A constant-time, Carr-Purcell-Meiboom-Gill (CPMG) transverse relaxation, R(2), dispersion experiment for carbonyl carbons was designed and executed to detect micros-ms time-scale dynamics of protein backbone carbonyl sites. Because of the large (ca. 55 Hz) C(alpha)-C' J-coupling, the carbonyl signal intensity is strongly modulated as the spacing between CPMG pulses is varied, in uniformly (13)C enriched proteins, unless care is taken to minimize the perturbation of the C(alpha) magnetization by the CPMG pulses. CPMG pulse trains consisting of either a band-selective pulse, such as RE-BURP, or rectangular (with an excitation null in the C(alpha) region of the spectrum) pulses were employed in order to minimize C' signal modulation by C(alpha)-C' J-coupling. The performance of these types of CPMG refocusing pulses was assessed by computer simulation, and by comparing dispersion profiles measured for (1) uniformly [(13)C,(15)N, (2)H] ((2)H at non-labile hydrogen sites) labeled, and (2) uniformly (15)N/selectively-(13)C' labeled samples of HIV-1 protease bound to a potent inhibitor, DMP323. In addition, because the uniformly (13)C/(15)N/(2)H labeled sample was well suited to measure (15)N and (1)H R(2) dispersion as well as (13)C' dispersion, conformational exchange in the inter subunit beta-sheet hydrogen-bond network of the inhibitor-bound protease was elucidated using relaxation dispersion data of all three types of nuclei.