CLICK-17, a DNA enzyme that harnesses ultra-low concentrations of either Cu+ or Cu2+ to catalyze the azide-alkyne ‘click’ reaction in water

Abstract To enable the optimal, biocompatible and non-destructive application of the highly useful copper (Cu+)-mediated alkyne-azide ‘click’ cycloaddition in water, we have isolated and characterized a 79-nucleotide DNA enzyme or DNAzyme, ‘CLICK-17’, that harnesses as low as sub-micromolar Cu+; or, surprisingly, Cu2+ (without added reductants such as ascorbate) to catalyze conjugation between a variety of alkyne and azide substrates, including small molecules, proteins and nucleic acids. CLICK-17’s Cu+ catalysis is orders of magnitude faster than that of either Cu+ alone or of Cu+ complexed to PERMUT-17, a sequence-permuted DNA isomer of CLICK-17. With the less toxic Cu2+, CLICK-17 attains rates comparable to Cu+, under conditions where both Cu2+ alone and Cu2+ complexed with a classic accelerating ligand, THPTA, are wholly inactive. Cyclic voltammetry shows that CLICK-17, unlike PERMUT-17, powerfully perturbs the Cu(II)/Cu(I) redox potential. CLICK-17 thus provides a unique, DNA-derived ligand environment for catalytic copper within its active site. As a bona fide Cu2+-driven enzyme, with potential for being evolved to accept only designated substrates, CLICK-17 and future variants promise the fast, safe, and substrate-specific catalysis of ‘click’ bioconjugations, potentially on the surfaces of living cells.


INTRODUCTION
The copper (Cu + )-dependent azide-alkyne cycloaddition ('CuAAC') reaction is perhaps the most versatile and widely used of the 'click' chemistries used for bioconjugations, and more generally for coupling together two molecules of interest, in this case, a terminal alkyne and an organic azide (1)(2)(3). The CuAAC reaction has found many applications, including bioconjugations involving living cells. Pow-erful and versatile as it is, the requirement for Cu(I) (Cu + ) ions in CuAAC can be problematic, owing to the concomitant generation of destructive reactive oxygen species (ROS) by Cu(I) (4)(5)(6). This, and the desirability for ∼100 M-1 mM concentrations of Cu(I) (toxic to most living cells), have inhibited the full-blown use of CuAAC in biological applications. A number of innovative approaches to minimize or abrogate the toxic effect of Cu(I), and of copper in general, have been developed in recent years. These include (i) utilization of specific ligands to bind and stabilize Cu(I); such ligands, including the water-soluble tris(3hydroxypropyltriazolylmethyl)amine (THPTA), accelerate the reaction by perturbing the Cu(II)/Cu(I) redox potential toward Cu(I) as well as serving as sacrificial oxidation substrates for the generated ROS (7-9); (ii) developing Cu(I)-chelating azides as participating reagents (10) and (iii) the use of enforcedly proximal azide and alkyne (11) or strained alkynes (12,13), both under copper-free conditions. These novel approaches afford substantial benefits; nevertheless, they continue to show collateral disadvantages--for instance, the relatively promiscuous reactivity (such as with thiols) of the strained alkynes used for copper-free AAC (14).
Biological transformations of any kind are most optimally facilitated by enzymes. If there were to exist an enzyme (protein or nucleic acid) capable of catalyzing the azide-alkyne cycloaddition (AAC) reaction with high efficiency under low-to-zero copper concentrations, it should prove to be a highly useful reagent for catalyzing this powerful coupling reaction, especially in biological contexts. However, AAC or CuAAC are not metabolic reactions per se; and out of the large number of naturally occurring copper-utilizing proteins, it is difficult to identify an obvious candidate that could be 'evolved' towards catalyzing CuAAC. Nevertheless, powerful methodologies exist, such as in vitro selection from vast, random sequence, singlestranded RNA or DNA libraries ('SELEX') (15)(16)(17), for de novo identification of RNA or DNA biocatalysts (ri-
The 5 -hexynyl moiety was linked to the 'forward' PCR primer sequence (5 -hexynyl-GGATC GTCAG TGCAT TGAGA-3 ) during automated DNA synthesis. The reverse primer had the sequence: 5 -TACCC GTTGC GGrATA CCACC-3 (where 'rA' shows the location of a single ribonucleotide placed within the DNA oligonucleotide). Azide-PEG 3 -biotin (Sigma) was chosen as a substrate based on its solubility in water and because, following CuAAC, azide-clicked 5 -hexynyl-DNA molecules would be tagged with biotin, enabling their separation from unreacted DNA via binding to streptavidin (Figure 1).
For 3 -end radioactive labelling, 50 pmol of a given oligonucleotide was dissolved in 10 mM Li-HEPES·(pH 7.4) buffer and denatured by boiling for 5 min. The cooled DNA was then 3 -32 P labeled using a standard protocol for use of Terminal Deoxynucleotidyl Transferase (Thermo Fisher Scientific, Vilnius, Lithuania). 1 l of [␣-32 P]-dCTP (Perkin-Elmer) was added, and the reaction mixture was incubated for 1 h at 37 • C. The 3 -32 P labeled oligonucleotide was then size-purified in 8% denaturing gels containing 7 M urea.
In vitro selection 5 M Cu + was the concentration chosen for the first round of in vitro selection, to encourage selection of DNAzymes that worked with ultra-low Cu + concentrations or required it not at all. 500 pmol of the DNA library was amplified by a large-scale (20 ml) PCR in order to incorporate the 'forward' 5 -hexynyl primer in one of the two strands. The 'reverse' primer contained an internal riboside nucleotide. 2.5 ml PCR reactions were used for round 2 and for all subsequent rounds of selection.
5 -Hexynyl-labeled single stranded DNA was recovered following NaOH-catalyzed ribonucleotide cleavage of one of the two component strands of the DNA duplexes, and then size-purification of the 5 -hexynyl-modified strand away from the cleaved complementary strand by preparative denaturing polyacrylamide gel electrophoresis.
The entirety of the recovered 5 -hexynyl labeled ssDNA pool, doped with a portion of the same strand pool 3labeled with 32 P using terminal transferase (ThermoFisher Scientific, Vilnius, Lithuania) was used for the first round of in vitro selection, with parallel selections being performed with and without added THPTA (25 M). The 5 -hexynyl labeled ssDNA was first heated to 100 • C in S buffer (50 mM HEPES, pH 7.4, 300 mM NaCl, 50 mM KCl, 20 mM MgCl 2 ), then cooled down to 22 • C over a period of 10 min in a thermal cycler. Azide-PEG3-biotin ('azide-biotin') to a final concentration of 2.5 mM was then added, followed by CuSO 4 to 5 M. The CuAAC reaction was initiated at 22 • C by addition of sodium ascorbate to 2.5 mM. The DNA solution was incubated at 22 • C in a thermal cycler for 2 h, following which the DNA was ethanol precip- itated, washed with 70% ethanol, air-dried, and dissolved in 5 l of 10 mM HEPES buffer, pH 7.4 containing ∼1.9 nmol dissolved streptavidin. The solution was loaded directly onto a denaturing gel (8% acrylamide, 7 M urea), and the streptavidin-retarded DNA bands, visualized by phosphorimagery (Typhoon 9410 Phosphorimager, Amersham Biosciences), were excised from the gel, eluted into TE buffer (10 mM Tris, pH 7.4, 0.1 mM EDTA) and collected by ethanol precipitation. The recovered DNA was PCR amplified, again using one 5 -hexynyl labeled primer and one internal ribonucleotide containing primer, followed by base treatment and gel purification (as above), to generate an enriched pool of 5 -hexynyl labelled ssDNA for next round of in vitro selection. Following 25 rounds of selection, the enriched DNA sequences were cloned into Escherichia coli and 40 clones picked for sequencing (Genewiz, Canada). Preliminary CuAAC experiments were done with individual DNA clones (all 5 -hexynyl labeled), and clone CLICK-17, 79-nt long, was picked as a promising catalytic candidate. CLICK-17 had the nucleotide sequence: 5 -GG ATC GTCAG TGCAT TGAGA TTATTATGCAACTC TATGGGTCCACTCTGTGAATGTGAC G GTGGT AT CCG CAACG GGTA-3 . A sequence-permuted variant of CLICK-17 DNA, named 'PERMUT-17', was used as a control DNA throughout; PERMUT-17 had the nucleotide sequence: 5 -GGATC GTCAG TGCAT TGAGA GACAT ACATG TTATC GGTAT GTTCG AGCTC TATAT CT CGT ACCCG TTGCG GATAC CACC.
Following incubation at 22 • C in a thermal cycler for the indicated reaction time, the DNA was ethanol precipitated, washed thoroughly with cold 70% ethanol, air-dried, redissolved and treated with streptavidin (2 l from a 5 mg/ml stock). The resulting solutions were loaded directly onto a denaturing 10% polyacrylamide gel. DNA bands in the gel following electrophoresis were visualized either by phosphorimagery or by ethidium bromide staining followed by fluorescence imagery (fluorescent gel images were recorded with a GENi gel imaging system, Syngene, Frederick, Maryland).
For in cis reactions of lysozyme-(N 3 ) n (see below) with ≡-CLICK-17, 1 M lysozyme-(N 3 ) n was mixed with 2 M ≡-CLICK-17 DNA, along with copper salts (or not) with or without ascorbate, as indicated in the text. Reactions were carried out in R buffer, at 22 • C, for 2.5 h.

In trans reactions
2 M DNA oligonucleotides were refolded in R buffer. 50 M 3-azido-7-hydroxycoumarin ('azide-coumarin', Jena Bioscience, Germany), from a fresh stock made in dimethylsulfoxide, was then added followed by a mixture of appropriate concentrations of CuSO 4 and sodium ascorbate in 10 mM Li-HEPES, pH 7.4 (for the Cu(I) reactions). For Cu(II) reactions, CuSO 4 alone (and, no ascorbate) in 10 mM Li-HEPES, pH 7.4, was added. The solutions were transferred to the wells of a 384-well Falcon plate (black, flat bottomed, from Corning) and incubated at 22 • C for 5 min. 5-Hexyn-1-ol ('hexynol') or propargyl alcohol was finally added to initiate the reaction. End-point fluorescence readings were carried out in a fluorescence plate reader (Infinite M200 Pro, Tecan). Excitation was at 403 nm, and emission was monitored at 480 nm.
For in trans reaction of lysozyme-(N 3 ) n (see below) with alkynated Alexa Fluor Dye 546 (≡-AFDye 54) catalyzed by CLICK-17, 20 M lysozyme-(N 3 ) n was mixed with 200 M ≡-AFDye 546, 4 M folded CLICK-17 DNA, along with copper salts (or not) with or without ascorbate, as indicated in the text. Reactions were carried out in R buffer, at 22 • C, for 2.5 h.
Nucleic Acids Research, 2020, Vol. 48, No. 13 7359 Lysozyme labeling with Azido-PEG2-NHS Ester 5 mg chicken egg lysozyme (Sigma-Aldrich) was dissolved in 950 l of 0.1 mM NaHCO 3 buffer, pH 8.3. 1 mg azide-PEG2-NHS ester (BroadPharm, San Diego) was dissolved separately in 50 l DMSO. The two solutions were mixed together and incubated for 2 h at 22 • C with gentle rotation of the tube. 50 mM ethanolamine was then added and the reaction solution incubated for a further 0.5 h at 22 • C to quench unreacted NHS ester. The reaction mixture was then passed through Mirocon-10k centrifugal filters (10k Da size cut-off, Cedarlane, Burlington). Two further washes, each with 400 l ddH 2 O, were carried out. The concentrated protein was dissolved in about 100 l 10 mM HEPES buffer (pH 7.4). The concentration of protein was determined by the BCA assay (BCA protein assay kit, Thermo Fisher Scientific, Toronto), and mass spectrometry was carried out to verify the azide-labeling.

Mass spectrometry
An Agilent 1200 HPLC couple to a Bruker maXis Impact Ultra-High Resolution tandem TOF (UHR-Qq-TOF) mass spectrometer was used. The software used was: Compass 1.5. The ionization mode used was positive electrospray ionization (+ESI), with gas temperature (180 • C); Gas Flow (8 l/min); nebulizer (2 bar); capillary voltage (4200 V); mass range (300-2500 Da). The calibrant used was: Agilent Tune Mix L. For HPLC, the column used was: Zorbax 300SB-C8 particle size 3.5 micron, 50 mm length × 2.1 mm diameter (Agilent Technologies). The column temperature was maintained at 30 • C. The HPLC Gradient Table is

Mass spectrometry of in cis ≡-CLICK-17 DNA-labeled lysozyme
Mass spectrometry was carried out in a Bruker maXis Impact Quadrupole Time-of-Flight LC/MS System. For unconjugated lysozyme-(N 3 ) n , 1 mg/ml solution was prepared in ddH2O, and loaded directly. For ≡-CLICK-17/lysozyme-(N 3 ) n conjugates, the DNA was first cleaved by Nuclease P1 (New England Biolabs, MA, USA) to leave one guanosine (the 5 -most deoxyguanosine of the CLICK-17 sequence) per conjugated DNA still attached to lysozyme. For mass-spectrometry, in cis conjugation of ≡-CLICK-17 and lysozyme-(N 3 ) n , 2 M ≡-CLICK-17 DNA, 2 M lysozyme-(N 3 ) n , were reacted together at 22 • C for 1 h in R Buffer supplemented with 4 M CuSO 4 and 100 M sodium ascorbate. The reaction mixture was passed through Microcon-10K centrifugal filter to desalt. The concentrated solution was diluted with ddH 2 O, supplemented with 1/10 volume of 10× nuclease P1 reaction buffer, and 0.1 l nuclease P1 per 50 l reaction. Digestion was carried out at 37 • C for 15 min, and the reaction quenched by heating at 75 • C for 10 min. The resulting solution was cleaned up by passing through a Microcon-10K centrifugal filter. After adjusting the sample volume to make the concentration of protein ∼1 mg/ml, samples were analyzed by mass spectrometry.

Cyclic voltammetry
The buffer solution for all CV measurements was: 20 mM HEPES, pH 7.4, and 20 mM MgCl 2 (R Buffer). The dissolved DNA was folded as described above. Prior to making CV measurements, CuSO 4 was added and rested for 5 min. All samples was degassed with argon. Cyclic voltammetry measurements were carried out in a three-electrode, single chamber glass cell with a CHI 1040A Electrochemical Analyzer (Austin, TX). A platinum wire and an Ag|AgCl|3 M NaCl electrode were used as counter electrode and reference electrode, respectively. A glassy carbon electrode was used as the working electrode. All CV measurements were performed in a Faraday cage at 22 • C, in R Buffer, which had been subjected to deoxygenation for at least 15 min. The scan rate for all CV measurements was maintained at 50 mV/s.

DNAzyme design and selection
The key design strategy for identifying catalytic DNA sequences out of a large random-sequence library is to select first for quasi-catalytic DNAs, that carry out a single catalytic turnover (i.e. work 'in cis'); and are in the process chemically modified (tagged) such that they can be sequestered away from non-active DNAs within the library. Our single-stranded DNA (ssDNA) pool was synthesized with a 5 -hexynyl attachment on each DNA ( Figure 1). These sequences were allowed to fold to form their distinctive secondary/tertiary structures in an aqueous buffer solution ('S Buffer': 50 mM HEPES, pH 7.4, 300 mM NaCl, 50 mM KCl, 20 mM MgCl 2 ), then supplemented with 5 M Cu + and freely diffusible azide-biotin for 2 h at 22 • C ( Figure 1). DNA sequences capable of catalyzing CuAAC in this very low-copper regime self-tagged with biotin, and were separated on that basis from the unreacted DNAs (the biotin 'tag' binds tightly to the protein streptavidin, which in turn retards the electrophoretic mobility of biotinylated DNAs in a gel). Supporting Figure S1 provides a schematic diagram for the entire SELEX procedure.
A total of 25 rounds in vitro selection were carried out. Two parallel selection experiments were conducted, one with Cu + as the catalytic cofactor, and another with THPTA/Cu + as the cofactor. The enrichment of CuAACactive DNAs through the selection rounds could be tracked by monitoring the overall conversion percentage of pool DNA in rounds 1-7 (Supporting Figure S2). The enrichment trends observed for both selections were roughly com-parable; by round 7, > 50% of the input 5 -hexynyl-DNA had reacted with the input azide-biotin within 2 h. By Round 25, pool enrichment had stabilized. Forty clones of the Round 25 pool were picked and sequenced, and the results are compiled in Supporting Figure S3. A number of the sequenced clones showed identical sequences, and five families of sequence were identified. Clones were picked from each family and prepared as 5 hexynyl-labeled-ssDNA ('≡-DNA'), then tested for in cis CuAAC-promoting activities. Testing was carried out in S buffer, used for the in vitro selection, and clone CLICK-17 was picked as a promising catalytic candidate. 5 -hexynyl-CLICK-17 ('≡-CLICK-17') was then studied intensively. It was found that ≡-CLICK-17's reaction kinetics in a relatively simple reaction buffer ('R buffer': 50 mM Li-HEPES, pH 7.4, 20 mM MgCl 2 ) were comparable to those in the higher ionic-strength S buffer. Consequently, all detailed characterization studies on ≡-CLICK-17 were carried out, except where explicitly stated to be otherwise, in R buffer.  Figure 2A). In all lanes (excepting lane 2), the reaction was terminated by ethanol precipitation, and the recovered and redissolved DNA was treated with streptavidin prior to loading into a non-denaturing PAGE gel. Identical values of 32 P counts were loaded into each lane. Biotinylation of DNA gave rise to two or more DNA bands of retarded electrophoretic mobility by virtue of streptavidin binding (indicated by brackets in Figure 2A; the different retarded bands represent different stoichiometries of these robust complexes--one streptavidin protein is capable of binding up to four biotinylated DNAs).

The in cis reaction of 5 -hexynyl-CLICK-17 DNA (≡-CLICK-17)
Although precisely equal amounts of DNA (corresponding to 32 P counts) were loaded in each lane, the different incubations show different levels of DNA in the different lanes of the gel. The first six lanes of the experiment show that in all incubations carried out with 100 M Cu + , significant degradation of the ≡-CLICK-17 DNA occurs, particularly in the absence of THPTA, presumably via ROS generation in the solution (4-6). Substitution of Li-HEPES by Tris as the buffering reagent lowers this degradation ('Tris buffer' lane in Figure 2A). The presence of magnesium is required in the reaction buffer, presumably for the catalytically relevant folding of ≡-CLICK-17. The absence of magnesium ions also leads to a high level of ≡-CLICK-17 degradation; presumably, Mg 2+ modulates the binding (and degradative effect) of Cu + to the DNA. Curiously, the complete absence of copper ('-CuSO 4 ') does generate a small amount of streptavidin-retarded product, presumably from very low levels of copper-free AAC catalyzed by the folded ≡-CLICK-17.
The most curious, yet reproducible, result is the high level of CuAAC product seen in the presence of 100 M CuSO 4 in the absence of any added ascorbate ('-sodium ascorbate', Figure 2A). This phenomenon was investigated in more depth, below.
The six lanes on the extreme right side of Figure 2A show the yield of the CuAAC product formed from ≡-CLICK-17 as a function of Cu + concentration. With as little as 200 nM Cu + , ∼50% biotinylation of ≡-CLICK-17 is reached within 0.5 h; with 1 M Cu + , conversion is ∼80% complete in this time frame. While DNA degradation is modest at ≤20 M Cu + , at 100 M Cu + , such degradation is heavy (as shown by the smears at the bottom of the gel).
To determine if the above results are specific to ≡-CLICK-17, as opposed to 5 -hexynylated ssDNAs in general, we carried out analogous experiments with a nucleotide-sequence permuted DNA isomer of ≡-CLICK-17 ('≡-PERMUT-17'; nucleotide sequence given in Methods). Figure 2B shows these results. Here, again, Tris buffer generates less DNA degradation than does HEPES buffer.  Figures S4 and S5). Supporting Figure  S4 shows that in the presence of as little as 5 M Cu + , the ≡-CLICK-17 reaction is 80% complete within 20 min; with 8 nM Cu + , >20% conversion occurs in 2 h. A striking observation about the Cu + -dependence of the ≡-CLICK-17 reaction is that highest catalytic activity is observed with ∼ 5 M Cu + . Reaction with 20 M Cu + is dramatically slower (slower than with 1 M or 5 M Cu + ). The most plausible explanation is that at ∼20 M Cu + , ≡-CLICK-17 forms an alternatively folded conformer that is either non-catalytic or poorly catalytic. Indeed, Figure 3, which plots in cis biotinylation rates for ≡-CLICK-17 and for ≡-PERMUT-17 for the first 16 min of reaction, shows that at 5 M Cu + , the ≡-CLICK-17 reaction is ∼2000-fold more rapid than the ≡-PERMUT-17 reaction. However, at 20 M Cu + , the ≡-CLICK-17 and ≡-PERMUT-17 rates are comparable, likely representing baseline CuAAC rates observable with 2 M of any 79-nt long ≡-ssDNA.
Supporting Figure S6 shows ESI mass spectrometry data that confirm that the streptavidin-retarded products on gels, such as shown in Figure 2A, are indeed the CuAAC adduct of ≡-CLICK-17 and azide-biotin. To facilitate ESI mass spectrometry of the conjugation product, the CuAACreacted mixture was subjected to complete digestion with Nuclease P1, an endonuclease that cleaves DNA into 5 -  To investigate whether folded CLICK-17 DNA can catalyze CuAAC in trans, we first explored suitable assays for making such a determination. We picked 5-hexyn-1ol ('hexynol') as a diffusible alkyne substrate on the basis of its similarity to the 5-hexynyl moiety within ≡-CLICK-17. However, azide-biotin here was not a particularly useful substrate, because its expected triazole product with hexynol would be a low molecular weight compound requiring specialized detection procedures. We therefore examined a structurally distinct, fluorogenic azide (3-azido-7hydroxycoumarin or 'azide-coumarin') as a potential substrate for CLICK-17. Experiments were carried out in R buffer with 2 M CLICK-17 DNA, 8 mM hexynol and 50 M (limiting) azide-coumarin. The progress of any reaction could be followed using fluorescence measurements (excitation at 404 nm and emission at 480 nm). A fluorescence calibration curve, quantitatively linking measured fluorescence to triazole product concentration, is shown in Supporting Figure S7.
Supporting Figure S8 shows that both azide-coumarin and hexynol were diffusible substrates acceptable to CLICK-17 for CuAAC catalysis in trans. The initial rate for the CLICK-17-catalyzed reaction with 5 M Cu + was 0. The above results suggest that for its catalytic action using Cu + , CLICK-17 likely provides privileged binding site(s) as well as significant redox stabilization for one or more catalytically important Cu + ion(s) within its active site. As the data on the non-catalytic PERMUT-17 DNA demonstrate, beyond such privileged binding, DNA sequesters and renders CuAAC-unavailable any residual Cu + in the solution under these reaction conditions. Not only does PERMUT-17 DNA kill CuAAC comprehensively; in addition, its likely non-specific sequestration of copper cannot be reversed or rescued by THPTA.
To optimize in trans catalysis by CLICK-17, we explored whether the inhibition imposed by higher (20 M) Cu + on the activity of 2 M CLICK-17 could be overcome by varying the CLICK-17 concentration in the fixed presence of 20 M Cu + . Supporting Figure S9 shows that, indeed, CuAAC levels obtained with 4 M CLICK-17 are >3-fold higher than those obtained with 2 M CLICK-17. The complex dependence of CuAAC rates as a function of CLICK-17 concentration (at fixed [Cu + ]) likely results from the distinct modes of Cu + binding to CLICK-17, i.e. tight binding to the active site as well as non-specific binding along the length of the DNA.

CLICK-17 is catalytic with added Cu 2+ in the absence of explicit reductants
The most striking observation from the in cis experiments with ≡-CLICK-17 (Figure 2A) was that in the presence of 100 M CuSO 4 , with no ascorbate (or other extrinsic reducing agent) added, CuAAC was still catalyzed to a high level. By contrast, under those same conditions, no in cis reaction was seen with the control alkynyl DNA, ≡-PERMUT-17 ( Figure 2B).
To further understand these observations, a thorough investigation was carried out with (unalkynated) CLICK-17 DNA, first, under in trans conditions. The Cu 2+ data shown in Figure 4B, however, were strikingly different. With 20 M CuSO 4 in the reaction (in the absence of added ascorbate), only CLICK-17 was able to catalyze the reaction (∼40% conversion after 2 h). All the other controls (Cu 2+ alone; Cu 2+ plus THPTA; and Cu 2+ plus PERMUT-17 DNA) failed to support the reaction.
A striking feature of the time-dependence of the CLICK-17/Cu 2+ combination is its sigmoidal rather than hyperbolic shape, indicative of an initial 'lag' or 'induction' phase. Such kinetics are consistent with the need to reduce Cu 2+ and the in situ generation of a sufficient concentration of catalytically obligatory Cu + , stabilized by CLICK-17, to promote CuAAC (a comparison of the relative efficacies of added Cu + versus Cu 2+ under optimized reaction conditions, where CuAAC proceeds to completion in 30-40 min, is shown in Supporting Figure S10).

In trans-and in cis-labeling of macromolecules (proteins) using the CLICK-17 DNAzyme
We wished to explore the utility of the CLICK-17 DNAzyme (i) to catalyze the conjugation of an azidoprotein in trans with an alkynated fluorescent dye; as well as (ii) to catalyze the in cis conjugation of a fluorescentlylabeled version of itself (≡-CLICK-17-Fluorescein) to the azido-protein. We wished to test for these activities under the very low Cu + and Cu 2+ concentration regimes that were shown, above, to be characteristic of CLICK-17's activity with small molecule substrates.
First, we labeled the protein lysozyme (14,305 Da) with Azido-PEG2-NHS Ester. The purified products were analyzed by LC-mass spectrometry (ESI). Supporting Figure S11 shows a near-complete conversion of the unlabeled lysozyme to three new products of MW 14,489, 14,674 and 14,859 Da, corresponding to derivatized lysozymes with one, two, and three appended azide functionalities, respectively (each label adds ∼185 Da to the protein). This mixture of azido-lysozymes (hereafter referred to as lysozyme-(N 3 ) 1-3 ) was then tested first, for in trans coupling with alkyne labeled Alexa Fluor Dye 546 (≡-AFDye 546).  of the two fluorescent bands in the gel as free ≡-AFDye 546 ('≡-Dye'). The upper fluorescent band, generated with 5-50 M Cu 2+ in the presence of CLICK-17 (but not with PERMUT-17 nor a 5-fold molar excess of THPTA), was therefore hypothesized to represent the protein-dye conjugate. Indeed, panel C, which overlaps the protein mobility data of panel A with the dye fluorescence data of panel B, shows that this hypothesis is correct. The gel band to the extreme right in all panels shows a positive control, CuAAC under standard conditions used in the field (100 M Cu + in the presence of 500 M THPTA). Although the yield of lysozyme-dye conjugate catalyzed by CLICK-17/Cu 2+ under these conditions is ∼40% of that achieved with the positive control, it must be emphasized that the CLICK-17 here is using Cu 2+ as its exclusive cofactor (in the absence of any explicit reductant). In the presence of Cu + , CLICK-17 performs comparably to the positive control. LC-MS analysis of the dye-lysozyme CuAAC conjugates obtained, as shown in Figure 5, are given in Supporting Figures S12 and S13.  folded structure, we see in Figure 6 (as also in Figures 2 and  3) that higher concentrations of Cu + (for example, 20 M in Figure 6) disfavor CLICK-17's catalysis. Again, this is because at such higher [Cu + ], CLICK-17 forms an alternative, less catalytic or non-catalytic fold.
We carried out ESI mass spectrometry analysis on the mixture of species formed in the above in cis catalyzed conjugation of ≡-CLICK-17 DNA with lysozyme-(N 3 ) 1-3 . Supporting Figure S14, panels A and B, show, respectively, schematic diagrams of unmodified lysozyme and lysozyme-(N 3 ) n (where n = 1). Panel C shows the expected in cis conjugated product of lysozyme-(N 3 ) 1 reacted with ≡-CLICK-17 (the full length of the conjugated CLICK-17 is shown as 'DNA'). To facilitate ESI mass spectrometry identification of this latter product, the DNA component of the protein-DNA conjugate was digested to completion with Nuclease P1, an endonuclease that cleaves DNA into 5deoxyribonucleotide monophosphates. The portion of the appended CLICK-17 DNA expected to be removed by the nuclease is indicated by a red bracket. The resulting residual structure incoporates the 5 -most deoxyriboguanosine residue of the CLICK-17 sequence conjugated to the lysozyme via the triazole and linker atoms, as indicated.

Specific nucleotide sequences and foldings are required for CLICK-17 and CLICK-16 catalytic activity
To further confirm that correct folding of CLICK-17 DNA is a key factor for its catalytic competence, we examined the catalytic activity of two other DNA clones from our original SELEX experiment, CLICK-16 and CLICK-20, both with sequences closely related to that of CLICK-17. Second, we tested a number of rationally designed mutants derived from CLICK-17, CLICK-16, and CLICK-20. Third, we subjected the nucleotide sequences of CLICK-17, CLICK-16 and CLICK-20 to the folding algorithm, Mfold, to try and correlate catalytic activity with their predicted folded structures (31).
Supporting Figure S16, upper, shows the nucleotide sequences of the SELEX-derived DNA clones CLICK-17, CLICK-16 and CLICK-20, as well of the three mutants, CLICK-17 T24G, CLICK-20 C39G and CLICK-20 G24T. 77-nucleotide long CLICK-16 differs from 79-nt CLICK-17 in a single point mutation and in lacking two adjacent thymines present in CLICK-17. 79-nt CLICK-20 varies from CLICK-17 in having three non-adjacent point mutations. The three further sequences tested restore some of the nucleotide differences between CLICK-17, CLICK-16 and CLICK-20, listed above. Supporting Figure S16, lower, shows that these oligonucleotides are notably different in their ability to utilize either Cu + or Cu 2+ to catalyze CuAAC in trans. Specifically, CLICK-16 is catalytic, though less so than CLICK-17; while CLICK-20, CLICK-20 C39G and CLICK-20 G24T are inactive under these reaction conditions. These data further emphasize that CLICK-17's catalytic properties are precisely dependent on its nucleotide sequence and its concomitantly folded structure. The nucleotide sequences of the catalytic CLICK-17 and CLICK-16 as well as the non-catalytic CLICK-20 were subjected to the Mfold program used for predicting the most stably folded structures of single-stranded DNAs (31).
Supporting Figure S17 shows the thermodynamically most stable predicted folds for the three DNAs. It is interesting to note that the catalytically active CLICK-17 and CLICK-16 share a common fold, whereas the inactive CLICK-20 shows a divergent folded structure.

Investigations into the mode of action of CLICK-17
The data shown above cumulatively encourage a hypothesis that CLICK-17 is a DNA enzyme with a precise, catalytically active fold, and an active site that offers one or more privileged bindings sites for Cu 2+ or Cu + . It is hypothesized that such a binding site/sites powerfully impacts the redox potential of Cu 2+ /Cu + in a positive direction, i.e. toward Cu + . Therefore, CLICK-17 shows catalytic activity (both in cis and in trans) with added Cu 2+ alone, in the absence of any explicit reductant such as ascorbate. Conceivably, one or more Cu 2+ ion/ions bound to the active site are then relatively easily reduced to the CuAAC-capable Cu + species not by a classic reductant such as ascorbate, but an available buffer solution component, such as HEPES.
To test, first, the redox potential of copper ion bound to CLICK-17, we carried out cyclic voltammtry (CV) experiments. Figure 7 shows the results. The presence of the DNAzyme, CLICK-17, significantly influences the redox property of the Cu(II)/Cu(I) couple in the HEPES buffer. While there are no discernible changes in either the reduction or the oxidation peak with added THPTA, the oxidation peak appears 301 mV more positive when CLICK-17 is introduced to the electrolyte. This corresponds to a ∼150 mV shift in the formation potential of Cu(II)/(I), indicative of the much improved stability of Cu(I). In addition, the much larger separation between the oxidation and reduction peaks indicates slower electron-transfer rates between the two forms as well. In comparison, the presence of the folded, uncatalytic DNA strand (PERMUT-17) induces no such substantial changes to the CV, i.e., the formal potential shift is not as significant; and, the oxidation peak becomes broad. The above observations are consistent with the binding of Cu(I)/Cu(II) species to the CLICK-17 DNAzyme as being both strong and unique.
To investigate whether alkyne homocoupling might be responsible for our observed CLICK-17/'Cu 2+ ' catalysis, a close investigation of the in cis reaction of ≡-CLICK-17 was made. In the in cis reaction, the only alkynes present are the hexynyl moieties covalently appended to CLICK-17 DNA. If alkyne homocoupling were indeed occurring, two ≡-CLICK-17 strands would be joined end-to-end to yield a product ('CLICK-17-≡-≡-CLICK-17') of twice the length and molecular weight of 2 x CLICK-17 (strictly, a molecular weight of [(2 x CLICK-17) -2]), which should show a predictable electrophoretic mobility that should be observable in a denaturing polyacrylamide gel. Figure 8, panel A, shows that in the presence of either Cu 2+ or Cu + , only streptavidin retarded bands but no additional DNA band corresponding to a 158-nt DNA product is present in the gel (lanes 4-6). DNA bands of very slightly higher molecular weight, running just above the unmodified ≡-CLICK-17 bands, are discernable with both Cu 2+ (lanes 4-6) and Cu + (lanes 10-12)--this is the triazole product formed by ≡-CLICK-17 with azide-biotin.
We have shown that binding to CLICK-17 DNA (but not to PERMUT-17 DNA) dramatically shifts the redox potential of Cu 2+ /Cu + towards Cu + . There nevertheless remains the requirement of an electron source for the reduction. It is clear that a powerful reductant such as ascorbate is not required by CLICK-17/Cu 2+ . Figure 8, panel b, shows the effectiveness of different buffer solutions in promoting CLICK-17-catalyzed in trans CuAAC by Cu + as well as by Cu 2+ . It is evident that while the Cu + reaction works in all three buffers tested (albeit less well in Tris (8)), the CLICK-17/Cu 2+ catalysis is promoted by HEPES and MOPS but not at all by Tris. It has been reported that HEPES can act as a mild reductant (43)--thus, it is plausible that it acts as the mild reductant necessary to convert sufficient CLICK-17-bound Cu 2+ to the rquired CuAAC-active Cu + .
To address a potential role for alcohol oxidation (33) in this catalytic system, a number of our substrates were indeed alcohols (hexynol, propargyl alcohol). Nevertheless, deliberate supplementations of the in trans reaction with exogenous methanol (5% and 15%, v/v, respectively) led to less effective CuAAC rather than to rate acceleration (Supporting Figure S16, panel A). Crucially, the sigmoidal shapes of the rate profiles (and unchanging lag-times of 20-30 min) persisted irrespective of the alcohol content of these reactions, discouraging the hypothesis of a determining role for alcohol oxidation in CLICK-17's Cu 2+ reaction.
We also tested for whether any residual azide anion or a contamination of amines in the commercially synthesized azide substrates that we have used, could act as reductants for CLICK-17's 'Cu 2+ ' reaction. Supporting Figure S18 shows that supplementations of up to 20 M of either the N 3 − anion or of methylamine did not change the catalytic rate profiles of CLICK-17, nor did they change the characteristic sigmoidal shapes of these profiles.
Given that CLICK-17 is catalytic both in cis (as ≡-CLICK-17) and in trans and is able to accept structurally divergent azide substrates (azide-biotin, azide-coumarin, and azide-labeled lysozyme) and alkynes (hexynol, propargyl alcohol, Alexa Fluor Dye 546), we attempted to de- termine a putative K M value for one of our substrates, hexynol. Supporting Figure S19 shows the reaction profiles of Cu + /CLICK-17-catalyzed CuAAC with a fixed concentration (50 M) of azide-coumarin and varying concentrations (0.1-20 mM) of hexynol. It was concluded that the K M for hexynol with respect to CLICK-17 was >20 mM.

DISCUSSION AND FUTURE PROSPECTS
The data reported herein demonstrate that the 79-nt DNA oligonucleotide, CLICK-17, is a bona fide enzyme for the copper-dependent azide-alkyne cycloaddition (CuAAC) reaction. Folded CLICK-17 is a metalloenzyme, able of harness ascorbate-generated Cu + , as well as added Cu 2+ in the absence of ascorbate, with buffer components as likely reductants for CLICK-17's catalyze CuAAC. Curiously, guanine bases in DNA and RNA are relatively easily oxidized and are ready sources of electrons for such phenomena as the recently discovered intrinsic photoreactivation of cyclobutane thymine dimers within DNA (44)(45)(46)(47). It is therefore conceivable that one or more guanines bases within CLICK-17's sequence may also contribute to the Cu 2+ to Cu + reduction. This will be the subject of further investigation.
We show, using cyclic voltammetry, that complexation of copper ions to CLICK-17 DNA (but not to a sequence-permuted DNA isomer, PERMUT-17) perturbs the Cu(II)/Cu(I) redox potential, by approximately 150 mV relative to that of copper ions in the presence of either HEPES buffer alone or containing THPTA. These data are consistent with folded CLICK-17 DNA (but not folded PERMUT-17 DNA) providing one or more high-affinity and catalytically relevant copper ion binding sites within CLICK-17's active site.
CLICK-17 is CuAAC-active with Cu 2+ in reaction solutions buffered by the mildly reducing HEPES and MOPS but not by Tris. Likely, HEPES and MOPS are sufficiently reducing for the CLICK-17's bound (and redox-potential shifted Cu 2+ ion/ions--see above) but not for Cu 2+ bound to PERMUT-17 or to THPTA. Above, we also show that neither alkyne homocoupling nor adventitious reduction of Cu 2+ to Cu + by primary alcohols--two alternative mechanisms invoked for Cu 2+ -dependent CuAAC observed in other experimental systems that have been described in the literature, is responsible for the observed Cu 2+ -dependent activity of CLICK-17.
We do not yet know any details of the ligand environment or environments provided by CLICK-17. CLICK-17 has 79 heterocyclic nucleobases (of which 23 are guanines); and, at pH 7.4, 79 full negative charges. Several studies have examined the preferred copper-binding sites in DNA. Sigel and Sigel (29) have described that the guanine N7 and O6 positions, together with a negatively charged phosphate oxygen, provides a tight multidentate binding site for Cu 2+ (with estimated formation constant of ∼10 4 M −1 ). The binding of Cu + to DNA may be stronger yet; association constants of ∼ 10 9 M −1 have been reported (28). A viable cytosine-Cu + -cytosine complex has also been reported in DNA (30). However, the ligand environment in CLICK-17's active site for binding one or more Cu 2+ /Cu + ions may be yet more intricate, conceivably involving Cu + /Cu + or Cu 2+ /Cu + bimetallic binding sites such as have been observed elsewhere (34).
Future work will focus on generating and studying truncated pieces of CLICK-17 that may be still capable of efficiently catalyzing CuAAC with both Cu + and added Cu 2+ . These, and other studies will help us to fully understand the mechanism of CLICK-17.
We have shown that CLICK-17 itself, though it is a prototypical CuAAC enzyme, is not especially substratespecific, and will accept a variety of small molecule as well as macromolecular azide (azide-biotin; azide-coumarin; azide-labeled lysozyme) as well as alkyne (≡-CLICK-17; hexynol; propargyl alcohol) substrates. This substrate generality is an unexpected and peculiar virtue of CLICK-17. However, we do expect to use SELEX to obtain Cu 2+dependent, CuAAC-catalyzing DNA enzymes that are highly substrate specific. A well-established variant of SE-LEX, called 'counter-SELEX' has widely been used (48,49) to select for DNAzymes and ribozymes that show high substrate-specificity. Indeed, we have earlier used counter-SELEX with great success for obtaining ribozymes that utilize thiamin as a cofactor for decarboxylating ␣-keto acid substrates (50).
CLICK-17, as a prototypical enzyme for the CuAAC 'click' reaction, is both larger in size (79 nt; ∼26 kD) and more expensive than a small molecule ligand such as TH-PTA (∼0.5 kD). However, automated DNA synthesis in recent years has become both highly efficient and low-cost; so that the cost differential between the two is likely not as large as may be imagined. Over and above size and cost issues, CLICK-17 and other DNA enzymes are likely to offer the following unique advantages.
First, a stretch of DNA such as CLICK-17, synthesized in an automated DNA synthesizer, is easily derivatized (during synthesis) with multiple non-DNA functionalities, including either a variety or multiple copies of fluorophores, for instance. Given the high efficiency and very low concentration of either Cu(I) or Cu(II) required for the in cis reaction catalyzed by CLICK-17 (whereby ≡-CLICK-17 itself becomes attached to an azide-containing protein), for instance, a very high level of fluorophore and/or other functional group attachment to the protein should be possible. Figure 9 a illustrates this concept. We have shown that azido-lysozyme, as a model protein (containing four disulfide linkages with potential copper-sequestering ability) can be efficiently labeled with multiple fluoresceinated CLICK-17 DNAs ( Figure 5) at very low M Cu 2+ (or Cu + ) concentrations.
It is anticipated that we will be able, using systematic deletion mutagenesis, to reduce the effective chain length of CLICK-17, by eliminating sequences that are extraneous to the catalytic sequence motifs present within the 79-nt CLICK-17. Figure 9B shows another of our conceptions of the potential usefulness of substrate-specific CuAAC DNAzymes, whereby different substrate-specific (as well as low Cu 2+ concentration-utilizing) DNAzymes could be used orthogonally. Thus, a 'DNAzyme 1' could substratespecifically couple, say, an azide-labeled cell surface determinant, 'a', with an alkyne-linked tag, 'c', exclusively. Whereas, a different DNAzyme, 'DNAzyme II', could couple a different cellular determinant, 'b', exclusively, with a tag, 'd' in the same reaction mixture. Selecting for such highly substrate-specific 'click' DNAzymes and ribozymes is a key goal for our own research.
It is our expectation that DNAzymes such as CLICK-17 may present a number of advantages over small molecule catalysts, such as THPTA, within a defined but restricted set of applications (such as operating in an aqueous medium and not organic solvents) yet presenting key benefits. Both the DNAzymes themselves and their catalytic activity In this schematic, the high substrate specificity that DNA enzymes are known to be capable of, is invoked. Here, a substrate-specific 'DNAzyme I' would specifically couple an azide-derivatized cell surface determinant ('a') with an alkyne-labeled tag, 'c', but not with an azide-derivatized cell surface determinant ('b') with an alkyne-labeled tag, 'd'. Conversely, a different enzyme, 'DNAzyme II' would exclusively couple an azide-derivatized cell surface determinant ('b') with alkyne-'d' and not 'a' with 'c'.
should be highly biocompatible (not least owing to the use of Cu 2+ instead of Cu + , and at such low concentrations as to be minimally injurious to cells). Thus, they should be particularly useful for CuAAC labeling of the surfaces of live cells as well as for labeling viral coats (such as on bacteriophages used for phage display). A major advantage we foresee for the utilization of very low--10-20 M--concentrations of Cu 2+ (and not Cu + generated in situ by a strong reductant such as ascorbate), is lower production (and lower concomitant damage from) ROS species. Being a macromolecule, the DNAzyme is furthermore likely to act as a superior sacrificial scavenger itself for any level of ROS species that may still be generated.