Approximately 65% (11/17) of cancer patients participating in an ongoing Phase I clinical trial with recombinant interleukin-2 developed nonneutralizing serum IgG anti-interleukin-2 antibodies within 1 month of initiating therapy. These antibodies could be detected using any of several standard techniques including immunoblots and enzyme-linked immunosorbent assays. Western blot analysis and retention experiments with protein A-Sepharose indicate that the antibodies are specific for interleukin-2. The interleukin-2 mutein utilized in this clinical trial (des-ala-ser125 r-IL-2) differs from the major species of the human T cell-derived lymphokine in that it lacks the N-terminal alanine of the native molecule, is not glycosylated, and possesses a serine-cysteine substitution at position 125. Another recombinant interleukin-2, identical to the mutein except that it retains the cysteine at position 125 (des-ala-cys125 r-IL-2), strongly competes with the mutein in competitive enzyme-linked immunosorbent assays, suggesting that the amino acid substitution is not responsible for the recognition of the molecule by serum antibodies. Conversely, nonrecombinant T cell-derived interleukin-2 fails to compete in these assays and is not retained by protein A-Sepharose columns when mixed with high-titer antiserum. These results suggest that the anti-interleukin-2 serum antibodies generated in the course of treatment do not react with the nonrecombinant lymphokine but recognize epitopes peculiar to recombinant forms which are not dependent on the amino acid substitution at position 125. The failure of the antibodies to neutralize the biological activity of recombinant interleukin-2 (IL-2) in lymphocyte proliferation assays and to bind to the native lymphokine suggests that they may not affect IL-2-dependent cellular immune functions in vivo.