Evolved polymerases facilitate selection of fully 2′-OMe-modified aptamers† †Electronic supplementary information (ESI) available: Methods, supplementary scheme, tables and figures. See DOI: 10.1039/c7sc03747c

Evolved DNA polymerases are used in selections with fully 2′-OMe modified libraries to identify aptamers with high affinity for HNE.


Introduction
Aptamers, RNA or single-stranded DNA (ssDNA) oligonucleotides that bind to a specic target, are useful affinity reagents and potentially promising therapeutics. [1][2][3] Aptamers are discovered by creating pools of oligonucleotides with randomized sequences, enriching the pools for members that bind to a desired target (e.g. selection), and then amplifying the enriched pools and subjecting them to additional rounds of selection. While both DNA and RNA aptamers may be developed relatively quickly and inexpensively, their use is limited by the inherent instability of natural oligonucleotides in biological solutions due to nuclease degradation. 4,5 Thus, there is signicant interest in the discovery of modied aptamers that are more resistant to nucleases, with 2 0 -OMe and 2 0 -F substituents having received the most attention. In particular, 2 0 -OMe substituents are desirable because they impart the highest level of nuclease resistance and are also relatively inexpensive as triphosphates. Moreover, while the discovery of aptamers with high affinity to desired targets is now routine, a second limitation in aptamer development is the identication of aptamers with high specicity, which is generally more challenging. This is particularly true with positively charged protein targets due to the dominance of non-specic electrostatic interactions with the negatively charged phosphate backbone. 6 Modications such as 2 0 -OMe substituents could potentially facilitate the formation of, or even directly engage in, more specic interactions, but this has not been extensively investigated.
Aptamers with 2 0 -modications have historically been generated via post-selection modication of a natural aptamer, 7,8 but their introduction oen interferes with the selected activity. The straightforward inclusion of 2 0 -modications in the selection process itself is challenging, because they generally interfere with polymerase recognition. While combinations of polymerases, polymerase mutants, and/or specic reaction conditions have been identied that allow for the inclusion of various types of modied sugars, 4,9,10 the inclusion of 2 0 -OMe modications has proven particularly challenging. The only known, successful example using 2 0 -OMe modications in SELEX was provided by Burmeister et al. 11 who reported that with precisely optimized conditions, the Y639F/H784A/ K378R triple mutant of T7 RNA polymerase can transcribe DNA into nearly fully 2 0 -OMe modied oligonucleotides, which along with reverse transcription by Thermoscript reverse transcriptase and amplication of the natural DNA, allowed selection of modied aptamers that bind vascular endothelial growth factor (VEGF). However, chemical synthesis was used to supplement the transcribed library for selection and the transcription reactions required the addition of unmodied GTP, indicating that at least the majority of sequences could not be transcribed in their fully modied form. Reports of additional T7 RNAP mutants with improved recognition of 2 0 -OMe substrates 12,13 are notable, but no report of their use to select for modied aptamers has yet been reported.
To enable the more facile discovery of 2 0 -modied aptamers, we recently reported the directed evolution of variants of the Stoffel fragment of Taq DNA polymerase that better recognize 2 0 -modied substrates. 14 One variant, SFM4-3, was found to efficiently and directly PCR amplify 2 0 -F-modied oligonucleotides, 14 and we used it to evolve 2 0 -F-purine modied aptamers that bind human neutrophil elastase (HNE), 15 a serine protease associated with numerous inammatory diseases. 16,17 Interestingly, HNE is positively charged, and the 2 0 -F substituents were found to decrease non-specic electrostatic interactions mediated by the negatively charged oligonucleotides in favor of a more specic mode of molecular recognition. In addition, we evolved two polymerase variants, SFM4-6 and SFM4-9, that more efficiently "transcribe" and "reverse transcribe" fully 2 0 -OMe modied oligonucleotides, respectively. 14 Here, we apply these evolved polymerases to selections for HNE binders using libraries of fully 2 0 -OMe modied sequences, identifying several with high affinity, specicity, and stability against nuclease degradation.

Results and discussion
To explore the use of our evolved polymerases for the more facile discovery of 2 0 -OMe-modied aptamers, we initiated selections for fully modied HNE binders using SFM4-6 and SFM4-9. A library of $3 Â 10 14 67-mer ssDNA fragments with a central 30-mer random region anked by primer binding sequences was rst transcribed into a 2 0 -OMe-modied library using SFM4-6 and a fully 2 0 -OMe-modied primer ( Fig. 1 and Scheme S1 †). The transcription product was incubated with TurboDNase to generate a DNA-free, single-stranded 2 0 -OMe oligonucleotide library of $6 Â 10 13 members, which was heated to 95 C and then cooled slowly to allow for secondary structure formation. The library was then subjected to selection for HNE binding (Fig. 1). Briey, the library was incubated with immobilized HNE, subjected to washing with binding buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 6 mM KCl, 2 mM MgCl 2 ), and the enriched library was recovered by eluting with hot formamide. Using SFM4-9, the recovered oligonucleotides were then reverse transcribed into DNA and PCR amplied with OneTaq DNA Polymerase (New England Biolabs). The process of transcription, selection, reverse transcription, and amplication was then repeated with the selection stringency progressively increased each round by including a negative selection step consisting of pre-incubation in an empty well and then in a well coated with bovine serum albumin, in addition to employing longer wash times and adding increasing amounts of yeast tRNA to the binding buffer (up to 3000 mg mL À1 ), to compete with non-specic 2 0 -OMe oligonucleotide binding (Scheme S1 and Table S1 †).
Aer ve rounds of selection, we sequenced 25 clones and found two that were enriched, 2mHNE-1 and 2mHNE-2 (Table 1 and Fig. S2 †). We then subjected the enriched library to an additional ve rounds of selection in which the stringency was further increased by increasing the ionic strength of the wash buffer (by adding up to 500 mM NaCl) and by adding increasing amounts of fetal bovine serum (FBS; up to 80%; see Table S1 †). Sequencing 30 library members revealed four additional sequences that were enriched, 2mHNE-3-2mHNE-6 ( Table 1).
The identied oligonucleotides were individually prepared via transcription with a fully 2 0 -OMe-modied primer containing a 5 0 -Alexa 488 uorophore and their affinity for HNE was characterized using microscale thermophoresis (MST) with binding buffer supplemented with 0.05% Triton X-100 and 0.1% PEG 8000 (Table 1 and Fig. S3 †). The round-ve aptamers 2mHNE-1 and 2mHNE-2 bound to HNE with K d values of 520 and 180 nM. The round-ten aptamers bound with higher affinity, with K d values ranging from 45 nM to 120 nM. For comparison, we prepared and analyzed the previously reported, fully natural HNE aptamer DNA-I, 18 which under identical conditions bound to HNE with a K d of 280 nM.
To determine whether the 2 0 -OMe substituents contributed to function, we prepared the fully natural counterparts (HNE-1-HNE-6) by PCR amplication of chemically synthesized DNA templates, with one primer that was labeled with a 5 0 -Alexa 488 uorophore and one that was labeled with a 5 0 -biotin. The PCR product was immobilized on streptavidin beads, which were then washed with NaOH to elute the desired ssDNA aptamers. K d values for HNE-1-HNE-6 were determined by MST to be 500, 520, 300, 340, 480, and 550 nM, respectively (Fig. S3 †). Thus, with the exception of the round 5 aptamer 2mHNE-1, the 2 0 -OMe substituents contribute signicantly to HNE binding.
As mentioned above, aptamers oen bind positively charged proteins, such as HNE (pI z 9), via non-specic electrostatic interactions, raising specicity as a concern. Indeed, previous studies revealed that DNA-I does not retain affinity for HNE in the presence of high salt concentrations. 15 We thus examined the specicity with which 2mHNE-5 binds to HNE compared to other proteins with various pI values, under both low (150 mM) and high (500 mM) salt conditions, using a uorescence-based plate assay. Briey, HNE, bovine serum albumin (pI z 4.7), thrombin (pI z 7), porcine pancreas elastase (pI z 8.5), chymotrypsin (pI z 8.8), or lysozyme (pI z 11) was non-specically immobilized to the surface of individual wells of 96-well microtiter plates; 2mHNE-5 or DNA-I, both 5 0 -Alexa 488 labeled, was added to a nal concentration of 10 nM; and aer the plates were washed with binding buffer (150 mM NaCl) or high salt binding buffer (500 mM NaCl and 66 mg mL À1 yeast tRNA), uorescence was measured (Fig. 2). Under low salt conditions, despite the concentration of the aptamers being signicantly below the K d values measured by MST, both 2mHNE-5 and DNA-I bound to HNE, but did not bind strongly to any of the other proteins (although both aptamers exhibited some affinity for lysozyme). However, under the high salt conditions, 2mHNE-5 specically bound HNE while DNA-I did not, demonstrating that the 2 0 -OMe modied aptamer binds to HNE via more specic interactions.
To explore the ability of the aptamers to bind HNE in serum, we repeated the above described low-salt plate binding assays with 2mHNE-5 or DNA-I, but with increasing amounts of FBS in the wash buffer (Fig. 3A). The results clearly show that DNA-I rapidly loses affinity for HNE as the percentage of FBS is increased. In contrast, 2mHNE-5 loses affinity more slowly and even retains the ability to bind HNE during washing with 100% FBS. We also directly measured binding in the presence of 80% FBS, and found that 2mHNE-5 clearly binds HNE better than does the DNA-I control (Fig. 3B).
To conrm that the 2 0 -OMe substituents do confer 2mHNE-5 with resistance to nuclease degradation, folded 2mHNE-5, HNE-5, or DNA-I was mixed with undiluted FBS (Fig. 4). Aer incubation at 37 C, PAGE analysis revealed the degradation of DNA-I, with little aptamer remaining aer 4 h. In contrast, even aer 24 h, no degradation of 2mHNE-5 was observed, clearly  demonstrating that the 2 0 -OMe modications do indeed provide signicant stabilization against nuclease degradation. HNE-5 was degraded even faster than DNA-I, likely due to the absence of secondary structure, suggesting that the 2 0 -substituents of 2mHNE-5 are required for proper folding.

Conclusions
There has been both academic and industrial interest in developing 2 0 -OMe aptamers, 4,11,19 but their identication has historically been challenging, because aptamer selections rely on polymerase recognition, and the modied triphosphates are not well recognized. In fact, to our knowledge, prior to this work no selections have been performed in which the modied libraries were produced using only 2 0 -OMe triphosphates. The directed evolution of SFM4-6 and SFM4-9 now allows for the "transcription" and "reverse transcription" of fully modied aptamers using only 2 0 -OMe modied triphosphates and we have used them here to select 2 0 -OMe modied aptamers that bind HNE. Importantly, the OMe substituents of the selected aptamers are required for high affinity and specic HNE binding. Likely, both the affinity and specicity afforded by the modications could be increased further by additional diversi-cation and selection, including negative selection against binding related proteins. Finally, the availability of SFM4-6 and SFM4-9 should enable the facile selection of other 2 0 -OMe aptamers and the exploration of the effects of the added OMe groups beyond imparting the aptamers with resistance to nucleases.

Conflicts of interest
There are no conicts to declare.