Solid-Phase Synthesis of Phosphorothioate/Phosphonothioate and Phosphoramidate/Phosphonamidate Oligonucleotides

We have developed a robust solid-phase protocol which allowed the synthesis of chimeric oligonucleotides modified with phosphodiester and O-methylphosphonate linkages as well as their P-S and P-N variants. The novel O-methylphosphonate-derived modifications were obtained by oxidation, sulfurization, and amidation of the O-methyl-(H)-phosphinate internucleotide linkage introduced into the oligonucleotide chain by H-phosphonate chemistry using nucleoside-O-methyl-(H)-phosphinates as monomers. The H-phosphonate coupling followed by oxidation after each cycle enabled us to successfully combine H-phosphonate and phosphoramidite chemistries to synthesize diversely modified oligonucleotide strands.


Oligonucleotide synthesizer
To combine phosphoramidite and phosphotriester and/or H-phosphonate chemistries we built, in collaboration with the Institute of mechanic and electronic workshops, two special synthesizers where we can change independently all parameters of the individual synthetic procedures and create any other procedures required. Usually two chemically different methods of oligonucleotide synthesis have been combined in dependence on monomers types.

Deblocking step
The detritylation step (deblocking) brought complication resulting in a low yield of the coupling step. The solution of 3% dichloroacetic acid 7 in DCM used for detritylation contained, depending on the content of water, variable amount of dichloroacetic anhydride 8. Its presence partially acylated hydroxyl groups of deprotected oligonucleotide 9 resulting in dichloroacetate-capped oligonucleotide 10 (Scheme 2) [3]. The anhydride 8 is in equilibrium with 7 in dry DCM solution, so that the use of deblocking solution with 150-200 ppm of water (Karl-Fisher) prevented acylation of hydroxyl groups. Scheme 2. Acylation during deblocking step.
Anhydride free deblocking mixture increased condensation yield of the next condensation step in case of phosphoramidite method, but showed another complication, based on the non-covalent binding of dichloroacetic acid 7 to the oligonucleotide backbone [4]. This complication concerns namely advanced phosphotriester and H-phosphonate chemistries, because the monomer is activated via mixed-anhydride with TIPSCl and DMOCP, respectively. These reagents may activate dichloroacetic acid, non-covalently bond to oligonucleotide chain so that partially dichloroacetate-capped oligonucleotide chain was formed. To overcome these problems, the increase of washing procedures was recommended after deblocking step [5] (e.g., the use of washing solvents with defined amount of water [6], or use other solvent mixture such as pyridine-methanol).
To monitor the coupling step during the synthesis we measured electric conductivity of the effluent from the reactor via flow cell during the deblocking of the dimethoxytrityl groups. The effluent contained dimethoxytritylium dichloroacetate ion pair which is conductive in DCM. The obtained peak was integrated, and the area compared to the previous step to check step-by-step yields.

Solid support
Before optimizing synthetic protocols for new monomers, we focused on comparison of two most commonly used solid supports, the LCAA-CPG (Long Chain Alkylamine Controlled Pore Glass) and TentaGel S NH2 (polystyrene-based resin grafted with polyethylene glycol with amino functional groups), where in both cases the nucleoside was attached to the free amino group through the hemisuccinate (Fig. 1). In contrast to CPG, TentaGel showed considerable particle size changes (swelling) in different solvents.  Therefore, we tested a series of activating agents [7] commonly used in Hphosphonate chemistry (Fig. 2), several solvents and their mixtures, bases, and concentration of reagents. In this case, we have tested monomers 3, and 6 as compared to standard H-phosphonates. All condensing agents were tested in different solvent systems and bases (Tab. 1.) at two concentrations of monomers  Another possible alternative for the oxidation of H-phosphonate/phosphinate linkages in oligonucleotides used was the Atherton-Todd reaction [11][12][13] We tried a water containing oxidation mixture, but also as in case of iodinemediated oxidation, the H-phosphonate linkage was oxidized almost quantitatively and H-phosphinate bond was hydrolyzed much faster than oxidized. For Atherton-Todd reaction it is known that an alcohol can also act as a nucleophile [14]. The

Amidation
Encouraged by success in previously described oxidative coupling and based on the fact that the primary or secondary amines are most commonly used as a nucleophile in Atherton-Todd reaction [12,13], we examined the preparation of electroneutral oligonucleotides with non-bridging P-N bond by oxidative amidation of P-H moiety in H-phosphinate linkage. Natural oligonucleotides with a phosphoroamidate arrangement were already prepared, and their hybridization properties studied [11,15,16]. We investigated differences between phosphoroamidate and phosphonoamidate linkages. Two amines, the primary N,N--dimethylethylenediamine which was already incorporated into phosphoroamidate oligonucleotides [15,16] and morpholine as a representative of secondary amines were selected. The morpholino six-membered ring could exhibit interesting steric properties if built into oligonucleotides. The mechanism of the amidation is essentially the same as for P-H to P=O oxidationfirst the P-Cl intermediate 15 is formed by reaction H-phosphinate dimer 11 with tetrachloromethane followed by the nucleophilic attack of an amine to produce the desired P-N dimer 17a,b (Scheme 6). The amidation mixture used for this reaction was more simple than the oxidation mixture for the preparation of P=O bond, because the amine which was acting in the reaction as a nucleophile is simultaneously hydrogen chloride scavenger.
We tested a various ratio of tetrachloromethane and primary or secondary amines  The reason for that could be seen in a low electron density on the nitrogen atom of the P-NH moiety which prevents the protonation of the amide nitrogen atom by dichloroacetic acid as the first step of the hydrolytic mechanism; the protonation is necessary for the subsequent nucleophilic attack with water molecule.
The results obtained on amidation of H-phosphonate and H-phosphinate linkages are summarized in Table 4.

The sulfurization (P-H → P-SH) of H-phosphonate and H-phosphinate linkages to
H-phosphonothioate and H-phosphinothioate bond, respectively, was carried out using two methods. The first, employing elemental sulfur in various solvents and in the presence of bases [17] Table 5, together with reaction time, yield, and additional treatment.

Summary of oxidation procedures
Study on coupling reactivity of monomers 3 and capability of the formed H- linkage to be oxidized, sulfurized, and/or amidated provided excellent results. Similar results we also obtained with monomers 6 whose 5-oxygen atom was replaced with sulfur. We did not observe a significantly different coupling efficiency, only the oxidation, sulfurization, and amidation of the linkage showed a slight increase of the unspecified by-products (Fig. 6).
In Table 6 all successfully used oxidation/sulfurization/amidation mixtures together with type of modified linkages are summarized. Table 6. Summary of oxidation/sulfurization/amidation mixtures and their abbreviations (R, R´ -depend on used amine).

Stability of modified internucleotide linkages
During the optimization of the condensation and oxidation steps, we observed a different behavior of both H-phosphonate and H-phosphinate types of internucleotide linkages. Therefore we decided to examine stability of linkages using a series of modified d(CT) dimers. All subsequent experiments were performed on HPLC/MS at regular time intervals, namely 3, 6, 9, and 12 hours. The regularity was ensured by an autosampler, which injected regularly the aliquot of the same samples.

Stability of non-bridging P-SH bond
We found that aqueous 0.05M sodium periodate desulfurated within 12 h almost   As another type of oxidation agent we tested 0.1M solution of iodine in wateracetonitrile-pyridine (1:1:1) mixture. Surprisingly, only desulfurization of both phosphorothioate and phosphonothioate linkages to phosphodiester and phosphonate bonds, respectively, proceeded in quantitative yields within 5 minutes without evidence of the cleavage of the internucleotide linkages ( Fig. 9). In contrast to sodium periodate, the use of iodine showed no difference between reactivity of phosphorothioate and phosphonothioate linkages. In view of these results, we exposed these oxidation conditions to the d(CT) dimer  Interestingly, in this case we observed no oxidation of sulfur in the 5'-position to sulfoxide moiety, even within 12 hours of treatment (Fig. 10). Our effort to find suitable conditions for sulfur oxidation at the 5'-position to obtain quantitatively sulfoxide-containing linkage failed.

Stability of non-bridging P-N bond
The pH stability of the prepared d(CT) dimers with non-bridging amide P-N bonds  Only non-bridging phosphoromorpholidate and phosphonomorpholidate bonds were cleaved from 38% and 98%, respectively (Fig. 11). The HPLC profile of crude d(CT) dimers with phosphoromorpholidate and phosphonomorpholidate linkages showed 2.5% and 21% of hydrolyzed products, respectively, resulting from cleavage of the P-N bond during detritylation step by action of DCAA and residual water content.
In contrast to a low acid-catalyzed hydrolytic stability of the P-N bond of 3-O- (Fig. 11), we found significantly increased stability (only 18% within 9 hours) of the P-N bond of linkage where the 5-oxygen is replaced with sulfur atom (Fig. 12). The higher hydrolytic stability could be seen in lower electron density on the morpholidate nitrogen atom which is less basic and thus its protonation as first step of the hydrolytic process is much more difficult. The HPLC profile of crude d(CT) linkage showed the presence of 16% of hydrolyzed product resulting from detritylation step, as mentioned above, and considerable amount of by-products that will increase after acid treatment (Fig.   13).

Stability of H-phosphonate/phosphinate ester linkage
We were also interested in the stability of non-oxidized  (Fig. 15). Both major products of the cleavage were recognized as   condensation and oxidation of the phosphoramidite unit has proved to be very beneficial and will be discussed later in the text. Table 7. Summary of internucleotide linkages stability.    The same procedure we used for preparation dT102 where the backbone was fully sulfurized to give rise to the phosphonothioate/phosphorothioate linkages by the sulfurization mixture (Ox-D) used at the end of the synthesis. The crude product was analyzed on IEC HPLC (Fig. 17) and the major peak was identified as the desired phosphoro/phosphonothioate linkages-containing dT102 (MALDI). Therefore we used oxidation mixture where methanol was replaced with 3hydroxypropionitrile. The formed 2-cyanoethyl ester has been shown to be stable within the subsequent synthetic cycles. Under these conditions and using capping steps we obtained a satisfactory yield of dT101 (Fig. 18). This approach provided even slightly worse IEC HPLC profile (Fig. 19) because of accumulation of failure sequences into one region (product shorter by one unit), but the yield of the desired dT102 was higher than at the use of fully H-phosphonate approach. The mentioned failure sequence was shortened by one phosphonate unit, indicating a partial cleavage of the phosphorothioate linkage, possibly during the capping step. However, this fact was not proved, since the model dimer did not show this phenomenon.  In the case of decamer dT103, three phosphoromorpholidate linkages originating from amidation of H-phosphonate bonds with Ox-F mixture, are present. IEC HPLC profile (Fig. 20) shows a large quantities of failure sequences.

Preparation of modified oligonucleotides
Even worse results were obtained at the preparation of dT104 (Fig. 21) containing three phosphonomorpholidate linkages originating from amidation of Hphosphinate bonds with Ox-F mixture. IEC HPLC profile (Fig. 21) shows again a large amount of failure sequences which came from acid-catalyzed hydrolysis of phosphonomorpholidate linkage which was repeatedly exposed to the deblocking step (deblocking solution contained 150-200 ppm of water). We also attempted to prepare two decamers using an amidation mixture where morpholine was substituted for N, N-dimethylethylenediamine, but in this case, we encountered analytical difficulties. None of the mobile phases used on IEC HPLC were able to separate these zwitter-ionic oligonucleotides. However, when N, N-

Synthetic Protocol C
Because of low yields at the preparation of phosphoroamidate/phosphonoamidate dT10 where synthesis was complicated by acidic hydrolysis of the amide linkages in the deblocking step, we changed the synthetic strategy to overcome these obstacles to obtain the desired oligonucleotides in a better yield. We found that the oxidation procedure of the phosphoramidite chemistry need not  Fig. 22; A, B) and dT104 ( Fig. 22; C, D) showed remarkable improvement of the synthesis. Table 10. Synthetic Protocol C H-phosphonate/H-phosphinate monomers condensation and oxidation/sulfurization/amidation at the end of syntetic cycle. The same washing procedures used as in Synthetic Protocol A (Table 8).  (Fig. 23). The comparison of HPLC profiles demonstrated that the application of Synthetic protocol B provided the highest yield of dT101. We also compared the efficiency of the sulfurization procedures. For IEC HPLC profiles of fully and partially sulfurized dT102 and dT105, respectively, see Fig. 24. Last but not least, we compared the effectiveness of the combined phosphoramidite method (CSO oxidation per cycle) with the H-phosphonate method (oxidation/ sulfurization/amidation after last cycle), where only H-phosphonate monomers were used (Fig. 25).  Table 11. The overview of oxidative methods used in Synthetic protocols for generation of modified internucleotide linkages.   Individual oligonucleotides were grouped (Tab. 12) by type of modification and the obtained curves were overlayed. First, we compared the melting points of modif DNA*DNA duplexes with those found for unmodified DNA*DNA duplex. Figure 26. Thermal characteristics (melting curves) of modif DNA*DNA duplexes compared with unmodified control DNA*DNA duplex: For structure of modif DNA ON1-ON15 see Table 12 (Group A-Group D).

PHYSICOCHEMICAL PROPERTIES OF ONS
The increasing number of odifications in the modified DNA strands ON1-ON3 decreased melting points of modif DNA*DNA duplexes (Tab. 12; Fig. 26 Group A).

CHARACTERIZATION OF OLIGONUCLEOTIDES
The 5´→3´ trityl-off synthesis of oligonucleotide was performed by phosphoramidite method (standard conditions recommended by the supplier) and