It has previously been demonstrated that LNA-modified templates can be read by Phusion High Fidelity DNA polymerase [25,27] and that LNA triphosphates can be used as a substrate by KOD DNA polymerase [22,23,27]. In principle, this paves the way for in vitro selection experiments using LNA-modified libraries. We have devised the following setup that allows us to amplify a DNA library with all adenosines replaced by LNA-A and then re-generate the single-stranded LNA-modified pool:

We wanted to know whether our setup was able to sustain an LNA-A modified library, or whether the foreign nature of the LNA moieties would result in loss of the LNA-containing members. We therefore performed seven rounds of mock in vitro selection in the absence of a specific target. By doing seven rounds we emulate a typical in vitro selection experiment, but any changes in the library’s composition are solely due to the intrinsic selection pressure of our setup.

By using different barcoded PCR primers for rounds 1, 4, and 7, we could pool samples from these rounds for 454 pyrosequencing. After appropriate sorting and filtering of the sequence data we obtained 19406 (round 1), 16427 (round 4) and 1529 (round 7) sequences corresponding to the 40-mer randomized region in our library. The lower number of sequences from round 7 is likely due to a dilution error.

According to our 454 sequencing data the overall LNA-A content dropped 2-fold from 20.5% in round 1 to 9.9% in round 4 and further 1.5-fold to 6.6% in round 7 (Figure 1A). Concomitantly, thymidine levels were near constant (at 28%), cytidine levels somewhat elevated (22.9% to 28.2%), and guanosine levels increased from 28.0% to 37.1%.

We speculated if successive LNA moieties pose a particular challenge in our setup and looked at the frequency of various lengths of poly-adenosine tracts. The dotted bars in Figure 1B indicate the distribution found in 1 million in silico-generated 40-mer sequences and acts as a reference. The distribution of adenosine tracts after round 1 corresponded well with the simulated distribution (Figure 1B, left panel). However, after round 4 we found a significant decrease of at least 35% in the number of poly-adenosine tracts (AA (2A), AAA (3A), AAAA (4A) etc.), whereas single adenosines (1A) increased in frequency by 28%. This tendency became even more pronounced after round 7. Notably, we only saw this trend with LNA-A, as the relative frequencies of thymidine, cytidine and guanosine tracts were practically unaffected between rounds 1, 4, and 7 (Figure 1B, right panels). This suggests that single LNA moieties are favoured over successive LNAs.

We also asked whether the changes in LNA content and context were specific to LNA-A. To this end, we compared libraries maintained with LNA ATP and LNA TTP. We used an oligo-capture strategy for purification of these LNA-modified libraries; details can be found in the Experimental section. Digestion of the libraries suggested slightly increasing exonuclease susceptibility and thus decreasing LNA content over the first three rounds, regardless of the type of LNA nucleotide (Figure 2). After three rounds of mock in vitro selection we cloned and Sanger sequenced between 27 and 43 clones from each library.

The library maintained with LNA ATP showed a 2-fold drop in total adenosines (20.7% to 10.3%; Figure 3A) in the randomized 40-mer region, and single adenosines increased in numbers, while tracts of 3 or more adenosines became absent (Figure 3B). This behaviour was very similar to that observed with LNA ATP in our 454 sequencing data (Figure 1), although the number of members evaluated by Sanger sequencing was much smaller.

The library maintained with LNA TTP showed a remarkable 4.5-fold drop in total thymidines (31.0% to 6.8%; Figure 4A). The ratio of single thymidines increased 2-fold, while tracts of two or more thymidines were greatly reduced or entirely absent. Conversely, the distributions of adenosine, guanosine and cytidine tracts corresponded well with their simulated distributions (Figure 4B). Overall, our data suggest a similar or stronger skewing of LNA content in a library maintained with LNA TTP compared to that maintained with LNA ATP.

Library and primers: 5'-GCTCTGTCGTCGTGGTCGGTCC-(N₄₀)-GGCCTTTTGTGTGTCGTTT, 5'-GCTCTGTCGTCGTGGTCGGTCC, 5'-AAACGACACACAAAAGGCC. Barcoding forward primers for 454 sequencing were 5'-CGTATCGCCTCCCTCGCGCCATCAGCTCTGTCGTCGTGG­TCGGTCC (rd. 1–3), 5′-CGTATCGCCTCCCTCGCGCCATCAGCTGTCGTCGTGGTCGGTCC (rd. 4–6), and 5'-CGTATCGCCTCCCTCGCGCCATCAGCGCTCTGTCGTCGTGGTCGGTCC (rd. 7). Synthesis of LNA-A containing libraries was in the forward direction, while synthesis of LNA-T containing libraries was in the reverse direction, using the complementary strand as template.

5 nM template and 0.5 µM of each primer was combined in 1× Phusion HF buffer with 200 µM of each deoxyribonucleotide triphosphate and 0.02 units/µL Phusion DNA polymerase (Finnzymes, Espoo, Finland). Typical PCR conditions were: 98 °C/5 min, 24 cycles (98 °C/5 s, 53 °C/10 s, 72 °C/5 min), 4 °C/hold. Sanger sequencing experiments used 0.5 nM template and 0.04 units/µL Phusion DNA polymerase over 20 cycles.

Reaction conditions were: 1 × KOD XL buffer, 3 mM MgSO₄, 0.2 mg/mL BSA, 0.2 units/µL KOD XL DNA polymerase (Novagen), 2.5 µM primer, 2.5 ng/µL dsDNA, and 0.25 mM of each nucleotide triphosphate as required; LNA triphosphates were synthesized at the Nucleic Acid Center. Thermocycling was: 95 °C/1 min, 15 cycles (95 °C/15 s, 60 °C/35 s, 72 °C/5 min), 4 °C/hold. Reactions were quenched with excess EDTA and purified from denaturing 6% polyacrylamide gels by crush-and-soak.

Aliquots of LNA-A-modified libraries after 1, 4 and 7 selection rounds were barcoded by PCR and pooled. 454 GS FLX-sequencing was by GATC Biotech (Solna, Sweden). Filtering parameters were adjusted for short amplicons by GATC Biotech. The amplicons were sorted by barcode and the randomized regions were extracted and analyzed using custom Python scripts. Incomplete regions or members with low 454 quality scores were removed. Tract distribution was simulated by generating and analyzing 1 million random 40-mers.

We used an oligo capture approach for libraries for Sanger sequencing: PCR by Phusion HF DNA polymerase was with a 5'-phosphorylated primer to enable the removal of the forward strand by lambda exonuclease. The resulting single-stranded template was then used with KOD DNA polymerase and the appropriate triphosphates. Only full-length products were recovered by hybridizing to a biotinylated capture probe specific for the 3' end of the product and subsequent capture on Streptavidin-coated paramagnetic beads. After careful washing the pure product strands were heat-eluted.

Double-stranded DNA was prepared by PCR with a 5'-phosphorylated primer and a 5' fluorescein-labeled primer. Digestion with lambda exonuclease (New England Biolabs, Ipswich, MA, USA) was with 6.7 units/µg double-stranded DNA (50 ng/µL final DNA concentration) at 37 °C for 25 min. The reaction was stopped by addition of 1 volume 50 mM EDTA or by heating to 75 °C for 15 min. Full digestion was verified by agarose gel electrophoresis and ethidium bromide staining. Fluorescence scanning was on a Typhoon phosphoimager (GE Healthcare, Piscataway, NJ, USA).

Primer extension products were slow-annealed over 30 min. or more with at least 2-fold molar excess of dual biotinylated oligomer complementary to the products’ 3' end (5'-[biotin]-AAACGACACACAAAAGGCC). The solution was adjusted to 1 × B&W (according to manufacturer’s instructions) and combined with Dynabeads M-280 Streptavidin (Invitrogen, Life Technologies Europe BV, Naerum, Denmark). The number of beads corresponded to an excess of biotin binding sites. After 15 min with gentle agitation the beads were washed first in 1 × B&W, then in 1 × SSC, before heat-elution of the non-biotinylated species into 40 µL 1 × SSC (75 °C, 5 min).

The DNA library, the LNA-containing libraries, and a synthetic LNA-A containing oligomer (5'-GGTCTGGTCCACACCCAGCCGCCaCCCaGGGaCGCaGCCaGGCaCGGCGGGCCTATAGT­GAGTCGTATTA; lowercase is LNA-A) were 5'-radiolabelled with T4 polynucleotide kinase (Fermentas) and gel purified. The oligos were incubated at 72 °C in digestion buffer (1 × KOD buffer #2, 3 mM MgSO₄, 0.2 mg/mL BSA) with or without 0.02 units/µL KOD DNA polymerase (Novagen, San Francisco, CA, USA). Reactions were quenched in 1 vol. ice-cold 95% formamide/50 mM EDTA and resolved on 13% denaturing polyacrylamide gels. Autoradiography was on a Typhoon phosphoimager (GE Healthcare).

PCR products were blunt-end cloned into SmaI-digested pUC18 or pUC19 (this offered multiple inserts per plasmid) and transformed into E. coli TOP10 (Invitrogen). Plasmids from re-streaked colonies were purified (Miniprep, Qiagen, Valencia, CA, USA) and sequenced by Eurofins MWG Operon (Ebersberg, Germany). Sequence analysis was with custom Python scripts.