Ultra-pure dNTPs were purchased from Amersham Pharmacia Biotech. The exonuclease-deficient Klenow fragment of E. coli DNA polymerase I (PolA^∗) was purchased from New England Biolabs, Taq DNA polymerase (TaqP) was from Roche Boerhinger.

1,5-Anhydro-2,3-dideoxy-2-(thymin-1-yl)-D-arabino-hexitol (hT) [12] was 5′-phosphorylated according to Tener's method [13]. The monophosphate was then activated as the morpholidate derivative and converted to the corresponding triphosphate by condensation with pyrophosphate according to reported procedures [14]. Crude hTTP was purified by DEAE-cellulose column chromatography (HCO₃⁻ form, gradient from 0 to 0.8 M triethylammonium bicarbonate). For DNA polymerase reactions, hTTP was further purified by reverse phase HPLC (C18 column). Triethylammonium salt of hTTP was finally converted to the sodium salt by passing through a Dowex 50W-X8 (Na⁺ form) column. All intermediates were characterized by NMR spectroscopy. Purity of hTTP was assessed by ³¹P NMR spectroscopy and mass spectrometry. ³¹P NMR (D₂O) δ: −20.18 (t, Pβ); −9.60 (s, Pα); −4,58 (s, Pγ).

The phosphoramidite HNA building blocks were synthesized as described before [15]. The oligonucleotides containing several hexitol motifs were synthesized according to previously reported procedures [9] and the correct mass was obtained as determined by negative-ion electrospray mass spectrometry.

The primer (P7: 5′-CAGGAAACAGCTATGAC-3′) was labelled at its 5′-end with [γ³²P]-ATP (4500 Ci mmol⁻¹) and T4 polynucleotide kinase (10 U μl⁻¹). Primer-template duplexes (1:1.5 ratio) were annealed in the kinase buffer by heating at 75 °C for 15 min and cooling slowly to room temperature for 1 h. The elongation reaction mixture contained the corresponding 2× polymerase buffer (10 mM Tris-HCl, 5 mM MgCl₂, 7.5 mM DTT (pH 7.5) for PolA^∗; 10 mM Tris-HCl, 1.5 mM MgCl₂, 50 mM KCl (pH 9), 1% Triton for TaqP), the annealed primer-template duplex (200 nM) and the enzyme at the concentrations indicated in the legends of Fig. 3. The reactions were initiated with the addition of an equal volume of a 2× dNTP solution, incubated for 3 min at 37 °C with PolA^∗ or for 15 min at 48 °C with TaqP, and quenched by addition of a gel loading buffer (0.2% bromophenol blue, 0.2% xylene cyanol FF, 80% formamide). The samples were heated at 75 °C for 2 min and loaded onto a denaturing polyacrylamide gel (20%, 7M urea). After electrophoresis, the reaction products were visualized by autoradiography or using a PhosphorImager (Molecular Dynamics).

For the PCR amplification experiments, 25 pmol of forward primer, 25 pmol of oligonucleotide (P2, P3, P4, P5 or P6 listed in Table 1), 1 ng of pTZ18$\Omega$thyA DNA template (see below) and 200 μM of each dNTP were mixed in reaction buffer (Roche) and Taq DNA polymerase (1 U) was added in a 50 μl final volume. The cycling parameters were: 5 min, 95 °C; 25 × (30 s, 95 °C; 30 s, 35 °C < T<55 °C; 30 s, 72 °C); 10 min, 72 °C. The samples (5 μl) were analysed by electrophoresis on a 1% agarose gel.

The pTZ18$\Omega$thyA plasmid was obtained by site directed mutagenesis of pTS0 (pTZ 18R containing the E. coli thyA gene at the HindIII site [16] to introduce a NheI restriction site at positions 429 of the coding sequence (unpublished data). The pTZ18$\Omega$thyA plasmid was digested with the NheI and NsiI restriction enzymes, then ligated to a NheI-NsiI 1.1 kb DNA fragment (spc⁺) conferring resistance to spectinomycin and streptomycin from pHP45$\Omega$ [17] to produce the plasmid pAK1. The pAK1 plasmid was digested with NheI and NsiI, purified by electrophoresis on a 1% agarose gel and ligated to the oligonucleotide 5′-^pCTAGTGCA-3′ to produce pAK2. The pAK2 plasmid bears a deletion of 10 nucleotides and cannot code for an active thymidylate synthase.

Equimolar amount (50 ng) of pAK1, cleaved with NheI and NsiI and purified on a 0.7% agarose gel, and pAK2, linearized with EcoRI, were mixed in a 10 mM Tris-HCl (pH 7.5), 100 mM NaCl buffer, then denatured for 5 min at 85 °C, and slowly cooled to room temperature over 1 hour. The heteroduplexes formed were desalted through a Microcon column (Millipore).

The 18-mer oligonucleotides used were listed in Table 1. Each oligonucleotide (2 pmol) was hybridised with the heteroduplexes described above (0.02 pmol) in annealing buffer (50 mM Tris-HCl, 10 mM MgCl₂, 10 mM DTT, 25 μg ml⁻¹ BSA pH 8.0) by heat denaturation for 5 min at 85 °C, followed by slow cooling to room temperature over 1 h. Ligation was performed in the annealing buffer supplemented with 1 mM ATP and 5U of T4 DNA ligase by incubating the mixtures at 16 °C overnight. An aliquot of ligation mixtures was used to transform CaCl₂ competent cells of the strain β1308 (ΔthyA::erm) [18]. After 30 min of incubation at 4 °C, the cells were shocked for 5 min at 37 °C and 1 ml of Muller–Hinton (MH) medium, containing 0.3 mM thymidine was added. After a 1-h incubation at 37 °C with agitation, the cells were centrifuged and washed twice in the same medium without thymidine. Part (1/10) of the transformation mixture was plated on MH agar plates containing thymidine (0.3 mM) and carbenicillin (100 μg ml⁻¹) and the remaining 9/10 on MH agar plates.

A selection scheme was designed to assay transmission of genetic messages from HNA oligonucleotides to a DNA replicon by the replication enzymes in a living bacterium. It consisted of transforming E. coli cells with plasmids bearing a HNA:DNA heteroduplex encoding a functional and selectable gene, such that propagation of that gene would require utilization of the HNA stretch as a template (Fig. 2A). Synthetic 5′-phosphorylated 18-mer mosaic HNA/DNA oligonucleotides encoding the active site region surrounding the catalytic Cys146 of E. coli thymidylate synthase were ligated to a gapped vector carrying an inactive allele of the E. coli thyA gene that precisely lacked the codons for that catalytic region (Fig. 2B). Since active thymidylate synthase is absolutely required for growth of E. coli in nutrient media devoid of thymine or thymidine, this system was expected to provide a tight selection for HNA copying into DNA in vivo.

The length of contiguous hexitol nucleotides in the mosaic oligonucleotides was set as the experimental variable. Prototrophic transformants (i.e. capable of growth without added thymidine) of the E. coli strain β1308 deleted for the thyA gene could be reproducibly obtained by ligating the carrier vector to mosaic oligonucleotides with 1-, 3- and 6-unit long hexitol nucleotide stretches (Table 1). As expected, ligation of the 15-mer DNA oligonucleotide lacking the Cys146 codon (P1) did not yield any prototrophic transformants. Omitting oligonucleotides, carrier vector, or ligase in heteroduplex assembly also yielded no prototrophic transformants. Similar amounts of thymidine prototrophic transformants were obtained for a mosaic oligonucleotide bearing one hexitol (P3) and the control 18-mer DNA oligonucleotide (P2).

A 2.5-fold drop in the yield of prototrophic transformants resulted from extending the hexitol stretch from one to three nucleotides, and a close to 200-fold drop from extending the hexitol stretch from one to six nucleotides. Sequencing of the plasmids conferring thymidine prototrophy demonstrated that the message conveyed in hexitol stretches was correctly copied into DNA in all cases. In particular, the thyA sequences resulting from ligation of the mosaic oligonucleotides P5 and P6, both of which encoding the wild-type amino acids of the ThyA catalytic region but containing two silent mutations, were found to match exactly the 6-long hexitol stretches from which they respectively originated. This result indicated accurate copying of hexitol templates. Thus, the decrease in the yield of active thyA genes among clones transformed using hexitol templates was probably caused by inefficient rather than erroneous copying by E. coli DNA polymerases. The inefficiency of ligating mosaic HNA/DNA oligonucleotides to carrier vectors cannot be excluded, however. The identity of the DNA polymerases transcribing HNA messages into DNA in E. coli is currently being investigated.

DNA synthesis in response to hexitol-containing templates was examined in vitro by primer elongation reactions catalyzed by the exonuclease-deficient Klenow fragment of DNA polymerase I (PolA^∗). Two duplexes were designed in which the templates contain either one hA nucleotide followed by four dC nucleotides (T1) or a stretch of five hA nucleotides (T2). The integrity of the synthesized templates was verified by mass spectrometry. Elongation products obtained with T1 in the presence of each of the four canonical nucleoside triphosphates as well as in the presence of dTTP and dGTP are shown in Fig. 3A. Under experimental conditions which minimize the frequency of base misincorporations (see Section 2 Materials and methods for details), dTMP was best incorporated opposite the hA nucleotide leading to nearly complete elongation of the primer. There was only minor incorporation of dCMP and dAMP and no detectable incorporation of dGMP. With a mixture of dTTP and dGTP, all the primer was elongated and a single product was formed which corresponded to the primer elongated by four nucleotides. These results indicated that PolA^∗ is able to copy a base linked to the hexitol motif with a strong preference for the complementary base and that once the hA:dT nucleotide pair is formed, the enzyme can readily copy several dCMP residues of the template. Applying the same experimental conditions to the template T2, there was no detectable incorporation of dAMP, dCMP or dGMP whereas polymerization of dTMP was quite efficient with three residues incorporated opposite the hexitol motifs. The base specificity was thus again preserved (Fig. 3B). At higher enzyme/template-primer ratio (2-fold) or higher dTTP concentration (500 μM), all the primer was utilized and a product was formed by condensing four dTMP residues (Fig. 3C).

Our observation that PolA^∗ could copy hexitol containing templates into DNA led us to test whether the polymerase was also endowed with HNA-dependent HNA polymerase activity. The hexitol thymidine triphosphate (hTTP) was obtained following a synthetic route that excluded the concomitant formation of dTTP, as confirmed by spectroscopic characteristics (see Section 2 Materials and methods). We examined the condensation of hTTP with templates containing one or several adenine hexitol motifs. With the template containing one hA nucleotide (T1), the incorporation of hTMP was not as efficient as the incorporation of dTMP but much better than the incorporation of dCMP or dAMP, with about half of the primer elongated (Fig. 3A). In the presence of both hTTP and dGTP, again about half of the primer was elongated. The major product corresponded to a primer elongated by two nucleotides. There were two other products with one and four nucleotides incorporated (Fig. 3A). With the five-long hexitol stretch (T2), incorporation of hTMP was not detectable (Fig. 3B). However, when the enzyme/template-primer ratio was doubled or the concentration of hTTP raised to 500 μM, elongation products were obtained with one and two hTMP incorporated (Fig. 3C). Polymerization products obtained with T2 by the action of TaqP in place of PolA^∗ in the presence of each of the four canonical nucleotides and hTTP are shown in Fig. 3D. As with PolA^∗, hTMP was incorporated opposite the hA nucleotide, albeit less efficiently than dTMP. At higher hTTP concentration (500 μM), the formation of the product with three incorporated hTMPs was increased.

We also tested the introduction of hexitol motifs in polymerase chain reactions, elaborating on the fact that primer elongation with deoxynucleotides was catalyzed by TaqP in response to hexitol motifs located in a template (see Fig. 4). We thus attempted to measure the templating efficiency of hexitol-containing primers by the production of nucleic acid through PCR amplification. The oligonucleotides designed for in vivo transmission of HNA-carried messages in E. coli (Table 1) were used for this purpose, with the E. coli thyA gene as the DNA template (590 bp amplification product). Mosaic oligonucleotides (18-mers) differing by the length of the hexitol stretch were used as upstream primers, and a standard DNA oligonucleotide (M13 forward primer) as the downstream primer. The annealing temperature for amplifications was systematically tested for each couple of primers (see Section 2 Materials and methods). When the upstream primer contained one hexitol (P3), the amplification was similar to that obtained with the control DNA upstream primer (P2) (Fig. 4A and B). With a stretch of three hexitol motifs in the upstream primer (P4), the quantity of nucleic acid product was slightly reduced (Fig. 4C). No product could be detected with upstream primers containing six hexitol motifs (Fig. 4D). This could be explained either by the instability of the duplex formed with DNA or by the inability of the TaqP to copy long HNA stretches. Overall, these results were in agreement with those obtained both in vivo and in the primer extension experiments.