Next we included the presence of a membrane monolayer in the simulation. Finally, we simulated the extrusion of the glucan chains from the CSC by relaxing the monomers of each of the extended chains one by one, while moving the modeled synthetic site in the plane of the membrane by increments matching the length of a glucose monomer. No water was explicitly included in the simulations due to the large computational time that would have been required to account for its movements. In addition, the free water potential in the presence of solutes and hydrophilic cell wall matrix polymers immediately outside the plant plasma membrane is unknown, making water difficult to model meaningfully for our purpose. In any case, it is likely that the main features of cellulose chain behavior shown here are similar to those that would be computationally predicted in the presence of water, which is expected to affect mainly the absolute value of movement dynamics. Given the evidence from simulations for pair-wise initial chain aggregation, the overall observations for a 6-chain protofibril are predicted to represent those in native cellulose Torin 1 fibrils containing 18–36 chains. The crosssection of the 6-chain protofibril was about 205 A ˚ 2. If a native cellulose fibril was formed from six such protofibrils, it would have a cross-sectional area of 1230 A ˚ 2 and a diameter of 4 nm in accordance with published data for some cell walls and the 4.1 nm average fibril diameter estimated from the FF-TEM replicas in this study. Alternatively, the initial pairwise glucan chain aggregation shown by the molecular simulations could support the formation from fewer than 36 glucan chains of (+)-JQ1 1268524-70-4 smaller cellulose fibrils that also exist in nature. The simulations showed a major role of van der Waals forces in chain interactions, which is consistent with previous modeling of forces driving the initial crystallization of cellulose. The time required for molecules to interact is 7–8 orders of magnitude faster than the time required for synthesizing a cellulose monomer. However, the 6-chain protofibril would not be fully crystalline, which remains consistent with evidence that cellulose crystallization may limit the rate of polymerization in bacteria and plants. The modeling showed that the membrane plays an important role in the initiation of glucan assembly by restricting chain motion to one plane and increasing the probability of contact. Once interchain contact of at least 6 monomers is made, glucan assembly further proceeds to form the protofibril through a ratchet-like mechanism involving hydrogen bonds. Importantly, the computational 
modeling showed that a disorganized group of glucan chains remained pooled at the base of the protofibril when membrane was present or absent in the simulations. Morphological evidence of a pool of glucan chains at the base of native cellulose fibrils was seen in FF-TEM images: a hemisphere of material existed at the ends of putative cellulose fibrils on the plasma membrane surface of TEs engaged in cellulose synthesis. Previously, FF-TEM was extensively used to visualize rosette CSCs in the PF of the fractured plasma membrane at the point where the TMH of multiple CESAs cross the plasma membrane.