This paper contributes to development of a microscopic approach to predicting stable conformations of proteins in solvent. We report results of the first attempt to combine Monte Carlo simulated annealing, a powerful conformational sampling technique, and the reference interaction site model (RISM) theory, a statistical-mechanical treatment for molecular fluids. In solvent the key function is the total energy defined as the sum of the conformational energy and the solvation free energy, and the RISM theory is employed to calculate the latter. Starting from an initial conformation given, our computer program samples many conformations and then finds the conformation with the minimum total energy. Met-enkephalin in the two different solvents, a model water and a simple, repulsive-potential system, are considered. In water the solvation free energy varies greatly from conformation to conformation, while in the simple solvent it remains almost unchanged against conformational changes. In water most of the conformations with larger solvation free energies are strongly rejected, and the number of probable conformations is drastically reduced, which is suggestive that Met-enkephalin is forced to take conformations favored by water far more rapidly than in gas phase and in the simple solvent. The set of stable conformations obtained in water are quite different from those in gas phase and the simple solvent: they are characterized by almost fully extended backbone structure with large fluctuations in side-chain structure, which are in qualitatively good agreement with those determined by the recent nuclear magnetic resonance (NMR) experiments.