Ab initio synthesis by DNA polymerases


  • Nadezhda V. Zyrina,

    Corresponding author
    1. Institute of Theoretical and Experimental Biophysics, Russian Academy of Sciences, Pushchino, Moscow Region, Russia
    • Correspondence: Nadezhda V. Zyrina, Institute of Theoretical and Experimental Biophysics, RAS, 142290 Pushchino, Moscow Region, Russia. Tel.: +7 4967 739421;

      fax: +7 4967 330553;

      e-mail: zyrina.nv@gmail.com

    Search for more papers by this author
  • Valeriya N. Antipova,

    1. Institute of Theoretical and Experimental Biophysics, Russian Academy of Sciences, Pushchino, Moscow Region, Russia
    Search for more papers by this author
  • Lyudmila A. Zheleznaya

    1. Institute of Theoretical and Experimental Biophysics, Russian Academy of Sciences, Pushchino, Moscow Region, Russia
    Search for more papers by this author


The polymerization of free nucleotides into new genetic elements by DNA polymerases in the absence of DNA, called ab initio DNA synthesis, is a little known phenomenon. DNA polymerases from prokaryotes can effectively synthesize long stretches of linear double-stranded DNA in the complete absence of added primer and template DNAs. Ab initio DNA synthesis is extremely enhanced if a restriction endonuclease or nicking endonuclease is added to the reaction with DNA polymerase. The synthesized ab initio DNA have various tandem repeats. Sequences similar to those of ab initio DNA products are found in many natural genes. The significance of ab initio DNA synthesis is that genetic information can be created directly by protein. The ab initio DNA synthesis is considered a non-specific synthesis in various DNA amplification techniques. In this review, we present the main studies devoted to this phenomenon and introduce possible mechanisms of this synthesis from our current knowledge.


In general, template-independent generation of genetic information by DNA polymerases is a known process. A number of error-prone DNA polymerases efficiently incorporate nucleotides in DNA lesions where template information is absent (Goodman, 2002). Interestingly, some high-fidelity DNA polymerases are also able to bypass DNA lesions in vitro (Hsu et al., 2004).

Another instance of template-independent nucleotide polymerization is terminal deoxynucleotidyl transferase-like activity, achieved by adding dNTPs to the 3′-OH terminus of a blunt-ended duplex DNA substrate (Clark, 1988). Template-independent nucleotide polymerization occurs upon template switching when DNA polymerases synthesize a nascent strand of DNA from two discontinuous template strands (Clark, 1991; Garcia et al., 2004). In this process, polymerases use the overhangs to juxtapose two unlinked templates. The formation of these overhangs may be the result of template-independent nucleotide addition by DNA polymerases.

However, the highly efficient DNA synthesis discussed in this review differs remarkably from the above described examples. It takes place in the absence of any added DNA. Although this phenomenon, called ab initio DNA synthesis, has been known for 50 years, incontrovertible evidence was only obtained in the last decade.

History of ab initio DNA synthesis research

Initiation of DNA synthesis in a typical replication mode requires a template DNA strand and a primer, a short oligonucleotide complementary to the template DNA region with a free 3′-OH terminus. However, in the 1960 and 1970s it was shown that some prokaryotic DNA polymerases are capable of providing the de novo synthesis of poly(dA-dT) and poly(dG)poly(dC) without any added primer or template DNA (Schachman et al., 1960; Okazaki & Kornberg, 1964; Burd & Wells, 1970). These studies were conducted with partially purified preparations of enzymes; the scientific community assumed that this synthesis might be due to contamination by DNA or other enzymes (Nazarenko et al., 1979) and these data have therefore not been given due attention. Only 30 years later was it convincingly demonstrated that highly purified thermophilic DNA polymerases Tli and Tth were able to synthesize about 50 kb of DNA without any template and primer (Ogata & Miura, 1997, 1998a). This phenomenon was called ‘creative’, or ab initio DNA synthesis. The possibility of DNA contamination in the reaction mixture, which may serve as a primer and/or template, was vigorously excluded. The synthesized double-stranded DNAs had mainly short repetitive and palindromic sequences, and GC content was about 25%. The reaction conditions (temperature, ionic strength, and pH) were extremely important for this reaction (Ogata & Miura, 1998b). Based on these findings, Ogata & Miura suggested that genetic information might be created directly by protein.

At the same time, another group described the primer/template-independent polymerization of dATP and dTTP into poly(dA-T) by highly purified thermophilic DNA polymerases Taq and Tth (Hanaki et al., 1997, 1998). The 5′–3′ exonuclease activity and the terminal deoxynucleotidyl transferase (TdT)-like activity seemed to be essential for this synthesis. The primer/template-independent polymerization appeared to proceed via two reactions, the slow formation of 16–19-nt-long oligo(dA-T) without primer/template and the rapid elongation of the oligo(d A-T) by self-priming. When the substrates were depleted, DNA polymerases degraded the high-molecular-weight polymer from the oligomers by their own exonuclease activity. The authors proposed that the elongation and the degradation reactions proceed simultaneously. More detailed studies revealed that the majority of the ab initio synthesized DNAs represented both repeated sequences and short homologous blocks or randomly synthesized sequences (Cheng & Calderon-Urrea, 2011).

Ab initio synthesis in the presence of restriction endonucleases

The new type of ab initio DNA synthesis was found by the Frank–Kamenetskii group (Liang et al., 2004). They showed that ab initio DNA synthesis was extremely enhanced if a thermostable restriction endonuclease (Tsp509I, TspRI, etc.) was added to the reaction with thermophilic DNA polymerase (Vent, Bst and 9ºNm). The high efficiency of this synthesis resulted from the exponential amplification involving digestion/elongation cycles: a longer DNA with numerous recognition sites for the restriction endonuclease was digested to short fragments, and the short fragments were used as seeds for elongation to synthesize longer DNA. DNA was synthesized with a short lag period of 4 min and the synthesis was almost complete in 1 h. More than 10 μg of DNA was synthesized by 1 unit of DNA polymerase and the yield of synthesized DNA was more than 90%! This reaction was faster than the ab initio DNA synthesis reported by Ogata & Miura (with the yield of synthesized DNA of 2.2% after a lag period of 1 h) (Ogata & Miura, 1997, 1998b). A clear relationship between the length of the DNA molecules synthesized and the activity of endonuclease was observed (Liang et al., 2004). The synthesized double-stranded DNA had a highly repetitive palindromic sequence. Every repeating unit (motif) consisted of one or two recognition sites for the restriction enzyme, separated by an additional small random sequence.

Later it was found that the ab initio synthesis could be carried out at lower temperatures (4–37 °C) by combining non-thermophilic restriction endonuclease or non-specific endonuclease DNAse I with the thermophilic DNA polymerases (Liang et al., 2006). Moreover, the ab initio synthesis by thermophilic DNA polymerases alone [Vent, Vent (exo−) and Bst] was more efficient at lower temperatures than at the optimal high temperatures (Liang et al., 2007; N.V. Zyrina, unpublished data).

Ab initio synthesis stimulated by nicking endonucleases

At approximately the same time we found that very intensive ab initio synthesis takes place in the presence of nicking endonuclease Nt.BspD6I (Fig. 1) (Zyrina et al., 2007). Similar to restriction endonucleases, nicking endonucleases recognize a short specific sequence in double-stranded DNA and cleave DNA at a fixed position relative to the recognized sequence. However, unlike restriction endonucleases, nicking endonucleases make a nick in only one, predetermined DNA strand. When nicking endonuclease Nt.BspD6I was added to a reaction mixture with the large fragment of Bst DNA polymerase, DNA products of over 40 kb in an amount of about 10 μg were synthesized, with the synthesis reaching a maximum already after 1.5 h. Bst polymerase was used to carry out template-independent DNA synthesis without nickase. However, the amount of the product became clearly detectable only after a reaction time of about 18 h.

Figure 1.

Ab initio DNA synthesis stimulated by Nt.BspD6I, 1-h incubation (adapted from Zyrina et al., 2007). (a) Reaction products in 1% agarose gel. (b) The same products in 12% denaturing polyacrylamide gel. Top numbers – activity of Nt.BspD6I; N – heat-inactivated Nt.BspD6I (1 U); N1 – incubation without Bst polymerase in the presence of 1 U of Nt.BspD6I. The DNA size marker is shown in bp.

The macromolecular structure and the characteristics of the sequence DNAs synthesized in the presence of Nt.BspD6I differed from those synthesized by DNA polymerases alone or in the presence of restriction endonucleases. Some of DNA molecules had a branched structure. The sequences of DNA were represented mainly by non-palindromic, differently oriented tandem repeats containing Nt.BspD6I recognition site (GAGTC) with the only additional nucleotide being nucleotide A or T (Fig. 2).

Figure 2.

The sequences of DNA synthesized in the presence of Nt.BspD6I (adapted from Zyrina et al., 2007). Color arrow, motif orientation; circle, insertion; triangle, deletion.

Efficient synthesis was also observed in the presence of nicking endonucleases Nt.AlwI, Nb.BbvCI, and Nb.BsmI (V.N. Antipova & N.V. Zyrina, unpublished data).

A surprising consistency became obvious: both restriction and nicking endonucleases strongly stimulate ab initio DNA synthesis, not only giving seeds for elongation but also somehow determining the sequence of a synthesized product.

Ab initio synthesis stimulated by DnaB helicase

New DNA molecules over 100 kbp long can be synthesized without preexisting matrices when helicase DnaB is added to a reaction mixture with thermophilic (Bst, Tth, Pfu) or mesophilic polymerases (T7 and Escherichia coli polI Klenow fragment) (Kaboev & Luchkina, 2004). The synthesized double-stranded DNA had single-stranded stretches. The sequence was A–T-rich and highly repetitive.

Taken together, the data presented demonstrate convincingly that prokaryotic DNA polymerases are able to synthesize repetitive and palindromic (or quasipalindromic) DNA sequences without preexisting matrices.

A hypothetical model of ab initio DNA synthesis by DNA polymerases

The use of highly purified components by different groups supports the hypothesis that endogenous priming oligonucleotides are generated by DNA polymerase per se without added primer and template DNAs (Ogata & Miura, 1997; Liang et al., 2004; Cheng & Calderon-Urrea, 2011). However, the initiation of ab initio DNA synthesis remains a mystery. Ogata & Miura (1997) proposed a model in which amino-acid side chains of DNA polymerase, which normally interact with the single-stranded region of a template and a primer, bind and utilize dNTPs in a specific order to form discrete DNA sequences. Ramadan and coworkers showed that the condensation of dNTPs to short oligonucleotides by DNA polymerases does actually occur (Ramadan et al., 2004).

The hypothetical mechanism of ab initio DNA synthesis can be described with several stages. In the initial step, DNA polymerase generates a pool of oligonucleotides with random sequences. At the following stage, oligonucleotides with specific sequences, which can ‘facilitate’ their own replication, are amplified. Palindromic sequences are preferable because they can form reversible hairpin structures at their 3′-termini, thus priming the DNA elongation (Fig. 3a, 1–4) (Ogata & Miura, 2000; Ogata & Morino, 2000). The following stage is the DNA elongation. Subsequent rounds of hairpin formation, elongation and slippage lead to the propagation of the palindromic sequences in the form of extended tandem repeats (Fig. 3a, 3–3b). Very long DNA stretches may be synthesized through multiple strand displacement reactions on the 3′-termini of formed hairpins (Fig. 3a, 4–4a) (Liang et al., 2004). A restriction endonuclease may somehow help DNA polymerase to select the sequence to be synthesized. A partial digestion of the repeats, which include palindromic sites recognized by the endonuclease, leads to the formation of a pool of shorter molecules serving as primers for DNA synthesis (Fig. 3b).

Figure 3.

The hypothetical mechanisms of ab initio DNA synthesis. (a) Hairpin formation and elongation. (b) Digestion of one motif by a restriction endonuclease. (c) DNA elongation and branching during ab initio DNA synthesis stimulated by nicking endonuclease. Numerals – processes; arrow – motif orientation, color lines – complementary DNA strands; dashed line – elongating strand; black arrow –hydrolysis. Bold type indicates a recognition site for restriction or nicking endonuclease.

These models are based on hairpin formation within 3′-termini due to their palindromic nature. However, these models do not entirely explain how non-palindromic repeats propagate. The formation of hairpins in non-palindromic sequences is not possible either in the middle or at the ends of the molecule. DNA elongation in this case may result from the slippage synthesis in which a repeat containing a loop forms and moves through the whole DNA strand (Schlotterer & Tautz, 1992). However, molecules need to remain linear after the slippage synthesis. The evidence of branched DNA molecules and the presence of oppositely oriented motifs within a single DNA molecule suggest an additional mechanism of DNA elongation (Fig. 3c) (Zyrina et al., 2007). Oppositely oriented motifs may result from the joining of the complementary ends of the displaced parental and nascent strands (Fig. 3c, 1). Complementary ends are formed by the TdT-like activity of polymerase. Differently oriented motifs allow 3′-OH hairpin formation (Fig. 3c, 2). Branched molecules appear as a result of intermolecular hybridization when single-stranded molecules with oppositely oriented motifs accumulate (Fig. 3c, 3).

Of note, all these models consider ab initio DNA synthesis from the stage at which certain priming oligonucleotides for DNA synthesis are already present. They show effective amplification of these oligonucleotides and the formation of long DNA molecules. How priming oligonucleotides are formed remains unclear.

A possible functional role of ab initio DNA synthesis

Tandem sequences consisting of short repeats occur in all genomes. A comparison of ab initio DNA sequences with those of the known natural DNAs revealed that very similar tandem repeats are present in coding and non-coding regions (Ogata & Miura, 1997; Liang et al., 2004; Cheng & Calderon-Urrea, 2011). This fact suggests that repeating sequences were synthesized by DNA polymerase in a template-independent manner.

Ab initio DNA synthesis by different enzymes polymerizing nucleic acids suggests that similar synthesis by primordial enzymes probably occurred early on Earth. It was proposed that the pool of coding sequences that emerged in the prebiotic world represented repeats of nucleotide oligomers (Ohno, 1987). This hypothesis was further developed in work concerned with ab initio DNA synthesis. Probably, primitive polypeptides with polymerase-like activity synthesized DNAs consisting of simple repetitive sequences (Ogata & Miura, 1998a). These molecules gradually ‘evolved’ into degenerate sequences during error-prone replication by primordial enzymes. Presumably, the digestion of nucleic acids played an important role in the early evolution of genetic material (Liang et al., 2004). Apart from increasing the amplification efficiency, restriction enzymes could serve as a factor in selectivity and diversity of sequences. The polymerization and digestion reactions could be carried out by either a protein or another functional molecule, thereby being providing the foundations of life.


The works discussed in the review convincingly establish ab initio DNA synthesis by DNA polymerases as the existing phenomenon. But what techniques can we apply these data to?

The major problem of numerous nucleic acid amplification methods is the accumulation of non-specific products, which hamper identification of specific sequences. This process may be a result of ab initio DNA synthesis by thermophilic DNA polymerases. Ogata & Miura (1997) found some DNA-like material during a PCR experiment in a control reaction without added primer and template DNAs. The utility of other nucleic acid amplification techniques (strand displacement amplification, rolling circle amplification, exponential amplification reaction, etc.) is also hampered by non-specific synthesis (Chan et al., 2004; Ehses et al., 2005; Inoue et al., 2006; Zyrina et al., 2007; Tan et al., 2008). One strategy for suppressing or eliminating non-specific amplification is based on improving the reaction conditions. However, the problem of nonspecific amplification still remains unsolved. Because a variety of amplification methods have proven to be very useful for diagnostics, a new strategy for eliminating non-specific amplification has yet to be developed.

Such a new strategy was offered for isothermal DNA amplification in the presence of nicking enzymes (Zyrina et al., 2012). It was based on the use of SSB proteins as inhibitors of non-specific ab initio synthesis. One of the proteins, T4 gp32, almost completely inhibited ab initio DNA synthesis by Bst polymerase alone or Bst polymerase with nicking endonuclease and did not suppress the synthesis of a specific product.

The investigation of ab initio DNA synthesis raises quite a number of interesting questions. The knowledge gained will increase our understanding of how DNA polymerases function and will also suggest future research in molecular biology. The results may be very useful to develop techniques requiring fast and inexpensive preparation of large amounts of DNA.


The authors acknowledge Prof. Dr O.N. Ozoline for helpful discussions on this paper and Mrs S.V. Sidorova for technical assistance. The work was supported by the Russian Foundation for Basic Research, grant no. 12-04-01399_a.