13C NMR chemical shift results as a function of pH for a series of carboxyl 13C-enriched saturated fatty acids (8-18 carbons) bound to bovine serum albumin (BSA) are presented. For octanoic acid bound to BSA (6:1, mol/mol), the chemical shift of the only FA carboxyl resonance (designated as peak c), plotted as a function of pH, exhibited a complete sigmoidal titration curve that deviated in shape from a corresponding theoretical Henderson-Hasselbach curve. However, FA carboxyl chemical shift plotted as a function of added HCl yielded a linear titration curve analogous to those obtained for protein-free monomeric fatty acid (FA) in water. The apparent pK of BSA-bound octanoic acid was 4.3 +/- 0.2. However, the intrinsic pK (corrected for electrostatic effects resulting from the net positive charge on BSA) was approximately 4.8, a value identical to that obtained for monomeric octanoic acid in water in the absence of protein. For long-chain FA (greater than or equal to 12 carbons) bound to BSA (6:1, mol/mol), chemical shift titration curves for peak c were similar to those obtained for octanoic acid/BSA. However, the four additional FA carboxyl resonances observed (designated as peaks a, b, b', and d) exhibited no change in chemical shift between pH 8 and 3. For C14.0 X BSA complexes (3:1 and 6:1, mol/mol) peaks b' and a exhibited chemical shift changes between pH 8.8 and 11.5 concomitant with chemical shift changes in the epsilon-carbon (lysine) resonance. In contrast, peaks c and d exhibited no change and peak b only a slight change in chemical shift over the same pH range. We conclude: the carboxyl groups of bound FA represented by peaks a, b, b', and d were involved in ion pair electrostatic interactions with positively charged amino acyl residues on BSA; the carboxyl groups of bound FA represented by peak c were not involved in electrostatic interactions with BSA; the similarity of the titration curves of peak c for BSA-bound octanoic acid and long-chain FA suggested that short-chain and long-chain FA represented by peak c were bound to the same binding site(s) on BSA; bound FA represented by peaks b' and a (but not d or b) were directly adjacent to BSA lysine residues. We present a model which correlates NMR peaks b, b', and d with the putative locations of three individual high-affinity binding sites in a three-dimensional model of BSA.