Converting Double-Stranded DNA-Encoded Libraries (DELs) to Single-Stranded Libraries for More Versatile Selections.

DNA-encoded library (DEL) is an efficient high-throughput screening technology platform in drug discovery and is also gaining momentum in academic research. Today, the majority of DELs are assembled and encoded with double-stranded DNA tags (dsDELs) and has been selected against numerous biological targets; however, dsDELs are not amendable to some of the recently developed selection methods, such as the cross-linking-based selection against immobilized targets and live-cell-based selections, which require DELs encoded with single-stranded DNAs (ssDELs). Herein, we present a simple method to convert dsDELs to ssDELs using exonuclease digestion without library redesign and resynthesis. We show that dsDELs could be efficiently converted to ssDELs and used for affinity-based selections either with purified proteins or on live cells.

Introduction

Identification of high-quality binders that can modulate the function of biological targets is important for the advancement of drug discovery as well as chemical biology research. The advent of DNA-encoded chemical libraries (DELs), originally proposed by Brenner and Lerner in 1992,[1] as well as Gallop and co-workers in 1993,[2] has made the expensive and complex high-throughput screening (HTS) accessible to many researchers. In a DEL, each library compound is coupled with a unique DNA tag, which encodes its chemical structure and also acts as an amplifiable template for structural elucidation. Since the spatial encoding is replaced with DNA encoding in DELs, it allows the synthesis, processing, and selection of hundreds of millions to many billions of compounds in a single mixture at the minute scale (Figure a), and it does not require sophisticated equipment and logistics for compound handling and the selection can be done in a short time frame.[3,4] In addition, DELs can access greater chemical space than biological display libraries, especially with the recent development of DEL-compatible reactions.[5−14] In the past decades, this technology has been widely adopted by the pharmaceutical industry and also showed potential as a powerful tool in academic research.[3,9,10,15−30]

Figure 1

(a) Scheme for DEL selection against immobilized protein targets. (b) HP-based encoding method for dsDELs. (c) Converting dsDELs to ssDELs may enable wider target scope and more versatile applications of DELs.

In 2009, GSK published a seminal work, in which they demonstrated the application of DELs on an industrial scale.[31] In this report, a short, covalently linked double-stranded DNA (dsDNA), named the “headpiece (HP)”, was designed as the starting point of library synthesis (Figure b). In each chemical synthesis step, the compounds are encoded by enzymatically ligating the respective DNA codes, also in the form of dsDNA, to the HP. Finally, a closing primer containing the distal primer-binding site for PCR amplification is added to complete the library synthesis/encoding. In addition, polymerase extension could be used for library encoding, which also resulted in dsDNA-encoded DELs (dsDELs).[32−34] The dsDNA tags may better protect the nucleobases from chemical modifications during library synthesis and are less likely to bind to proteins due to the more rigid structure.[31] In fact, today, the majority of DELs are encoded with dsDNA tags. Although having been used in early studies,[2,35] DELs encoded with single-stranded DNAs (ssDELs) are less abundant and they are often prepared with special encoding methods, such as DNA-templated synthesis (DTS),[36] DNA routing,[37] and DNA-directed assembly.[38,39] ssDNAs can also be used to assemble the encoded self-assembling chemical (ESAC) libraries[40−42] and DNA-encoded dynamic libraries.[43−51] Moreover, ssDELs are amendable to a number of selection methods suitable for nonimmobilized proteins, such as the interaction determination using unpurified proteins (IDUP),[52,53] the cross-linking-based methods,[54−60] and the selections against membrane proteins on/inside live cells (Figure c).[61,62] In these methods, the ssDNA tag provides a site for hybridization with an additional DNA to covalently connect with the target upon binding. These methods may expand the target scope of DELs to more complex targets in a more biologically relevant environment.[63,64] In addition, a recent report showed that ssDELs had higher enrichment folds than dsDELs, especially for low- to moderate-affinity binders.[65] Collectively, it is desirable to convert dsDELs, which are more widely available, to ssDELs that can be used for more diverse applications. Here, we show that dsDELs could be converted to ssDELs without library redesign and resynthesis using nuclease digestion. The resulting ssDELs can be used in affinity-based selections with purified proteins in solution or on live cells.

Results and Discussion

Previously, several strategies have been developed for converting dsDNAs to ssDNAs, including affinity separation,[66] asymmetric PCR,[67,68] denaturing electrophoresis,[69] and selective enzymatic digestion.[70,71] For converting dsDELs, ideally the method should (a) be compatible with the architecture of the majority of dsDELs, (b) not degrade ssDNAs that contain the DNA codes and primer-binding sites, (c) have a mild condition without damaging/modifying the library compounds, and (d) be technically simple with minimal loss of the precious library material. Thus, we chose to use DNA exonuclease digestion to selectively remove one strand of the dsDNA tag (Figure c), and the resulting ssDELs could be purified with a simple ethanol precipitation step. Initially, we tested T7 exonuclease and exonuclease III, as they can progressively digest one strand of DNA duplexes.[72,73] Unfortunately, we observed that both enzymes overcame the unnatural linkage and digested both of the DNA strands of HP (Figure S1). Next, we switched to Lamdba exonuclease (λ-exo), which does not have ssDNA activity but usually requires a 5′-phosphate group for digestion (Figure a).[74] As shown in Figure b, an 18-base pair HP (HP-1; 5 μM), which contains an amino group as the starting point of library synthesis and a 2-nt sticky end for adding the encoding tags, was treated with λ-exo (5 u/μL; 37 °C). The digestion mixture was ethanol-precipitated and resolved with denaturing electrophoresis, which showed the digestion of one DNA strand of HP-1 (Figure b). The DNA bands were excised, extracted, and characterized with mass spectrometry (MS). The results confirmed that the lower band was mostly the single-stranded digestion product of HP-1 with one phosphorylated nucleobase (an adenosine) remaining; a small amount of the product with all of the nucleobases before the unnatural linker digested was also observed (Figure c). Based on our previous studies, the remaining base will not interfere with DNA hybridization and DNA-templated cross-linking;[75,76] thus, we reason that there is no need to push for complete base removal and the mixed products could be used for ssDEL-based selections and other applications.[63] In addition, the processive digestion pattern of λ-exo[77] was clearly observed and a near-complete digestion was achieved after 16 h.

Figure 2

(a) Structure and DNA sequence of HP-1; Lamdba exonuclease (λ-exo) selectively digests the 5′-phosphorylated strand in HP-1 to yield ssDNA products. (b) Denaturing electrophoresis (20% TBE-urea) analysis of the digestion reactions. Conditions: 5 μM HP-1 in λ-exo buffer (5 μL); 37 °C at varied reaction times; λ-exo: 5 units/μL. (c) MS analysis of the digestion products extracted from each lane in (b). The product peaks are highlighted with an asterisk, and the expected/observed masses (m/z) and the structures/sequences are shown as marked.

We tested whether the ssDNA product could be further digested by λ-exo under prolonged reaction times. As shown in Figure a, gel analysis showed a >95% digestion yield after 6 h and no overdigestion was observed after 72 h. Moreover, we used quantitative PCR (qPCR) to quantify the DNA copy numbers in the reaction mixture at different time points; the results showed a 90.7% recovery yield after 48 hours of digestion (Figures b and S2). In the synthesis of dsDELs, after the encoding DNA tags are ligated to the headpiece, a closing primer will be added to install the distal primer-binding site and complete the library encoding (Figure b). In most cases, the 5′-end of the closing primer is a hydroxyl group without phosphorylation; thus, it is more desirable if the 5′-unphosphorylated headpiece DNAs could also be digested. Previously, the Zhao group showed that DNA duplexes with 5′-hydroxy or other modifications could be digested by λ exo, suggesting that a 5′-phosphate may not be absolutely required for λ exo.[77] To test this, a series of unphosphorylated headpiece DNAs (duplex region: 16–20-bp), containing 5′-overhang (2-, 10-, 20-, and 30-nt), blunt end, or 3′-overhang (2-, 10-, 20-, and 30-nt), were prepared and subjected to λ exo digestion (Figure a). The results showed that λ exo efficiently digested the unphosphorylated DNAs with 3′-overhangs or blunt end (Figures b,c), but it could only digest the duplexes with a short 2-nt 5′-overhang (Figure d), which is consistent with the known activities of the enzyme.[74] In practice, both 3′- and 5′-overhang are being used in dsDELs, depending on the closing primer;[78−85] therefore, the λ-exo-mediated digestion may be directly applied to 3′-overhang dsDELs, whereas a primer extension step may be needed for dsDELs with 5′-overhang to fill the single-stranded region.

Figure 3

(a) HP-1 was digested with λ-exo at different time durations (up to 72 h); the reaction mixtures were analyzed with electrophoresis (20% TBE-urea), and the digestion yields were calculated (details in Figure S2). (b) qPCR quantitation of the DNA copy numbers of the reaction mixtures. See qPCR details and the standard curves in Figure S2; all data are based on biological triplicate. Conditions: HP-1; 4 nM; λ-exo (5 u/μL, 20 μL) in lambda nuclease buffer at 37 °C.

Figure 4

(a) Series of unphosphorylated headpieces (duplex region: 16–20 bp) were subjected to λ-exo digestion; n indicates the number of overhanging bases. See the Supporting Information for details. (b–d) Denaturing gel analysis of the digestion reactions. The conditions are the same as in Figure , and the duplex substrates used are as marked on each lane.

Next, to verify the feasibility of λ exo digestion in a library format, we prepared a model library containing five compounds on an 18-bp headpiece duplex with a 2-nt 3′-overhang (dsDEL-1; Figure a). The mixture was treated with λ exo at 37 °C for 20 h and analyzed with UPLC-MS. As shown in Figure c, the LC trace showed three major product peaks and the MS analysis of the peaks have identified the ssDNA products of all five compounds. Similar to Figure c, both the products with one nucleobase remaining and with all bases digested (prior to the unnatural linker) were identified (Figure c). Furthermore, to mimic the DNA tag length of typical dsDELs, another model library was prepared by conjugating the five compounds to a 70-bp, unphosphorylated headpiece (dsDEL-2; Figure a). After λ exo digestion, UPLC-MS analysis again showed the ssDNA products of all five compounds (Figure d). Collectively, these results demonstrated that lambda exonuclease could be used to convert dsDELs to ssDELs.

Figure 5

(a, b) Two model libraries (dsDEL-1 and deDEL-2) with different tag lengths were prepared and subjected to λ exo digestion. (c) Digestion reaction of dsDEL-1 was analyzed with UPLC-MS; the LC trace showed three major peaks, and the respective MS data of each peak are shown below. (d) Digestion reaction of dsDEL-2 was also analyzed with UPLC-MS; MS results are shown. The ssDNA digestion products are as marked; dAp: with an adenosine phosphate remaining.

Recently, cross-linking-based approaches have been developed to perform DEL selections against nonimmobilized targets in solution, in cell lysates, or with live cells.[54−60] These methods require an ssDNA region close to the library compound for hybridization with a reactive DNA. Upon ligand binding, the reactive DNA covalently captures the protein, thus establishing a stable connection between the ligand and the target. We further tested whether the ssDNA product generated by λ exo digestion could be used in ligand-directed target cross-linking.[75] As shown in Figure a, we prepared an 18-bp headpiece DNA (HP-2) conjugated with a 4-carboxybenzene sulfonamide (CBS), a moderate-affinity ligand for carbonic anhydrase II (CA-II, Kd = 3.2 μM).[86]HP-2 was digested with λ exo to generate an ssDNA (binding probe; BP-1). Next, two complementary DNA strands carrying a 5′-biotin and a 3′-photo-reactive group phenylazide (capture probes; CP-1 and CP-2) were hybridized with BP-1. CP-1 and CP-2 have a 3 and 6-nt spacer, respectively, to extend the photo-cross-linker to reach the protein target. The BP-1/CP duplexes were incubated with CA-II (2 μM). After brief UV irradiation (365 nm, 30 s.), the mixtures were analyzed with Western blotting. The results clearly showed the capture of CA-II with the detectable bands matching the molecular weight of the DNA-CA-II conjugate (lane1 and lane 4; Figure b). No capture was observed without BP-1 or UV irradiation, indicating that the cross-linking was specific.

Figure 6

(a) Ligand-directed labeling of the target CA-2 using BP-1, generated by λ exo digestion, and CP-1/CP-2. (b) Western blots of the labeling experiments. Conditions: DNA probes: 2 μM each; CA-II: 1 μM; reaction buffer: 1× phosphate-buffered saline (PBS) and 0.1 M NaCl; UV: 365 nm, 30 s, 0 °C; λ exo digestion, 20 h at 37 °C. IB: immunoblotting. (c) Selection scheme of dsDEL-3 against CA-2.[55,56] (d) Sanger sequencing results before and after the selection. The coding regions are highlighted. See the Supporting Information for experimental details.

Next, we proceeded to model library selections. Two model libraries were prepared, in which the CBS ligand was encoded with a “GTGTGA” codon and mixed with the nonbinding background (with an amino group) encoded with a “TGACCT” codon at a ratio of 1:10 (dsDEL-3) (Figure c). dsDEL-3 was digested by λ exo, hybridized with a capture probe with a 6-nt spacer, and selected against a biotinylated CA-II following previously reported procedures.[55,56] In brief, after the library was incubated with the target and UV irradiation, the CA-2/DNA complexes were affinity-purified with streptavidin beads. After washes, the selected library was eluted, PCR-amplified, and analyzed with Sanger sequencing. As shown in Figure d, although the postselection library still had mixed sequences at the coding region, enrichment of the CBS-encoding GTGTGA sequence could be observed, indicating that λ-exo-digested dsDELs are amenable to cross-linking-based selections.

Finally, we conducted live-cell-based selections using a λ-exo-converted DEL. Carbonic anhydrase 12 (CA-12) is a membrane carbonic anhydrase implicated in malignant cancers, and it can be inhibited by arylsulfonamide ligands.[87] Previously, we used ligand-directed photo-cross-linking to enable the selections of both DELs and regular, non-DNA-encoded chemical libraries against CA-12 on live cells.[88,89] Neri and co-workers also reported cell-based selections against CA-9, another membrane carbonic anhydrase isoform using dual-pharmacophore DELs.[90] CBS is a known ligand for CA-12 (Kd = 0.97 μM)[88] and can be used as a positive control in the selection.[88,89] We first tested whether the CBS-conjugated ssDNA, obtained from λ exo digestion, could selectively capture the CA-12 protein on live cells. As shown in Figure a, A549 cells were cultured under hypoxia condition to increase CA-12 expression (Figure S3).[91,92]HP-2 was converted to the ssDNA BP-1 (Figure a) and then hybridized with CP-2. The BP-1/CP-2 duplex was directly incubated with A549 cells (4 °C, 1.5 h). After UV irradiation (365 nm, 30 s) and washes, the biotinylated proteins were affinity-purified using streptavidin beads and analyzed with Western bolt, and the results showed the specific capture of CA-12 (Figure b). Next, we prepared a model library containing a CBS-conjugated headpiece and 200-fold excess of the nonbinding background (dsDEL-4; Figure c). The CBS-conjugated headpiece and the background headpiece have orthogonal primer-binding sites.[54] The library was digested with λ exo, hybridized with the photo-reactive CP-4 (6-nt spacer), and then incubated with A549 cells overexpressing CA-12. After washes to remove the nonbinders, the binders were eluted under strong denaturing condition (90 °C, 10 min.).[88] The selected library was PCR-amplified and the enrichment fold of CBS was quantified by qPCR following previous reports.[54,55] The results showed a 13.7-fold enrichment of CBS (average of two biological replicates; Figure d). In contrast, a control selection with MEF cells with a low CA-12 expression level did not show significant enrichment of CBS. Collectively, these results demonstrated that λ-exo-converted dsDEL is compatible with cross-linking-based selections on live cells.

Figure 7

(a) Ligand-directed labeling of CA-12 on live A549 cells using BP-1, generated by λ exo digestion of HP-2, hybridized with CP-2 (with a 6-nt spacer). (b) Western blot analysis of the affinity pulldown experiments. Conditions: DNA, 5 μM; cell number, 100 million; incubation buffer, 1× PBS with 0.05 M NaCl; UV, 365 nm, 30 s, 0 °C; λ exo digestion, 20 h at 37 °C. Lane 1: BP-1/CP-2 with hv; lane 2: no BP-1; lane 3 no hv. (c) dsDEL-4 (a 1:200 mixture of CBS-conjugated headpiece and an amino-headpiece) was digested with λ exo, hybridized with CP-4, and used for selections against A549 and MEF cells, respectively. (d) After washes and elution, the selected library was analyzed with qPCR to determine the enrichment fold of the binders.[54] See the Supporting Information for detailed procedures.

Conclusions

In summary, we have described a method to convert dsDELs to ssDELs using lambda exonuclease, therefore making the library amendable to the applications and selection methods that require an ssDNA region, including ligand-directed target capture and cross-linking-based selection on live cells. Typically, λ exo preferably digests the 5′-phosphorylated strand in a DNA duplex; however, the results from Zhao and co-worker[77] and in Figure showed that λ exo also had reactivity toward 5′-unphosphorylated DNA strands, although the reactivity might be relatively lower.[77] The unnatural PEG linker in the headpiece DNA impeded λ exo digestion and the ssDNA strand that contains the compound, and the DNA codes remained intact. A relatively strong condition is needed to drive the digestion to completion, but again the unnatural linker appeared to be quite stable and prevented overdigestion. Moreover, this method avoids the need for library redesign and resynthesis, and the existing dsDELs may be straightforwardly converted to ssDELs for various applications. Considering that nearly all commercial and open-source DELs are encoded with dsDNA tags, we anticipated that this method will facilitate the applications of large-scale DELs to a wider range of biological targets.

Experimental Procedures

DNA Modification and Purification

Modified oligonucleotides with amine and biotin were purchased from Hitgen, Inc. or BGI Genomics Co. Ltd and used after HPLC purification. For other modifications, most small molecules were coupled with amine-modified oligonucleotides through amidation reactions and then purified by reversed-phase HPLC using a gradient of acetonitrile (5–80%) in 100 mM TEAA (pH 7.0), followed by lyophilization. Oligonucleotides were quantitated by a BioTek Epoch UV–vis spectrometer based on the calculated extinction coefficient at 260 nm. Oligonucleotides were characterized by an Agilent 1290 Infinity II UPLC coupled with ESI-MS.

Oligonucleotide Purification by Ethanol Precipitation

To an aqueous DNA solution (ideally the sample volume is less than 400 μL), 0.1 volume of 3 M NaOAc (pH 5.0) was added to adjust the sample pH to 5.0, followed by the addition of 1/40 volume of 10 mg/mL glycogen (Aldrich, final glycogen concentration: 133.33 μg/mL) and 2.5 volume of absolute ethanol. The solution was maintained at −80 °C for at least 1 h and then centrifuged at 14,000g for 30 min at 4 °C. The supernatant was discarded, and the pellet was dried by a SpeedVac. The recovered sample was dissolved in an appropriate buffer for subsequent analysis or experiments.

Lamdba Exonuclease Digestion

5 μM library or the DNA probes, 5 μL of 10× Lamdba exonuclease buffer, 20 μL (5 units/μL) of Lamdba exonuclease were mixed. The mixture was supplemented with water to a final volume of 50 μL. Digestion reactions were maintained at 37 °C for the specified time durations, before subjected to ethanol precipitation. The samples after ethanol precipitation were directly analyzed with Western blot or mass spectrometry.

Cell Culture

A549 cells with elevated CA12 expression were obtained with hypoxia cultivation[93] (AnaeroPack; Mitsubishi Gas Chemical) at 37 °C for 36 h in DMEM (10% FBS, 100 unit/mL penicillin, and 100 g/mL streptomycin). MEF cells were cultured in DMEM supplemented with 10% FBS, in a 5% CO2 humidified incubator at 37 °C.