We selected a set of G. max (Gm) CesA isoforms that phylogenetically cluster with Arabidopsis thaliana CesA1, CesA3 and CesA6, respectively, and that are widely and strongly expressed in unlignified soybean tissues (Figure 1, Figure 1—figure supplement 1A and B). Existing co-expression data from ATTED-II https://atted.jp/ (Obayashi et al., 2022) indicate that the enzymes are co-expressed with other primary cell wall genes implicated in pectin, arabinogalactan, and galactoglucomannan biosynthesis (Figure 1). We therefore conclude that the selected CesAs indeed represent primary cell wall CesAs.

To biochemically and structurally characterize the GmCesAs, we followed a similar heterologous expression protocol as established previously for hybrid aspen CesA8 (PttCesA8) (Purushotham et al., 2020). In short, the GmCesAs were expressed with N-terminal poly-histidine tags in Sf9 insect cells and purified by metal affinity and size exclusion chromatography in the detergent glyco-diosgenin (GDN; Materials and methods). Size exclusion chromatography separated all GmCesA isoforms into high and low molecular weight species (Figure 2A–C). Cryogenic and negative stain EM analyses identified these species as GmCesA trimers and monomers, respectively (see below). Of note, compared to GmCesA1 and GmCesA3, the yield of trimeric GmCesA6 was more variable, with some preparations producing primarily monomeric species. This suggests that GmCesA6 is less stable in a detergent-solubilized state compared with the other isoforms.

To test whether the detergent-solubilized GmCesA trimers dissociate into monomers over time, the purified GmCesA trimers were reinjected onto the size exclusion chromatography column after an overnight incubation on ice. For all species, the reinjected material eluted as a trimeric complex, indicating that assembled trimers are stable and do not interconvert with monomers within this timeframe (Figure 2A–C). The co-purifying monomers likely arise from incompletely assembled trimers or oligomer dissociation during purification.

Cellulose biosynthetic activity of the purified CesAs was quantified by measuring the incorporation of ³H-labeled glucose into insoluble cellulose, followed by scintillation counting, as previously described (Purushotham et al., 2016). As shown in Figure 2D, the relative activities of the monomeric and trimeric GmCesA fractions are comparable for each isoform, demonstrating that both species are catalytically active in vitro. Between the different isoforms, GmCesA1 exhibits greatest product accumulation (Figure 2D). In all cases, the in vitro synthesized polymer is readily degraded by a cellulase, indicating the formation of authentic cellulose. No product was obtained in the presence of EDTA, in agreement with previous observations (Purushotham et al., 2016).

To further assess catalytic differences between the GmCesA isoforms, we determined their pH optima for catalytic activity (Figure 2E). All GmCesA isoforms show greatest catalytic activity at neutral to mild-alkaline pH. GmCesA3 exhibits an activity optimum at pH 7 with a sharp decline at pH 8 and 9. In contrast, the activities of GmCesA1 and GmCesA6 peak at pH 8, with a slight decline at pH 9 (Figure 2E). Quantifying the release of UDP during biosynthesis reactions using a ‘UDP-Glo’ glycosyltransferase assay (Das et al., 2016) reveals Michaelis–Menten constants for GmCesA1 and GmCesA3 with respect to UDP-Glc of 0.44 mM and 0.18 mM, respectively. The apparent V_max value is ~eightfold higher for GmCesA1 compared to GmCesA3 (Figure 1—figure supplement 1C). The activity of GmCesA6 was too weak to be analyzed by this method. Further, we analyzed inhibition of the isoforms by UDP, which competitively inhibits PttCesA8 and related GT-2 enzymes (Gow and Selitrennikoff, 1984; Omadjela et al., 2013; Purushotham et al., 2016; Tlapak-Simmons et al., 2004). GmCesA1’s apparent IC₅₀ for UDP is about 0.8 mM, whereas this concentration is increased to about 1.3–1.5 mM for GmCesA6 and GmCesA3, respectively (Figure 2F).

Cryo-EM analyses of the high molecular weight GmCesA fractions (Figure 2A–C) revealed their organization into homotrimeric complexes (Figure 3A, Figure 3—figure supplements 1 and 2, and Supplementary file 1). For all isoforms, two-dimensional classification identified trimeric particles similar to PttCesA8 and cotton CesA7 (GhCesA7) (Purushotham et al., 2020; Zhang et al., 2021a). Particle classification in three dimensions followed by non-uniform refinement with applied C3 symmetry and local refinement generated cryo-EM maps ranging in resolution from about 3.0–3.3 Å.

Overall, the CesAs contain a cytosolic catalytic domain that interacts with the channel-forming TM region via three amphipathic IF helices (Figure 3B). The helices surround the entrance to the TM channel with the Trp residue of the QxxRW motif at its portal. As previously described for hybrid aspen PttCesA8 and GhCesA7, the CesA trimers are stabilized by the PCR domain that is inserted into CesA’s catalytic domain. (Purushotham et al., 2020; Zhang et al., 2021a).

The triangular PCR arrangement in a CesA complex positions the side chains of conserved Lys and Arg residues towards the threefold symmetry axis. These residues include Arg449, Lys452, and Arg453 in GmCesA6 and coordinate unidentified ligand(s) on the membrane distal and proximal side of the PCR triangle. For all three CesA isoforms, the ligands’ shapes are similar, suggesting that they represent the same small molecule (Figure 3—figure supplement 2B). On the membrane proximal side, the density extends by about 9 Å towards the membrane, perhaps representing a nucleotide bound in different poses, as previously suggested for PttCesA8 (Purushotham et al., 2020).

CesAs contain seven TM helices of which helices 1–6 create a cellulose conducting channel. In the previously described PttCesA8 and GhCesA7 complexes, TM helix 7 of one protomer packs against TM helices 5 and 6 of a neighboring CesA subunit (Figure 3C). In new soybean CesA structures, however, this helix is more flexible, as evidenced by weaker map quality (Figure 3—figure supplement 2A), and shifted to the periphery of the trimer. In the new position, the helix primarily mediates contacts with TM helix 5 and the C-terminal segment of TM helix 3 of the same CesA protomer (Figure 3B and C). The helix has been modeled for GmCesA1 and GmCesA6, while its density is detectable at a similar position but too discontinuous for modeling in the GmCesA3 map.

The displacement of TM helix 7 away from the TM channel of the neighboring subunit opens a lateral lipid-exposed window in the neighboring subunit’s channel architecture (Figure 3C). The window is formed by the N-terminal region of IF helix 3 and TM helices 4 and 6. About midway across the membrane, the opening is roughly 6 Å wide, for example between the side chains of Ile901 in TM helix 4 and Trp1018 in TM helix 6 (Figure 3C). The lateral window likely exposes the translocating nascent cellulose polymer to the hydrophobicity of the lipid bilayer. A similar lipid exposed polysaccharide translocation pathway has recently been described for hyaluronan synthase (Maloney et al., 2022).

Plant CesAs contain two structurally and functionally unresolved domains, which are the N-terminal domain (NTD) and the CSR. The NTD has been resolved at lower resolution for PttCesA8, where it forms a helical stalk extending from the catalytic domains into the cytosol (Purushotham et al., 2020). We observe a similar stalk-like extension in some of the trimeric GmCesA3 particles (Figure 3—figure supplement 1). However, the NTD is only resolved for a small subset of GmCesA3 particles and not resolved at all for the other GmCesAs. This suggests that it can adopt multiple conformations, with the stalk being one of them.

The CSR has been proposed to be intrinsically disordered (Scavuzzo-Duggan et al., 2018) and only its short N- and C-terminal helical segments are visible in the cryo-EM maps. For GmCesA1, however, at lower contour levels, additional CSR density is evident, extending from corners of the catalytic domains of the GmCesA1 trimer (Figure 3D). Viewed from the cytosol, the extra density resembles a hook extending by about 20 Å clockwise and tangentially along the trimer’s corners at the water-lipid interface. At this position, the CSR’s N-terminal conserved cysteine residue(s) postulated to be acylated (Kumar et al., 2016) reside near the membrane interface (Figure 3D and E).

We next tested whether GmCesA homotrimers of different isoforms would interact in vitro. To this end, we individually expressed and purified trimers of poly-His tagged GmCesA1 and GmCesA6 (His-CesA1 and His-CesA6) and TwinStrep-tagged GmCesA1 and GmCesA3 (Strep-CesA1 and Strep-CesA3) (Figure 4). Cross-isoform interactions were tested by tandem purifications over Ni-NTA and Strep-Tactin affinity matrices. His-CesA1 can be distinguished from Strep-GmCesA3 by Coomassie stained SDS-PAGE due to size differences, whereas all other species comigrate. Therefore, western blotting together with Coomassiestained SDS-PAGE was performed to evaluate the co-purification results.

An equimolar mixture (based on UV absorbance) of His-CesA1 and Strep-CesA3 was incubated for 180 min on ice and sequentially purified using (1) Ni-NTA resin and (2) Strep-Tactin beads (Figure 4A). The Coomassie stained SDS-PAGE resolved both GmCesA species in the initial mixture, after elution from the Ni-NTA resin, as well as upon elution from the Strep-Tactin beads. The identity of the bands as His-CesA1 and Strep-CesA3 was confirmed by western blotting. We observed no non-specific binding of His-CesA1 to Strep-Tactin beads or Strep-CesA3 to Ni-NTA resin (Figure 4E).

Similar experiments with combinations of Strep-CesA3 and His-CesA6 (Figure 4B) and Strep-CesA1 and His-CesA6 (Figure 4C) yielded comparable results, although the co-eluting species cannot be distinguished by Coomassie staining alone, due to comigration. None of the species showed detectable non-specific binding to the affinity resins in the absence of the corresponding tags (Figure 4E–G).

As an additional control, we analyzed whether differently tagged homotrimers of the same isoform also interact with each other. To this end, a mixture of His-CesA1 and Strep-CesA1 was subjected to tandem affinity purification as described above (Figure 4D). We failed to detect any co-purification of Strep-CesA1 when applying Ni-NTA as the first affinity chromatography step. Similar results were obtained for Strep- and His-tagged combinations of homotrimers of CesA3 or CesA6 (Figure 4—figure supplement 1A–D). Qualitatively, our interaction data are consistent with previously published co-immunoprecipitations of primary and secondary CesA isoforms (Gonneau et al., 2014; Timmers et al., 2009).

Similar tandem purification experiments were also performed with the monomeric CesA fractions obtained from size exclusion chromatography (Figure 2A–C). As observed for the homotrimeric complexes, monomeric His-CesA1 co-purifies with monomeric Strep-CesA3, demonstrating that trimeric assemblies are not necessary for the observed interactions (Figure 4—figure supplement 1E and F).

The biological functions of CesA’s NTD and CSR are currently unknown. The NTD’s RING-like domain has been shown to form dimers and trimers in vitro (Kurek et al., 2002; Purushotham et al., 2020), raising the possibility that it could form inter-trimer complexes accounting for the observed isoform interactions.

To test this hypothesis, N-terminally truncated constructs of GmCesA1 and GmCesA3 were expressed and purified as described for the full-length variants. The constructs lack the first 259 (GmCesA1) and 242 (GmCesA3) residues yet purify as trimers (besides monomers) and exhibit in vitro catalytic activity similar to the full-length constructs (Figure 5A). A tandem affinity purification of a mixture of the truncated His-CesA1 and Strep-CesA3 isoforms demonstrates their interaction in the absence of the NTD, similar to the full-length enzymes (Figure 5B and C and Figure 4A).

To test whether the CSR is involved in isoform interaction, the domain was replaced in the N-terminally truncated GmCesA1 and GmCesA3 constructs with a flexible loop of 20 residues (see Materials and methods), thereby generating GmCesA constructs devoid of the NTD as well as the CSR (Figure 5D). Negative stain EM of the purified truncated GmCesA1 particles demonstrates their trimeric assembly, as observed for the full-length and NTD truncated versions (Figure 5D). Upon deletion of the CSR, the GmCesA1 and GmCesA3 isoforms no longer co-purify in vitro, suggesting that the region is indeed required for inter-isoform interaction (Figure 5E). Further, an N-terminally and CSR truncated GmCesA3 construct shows only minor, most likely non-specific, interaction with full-length GmCesA1, suggesting that GmCesA isoforms interact primarily via their CSR domains (Figure 5F).

Negative stain electron microscopy analyses of the individual GmCesA homotrimers revealed monodisperse particle distributions (Figure 6A–C). The size and shape of the particles is consistent with the GmCesA homotrimers observed by cryo-EM and remains unchanged over a course of a week when incubated on ice.

To visualize the interaction of the isoforms, all three homotrimeric isoforms were combined at equal molar ratio and subjected to size exclusion chromatography after incubation overnight on ice. Negative stain EM analysis of high molecular weight fractions eluting after the void volume revealed GmCesA clusters of varying stoichiometries, ranging from 2 to >10 particles (Figure 6D). The clusters likely arise from trimer-trimer interactions in different orientations and vary in diameter from about 50–100 nm. No clustering was observed for any of the individual GmCesA isoforms alone, even after prolonged incubations on ice.

Further, cryo-EM was used to analyze a mixture of individually purified GmCesA1 and GmCesA3 trimers. The trimers were combined and incubated on ice for 60 min prior to cryo grid preparation. Focusing on particles larger than a CesA trimer, two-dimensional classification of the obtained particles revealed shapes resembling arrangements of two GmCesA trimers. The identified particles are consistent with side-by-side arrangements of two GmCesA trimers in opposite orientations. By this organization, the catalytic domains of the trimers are co-planar and the micelle-embedded TM segments are above and below the plane (Figure 6E).

To support this interpretation, two cryo-EM volumes of a GmCesA1 trimer were manually arranged side-by-side but in inverted orientations. The individual volumes were placed such that the CSR densities of two CesA subunits of each trimer would contact each other. The obtained dimer-of-trimer volume was then used to calculate two-dimensional projections for comparison with the experimentally obtained 2D class averages. Indeed, the generated ‘upside down and side-by-side’ arrangement of GmCesA trimer volumes resembles the experimental class averages (Figure 6F and G). Steric interferences of the micelle-embedded TM regions likely prevent the physiological parallel arrangement of two GmCesA trimers under the experimental conditions.

We investigated whether the cross-isoform interaction of GmCesA trimers affects their in vitro catalytic activities. To this end, in vitro cellulose biosynthesis was quantified radiometrically from reactions containing one GmCesA isoform at a constant concentration (20 µM) and increasing concentrations of a different Gmdifferent CesA (1–20 µM). As a reference, cellulose biosynthetic activities were also determined for each isoform alone at the concentrations used in the combined assays. As shown in Figure 7A–F, for all isoform combinations, the measured activities exceed the theoretical activities (calculated by adding the individually measured activities) at least one to two-fold, depending on the isoform combination. This suggests synergistic cellulose biosynthesis in the presence of two CesA isoforms. Performing the titration experiment with samples of the same isoform does not reveal any synergy, consistent with the lack of interaction between trimers of the same CesA isoforms (Figures 7G and 4D and Figure 7—figure supplement 1).

Additionally, comparing the measured and additive activities obtained after combining all three GmCesA isoforms (at 6.6 µM each) reveals an experimental activity about threefold above the additive value. This activity level may arise from different dimeric arrangements of GmCesA trimers (1+3, 1+6, and 3+6) and/or the formation of larger complexes of different isoform trimers (1+3 + 6; Figure 7H).

GmCesA1 exhibits higher in vitro catalytic activity compared to GmCesA3 and GmCesA6 (Figure 2D and Figure 1—figure supplement 1B). To test whether the observed synergistic effects are due to altered catalytic activity of only one isoform or both, one GmCesA isoform was inactivated after purification by incubation with the oxidant sodium tetrathionate (Tie et al., 2004). While the inactivated CesAs exhibit activity levels comparable to EDTA-treated negative controls, they remain trimeric and interact with other isoforms as observed for the unmodified versions (Figure 7—figure supplement 2). Inactivation could be due to modification of a conserved cysteine residue in CesA’s catalytic pocket (such as Cys558 or Cys630 in GmCesA6).

Performing the above-described activity assays with pairs of inactive and active GmCesA trimers demonstrates that all three isoforms exhibit increased catalytic activity when combined with an inactive trimer of a different isoform (Figure 7—figure supplement 3). This suggests that intertrimer interactions impact the catalytic activity of all isoforms, perhaps by altering the accessibility of the catalytic pocket (discussed below).

A pre-computed cluster of loose CesA homologues (HOM05D000074) was downloaded from Dicots PLAZA 5.0 (Van Bel et al., 2022), from which soy (Glycine max), Arabidopsis, cotton (Gossypium hirsutum), poplar (Populus trichocarpa), tomato (Solanum lycopersicum), Physcomitrium, and Amborella protein sequences were isolated. The sequence of CesA8 from Populus tremula x tremuloides (PttCesA8) was also added at this stage for later reference. A region corresponding to the Pfam 'Cellulose synthase' domain HMM profile (PF03552) was then extracted using HMMER (Eddy, 2011) with an E-value cutoff of 1×10⁻⁸⁰ and a bespoke Python script. The extracted sequences were aligned using MAFFT v7.490 (Katoh and Standley, 2013) before an initial near-maximum likelihood phylogeny using FastTree under default settings. This tree was used to extract bona fide CesA sequences from Csl sequences using Figtree. The corresponding rows of the alignment were then extracted alongside A. thaliana CslD5, which was used as an outgroup in the following analysis. Model selection was carried out using ProtTest version 3.4.2 (Darriba et al., 2011) before calculating the final maximum-likelihood phylogeny using RAxML version 8.2.12 (Stamatakis, 2014) under a JTT + I + Γ model with 100 rapid bootstrap pseudo-replicates. The tree was rendered in Figtree with subsequent labelling in Inkscape.

The primary cell wall CesA1, CesA3 and CesA6 genes from soybean (Glyma.06G069600, Glyma.12G237000, and Glyma.02G080900) were synthesized (Gene Universal) with an N-terminal 12 x His-tag coding sequence and cloned into NotI and HindIII restriction sites in the pACEBac1 vector. The N-terminally TwinStrep-tagged GmCesA1 and GmCesA3 constructs were generated by QuikChange mutagenesis from the pACEBac1-12xHisCesA vectors. The N-terminally deleted GmCesA constructs (ΔNCesAs) were generated by PIPE cloning from the full-length constructs, resulting in plasmids pACEBac1-ΔNCesA1R260, pACEBac1-ΔNCesA3V243, and pACEBac1-ΔNCesA6M248. The CSR regions of GmCesA1 (residues 654–713) and GmCesA3 (residues 648–709) were replaced with the sequence ASGAGGSEGGGSEGGTSGAT (Baytshtok et al., 2017) in the -ΔNCesA1R260 and -ΔNCesA3V243 backgrounds by gene synthesis.

A-multi GmCesA expression cassette was generated for the co-expression of 12xHis-GmCesA1, TwinStrep-GmCesA3 and 3xFlag-GmCesA6 using homing endonuclease/BstXI multiplication according to the protocol detailed in the MultiBac Multi-Protein Expression in Insect Cells manual (Geneva Biotech). Briefly, the TwinStrep-GmCesA3-pACEBac1 vector (acceptor) was single restriction digested with BstXI. The His-GmCesA1-pACEBac1vector was double restriction digested with I-CeuI and BstXI to excise the promoter and terminator containing His-GmCesA1 (donor). The resulting fragment was inserted into the BstXI-digested TwinStrep-GmCesA3-pACEBac1 vector to generate the TwinStrep-GmCesA3_His-GmCesA1-pACEBac1 construct. The above cloning procedure was repeated one more time to insert the 3xFlag-GmCesA6 (donor) into the GmCesA1 and GmCesA3 containing vector (acceptor) to generate the multi-GmCesA expression construct TwinStrep-CesA3_His-GmCesA1_3xFlag-GmCesA6-pACEBac1.

An aliquot of 3 μL of 100 ng/μL pACEBac1-GmCesA plasmid was used for transformation into 50 μL chemically competent DH10MultiBacTurbo E. coli cells. Bacmids were isolated from white colonies on a Bluo-Gal agar plate and transfected into Spodoptera frugiperda 9 (SF9) cells. P0, P1, and P2 baculovirus was generated according to the Joint Centre for Innovative Membrane Protein Technologies (JCIMPT) protocol.

For GmCesA expression, Sf9 insect cells were infected with 15 mL P2 baculovirus per 400 mL at a density of 3×10⁻⁶ cells per mL and grown at 27 °C for 48–72 hr in an orbital shaker. Cells were then harvested by centrifugation at 5,000×g for 10 min at 4 °C. Cell pellets were resuspended in buffer A (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5 mM sodium phosphate, 5 mM sodium citrate, and 1 mM TCEP) supplemented with 1% lauryl maltose neopentyl glycol (LMNG, Anatrace), 0.2% cholesteryl hemisuccinate (CHS, Anatrace), and protease inhibitors PIC (0.4 mM AEBSF, 2 μM aprotinin, 30 μM pepstatin, 7.5 μM pepstatin, 40 μM bestatin, 35 μM E-64, and 4.5 mM benzamidine hydrochloride) and lysed in a glass dounce homogenizer  (~30 strokes). The lysate was solubilized at 4 °C for 1 hr on a rocker. After separation of insoluble material by centrifugation at 200,000×g for 45 min, 5 mL of Ni-NTA resin (HisPur Ni-NTA Resin, Thermo scientific) and 20 mM imidazole was added to the supernatant and incubated for 1 hr at 4 °C on a rocker. After batch binding, the resin was packed into a gravity flow column and then sequentially washed twice with ten column volumes each of buffer A containing 40 mM imidazole, 0.02% glyco-diosgenin (GDN, Anatrace) and PIC, followed by ten column volumes of buffer A containing 1 M NaCl and 0.02% GDN (wash 3) and PIC. The final wash step (wash 4) was with ten column volumes of buffer A containing 0.02% GDN, PIC and 60 mM imidazole. The GmCesAs were eluted with six column volumes of buffer A containing 0.02% GDN, PIC and 400 mM imidazole.

The TwinStrep-tagged GmCesAs were affinity purified by incubating the membrane extract with 5 mL Strep-Tactin sepharose at 4 °C for 1 hr on a rocker. After batch binding, the resin was packed into a gravity flow column and washed as described above for Ni-NTA affinity column except the buffers lacked imidazole. The TwinStrep-tagged GmCesA was eluted using 6 column volumes of buffer A containing 0.02% GDN and 5 mM desthiobiotin.

His- and TwinStrep-tagged GmCesAs were further purified by size-exclusion chromatography (SEC) using Superose 6 Increase 10/300 GL column (Cytiva) equilibrated in buffer A containing 0.02% GDN without any protease inhibitors. The purified GmCesA trimers and monomers were immediately used for activity assays and pulldown experiments or flash-frozen in liquid nitrogen and stored at –80 °C.

After SEC, either the trimeric or monomeric fractions of His-GmCesA and TwinStrep-GmCesA were pooled and used for tandem affinity chromatography using Ni-NTA and Strep-Tactin sepharose beads. His-CesA1 and TwinStrep-CesA3 trimers were mixed at 150 μg/mL concentration and incubated at 4 °C for 3 hr. The mixture was first loaded onto 200 μL bed volume Ni-NTA beads for 1 hr at 4 °C in the presence of 20 mM imidazole. After collecting the flowthrough, the beads were washed with 10 bed volumes of buffer A containing 0.02% GDN and 20 mM imidazole for three washes followed by elution in 5 bed volumes of buffer A containing 0.02% GDN and 400 mM imidazole.

The Ni-NTA eluted material was next bound to 200 μL Strep-Tactin sepharose beads for 1 hr at 4 °C in a rotator shaker. After batch binding, the beads were packed into a gravity flow column. The flow-through was collected and the column was subsequently washed three times with 10 column volumes of buffer A containing 0.02% GDN. Bound proteins were eluted with buffer A containing 0.02% GDN and 5 mM desthiobiotin. All fractions were analyzed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting using anti-His (QIAGEN, #34650) and anti-Strep (MilliporeSigma) primary antibodies and a DyLight 800-coupled anti-mouse secondary antibody (Rockland, #610-145-002) for detection.

The same protocol was used to study the interactions of GmCesA1 and GmCesA6 or GmCesA3 and GmCesA6 using Strep-tagged GmCesA1 and His-tagged GmCesA6, Strep-tagged GmCesA3 and His-tagged GmCesA6, respectively. Control binding experiments were performed by loading the His-tagged GmCesA species onto Strep-Tactin sepharose beads and the Strep-tagged GmCesAs onto the Ni-NTA beads. In each case, the beads were washed as described above.

GmCesA was inactivated by treating GmCesA trimers with 10 mM sodium tetrathionate overnight at room temperature (24 °C). The treated sample was purified over a size exclusion column to remove the excess sodium tetrathionate. Binding experiments with inactivated and untreated CesAs were performed using Ni-NTA beads as described above.

Freshly purified or aliquots of flash-frozen enzyme thawed on ice were used for activity assays. In general, activity assays were performed as described earlier (Purushotham et al., 2020). Radiometric quantification of in vitro synthesized cellulose was performed by combining 5 μM GmCesA, 5 mM UDP-glucose (UDP-Glc), and 0.34 μM (12.5 mCi/L) UDP-[³H]-Glc in buffer containing 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 20 mM MgCl₂, 5 mM sodium phosphate, 5 mM sodium citrate, and 1 mM TCEP. The reaction mixtures were incubated for 45min at 37°C. After incubation, the entire reaction mixture was spotted on Whatman Grade 3 MM chromatography paper. Free substrate was removed by descending paper chromatography in 60% ethanol. The radioactivity retained at the origin was quantified by scintillation counting.

The pH optima of GmCesAs were determined by incubating the GmCesAs in MMT buffer, consisting of DL-malic acid, MES and Tris base in the molar ratios 1:2:2-DL-malic acid:MES:Tris base. The desired pH was adjusted with NaOH or HCl. Activity assays were performed as mentioned above in technical triplicate from two biological replicates.

Cellulase digestions were performed by adding 5 U of endo-β–1,4-glucanase (Trichoderma longibrachiatum; Megazyme: E-CELTR) directly to the reaction mixture. Following the in vitro synthesis reaction, cellulase treatment was performed for 3 hr at 37 °C. Cellulose quantification by scintillation counting was performed as mentioned above to quantify the product.

Steady-state kinetic analyses were performed in triplicate using the UDP-Glo glycosyltransferase Assay kit (Promega) to monitor the released UDP according to the manufacturer’s instructions. For measuring enzyme kinetics, the reaction mixtures containing 0.33 and 3.3 nM trimeric GmCesA1 or GmCesA3, respectively, and 0–1 mM UltraPure UDP-Glc (Promega) were added to 20 mM Tris-HCl buffer, pH 7.5, 100 mM NaCl, 5 mM sodium phosphate, 5 mM sodium citrate, 20 mM MgCl₂, 1 mM TCEP, and 0.02% GDN and incubated in a final volume of 25μL for 1 h at 30 °C. Afterwards, the reaction mixture was mixed with an equal amount of UDP-Glo reagent (Promega) in a 96-well Nunclon Delta-Treated flat-bottom microplate (Thermo Fisher Scientific) and incubated for 1 h at room temperature before measuring luminescence using a GloMax Explorer plate reader (Promega). A standard curve was used for quantification of the UDP produced. Kinetic values were obtained using the nonlinear regression function in GraphPad Prism.

UDP inhibition was analyzed using radiometric quantification of in vitro synthesized cellulose, as described above by titrating 0.01–7.5 mM UDP in the reaction. Substrate concentrations for the individual reactions were 1.4 mM, 0.5 mM, and 2.3 mM for GmCesA1, GmCesA3, and GmCesA6, respectively. Inhibition constants (IC₅₀) for each GmCesA were obtained by data analysis in GraphPad Prism.

Activity synergism between different GmCesA isoform trimers were studied by mixing two GmCesA isoforms, one at a constant concentration of 20 μM and one at increasing concentrations from 1 to 20 μM. The activities of the individual isoforms at the respective concentrations were also measured. In vitro synthesized cellulose was quantified by scintillation counting, as described above.

After size exclusion chromatography, the freshly purified protein fractions were pooled. The protein quality was monitored by negative stain EM. The proteins were diluted to 0.01 mg/mL and 4 μL was applied to a glow discharged Formvar/Carbon grid (Electron Microscopy Sciences) for 30 s, followed by 2 x washes with 4 μL H₂O. The grid was negatively stained with 4 μL 0.75% Uranyl Formate (UF) in H₂O for 30 s. Excess UF was removed by blotting with filter paper and the grid was air dried. Images were taken on a Tecnai F20 at the Macromolecular Electron Microscopy Core (MEMC) facility at the University of Virginia.

For cryo grid preparation, the protein samples were concentrated until NanoDrop readings reached 3 mg/ml (1.8 mg/mL using an extinction coefficient of ~200,000/ (M cm)). 2.5 µL were applied to a C-flat 300 mesh 1.2/1.3 copper grids (Electron Microscopy Sciences), glow-discharged in the presence of amylamine at 25 mA for 45 s, and blotted with a Vitrobot Mark IV (FEI, Thermo Fisher Scientific) with force 4 for 6 s at 4 °C, 100% humidity, and frozen in liquid ethane.

Cryo-EM data were collected at the MEMC at the University of Virginia on a Titan Krios microscope operated at 300 keV and equipped with a Gatan K3 direct electron detector positioned post a Gatan Quantum energy filter. On average, a total of ~6000 movies were collected from one or two grids in counting mode at a magnification of 81 K, pixel size of 1.08 Å, and defocus range from –2.2 to –1.2 µm with step size of 0.2 µm. The total dose was 50 e^-/Ų. Movies with 40 frames were collected at 5.17 s/movie rate.

Cryo-EM data processing was done in cryoSPARC v4 (Punjani et al., 2017). The general workflow is described as supplemental information, Figure 3—figure supplement 1. Based on improved map qualities, all volumes were generated imposing threefold symmetry (C3). Maps generated without symmetry assignment (C1) were used to visualize the unidentified ligand coordinated by the PCR domains at the particle’s symmetry axis. Initial protein models were generated in AlphaFold and manually adjusted in Chimera and Coot (Emsley and Cowtan, 2004; Pettersen et al., 2004). Models were refined in Phenix.refine (Afonine et al., 2012). The following regions were omitted from the constructs due to weak or missing map density: GmCesA1:1–260 (NTD), 654–717 (CSR), 954–978 (gating loop), 1064–1078 (C term); GmCesA3: 1–251 (NTD), 648–712 (CSR), 948–974 (gating loop), 1029–1079 (C term and TM 7); and GmCesA6: 1–248 (NTD), 646–712 (CSR), 952–971 (gating loop), 1060–1078 (C term). Cellobiose was modeled at the acceptor positions of GmCesA3 and GmCesA6. The corresponding density in the GmCesA1 map was too weak for interpretation. Structural representations were generated in ChimeraX or Pymol (Pettersen et al., 2021; PYMOL, 2025).

CesA containing expression constructs are available upon request.