Carcinoembryonic antigen (CEA), a 180,000 dalton cell surface glycoprotein expressed on tumors of the colon, breast, ovary, and lung, has seven predicted immunoglobulin-like domains (N-A1-B1-A2-B2-A3-B3), most of which are recognized by distinct monoclonal antibodies. To study the individual domains, we have prepared several of the domains (N, A3, and A3-B3) by solid-phase peptide synthesis. The syntheses were performed by the Fmoc method using single couplings, elevated temperatures for both the coupling and deblocking reactions, and a flexible solvent system for the coupling reactions. The syntheses were accomplished on an in-house built synthesizer which allowed for temperature control and flexible solvent control during the course of the coupling reactions. Due to the large size of the peptides (84-184 residues), it was anticipated that the overall purity of the final product would not exceed 60% even for an average coupling yield of 99.5%. Therefore, several of the peptides were synthesized with a His6 "tail" at the amino terminus, allowing for purification on a Ni-NTA chelate column. For the most part, the purified peptides exhibited single sharp peaks by RP-HPLC, migrated at their expected molecular weights by gel permeation chromatography, gave correct masses by electrospray ionization or matrix-assisted laser desorption ionization time of flight mass spectrometry, gave the expected amino acid analyses, N-terminal sequences, and tryptic maps, and bound their appropriate monoclonal antibodies. The N-domain was extremely hydrophobic, requiring 6M guanidinium hydrochloride for solubilization, the A3 domain was soluble in dilute acid, and the B3 domain had an intermediate solubility. The affinity constants of the A3 domain and several mutants (also made by peptide synthesis) are reported, along with characterization of the 178 amino acid two-domain peptide, A3-B3. Although there is no evidence for proper folding of these domains by NMR, their ability to bind monoclonal antibodies with high affinity suggests that this is a plausible approach for producing individual domains of CEA.