The far-ultraviolet circular dichroism spectrum of the alpha beta-tubulin dimer analyzed by six different methods indicates an average content of approximately 33% alpha helix, 21% beta sheet, and 45% other secondary structure. Deconvolution of Fourier transform infrared spectra indicates 24% sheet, 37% (maximum) helix, and 38% (minimum) other structure. Separate alignments of 75 alpha-tubulin, 106 beta-tubulin, and 14 gamma-tubulin sequences and 12 sequences of the bacterial cell division protein FtsZ have been employed to predict their secondary structures with the multiple-sequence method PHD [Rost, B., & Sander, C. (1993a) J. Mol. Biol. 232, 584-599]. The predicted secondary structures average of 33% alpha helix, 24% beta sheet, and 43% loop for the alpha beta dimer. The predictions have been compared with sites of limited proteolysis by 12 proteases at the surfaces of the heterodimer and taxol-induced microtubules [de Pereda, J. M., & Andreu, J. M. (1996) Biochemistry 35, 14184-14202]. From 24 experimentally determined nicking sites, 18 are at predicted loops or at the extremes of secondary structure elements. Proteolysis zone A (including acetylable Lys40 and probably Lys60 in alpha-tubulin and Gly93 in beta-tubulin) and proteolysis zone B (extending between residues 167 and 183 in both chains) are accessible in microtubules. Proteolysis zone C, between residues 278 and 295, becomes partially occluded in microtubules. The alpha-tubulin nicking site Arg339-Ser340 is at a loop following a predicted alpha helix in proteolysis zone D. This site is protected in taxol microtubules; however, a new tryptic site appears which is probably located at the N-terminal end of the same helix. Zone D also contains beta-tubulin Cys354, which is accessible in microtubules. Proteolysis zone E includes the C-terminal hypervariable loops (10-20 residues) of each tubulin chain. These follow the two larger predicted helical zones (residues 372-395 and 405-432 in beta-tubulin), which also are the longer conserved part of the alpha- and beta-tubulin sequences. Through combination of this with other biochemical information, a set of surface and distance constraints is proposed for the folding of beta-tubulin. The FtsZ sequences are only 10-18% identical to the tubulin sequences. However, the predicted secondary structures show two clearly similar (85-87 and 51-78%) regions, at tubulin positions 95-175 and 305-350, corresponding to FtsZ 65-135 and 255-300, respectively. The first region is flanked by tubulin proteolysis zones A and B. It consists of a predicted loop1-helix-loop2-sheet-loop3-helix-loop4-sheet fold, which contains the motif (KR)GXXXXG (loop1), and the tubulin-FtsZ signature G-box motif (SAG)GGTG(SAT)G (loop3). A simple working model envisages loop1 and loop3 together at the nucleotide binding site, while loops 2 and 4 are at the surface of the protein, in agreement with proteolytic and antigenic accessibility results in tubulin. The model is compatible with studies of tubulin and FtsZ mutants. It is proposed that this region constitutes a common structural and evolutionary nucleus of tubulins and FtsZ which is different from typical GTPases.