Experimental support for the model was found in cells synthesizing cellulose through use of improved methods for FF-TEM. We visualized the surface of plasma membranes adjacent to the forming plant cell wall–the location where the cellulose fibril crystallization process is expected to take place. In high resolution replicas of cells that had been actively engaged in cellulose synthesis, we observed a hemispherical accumulation of material at the apparent origination point of cellulose fibrils, with the tops of rosette CSCs also visible nearby on the surface of the plasma membrane. The hemispherical accumulation of material may represent an uncrystallized aggregate of glucan chains at the base of the forming cellulose fibril, as was predicted computationally. Although in nature each glucan chain likely exits through a small protein channel within one CESA of the rosette CSC, in these simulations, the membrane itself served to initially separate the forming glucan chains. The growth of the chains was modeled by relaxing the monomers of each chain one by one while the chain was moved up in increments of 5A ˚, corresponding to the length of one glucose monomer as if added one-by-one in the natural polymerization process. In this configuration, the nascent glucan chains interacted with and extended over the membrane through the hydrogen bonds, van der Waals interactions, thus preventing intrachain folding. To shorten the calculation time, the chains were forced to meet by VE-821 approaching step by step the first monomers produced. A minimum of ten free monomers was necessary before the nascent chains reached each other. Once contact was made, the same two-by-two and radial organization arose as observed with the pre-organized protofibril. In addition, the protofibril also had a tendency to bend at an early stage of its assembly. This proves that the supramolecular organization of the base of the protofibril did not depend on the configuration of the modeling, but only on the interchain and membrane-chain interactions. Interestingly, a minimum of six interacting monomers was required for the formation and maintenance of the protofibril above the loosely organized pool on the surface. Below this number, the interactions with the membrane promoted the disassembly of the protofibril. This result suggests that during the first assembly steps some hydrogen bonds, O–H interactions Paclitaxel 33069-62-4 between glucan chain monomers, are competing with other hydrogen bonds, H–P and H–N interactions between glucan and membrane molecules. The model membrane did not contain protein. Since proteins at the surface of the natural plasma membrane are expected to replace H–P interactions with weaker H–N interactions, they will favor interchain interaction and protofibril assembly. Interchain assemblies predicted by the simulation without proteins would be even more favorable with the presence of proteins. A closer look at the base of the 
protofibril at a later stage shows that, in this location, cellulose chains remained disorganized and looped out while staying agglomerated on the membrane through hydrogen bonds, H–P and H–N interactions between glucan monomers and the membrane. To obtain independent evidence for the existence of a zone of disorganized glucan chains at the site of extrusion before cellulose fibril formation.