The nucleobase tautomer stability was investigated here for guanine derivatives, similar to our investigations of adenine derivatives (Potowski et al., 2021) and the expanded Hachimoji genetic alphabet (Eberlein et al., 2020), since it is a precondition for unambiguously reading the DNA barcode. Applying the same computational methodology as in Potowski et al., 2021, which is a refined protocol compared to our earlier Hachimoji investigations (Eberlein et al., 2020), on other guanine derivatives than presented in this work, we calculated all possible tautomers of guanine Ia-c, 7-deazaguanine IIa-c, 8-aza-7-deazaguanine IIIa-c and 8-azaguanine IVa-c (Figure 2A, Supplementary Tables S1-S3) with respect to their thermodynamic stability. According to this analysis, the reaction free energy between the Watson-Crick tautomers (I-IVa) and both alternate tautomers (I-IVb,c, Figure 2A, Supplementary Tables S1-S3), which translates into tautomer populations, is increased in II-IVb and reduced in II-IVc compared to natural guanine. With the smallest energy difference of more than 5 kcalmol⁻¹ being observed for the tautomer Ib of the natural guanine, the resulting mismatching tautomer fractions are negligible for all non-natural species as their free energies are even larger. This is a remarkable result as the Hachimoji guanine derivatives are by far less stable from the tautomer perspective (Eberlein et al., 2020). Therefore, 7-deazaguanine, 8-aza-7-deazaguanine or 8-azaguanine containing oligonucleotides should be well suited for DNA encoding chemistry with Watson-Crick tautomer fractions of at least 99.99%. Thus, the choice of a specific nucleobase for encoding chemistry depends on (commercial) nucleoside availability and chemical stability.

Ligation of chemically modified 7De8a-dG-containing DNA barcodes by T4 ligase and correct reading of the resulting DNA template by polymerases are essential for DEL synthesis. We performed test ligations with T4 ligase based on our previously published coding strategy (Supplementry Figures S2B,S1) that uses an adapter oligonucleotide composed of inosine and stable abasic sites opposite the chemically stabilized code, followed by PCR amplification and DNA sequencing. The sequences of all oligonucleotides used in ligation and amplification experiments are given in Supplementary Table S5. Agarose gel electrophoresis using a 3% or 4% agarose gel was used to analyze the ligation experiments (Supplementary Figure S2). Amplicon sequencing of the fully encoded DNA-conjugates showed that the Taq polymerase correctly copied the nucleotides (Figure 2C, Supplementary Table S6). We then compared the amplification efficiency of 7De8a-dG-containing ligation products with one native DNA ligation product by qPCR (Figures 2D, Supplementary Figures S3–S9). Compared to the native DNA, the chemically modified template required a total of two to four cycles more to reach the log-linear phase of amplification. Lower amplification efficiency, especially prior reaching the log-linear phase, may at least partially be explained by the use of a universal adapter sequence opposite the chemically stabilized code. Previously, introduction of abasic sites into DNA templates has been shown to lead to much reduced amplification efficiency (Sikorsky et al., 2007).

DNA stability under diverse reaction conditions is an essential precondition for the synthesis of functional DELs. For that reason, we recently investigated the stability profile of different CPG-bound pyrimidine DNA oligonucleotides and mixed pyrimidine/purine DNA oligonucleotides in the presence of acidic solutions, diverse metal salts and organic reagents (Table 1) (Potowski et al., 2019). Especially purine containing sequence 3 revealed low stabilities under various conditions e.g. acids or oxidants. Based on these findings and inspired by an early account describing encoded solid-phase peptide chemistry (Needels et al., 1993), we proposed that the substitution of native purine nucleobases in a given sequence by 7-deazaadenine would lead to a three-letter DNA barcode 4 with enhanced chemical stability. Indeed, exchanging 2′-deoxyadenosine with 7De-dA 1 (oligonucleotide 4) led to a much-improved stability profile which was comparable with the stability of pyrimidine DNA sequences (Potowski et al., 2021). Here, we aimed at going a step further to arrive at a chemically stabilized four-letter DNA barcode 6. Therefore, we kept the 7De-dA 1 and replaced in addition the native 2′-deoxyguanosine by the modified 7De8a-dG 2. To our delight, this chemically modified four-letter oligonucleotide 6 showed a higher stability against the tested conditions as compared to the native DNA 3, too, see Table 1 for the screening results. Especially metal ions were much better tolerated by sequence six than by the native DNA 3 (Table 1, entries 2,9,12). However, with the modifications of the 2′-deoxyguanosine we did not reach the same overall level of stability as observed with the previously described 7De-dATC sequence 4. The stability screen on both 7De8a-dG-containing sequences 5 and 6 revealed that still some DNA degradation occurred under acidic or oxidative conditions. For instance, the incubation of these sequences with a 3.7% aqueous hydrochloric acid solution for 22 h at room temperature led to ca. 50% depurination, while 7De-dA-containing DNA 4 was left completely intact under these conditions, yet, native DNA 3 suffered almost complete degradation by depurination (Table 1, entry 1).

Encouraged by the above shown results regarding stability and biological compatibility we started to transfer the previously established reactions on CPG-bound DNA to the chemically stabilized four-letter barcode containing 7De-dA as well as 7De8a-dG (Figure 3, Supplementary Table S7) (Potowski et al., 2021). The first set of reactions belongs to the class of isocyanide multi-component chemistries (Figure 3A). The Ugi four-component reaction (U-4CR), the Ugi-azide four-component reaction (UA-4CR), and the Groebke-Blackburn-Bienaymé three-component reaction (GBB-3CR) were readily performed on a CPG-bound 16mer 7De-dATC7De8a-dG-DNA-encoded aldehyde 13, resulting in near-quantitative product conversions to target conjugates 17, 20, and 22 with less than 5% DNA degradation (Figure 3A). However, translation of the Ugi-aza-Wittig four-component (U-4CR/aza-Wittig) resulted in a high degree of DNA degradation (44%) and a conversion of 53% to the desired product 26 (Figure 3A). Next, we explored protic acid-promoted reactions like the (R)-(-)-BNDHP promoted Biginelli reaction (Figure 3B). This reaction proceeded well with urea to the desired product 29 with a high conversion of 86% and no detectable DNA degradation. Notably, also the more challenging phenyl urea that requires higher amounts of the acid worked smoothly without DNA damage (see SI). The (R)-(-)-BNDHP promoted Povarov reaction led to the desired product 32, too. However, in this case a low degree of DNA degradation was observed. The trifluoroacetic acid-promoted Pictet-Spengler reaction allowed for synthesis of the desired β-carboline 35 from a DNA-tryptophane conjugate 33 with 85% conversion. As expected after the stability screen, 40% of the DNA were degraded during the reaction. The intact DNA-Pictet-Spengler conjugate 34 could be isolated by preparative RP-HPLC. Last, we investigated the usability of Lewis acid-promoted reactions on the chemically stabilized barcode (Figure 3C). The Cu(I)/bipyridine-promoted Petasis reaction, the Zn(II)-promoted aza-Diels-Alder reaction as well as the Yb(III)-promoted Castagnoli-Cushman reaction worked smoothly without any DNA degradation and led to the desired DNA-coupled products 39, 41 and 44 with moderate to good conversions. The Yb(III)-mediated pyrazole synthesis using aryl hydrazines resulted in the desired product 45 with a conversion of 17% and DNA degradation of 31% (Figure 3C). The Au(I)-promoted pyrazoline containing spiroheterocycle 48 synthesis starting from DNA-aldehyde conjugate 13 with alkynol 46 and hydrazide 47 was successful with moderate conversion and no detectable DNA damage. Also, the Au(I)-promoted pyrazoline synthesis from 16mer 7De-dAT7De8a-dGC-alkyne conjugate 49 with isobutyraldehyde 50 and hydrazide 47 led to the desired product 51 with an excellent conversion, and without DNA degradation. However, the more demanding Au(I)-promoted pyrazole synthesis in glacial acetic acid at 60 °C resulted in a higher degree of DNA damage (32%). But still the isolation of intact DNA-pyrazole conjugate 52 was possible by preparative RP-HPLC.

Beside the synthesis of CPG-bound DNA conjugates, the protective group removal in aqueous solution plays an important role in DEL synthesis, too. Therefore, we explored the Boc deprotection of an isolated 16mer 7De-dAT7De8a-dGC-tetrahydroisoquinoline conjugate 32 with 10% TFA in water. However, the reaction conditions led to DNA degradation by exclusive depurination of the 7-deaza-8-aza-2′deoxyguanosine, as analyzed by MALDI MS, while the 7-deazaadenine remained intact. This clear-cut result was in line with the substantial amounts of depurination already observed in the Pictet-Spengler reaction that was performed with 5% of TFA for 20 h in an organic solvent on a CPG-bound and fully nucleobase-protected oligonucleotide.