A New 1,5-Disubstituted Triazole DNA Backbone Mimic with Enhanced Polymerase Compatibility

Triazole linkages (TLs) are mimics of the phosphodiester bond in oligonucleotides with applications in synthetic biology and biotechnology. Here we report the RuAAC-catalyzed synthesis of a novel 1,5-disubstituted triazole (TL2) dinucleoside phosphoramidite as well as its incorporation into oligonucleotides and compare its DNA polymerase replication competency with other TL analogues. We demonstrate that TL2 has superior replication kinetics to these analogues and is accurately replicated by polymerases. Derived structure–biocompatibility relationships show that linker length and the orientation of a hydrogen bond acceptor are critical and provide further guidance for the rational design of artificial biocompatible nucleic acid backbones.


■ INTRODUCTION
The replacement of natural phosphodiester (PO) bonds in DNA or RNA by artificial internucleoside linkages can generate remarkable biomimetic oligonucleotides (ONs) with applications as therapeutics, 1 xenobiotic genetic polymers, 2−4 aptamers, 3,5−7 and synthetic genes. 8−10 Biological integrity and favorable biophysical properties are critical, and good hybridization properties, mismatch discrimination, and compatibility with certain enzymes (e.g., RNase H) play crucial roles for antisense oligonucleotides. 11−15 Other applications of backbone-modified, bioactive oligonucleotides include modified CRISPR-Cas9 systems, 16 and compatibility with the cellular gene replication and expression machinery.
In general, replication-competent artificial DNA backbones can be used for gene synthesis, 8,9 sequencing, 17 or nucleic acid detection. 18 A detailed study revealed several molecular characteristics of artificial backbones that are required for compatibility with DNA polymerases during replication. 19 Such polymerase compatible artificial backbones comprise 5′-S-phosphorothioesters, 20 phosphorothioates, 21 disulfides, 22 boranophosphates, 23 phosphoramidates, 10,24,25 amides, 19,26 ureas, 18 squaramides, 18 and triazoles. 8,9,19,27−29 Among these, the triazole linkage (TL) represents a powerful and versatile chemical moiety that can be readily formed by the Cu Icatalyzed azide−alkyne cycloaddition (CuAAC) reaction, resulting in 1,4-disubstituted 1,2,3-triazoles. 30 Examples include TL1, 31 TL4, 27 TL5, 32 and TL6 29 (Figure 1). 33,34 Further modification can be achieved by alkylation of the triazole as exemplified by cationic TL7, 19 which is made by methylation of TL6. Many applications of such triazole backbones exist. For example, the CuAAC click ligation of 3′-azido ONs with 5′-alkyne adaptor ONs to form TL1 was recently described for the preparation of next-generation sequencing libraries. 17 However, the authors reported low replication efficiencies by several polymerases through TL1. Furthermore, click ligation has been used to assemble long DNA templates with isolated TL6 modifications which can be replicated 9,29 or transcribed 8,9 by polymerases in bacterial 9 or mammalian 8 cells while retaining high fidelity readthrough. 8,9,19,29 However, TL6 suffers from reduced binding affinity to a complementary DNA or RNA target 35−38 and induces a TL6-dependent slowdown in PCR replications. 9 In search of the ideal nucleic acid triazole linkage, we recently developed a 1,5-disubstituted 1,2,3-triazole internucleoside linkage which was prepared by Ru II -catalyzed azide−alkyne cycloaddition (RuAAC) (TL3; Figure 1). 38 Altogether, a diverse toolbox of TLs has been reported over the past two decades, all having distinct triazole orientations and linker lengths as indicated in Figure 1. However, whether the more recent triazole isoforms, such as the 1,5-disubstituted TL3, retain compatibility for replication by DNA polymerases had not yet been determined.
Here we report the synthesis of a novel, 1,5-disubstituted TL connecting two nucleosides through 6 bonds (TL2; Figure 1). We directly compare the effect of RuAAC vs CuAAC to form TLs from the same precursors (TL2 vs TL5), and we study the ability of oligonucleotides containing these TLs to form duplexes with a DNA or RNA target in comparison to reported triazoles TL1, TL3, TL4, and TL6. 38 Moreover, we test how efficiently polymerases are able to read through TLs 1−7 in replication templates. For the first time, we assess replication through a 1,5-disubstituted 1,2,3-triazole (TL2 and TL3). Evaluation of the molecular characteristics of the TLs with respect to conferred biocompatibility provides guidance for further biomimetic nucleic acid linkage designs.

■ RESULTS AND DISCUSSION
Synthesis of TL-Linked Dinucleosides. While the 1,5substitution of a TL (TL3) dinucleotide has been recently reported, 38 the corresponding RuAAC ligation of azide-and alkyne-modified ONs to form such triazole isoforms has not yet been developed. Thus, we opted for a dinucleoside strategy to incorporate TL2 into ONs. A suitable dinucleoside phosphoramidite building block containing the 1,5-disubstituted triazole results from RuAAC between alkyne 2 and azide 3 39 (Scheme 1A). A previously described method to selectively alkylate the 5′-OH of thymidine in a single step 40 was unsuccessful and resulted in alkylation of the thymine nucleobase instead of the desired 5′-OH ( Figure S1). Thus, silyl-protected thymidine 1 41 served as a substrate for propargylation of the 5′-OH to give alkyne 2. RuAAC with azide 3 39 gave dimer 4 with the correct 1,5-substitution ( Figure S2). TBAF-mediated deprotection of 4 gave alcohol 5, which was phosphitylated to dimer phosphoramidite 6 for ON synthesis.
Duplex Structure. Despite some destabilizing effects, TL1, TL3, TL4, and TL6 are known to retain global B-or A-form duplex structures with the DNA or RNA target, respectively. 38 In the current study, duplexes formed by TL2-and TL5modified ON2 and ON5 with complementary DNA or RNA were analyzed by circular dichroism (CD) spectroscopy ( Figure 2). The duplexes of ON2 with DNA and RNA show almost identical spectra to the equivalent duplexes formed by unmodified ON7. Duplexes of ON5 with DNA or RNA also retain global B-or A-form helical structures, respectively.
Enzymatic Read-through Studies. The ability of TLmodified ONs (TLs 1−6) to form duplexes with DNA and RNA targets while maintaining global structural integrity suggests that these TLs could be promising biocompatible DNA backbone mimics. Indeed, TL1, TL6, and methylated TL7 have been extensively studied as modifications in templates for replication by DNA and RNA polymerases. 8,9,16,19,29,42 Moreover, a previous study reported encouraging results for TL4 within modified primers in qPCR. 27 However, no read-through compatibilities of 1,4disubstituted TL5 or 1,5-disubstituted TLs (TL2 and TL3) have been reported. Thus, we compared TLs 1−7 in primer extension studies ( Figure 3 and Figure 4). We synthesized unmodified templates and TL-modified templates, each containing a single TL (TLs 1−7), using phosphoramidite 6 (TL2), 9 (TL5), or previously reported phosphoramidites (TL1, 19 TL3, 38 TL4, 27 TL6, 43 and TL7 19 ). We then assessed the polymerase compatibilities of TLs 1−7 by enzymatic extension of fluorescein (FL)-labeled primers followed by analysis of the FL-labeled products after gel electrophoresis. The sequences of the templates and primers can be found in Table S1.
Klenow Fragment. Extension of an FL-labeled primer by the large fragment of DNA polymerase I (Klenow fragment) using unmodified Template 0 (PO) resulted in formation of the expected full-length product, which was confirmed by mass spectrometry (MS) analysis (FL-replicon 0**; Figure 3A,B). Primer extension through the TL1-modified Template 1 gave a band that migrated slightly faster than FL-replicon 0 and a mass corresponding to a replicon with an A-deletion (FLreplicon 1**; Figure 3A,B). The fact that read-through of TL1 Values were obtained from the maxima dA 260 /dT vs T for 3 μM of each ON in 10 mM phosphate buffer, 200 mM NaCl, pH 7.0. The graphs of dA 260 /dT vs T are shown in Figure S4. ΔT m are relative to the unmodified control ON7. b Sequence of DNA target = 5′-GCTGCAAACGTCG-3′. c Sequence of RNA target = 5′-GCUG-CAAACGUCG-3′. dTL indicates the site where a PO is replaced by TLs 1−6. Journal of the American Chemical Society pubs.acs.org/JACS Article by Klenow fragment results in an A-deletion mutation which is situated next to the TL1 was previously shown by sequencing of the replicons. 19 Templates bearing 6-bond TLs 2−4 resulted in the formation of the expected full-length products as confirmed by MS (FL-replicons 2**, 3**, and 4**; Figure  3A,B). However, replication of the TL3 template also produced a truncated product, shown by MS analysis to be a consequence of premature extension termination directly before the triazole (FL-replicon 3*; Figure 3A,B). The longer 7-bond TL5 in Template 5 was replicated but resulted in a product with an A-deletion (FL-replicon 5**; Figure 3A,B), and additional faint truncation bands were visible on the gel. Extension through TL6 produced the expected full-length product (FL-replicon 6**; Figure 3A,B) and truncated products, one of which results from termination directly before TL6 (FL-replicon 6*; Figure 3A,B). The presence of  Table S1. For full mass spectra and peak assignments see Figures S5−S7 and Table S2.  Table S1. For full mass spectra and peak assignments see Figures S8−S10 and Table S3. cationic methylated TL7 in Template 7 is catastrophic to the primer extension and exclusively results in a truncated product from extension termination directly before the methylated triazole (FL-replicon 7*; Figure 3A,B). This confirms previous reports for this TL. 19 Hot-Start Taq Polymerase. It is known that replication by Taq polymerase is slowed down at high DNA concentrations such as those used in the primer extension assays. 44 Moreover, a previous study showed that the replication of an unmodified template can be significantly slowed down under these conditions (up to 120 min). 19 Thus, primer extensions along the TL-modified templates were performed with hot-start Taq polymerase at 60°C for an extended time of 160 min ( Figure  4). Replication of unmodified Template 0 (PO) resulted in the exclusive formation of the expected full-length product containing a Taq-characteristic 3′-A overhang as determined by MS (FL-replicon 0**; Figure 4A,B). Replication of the TL1-modified template resulted in minor amounts of a product for which MS suggested deletion of an A-nucleotide (FL-replicon 1**; Figure 4A,B). The major products from TL1 were truncated ONs where replication stopped directly before or after the triazole (FL-replicons 1*; Figure 4A,B). This supports the mutagenic effect of TL1 in causing deletions around the artificial linkage with a Taq polymerase. 19 Remarkably, replication through TL2 yielded a clean fulllength product including the expected Taq-characteristic 3′-A overhang (FL-replicon 2**; Figure 4A,B). No truncated products were detected on the gel or by MS. In contrast, a significant amount of truncated product was observed for the TL3 template where extension terminated directly before the triazole (FL-replicon 3*; Figure 4A,B). Extension through TL4 resulted in the expected full-length product without truncation (FL-replicon 4**; Figure 4A,B). In contrast to the previous primer extension experiment with Klenow fragment, replication through TL5 using hot-start Taq polymerase mainly resulted in truncated products from termination directly before or after the triazole (FL-replicons 5*; Figure 4A,B). Replication of the TL6 template produced the correct fulllength construct (FL-replicon 6**; Figure 4A,B) and a truncated product (FL-replicon 6*; Figure 4A,B). As seen for Klenow fragment, methylated TL7 completely blocks extension (FL-replicon 7*; Figure 4A,B). Across the Klenow and hot-start Taq primer extension assays, TL2 and TL4 stand out as being the only TLs resulting in the clean formation of the expected full-length product without any detectable mutations or truncated fragments observed on gel or by MS. The excellent read-through compatibility of TL2 by Klenow fragment and hot-start Taq polymerase was confirmed with an alternative sequence (Template 9; Figures S11, S12 and Table  S4).
qPCR Kinetics. To assess the time dependencies of TLs 1− 7, the modified templates were used in qPCR experiments with various extension times. Impaired read-through of a given template results in an increase of the threshold cycle (C t )the number of qPCR cycles after which a signal overcomes a set threshold. If impairment can be overcome with time, the resulting C t value will decrease with longer extension times if within the range of the time dependency. Amplification of unmodified templates can be considered as being time independent under the conditions of the study, as all internucleoside linkages are natural POs which are formed rapidly. C t values of qPCR reactions with extension times ranging from 15−240 s were determined for templates with TLs 1−7 using hot-start Taq polymerase (Figures S13−S18). The differences between the determined C t values of the modified templates to the C t values of the respective isosequential unmodified PO templates (ΔC t ) are plotted for each extension time in Figure 5A. Except for TL2 and TL7, all TLs show time dependency for the read-through of the artificial linkage. The apparent time independency of TL7 has been previously explained by a thermodynamic barrier created by TL7 that can only be overcome when transitioning from extension (60°C) to denaturation (95°C) in each cycle. 19 Most notably, the shorter 5-and 6-bond TLs 1−4 are less time dependent than the longer 7-bond TL5 and TL6. Indeed, ΔC t 's of TL5 and TL6 continued to decrease up to 240 s, the highest extension time tested in this study. The ΔC t 's of TL1, TL3, and TL4 decrease with longer extension times, but the  Table S1, Figure S23, and Figure S24.
fitted curves flatten toward an extension time of 240 s. In contrast, the ΔC t of TL2 showed no time dependency within the tested conditions. Together with the superior results from the primer extension experiments, this suggests that the observed time independency of TL2 does not originate from transitions between the cycles as seen for TL7. Remarkably, there is no significant kinetic barrier imposed by TL2 for readthrough of hot-start Taq polymerase under the tested conditions. DNA Sequencing. To verify that read-through of the TLmodified templates results in correct replication, PCR products from hot-start Taq polymerase-mediated amplification were sequenced. While TL1, TL6, and TL7 have been previously studied for replication fidelity, 19 the question whether TLs 2− 5 can be replicated correctly in PCR remains unanswered. Thus, PCR amplicons generated using hot-start Taq polymerase and templates containing TL2, TL3, TL4, or TL5 were cloned into a vector and transformed into Escherichia coli. Several colonies were randomly picked, and the recovered vectors were analyzed by Sanger sequencing ( Figure 5B). The 6-bond triazole backbones (TLs 2−4) resulted in the correct read-through producing the expected sequence for all colonies picked (n = 4). In contrast, the longer 7-bond TL5 resulted in a deletion mutation around the artificial linkage for which the T TL5 T motif has been read as a single T in four out of five colonies picked.
Structure−Biocompatibility Relationship. Analysis of the molecular properties of TLs 1−7 provides insight into the structure−biocompatibility relationship among the TLs for replication ( Table 2). Most noticeable are the differences in TL backbone lengths. For 5-bond TL1, whose number of bonds is closest to the natural PO, inefficient read-through by hot-start Taq was observed confirming previous reports. 17,19 Moreover, TL1 is mutagenic and results in single point deletions around the triazole when read by Klenow fragment or Taq polymerases. 19 The read-through efficiencies of 6-bond TLs 2−4 gave different results in the primer extension assay. However, all 6-bond TLs imposed the lowest kinetic barriers on polymerase read-through and all amplicons from PCRs with hot-start Taq polymerase had the correct sequences. The longer 7-bond TL5 expressed poor read-through efficiency in the primer extension assay with hot-start Taq polymerase and resulted in deletion mutations in PCR amplicons. In contrast, 7-bond TL6 is known to be read-through correctly resulting in the expected amplicon sequences. 19 Previous studies suggested that flexibility of the artificial backbone is required for readthrough. 19 There is an overall increase of flexibility along the TL backbone of 1,5-disubstituted TLs (TL2 and TL3: 2 × sp 2 ; Table 2 and Figure 1) compared to 1,4-disubstituted TLs (TL1 and TLs 4−6: 3 × sp 2 ; Table 2 and Figure 1). Moreover, the poor biocompatibility of 5-bond TL1 might be attributed to the combination of increased rigidity and highest duplex destabilization, defeating the positive contributions from being closest to the natural PO linker length. However, the flexibility model does not explain the lower replication efficiency of 6bond TL3 (2 × sp 2 ; Table 2 and Figure 1) compared to 6bond TL4 (3 × sp 2 ; Table 2 and Figure 1) and 6-bond TL2 (2 × sp 2 ; Table 2 and Figure 1) in the primer extension studies (TL2, TL3 and TL4; Figure 3A and Figure 4A). Taq polymerase mainly interacts with the DNA through hydrogen bonding with the phosphates, 45 and a crucial factor in readthrough of artificial backbones is the ability to donate an electron pair as a hydrogen bond acceptor. 19 As such, undesirable orientation of the hydrogen bond acceptor in TL3 is a plausible reason for its hampered read-through compatibility. Moreover, N3-methylation in TL7 is catastrophic, further suggesting that a hydrogen acceptor is important for replication. However, it is likely that the additional charge inversion and steric hindrance contribute to the replication incompatibility of TL7, too.
Taking these observations together, we hypothesize that the ideal TL results from a combination of linker length, availability of a hydrogen bond acceptor, and overall flexibility. Extending this to any artificial DNA backbone, we propose that the artificial backbone should be 5−6 bonds in length and provide an accessible hydrogen bond acceptor, while maintaining an overall high degree of flexibility. The new 1,5-disubstituted TL2 combines these molecular requirements and possesses outstanding biocompatibility. The proposed molecular requirements also agree with our previously described model, which was derived from read-through compatibility tests of a range of backbone linkages including amides, phosphoro(di)thioates, phosphoramidates, and squaramides. 18,19,24,25 It is important to emphasize that the observed read-through compatibilities are dependent on the polymerase and similar polymerase dependencies were reported for other PO mimics. 18,19 ■ CONCLUSIONS In conclusion, triazole linkages are interesting phosphodiester mimics for applications in synthetic biology, biotechnology, and other areas. A profound understanding of the structure− biocompatibility relationship among different TLs is crucial for their successful application in biochemical and biological systems. Here we report the synthesis of novel 1,5disubstituted TL2 by RuAAC and directly compare this to its 1,4-disubstituted equivalent TL5, which is formed by CuAAC. In duplex melting experiments, TL2-modified ON2 forms more stable duplexes with DNA and RNA targets than TL5-modified ON5. Moreover, the read-through kinetics and amplification of TL2 by Klenow fragment and Taq DNA polymerase exceed all other TLs tested. To the best of our knowledge, TL2 has the fastest read-through kinetics and the highest efficiency of full-length product formation by Taq polymerase among the TLs reported to date. Further studies will be necessary to determine if the excellent read-through compatibility of TL2 by Taq can be translated into a cellular environment and be accommodated by RNA polymerases as  TL2  6  2  not significant  yes  TL3  6  2  medium  yes  TL4  6  3  medium  yes  TL5  7  3  High  no, deletion  TL6  7  3  High  yes 19  TL7  7  3  very high  no, deletions 19 a Defined as the minimal number of bonds connecting the 5′-and the 3′-furanose rings. b Along the internucleoside linkage. c Defined as having the correct sequence after read-through has occurred. previously demonstrated for TL6. 8,9,29 Moreover, direct comparison between TL2 and TL5 revealed a remarkable difference in read-through accuracy. While TL2 was read correctly by hot-start Taq polymerase, TL5 resulted in a deletion mutation around the triazole linkage. We have discussed the structure−biocompatibility relationship of several TLs, and we provide a rationale to assist future designs of artificial backbone mimics in replication templates. Importantly, the structure−biocompatibility relationship presented in this study is derived from ONs with TLs between two thymidines, providing a rather challenging sequence context for polymerase read-through. Efficiencies can vary with the nucleobase sequence and generally improve with greater duplex stability (a higher GC-content) around the triazole. Indeed, enhanced read-through compatibility was previously observed for a C TL6 C motif. 19,29 Hence, we anticipate that further investigations using different sequences will provide additional validation of the proposed model. The application of RuAAC to form TL2 exemplifies a powerful strategy to increase TL flexibility through formation of 1,5disubstituted 1,2,3-triazoles. In contrast to the widely adopted CuAAC reaction, the feasibility of the RuAAC reaction to efficiently ligate azide-and alkyne-modified ONs in high yield, and without inducing DNA damage, remains elusive. However, in light of recent advances to obtain 1,5-disubstituted 1,2,3triazoles via water and air compatible Ni-catalyzed click chemistry, 46