“…In contrast to eRF1, Class 2 RFs (RF3 and eRF3) do not act as ribonucleoproteins+ Instead they are GTPbinding proteins (Grentzmann et al+, 1994;Mikuni et al+, 1994;Zhouravleva et al+, 1995) that hydrolyze the triphosphate when bound to the ribosome (Frolova et al+, 1996;Freistroffer et al+, 1997;Grentzmann et al+, 1998;Pel et al+, 1998)+ Interaction with eRF1 is essential for eRF3 GTPase, via contacts between their C-termini (Ebihara & Nakamura, 1999;Merkulova et al+, 1999)+ Such eRF1•eRF3 complexes are evident both in vivo and in vitro (Stansfield et al+, 1995;Zhouravleva et al+, 1995;Paushkin et al+, 1997)+ Thus, although eRF3 is probably not a natural RNA-binding protein, we still expect surfaces with potential RNA affinity on the protein+ RNAs with high affinities can be selected to virtually every protein, even toward peptide domains that lack natural sites for ribonucleotides or other polyanions (Gold et al+, 1995)+ Our selection yielded RNAs that definitely contact eRF3+ These are the Class II RNAs like RNA 27 (Figs+ 1 and 3) that have a unique multihelix-junction structure and complementary affinities for eRF1 and eRF3 alone (Fig+ 2)+ We suggest that these RNAs actually bridge the eRF1•eRF3 interface in the rightward (C domain; carboxyl-terminal) domain of the eRF1 structure (Song et al+, 2000)+ Such a distinct site would be consistent with the observed unique primary and secondary structures, which differ from Class I and eRF1 aptamers (Figs+ 1 and 3)+ However, Class II RNAs do not inhibit RF activity (Table 2)+ Because Class II RNAs were selected against the heterodimer, their binding may be consistent with a functional eRF1•eRF3 interface+ Thus, they may allow functional interactions between the RF proteins, as well as continued release function by the relatively distant NM domain+…”