The size and chemical nature of the stabilizing cap at microtubule (MT) ends has remained enigmatic, in large part because it has been difficult to detect and measure it directly. By pulsing steady-state suspensions of bovine brain microtubules (MTs) with trace quantities of [gamma(32)P]GTP and sedimenting the MTs through 50% sucrose cushions to reduce background contaminating (32)P to negligible levels, we were able to detect a small number of (32)P molecules that remain stably bound to the MTs (a mean of 25.5 molecules of (32)P per MT). Analysis of the chemical form of the stably bound (32)P by thin-layer chromatography revealed that it was all (32)P-orthophosphate ((32)P(i)). The (32)P(i) was determined to be located at the MT ends because colchicine and vinblastine, drugs that suppress tubulin incorporation into the MT by binding specifically at MT ends, reduced the quantity of the stably bound (32)P(i). Taxol, a drug that stabilizes MT dynamics by binding along the MT surface rather than at the ends, did not affect the stoichiometry of the bound (32)P(i). If the bound (32)P is equally distributed between the two ends, each end would contain 12-13 molecules of (32)P(i). Beryllium fluoride (BeF(3-)) and aluminum fluoride (AlF(4-)), inorganic phosphate analogues, suppressed the dynamic instability behavior of individual MTs and, thus, stabilized them. For example, BeF(3-) (70 microM) reduced the MT shortening rate by 2.5-fold and decreased the transition frequency from the growing or the attenuated state to rapid shortening by 2-fold. The data support the hypothesis that the stabilizing cap at MT ends consists of a single layer of tubulin GDP-P(i) subunits. The data also support the hypothesis that the mechanism giving rise to the destabilized GDP-tubulin core involves release of P(i) rather than hydrolysis of the GTP.