of partners increases; the larger the ensemble, the deeper the valleys in the adaptive landscape. The most interactive structures and functions are thus the most likely to be preserved in their original forms—they are effectively frozen in time.

The Earliest Genomic Tags. We now consider replication in the RNA world (Gilbert, 1986), an era that predates the evolution of either DNA or templated protein synthesis. In these simpler times, enzymes made of RNA replicated genomes made of RNA. Nonetheless, early replicases and genomes had to confront many of the same problems faced by contemporary RNA genomes. Two of these problems are immediately evident:

• Specificity of replication. How did genomes that were destined for replication distinguish themselves from junk RNA or catalytic RNAs that should not be replicated?
• The "telomere problem" (Watson, 1972). How did ancient genomes and replicases prevent loss of 3' terminal sequences during successive rounds of replication?

The replicative strategy of the contemporary bacteriophage Qß suggests a single, powerful solution to both problems. Qß is a (+)-strand RNA phage. As shown in Figure 3, at the very 3' terminus of Qß is a tRNA-like structure, which ends in the sequence CCA. The 3'-terminal tRNA-like structure in the Qß (+)-strand genome serves as a recognition element for the replicase, which initiates synthesis of the (-)-strand at the penultimate C of the CCA terminus. The tRNA-like structure thus ensures specificity of replication. Furthermore, the 3'-terminal CCA of Qß can also function, at least in principle, like a modern telomere: loss of part or all of the CCA sequence could be restored by the CCA-adding enzyme, tRNA nucleotidyltransferase. Regeneration of a CCA terminus

Figure 3 Qß bacteriophage genome. The genome of Qß bacteriophage is a single-stranded RNA with a 3'-terminal tRNA-like genomic tag. The genome also serves as a messenger for four phage proteins, including the catalytic subunit II of Qß RNA replicase.