The Sup35 prion protein of budding yeast has been reported to undergo phase separation to form liquid droplets both at low pH in vitro and when energy depletion decreases the intracellular pH in vivo. It also has been shown using purified proteins that this phase separation is driven by the prion domain of Sup35 and does not re-quire its C-terminal domain. In contrast, we now find that a Sup35 fragment consisting of only the N-terminal prion domain and the M-domain does not phase separate in vivo; this phase separation of Sup35 requires the C-terminal domain, which binds Sup45 to form the translation termination complex. The phase-separated Sup35 not only colocalizes with Sup45 but also with Pub1, a stress granule marker protein. In addition, like stress granules, phase separation of Sup35 appears to require mRNA since cycloheximide treatment, which inhibits mRNA release from ribosomes, prevents phase separation of Sup35. Finally, unlike Sup35 in vitro, Sup35 condensates do not disassemble in vivo when the intracellular pH is increased. These results suggest that, in energy-depleted cells, Sup35 forms supramolecular assemblies that differ from the Sup35 liquid droplets that form in vitro.