower than the average C4 frequency (compare yellow highlighted region in Fig. 4B with Fig. 4C). Because the C4 chemical shift is sensitive to the χ torsion angle around the C5–C6 bond, the lower shift suggests that the expansin binding site has a slightly altered interior cellulose conformation from that of the majority of cellulose. This change in cellulose conformation may be related to the presence of XyG near the expansin binding site: XyG is apparently able to modify cellulose crystallization during microfibril formation in model systems (25, 26). Another possibility is that expansin binding itself modifies the cellulose conformation, but the molecular dynamics simulation (above) gave no hint of such an effect.

These results demonstrate that expansin’s potency for CW loosening coincides with its selective binding to a specific and quantitatively minor site in the CW comprised of a slightly altered cellulose conformation in close proximity to XyG. This conclusion extends and deepens inferences from competitive binding studies (5) showing that the cellulose binding site for expansin is distinct from and less abundant than that for a series of cellulose binding modules that bind to the crystalline surface or disordered regions of cellulose. Those studies, which were based on binding isotherms, were made with pure cellulose (Avicel and bacterial cellulose) to avoid masking by the dominant but unproductive binding of expansin to acidic polysaccharides in the more complex plant CW. DNP-enhanced NMR, in combination with the use of expansin mutants, was able to circumvent this problem in a more informative manner, to reveal unprecedented details about expansin’s target for CW loosening.

One limitation of the current study stems from the use of the bacterial expansin, EXLX1, which is less active in wall-loosening assays than are plant expansins (4). This limitation stems from the difficulty of expressing plant expansins heterologously, which is a precondition for producing the labeled and mutated versions of expansin needed for our analysis. Because EXLX1 is homologous to plant expansins (4) and induces CW extension with characteristics very similar (although diminished) to plant expansins yet different from other wall loosening mechanisms (1, 27), we expect them to share the same target (see below as well), but verification of this point will require further testing.

Implications for CW Structure and Growth.

These results with EXLX1 dovetail nicely with other recent studies that are leading to revisions in our concepts of primary CW structure in plants, particularly with respect to the prevalent notion that XyGs function as tethers to link microfibrils into a load-bearing network (3, 28, 29). For instance, analysis of CW mechanical changes induced by substrate-specific endoglucanases (30) showed that most of the XyG in the CW was irrelevant to wall mechanics and that key mechanical junctions in the CW may be restricted to relatively few, inaccessible regions, dubbed “biomechanical hotspots,” where XyG and cellulose may be closely intertwined. The expansin-bound sites characterized here seem to be likely candidates for these regions, an idea that gains support from the observation that CWs lacking XyG are substantially diminished in their response to expansin (31). The idea that the majority of XyG does not contribute to CW mechanics in a major way is likewise supported by the relatively minor morphological and biomechanical phenotype of a genetic mutant of A. thaliana deficient in XyG (31, 32). Additionally, recent SSNMR data suggest that XyG interaction with cellulose in the CW is more limited than previously conjectured and may occur more by entrapment within microfibrils or groups of microfibrils than by coating the microfibril surface (13). Our characterization of expansin’s functional binding site is consistent with these emerging concepts of the structure of the growing CW and strengthen the concept that CW loosening by expansin and other wall-loosening agents may be restricted to limited, specific structures in the CW.

Finally, we note that the approach demonstrated here of DNP-enhanced NMR combined with tailored isotopic labeling is generally applicable to structure elucidation of protein binding to other macromolecular complexes such as cytoskeletons and extracellular matrices.

Materials and Methods

Protein and CW Samples.

The ¹³C, ¹⁵N-labeled expansins (EXLX1 from B. subtilis) were expressed in E. coli and purified as described before (6), except that M9 minimal medium containing ¹³C-labeled glucose and ¹⁵N-labeled ammonium chloride were used as the only carbon and nitrogen sources. The ¹³C-labeled CWs were produced by growing A. thaliana seedlings in the dark for 14 d in liquid medium containing 0.5% ¹³C-glucose (13), followed by isolation and extraction of the CW without drying. The harvested seedlings were frozen and ground in liquid nitrogen, and the material was treated sequentially with 1.5% (wt/vol) SDS (3 h) to remove intracellular proteins, porcine pancreatic α-amylase (5,000 units in 30 mL of 50 mM Mes buffer at pH 6.8, 12 h at 37 °C) to remove starch, Pronase (200 units in 20 mL of MES buffer at pH 7.5, 12 h at 40 °C) to further digest proteins, xyloglucanase [3 mg/mL from Aspergillus aculeatus (33) in 50 mM sodium acetate at pH 5.0], and 1 M NaOH with 20 mM NaBH₄ to reduce the amount of XyG and pectins. All solutions contained 0.02% NaN₃ to inhibit microbial growth, and multiple washes with deionized and distilled water were used between steps. The CWs were not dried or extracted with organic solvents to preserve native structure. Expansins were bound to the CW at a 1:100 ratio (dry mass) for the RKK and WWY samples and ∼1:60 for the WT sample. Expansin-bound CWs were centrifuged through a 40-μm filter to reach a water content of 40 wt%. Sugar composition analysis indicates that the resulting CWs contain 42% cellulose, 22% hemicellulose, and 36% pectin (SI Appendix, Table S1). The treatment with xyloglucanase and 1 M NaOH, which effectively enriched the CW cellulose content by removing 75% of the pectin and hemicellulose, would not be expected to alter cellulose structure or to dissociate tightly held complexes of cellulose and matrix polymers (28, 30).

For DNP experiments, the above CW samples were switched to a solution of 40% d₈-glycerol/D₂O/H₂O solution containing 35 mM TOTAPOL. A 43 mM stock solution of TOTAPOL was first prepared by dissolving 0.87 mg of TOTAPOL in 50 μL of d₈-glycerol:D₂O:H₂O (61:29:10 volume ratio) solution. Twenty to twenty-five milligrams of 40%-hydrated and expansin-containing CW was slowly dehydrated to ∼13%, then rehydrated with the TOTAPOL stock solution to reach a final hydration level of 40% and a TOTAPOL concentration of 35 mM. The final volume ratio of d₈-glycerol:D₂O:H₂O was 50:38:12. The mixture was stirred thoroughly and packed into a 3.2 mm sapphire rotor. The three CW samples contained 0.14 mg of WWY and RKK expansins and 0.23 mg of WT expansin. A control CW sample without TOTAPOL and with 36 mg of CW and 0.36 mg of expansin was measured at high temperature without DNP. A second control sample of pure expansin without the CW was used to compare the pure-expansin spectra with the NC-filtered spectra of expansin-containing CWs.

CW Sugar Composition Analysis.

The ¹³C-labeled Arabidopsis walls (∼3 mg) for the DNP NMR experiments were treated with 1 mL of 2 M trifluoroacetic acid (TFA) at 121 °C for 1 h to hydrolyze the matrix polysaccharides. The supernatant was dried under airflow and resuspended in distilled deionized water (ddH₂O). The 2 M TFA insoluble residue (predominantly cellulose) was hydrolyzed with 1 mL of 72% H₂SO₄ at room temperature for 1 h. The slurry of 72% H₂SO₄ was diluted with ddH₂O to 2 M concentration, incubated at 100 °C for 2 h with frequent vortexing (every 15 min), and neutralized with ammonium hydroxide (NH₄OH). Sugar composition of hydrolysates was assessed by high-performance anion-exchange chromatography with pulsed amperometric detection analysis. Briefly, TFA and H₂SO₄ hydrolyzed samples were filtered through a 0.1-μm filter (Amicon), and monosaccharides were separated on a Dionex Carbo PA-20 column at a flow rate of 0.5 mL⋅ min⁻¹ with 0.1 M NaOH isocratic elution for the first 15 min and then with linear gradient elution for 20 min starting with 0.1 M NaOH to a final mixture of 0.1 M NaOH and 0.1 M sodium acetate (1:1. vol/vol) and detected with a pulsed amperometric detector.

SSNMR Experiments.

All SSNMR spectra were measured at a magnetic field of 14.1 T, corresponding to ¹H and electron Larmor frequencies of 600 MHz and 395.18 GHz, respectively. Spectra without DNP were measured between 248 and 293 K, whereas DNP-enhanced spectra were measured at 102 K. The DNP-enhanced NMR experiments were carried out at Bruker Billerica laboratories on an Avance III SSNMR spectrometer equipped with a 3.2-mm triple-resonance low temperature MAS probe (34, 35). MW irradiation was generated by a second-harmonic gyrotron operating with a MW power of 9 W at the sample. The samples were spun at 8.5 kHz, and the temperature with MW on was calibrated by using KBr T₁ measurements (36). Non-DNP–enhanced experiments were conducted on an Avance II wide-bore SSNMR spectrometer. The ¹³C chemical shifts were referenced to tetramethylsilane either externally through adamantane or through the ¹H chemical shift by using the ratio of the gyromagnetic ratios of ¹H and ¹³C.

One-dimensional NC-filtered ¹³C spin diffusion experiments were conducted on expansin-containing CWs. The ¹³C magnetization was prepared with ramped ¹H-¹³C CP. A ¹⁵N-¹³C REDOR pulse train (t_mix) of 1.9 ms, corresponding to 16 rotor periods, was used to select the Cα and CO signals of expansin and remove all CW ¹³C signals (SI Appendix, Fig. S3A). A variable mixing time of up to 3 s transferred the protein ¹³C magnetization to the neighboring polysaccharides. For the buildup curves in Fig. 2F and SI Appendix, Fig. S6, intensities were calculated as .

For the RKK sample, a 2D protein-polysaccharide ¹³C correlation spectrum was measured. A t₁ period after the ¹⁵N-¹³C REDOR period encoded the protein signals, which were correlated to the polysaccharide signals after a long mixing period of 3 s (SI Appendix, Fig. S3B). A 2D protein-edited polysaccharide-polysaccharide ¹³C correlation spectrum was also measured by first transferring the NC-filtered protein ¹³C magnetization to the polysaccharides with a 3-s mixing time, then correlating the polysaccharide signals with each other with a 30-ms mixing time (SI Appendix, Fig. S3C). Only polysaccharides in close contact with RKK expansin exhibit intramolecular cross-peaks in this 2D spectrum.

Molecular Dynamics Simulations.

The initial model of the expansin–CW complex was built by superimposing the carbohydrate substrate in the crystal structure of the EXLX1–cellohexaose complex [Protein Data Bank (PDB) ID code: 4FER, chain B] (5) onto an edge glucan chain on the (100) face of a cellulose I_β microfibril, constructed from 25 chains, each 20 glucose units long, pr-equilibrated from its crystal structure (24) in water. We note that the number of chains in a cellulose microfibril and their arrangement (cross-sectional shape) are not well established at this time (37, 38); the configuration chosen here represents a somewhat arbitrary structure between the extremes of numbers of chains and shapes that are commonly postulated in current literature; these details of microfibril structure are unlikely to affect the conclusions drawn from this simulation. The initial configuration was selected because it produces the least steric clashes between the protein and the substrate, compared with choosing the other protein chain in the PDB, or other glucan chains in the microfibril. Counterions were added to allow for particle-mesh-Ewald summation (39, 40). This complex structure was placed in a 129 × 81 × 64 Å periodic water box and subjected to molecular dynamics for step-wise heating at 100, 200, and 300 K each for 20 ps, followed by 40 ps of equilibration. Finally, a 40-ns production trajectory at 300 K within the microcanonical NVE ensemble was collected for analysis. The CHARMM package (41) was used for calculations, in conjunction with force fields of CHARMM27 for the protein (with CMAP correction) (42, 43), C35 for cellulose (44), and TIP3 for water (45).