When Margarita Salas and Eladio Viñuela started their work on the Bacillus subtilis phage ϕ29, back in 1967, it could not be foreseen that this phage would code for such an exceptional enzyme, the ϕ29 DNA polymerase. Left to right: Luis Blanco, Miguel de Vega, Margarita Salas and José M. Lázaro. For an optimal performance, most replicative DNA polymerases (hereafter replicases) require the assistance of accessory proteins to confer the required processivity to the replication process and to unwind parental duplex DNA to allow replication fork progression (1). In addition, the inability of DNA polymerases to start de novo DNA synthesis imposes, in most organisms, the necessity for a short RNA molecule, made by a primase, to provide the 3'-OH group required to initiate DNA elongation. This requisite for a primer creates a dilemma to replicate the ends of linear genomes: once the last RNA primer for the lagging strand synthesis is removed, a portion of ssDNA at the end of the genome will remain uncopied. Bacteriophage ϕ29 has solved that issue by virtue of the unique catalytic features of an outstanding enzyme, the ϕ29 DNA polymerase. ϕ29 DNA polymerase (ϕ29 DNApol) is a monomeric enzyme of only 66 kDa, but fully responsible for viral DNA replication (2). Based on amino acid sequence similarities and its sensitivity to specific inhibitors, ϕ29 DNApol was included in the B family (eukaryotic-type) of DNA-dependent DNA polymerases (3). As any of them, it accomplishes sequential template-directed addition of dNMP units onto the 3'-OH group of a growing DNA chain, with insertion discrimination values ranging from 104 to 106 and with an efficiency of mismatch elongation 105–106-fold lower than that of a properly paired primer terminus (4). In addition, ϕ29 DNApol catalyses 3'–5' exonucleolysis, that is, the release of dNMP units from the 3' end of a DNA strand (5), as it occurs in most DNA replicases. Also fitting the norm, the 3'–5' exonuclease activity of ϕ29 DNApol (catalytic constant of 500 s−1) shows a preferential degradation of a mismatched primer terminus, in agreement with a role in proofreading of DNA insertion errors, thus enhancing the replication fidelity 102-fold (6, 7). However, ϕ29 DNApol is unique, as it is endowed with three distinctive properties that make it different from the rest of the replicases. One of the main peculiarities of ϕ29 DNApol resides in its ability to initiate DNA replication by using a terminal protein (TP) as primer (8), thus bypassing the need for a primase. It was amazing for us to find that the ϕ29 DNApol was able to use as primer the OH group of a specific amino acid in the TP, giving rise to a TP-dAMP product that most likely represented a template-directed initiation reaction, presumably dictated by the 3'-terminal T found at both ϕ29 DNA ends. It was important to find that ϕ29 DNApol could catalyze both the TP-primed initiation and full elongation of single-stranded oligonucleotides corresponding to the 3'-ends of ϕ29 DNA in spite of the strong 3'–5' exonuclease activity of ϕ29 DNApol, suggesting that the 3'-ends were being protected and selected as part of a templating strand. Unexpectedly, when Juan Méndez, a Ph.D. student, mutated the 3'-end terminal nucleotide (T) of the template, TP-dAMP formation was unaffected but elongation was fully inhibited. Strikingly, mutations at the second 3'-nucleotide of the template (also a T) did produce the complementary change in the TP-dNMP complex formed, but also inhibited elongation. These experiments were fully reproducible, demonstrating that the second 3'T was the director of the initiation reaction. Then, we wondered whether these results were an artefact of the in vitro system, or if there would be a mechanism allowing both elongation and the recovery of the information of the 3'-terminal nucleotide. Once demonstrated that a 3'-terminal reiteration of nucleotides was a strict requirement for elongation, we proposed the so-called sliding-back mechanism, extrapolatable to other phages, animal viruses such as adenovirus and hepadnaviruses, mitochondrial plasmids, and linear chromosomes and plasmids of Streptomyces, whose genomes contain a TP covalently linked to the 5' ends, together with various types of sequence iterations, constituting the replication origins. According to our model, ϕ29 DNApol and free TP form a 1:1 complex that interacts with the replication origins, located at both ends of its linear genome (TP-DNA). The DNA polymerase then catalyses the covalent linkage between the dAMP and the OH group of the TP Ser232 residue (TP-deoxynucleotidylation), giving rise to the initiation product TP-dAMP, this reaction being directed by the second nucleotide of the 3' reiteration (3'TTT…5'). This initiation complex slides-back one position, and the second nucleotide acts again as template to direct the incorporation of a second dAMP residue to form TP-AA, thus recovering the information of the 3'-terminal nucleotide and maintaining the integrity of the ϕ29 DNA ends (reviewed in 8, 9). Then, we found that the ϕ29 DNApol/TP heterodimer does not dissociate after initiation or sliding-back. Instead, there is a transition stage in which the DNA polymerase synthesizes a five-nucleotide-long DNA molecule while complexed with the primer TP, undergoes some conformational change during incorporation of nucleotides 6–9 (transition) and finally dissociates from the primer TP when nucleotide 10 is incorporated into the nascent DNA chain (elongation mode) (10). Then, the same DNA polymerase molecule resumes full-length TP-DNA synthesis, implying that the same active site is sequentially used both for TP-deoxynucleotidylation and DNA polymerization. Compared to the complexity of other in vitro replication systems, it was amazing to discover that efficient synthesis of full-length (19,285 bp) ϕ29 TP-DNA could be accomplished in vitro with only the presence of TP and ϕ29 DNApol (2). Soon after, we showed that the efficiency of this minimal replication system relied on two other unique catalytic features of ϕ29 DNApol: 1) an extremely high processivity (>70 kb, the highest described for a DNA polymerase), allowing replication of the entire genome from a single binding (and priming) event, without the assistance of processivity factors; 2) unlike most replicases, ϕ29 DNApol efficiently couples DNA polymerization to strand displacement, precluding the need for DNA unwinding proteins, as helicases. These three aforementioned exceptional properties: use of a TP as primer, intrinsic high processivity and intrinsic strand displacement ability, contribute to a successful enzyme design to support the symmetric DNA replication mode of bacteriophage ϕ29, by which the two DNA strands are synthesized continuously from a single priming event occurring at both termini of the linear molecule (11). By the end of 80s, the lack of a ϕ29 DNApol crystallographic structure that could provide functional insights prompted us to map the different enzymatic activities of this enzyme by site-directed mutagenesis of individual residues contained in regions of high amino acid similarity, as well as by the construction of deletion mutants (12). Thus, sequence alignments and site-directed mutagenesis served to identify the catalytic and ssDNA ligand residues responsible for the 3'–5' exonuclease activity, located at the N-terminal one-third of the enzyme (exonuclease domain). Based on these results, and in the conservation of the catalytic residues, we firstly proposed the hypothesis of an evolutionary conserved 3'–5' exonuclease active site among distantly related DNA-dependent DNA polymerases. Such an active site would be formed by three N-terminal amino acid motifs, which we named ExoI, ExoII and ExoIII, invariantly containing four carboxylate groups that bind two metal ions and one tyrosine residue involved in orienting the attacking water molecule (13), as it had been shown to occur in E. coli Pol I (14). Additionally, site-directed mutagenesis in ϕ29 DNApol pioneered the functional analyses of specific amino acids at motifs YxGG, Dx2SLYP, Kx3NSxYG, Tx2GR, YxDTDS and KxY, highly conserved at the C-terminal two-thirds of eukaryotic DNA polymerases from family B. These studies demonstrated the overlapping between polymerization and protein-primed initiation domains, and served to identify the amino acids involved in metal binding and catalysis, as well as DNA, TP and dNTP ligands (reviewed in 12). However, the exhaustive mutational scrutiny carried out throughout ϕ29 DNApol did not provide a structural rationale for both the intrinsic processivity and strand displacement capacity of the enzyme. The recent crystallographic resolution of the ϕ29 DNApol structure, carried out in collaboration with Tom Steitz's lab (Yale University), confirmed our previous mutational results. Thus, ϕ29 DNApol consists of an N-terminal exonuclease domain, containing the 3'–5' exonuclease active site, and a C-terminal polymerization domain that, like other DNA polymerases, is subdivided into the universally conserved palm (containing the catalytic and DNA ligand residues), fingers (containing the dNTP ligands) and thumb (which confers stability to the primer) subdomains (15). In addition, the structure of ϕ29 DNApol provided for the first time a topological basis for its intrinsic strand displacement capacity and processivity. 3D-structural comparisons indicated that the main difference between ϕ29 DNApol and other family B DNA polymerases was the presence of two additional subdomains, both corresponding to sequence insertions that we had previously identified as specifically conserved in the protein-primed subgroup of DNA polymerases and called as TP regions, TPR1 and TPR2 (16, 17). TPR1 lies at the edge of the palm and contacts TP during initiation and the upstream dsDNA during elongation, whereas TPR2 contains a β-hairpin structure located just facing the apex of the thumb subdomain (15). Palm, thumb, TPR1 and TPR2 subdomains form a doughnut-shaped structure that encircles the upstream duplex DNA at the polymerization active site, constituting an internal clamp that provides the enzyme with the maximal DNA-binding stability required for its intrinsic processivity (Fig. 1), mimicking and making unnecessary the sliding clamp used in other replisomes (15). Crystallographic resolution of the ϕ29 DNApol/TP heterodimer showed the upstream duplex DNA “tunnel” now occupied by the TP priming domain, which contains the priming Ser232, and whose overall dimensions and its negative charge mimic DNA in its interactions with polymerase during initiation. This explains why DNA synthesis by the heterodimer cannot begin at internal sites of the bacteriophage genome, as the upstream 3' template would sterically clash with TP (18). Despite the fact that the upstream DNA “tunnel” forms a closed ring, it must be opened as ϕ29 DNApol can replicate circular ssDNA templates. Strand-displacement amplification by ϕ29 DNApol, a structural basis. Left: Schematic representation of the continuous, strand displacement-mediated DNA amplification process performed by ϕ29 DNApol. By adding random primers, ϕ29 DNApol (orange spheres) can start DNA replication at multiple sites, simultaneously. Once reached a downstream primer, replication proceeds by coupling polymerization to strand displacement, rendering single-stranded branches available to bind more primers. This continuous DNA synthesis process will result in exponential and isothermal DNA amplification. Right: Detailed model of ϕ29 DNApol working at a replication fork (squared region at the left). Based on 3D-structural information, the TPR2 insertion of ϕ29 DNApol would contribute to a full encirclement of the DNA substrate, conferring an outstanding processivity, promoting the opening of the parental DNA strands, and the subsequent threading of the template into the active site tunnel. Thus, the TPR2 insertion acts as a wedge to couple polymerization to strand displacement. Reproduced with permission from Rodríguez et al. (2005) Proc. Natl. Acad. Sci. USA 102, 6407–6412, ©National Academy of Sciences, USA. On the other hand, TPR2, palm and fingers subdomains, together with the exonuclease domain, surround the downstream template strand, forming another tunnel whose narrow dimensions are incompatible with dsDNA binding (Fig. 1). Thus, unwinding is required to allow threading of the template through this tunnel to reach the polymerization site, using the same topological mechanism as hexameric helicases, and providing a structural basis for the strand displacement activity of ϕ29 DNApol (15, 19). Two distinctive properties displayed by ϕ29 DNApol, strand displacement capacity and processivity, together with its high polymerization fidelity due to the combination of a high dNMP insertion discrimination and a strong proofreading activity, led us to envisage ϕ29 DNApol as an ideal tool to achieve strand displacement amplification. Thus, the two aspects that were frequent limitations in the existing amplification procedures, fidelity of synthesis and length of the amplified products, would no longer be an issue. Firstly, we set up one of the most efficient in vitro DNA replication systems that, in addition to TP (acting as the sole primer) and DNA polymerase, also required ϕ29 SSB (to preclude the appearance of short palindromic DNAs) and ϕ29 DBP (for optimal activation of the origins). Using this “viral system,” ϕ29 DNApol could amplify limited starting amounts of ϕ29 TP-DNA molecules by three orders of magnitude after 1 hour of incubation at 30 °C (20). The quality of the amplified material was demonstrated by its ability to produce phage particles after transfecting B. subtilis, identical to that of the TP-DNA obtained from virions, establishing the basis for the development of heterologous DNA amplification procedures dependent on the phage ϕ29 replication machinery (20). In a more simple setup, the ability of ϕ29 DNApol to use circular ssDNA as template allows asymmetric rolling-circle DNA amplification, producing very long single-stranded concatemeric DNA molecules containing more than 10 copies of the original circular template (11, 21). In perhaps one of the most efficient procedures for isothermal dsDNA amplification (22, 23), developed by Amersham Biosciences/Molecular Staging, ϕ29 DNApol is combined with random hexamer primers to achieve isothermal and faithful 104–106-fold amplification via strand displacement of either circular [Templiphi™ (www.thempliphi.com)] or linear [Genomiphi™ (www.genomiphi.com)] genomes, yielding high-quality amplification products (Fig. 1) that can be either digested or sequenced directly without any further purification steps. Our research on ϕ29 DNApol, conducted over 25 years, has allowed us to exploit the enzymatic potential of this small viral enzyme, a good example of basic research applied to biotechnology. Because of that, ϕ29 DNApol is not only our favorite enzyme, but also the enzyme of choice of many researchers to amplify DNA. This article has been funded by the Spanish Ministry of Education and Science and by an institutional grant from Fundación Ramón Areces to the Centro de Biología Molecular “Severo Ochoa.”