In vitro selection experiments have produced nucleic acid ligands (aptamers) that bind tightly and specifically to a great variety of target biomolecules. The utility of aptamers is often limited by their vulnerability to nucleases present in biological materials. One way to circumvent this problem is to select an aptamer that binds the enantiomer of the target, then synthesize the enantiomer of the aptamer as a nuclease-insensitive ligand of the normal target. We have so identified a mirror-image single-stranded DNA that binds the peptide hormone vasopressin and have demonstrated its stability to nucleases and its bioactivity as a vasopressin antagonist in cell culture.The enantiomer of an autogenously folding macromolecule will assume the mirror-image structure of the parent molecule, as has been documented for proteins and nucleic acids (1, 2). The corollary that the enantiomers of two complex-forming molecules will interact identically also has been demonstrated, for example by inverting all chiral centers in both an enzyme and its substrate (3). The principle has application to a problem in biotechnology, namely, that the peptide or nucleic acid ligands generated by phage display or in vitro selection are susceptible to enzymatic degradation; for example, a DNA oligonucleotide injected i.v. is degraded with a half-life of Ϸ5 min (4-7). The principle of chiral inversion can be coupled to the powerful reiterable selection methods to generate ligands that are insusceptible to degradative enzymes. First, a peptide or nucleic acid ligand is generated that binds the synthetic enantiomer of the target molecule. Then, the enantiomer of the selected ligand is synthesized from D amino acids or L nucleotides. The resulting reflected ligand will then bind and possibly inhibit the target molecule ( Fig. 1). This ''selectionreflection'' strategy has been used recently in the identification of a D peptide ligand of a Src homology 3 domain (8) and L-RNA ligands of arginine and adenosine (9, 10).In our study, the vertebrate hormone arginine vasopressin (VP), a 9-residue cyclic L peptide, was chosen as a target for its ease of bioassay, its ease of synthesis in all-D form, and its cyclic structure that restricts conformational degrees of freedom relative to a linear peptide and may thus allow selection of a tighter binding ligand. A stable ligand of VP could be useful as a diagnostic reagent or as a therapeutic antagonist in diseases associated with excessive levels of VP in blood or cerebrospinal fluid. DNA was chosen as ligand material because much higher pool diversity can be achieved than for peptide selection, and the chemical stability of DNA is superior to that of RNA.We have identified a mirror image DNA that binds VP and have shown that this ligand is bioactive, specifically antagonizing the VP response of cultured kidney cells. Selectionreflection thus joins the approach of using modified RNA polymerase substrates (11-14) as a viable means of coupling the power of in vitro selection to the production of sta...