Toward conditional control of Smac mimetic activity by RNA‐templated reduction of azidopeptides on PNA or 2′‐OMe‐RNA

Oligonucleotide templated reactions can be used to control the activity of functional molecules based on the presence of a specific trigger sequence. We report an RNA‐controlled reaction system to conditionally restore the N‐terminal amino group and thus binding affinity of azide‐modified Smac mimetic compounds (SMCs) for their target protein X‐linked Inhibitor of Apoptosis Protein (XIAP). Two templated reactions were compared: Staudinger reduction with phosphines and a photocatalytic reaction with Ru(bpy)2(mcbpy). The latter proved faster and more efficient, especially for the activation of a bivalent SMC, which requires two consecutive reduction steps. The templated reaction proceeds with turnover when 2′‐OMe‐RNA probes are used, but is significantly more efficient with PNA, catalyzing a reaction in the presence of low, substoichiometric amounts (1%‐3%, 10 nM) of target RNA.


| INTRODUCTION
Templated reactions provide a powerful toolbox for the conditional control of chemical reactions and molecule activity. Utilizing the unique Watson-Crick-based recognition properties of short oligonucleotide probes, reactive groups of attached payloads can be arranged in spatial proximity, allowing a reaction to occur efficiently at low concentrations of reactants when the rate of non-templated bimolecular reactions is extremely low. A vast number of different nucleic acidencoded chemistries are available today, comprising ligations, bond cleavage reactions and functional group interconversions. [1,2] Although many research groups have designed templated reaction systems for a DNA-or RNA-controlled activation of fluorescent reporter dyes, much fewer examples have been reported that utilize oligonucleotide-templated reactions as a means to control the bioactivity of drug-like molecules. [3] This is due to a critical design constraint of such systems: the reaction must be able to create a pharmacophore, ideally from a completely inactive precursor. This is difficult to achieve in a single reaction step. Most templated drugrelease systems have therefore relied on dissociative chemistries that produce a bioactive molecule by bond cleavage, such as ester hydrolysis. [4][5][6][7][8][9] A disadvantage of these systems is that they are prone to premature fragmentation and not all drug candidates can be effectively inactivated by attachment to a cleavable linker. Few examples have demonstrated that activity can instead also be created from bond formation, such as acyl transfer. [10][11][12][13] Nucleic acid templated chemistries offer the possibility of turnover. [14][15][16] Reaction systems that allow a single template molecule to instruct the formation of several product molecules have been developed with the aim to increase the sensitivity of nucleic acid detection.
Such amplification systems could also be of interest for approaches in which nucleic acid templates control the bioactivity of a payload attached to nucleic acid molecules. In an effort to explore new options for templated drug activation, we considered Smac mimetic compounds (SMCs). These proapoptotic drugs are modeled after the N-terminal AVPI tetrapeptide motif of Smac, the natural antagonist of the Inhibitor of Apoptosis Protein (IAP) family, which is overexpressed by many human cancers. [17,18] Like their parent, SMCs bind to IAP baculoviral IAP repeat (BIR) domains to prevent interaction with target caspases and cause rapid autoubiquitylation and proteasomal degradation. X-ray and NMR-based structure-activity relationship studies revealed a central importance of their N-terminal amino group for SMC affinity: It is involved in multiple, essential hydrogen bridges with Asp-214, Glu residues 219 and 324 and Gln-319 residues of a small binding pocket in both BIR2 and BIR3 ( Figure 1A). [19][20][21] Mono N-methylation, which is often found in SMCs to enhance pharmacokinetic properties, is tolerated, but N,N-dimethylation, acylation or substitution with a methyl group increases the K d by several orders of magnitude, rendering the molecule inactive. [19,22] We speculated that this property could be exploited to put SMC activity under the control of an oligonucleotide-templated reaction. We envisioned that the introduction of an azide will enable a conditional "on switching" of SMC activity ( Figure 1B) by reduction to the amine. Azides accept but cannot donate hydrogen bonds and should, therefore, not be able to interact with the critical residues in the binding pocket. Moreover, azides are stable to conditions typically found in biological systems, offering prospects for a sequence-dependent "uncaging" of SMC bioactivity: Controlled by a conjugated oligonucleotide, the activating azide reduction would only occur in cells that express significant amounts of a specific template sequence. This could be used to directly counteract the effects of IAP overexpression by utilizing their mRNA to produce the inhibitor.
Herein we report the RNA-templated conversion of azidofunctionalized peptides to Smac mimetic molecules. We compared two reactions, the templated Staudinger reaction with phosphines and Ru 2+ complex-catalyzed photoreduction. Both reactions have previously been successfully applied in templated chemistries, to restore fluorescence of quenched fluorophores or to release bioactive molecules such as isopropyl-β-D-thiogalactopyranoside (IPTG) or estradiol via cleavage of p-azido(methyl)benzyloxycarbonyl linkers. [23][24][25][26][27] In contrast, we explore templated azide reduction as a way to unleash a peptide's N-terminal amino group without involving linkers. For recognition of the RNA template, the azidopeptides were appended to peptide nucleic acids (PNA) and 2 0 -OMe oligoribonucleotides. The comparison of both recognition units showed surprising differences in reactivity when the reactive probes were applied in excess to the RNA template.

| MATERIALS AND METHODS
Building blocks, resins and coupling agents were obtained from the commercial suppliers stated in the procedures or listed in the supporting information. Prior to coupling, L-2-azido propionic acid CHA salt (ChemPep, Wellington, Florida) was converted into its acid form by extraction from dichloromethane (0.2 M) and hydrochloric acid (0.5 M). HPLC-purified RNA templates and amino-modified 2 0 -OMe-RNA oligonucleotides were purchased from biomers.net (Ulm, Germany). Tris(2-carboxyethyl)phosphine dimethyl ester (dmTCEP) was prepared according to a literature procedure. [28] 2.1 | Protein expression and purification X-linked IAP (XIAP) L-BIR2-BIR3 protein (residues 120-356) containing an N-terminal 6x His-tag was expressed in Escherichia coli BL21(DE3) grown at 37 C in LB medium supplemented with Kanamycin (50 μg/ mL). Upon reaching an OD 600 of 0.6, protein expression was induced by addition of 0.4 mM IPTG and 100 μM ZnAc 2 for 20 hours at 20 C.

| Fluorescence polarization binding assay
Relative binding affinity (IC 50 ) was measured as described by Nikolovska et al. [29] Fluorescein-labeled reference binder Smac-1F was synthesized according to the procedure described in the literature. To each well of a black, nonbinding 96-well plate, 1 nM of Smac-1F and 15 nM L-BIR2-BIR3 construct in assay buffer (100 mM KH 2 PO 4 /K 2 HPO 4 , 110 mM NaCl, 5 mM DTT, 100 μg/mL bovine γ-globulin, 0.02% NaN 3 ) was added and mixed with different concentrations of Smac mimetics. The plate was incubated for 2 hours in the dark at room temperature and the anisotropy values were determined using a plate reader (Ex./Em. λ = 485/535 nm).

| Melting curve analysis
position.
The modified peptides were then tested in a fluorescence polarization-based competitive binding assay with a recombinantly expressed XIAP protein, comprising the BIR2 and BIR3 domains relevant for SMC binding (residues 120-356) and a fluorescent reference binder Smac-1F. [29] Gratifyingly, the H 2 N-and MeHNmodified sequences had similar affinities (Table 1, see also  Replacing the amine with an azide was only slightly less inactivating. Azido-SMC 3 showed an IC 50 value of 1.5 mM, while 7 had a half-maximal affinity of 0.6 mM. This confirmed that hydrogen bonding between the BIR domains and the N-terminal azide is not taking place. Conceivably, the larger size of N 3 vs NH 2 may also contribute to the inactivating effect, as the P1 binding pocket is relatively small. An additional experiment, in which 3 was treated with 10 eq. TCEP for 1 hour prior to the binding assay, verified that azido-SMC affinity to XIAP can be fully restored by Staudinger reduction ( Figure S2).
It should be noted that the binding affinity of a PNA-or 2 0 -OMe-RNA-appended SMC may be different to that of the free inhibitor, due to the steric bulk added by the oligomer. However, previous publications have demonstrated that such conjugates are indeed active and able to activate caspases in cell lysate [10,11] or induce apoptosis in human cancer cell lines. [31,32]

| Templated Staudinger reduction
Next, we prepared a set of PNA conjugates to test the reductive activation in a templated reaction ( Figure 2). PNA was used due to its high binding affinity, excellent sequence specificity and compatibility with solid phase peptide synthesis. [33,34] The probes were designed to be complementary to a stretch of mRNA coding for the X-linked Inhibitor of Apoptosis Protein (XIAP). This potent member of the IAP family mediates apoptosis resistance in many human cancers by directly inhibiting caspases 3, 7 and 9. [35] To increase solubility of the relatively hydrophobic conjugates, Lys and Glu residues were added to the N-or C-terminal end of the PNA, respectively. The combination of a basic and an acidic amino acid creates a net neutral zwitterion, which should not induce non-specific electrostatic interactions with the RNA template, but still mitigate PNA aggregation by providing additional options for solvation.
For the azido-SMC-bearing probes, the MV1-based sequence was used. It binds more strongly to XIAP and is potentially more metabolically stable due to the presence of unnatural amino acids. We explored both monovalent (N3) and bivalent ((N3) 2 ) azido-SMC-PNA conjugates. Dimerized Smac mimetics possess significantly higher affinity for IAPs than their monovalent counterparts, as they can concurrently interact with two BIR domains, mimicking the mode of action of their homodimeric parent Smac. The reduced form of (N3) 2 mimics the structure of BV6, a bivalent version of MV1. [30] As reducing probe, we tested dmTCEP, a more reactive and cell-permeable derivative of TCEP, [36] and the ruthenium complex Ru(bpy) 2 (mcbpy), both conjugated with PNA. [37] The latter acts as photoredox catalyst when excited by blue light (λ max = 452 nm), which adds an extra element of temporal control to the templated reaction that may prove useful in a further application of the reaction system in a biological context. Ru(bpy) 2 (mcbpy) also requires stoichiometric amounts of a sacrificial electron donor such as sodium ascorbate (NaAsc) or NADPH to be reactive. Winssinger and coworkers recently demonstrated that the cellular redox pool provides high enough concentrations of these molecules to allow Ru(II)-catalyzed reactions to proceed within living cells. [27,38] To reduce oxidation to the phosphine oxide, dmTCEP was coupled to the ε-amino group of 1 Lys under base-free conditions using DIC/Oxyma, while the ruthenium complex was introduced as NHS ester. To assess the role of proximity between reducing agent and azide, PNA probes were complementary to three different positions on the template, creating a gap of one to three unpaired nucleotides.
In the phosphine-based templated Staudinger reduction system, a twofold excess of reducing agent probe was used to allow full conversion of both monovalent and the bivalent azidopeptide conjugates N3 and (N3) 2 , respectively. On equimolar amounts of XIAP template, reduction of monovalent N 3 -SMC-PNA N3 was fast, yielding more than 50% product within 5 minutes regardless of the gap size ( Figure 3A and Table S1). Importantly, in the absence of RNA template only about 5% product was obtained after 6 hours. Comparing the maximum reaction speed v max , the RNA template accelerated the reduction of N3 by more than 700-fold. Two observations are noteworthy. First, after a fast, initial phase, the reaction kinetics slowed down. This may be due to the increased number of products poisoning the templated reaction by increasing the probability of unproductive alignments after strand exchange reactions. Second, the templated reaction did not reach full conversion and leveled at 70% to 80% yield. This is due to the emergence of two side-products ( Figure S3). Their amounts were too low to allow identification. We Comparing the different dmTCEP probes, the lowest total yield was obtained with the combination of 11 and N3. Although the larger gap size between the annealing sites of these probes on the template is an obvious explanation for this, it should be noted that the PNA: RNA duplex of 11 also possesses a higher T m (55.1 C) than 9 (46.2 C; Table 2, Figure S7). This can affect strand exchange reactions, as will be discussed later.
Next, we tested templated reactions of bivalent azidopeptide conjugate (N3) 2 . The templated reduction of (N3) 2 with dmTCEP-PNA 9-11 will first lead to an intermediate in which only one of the two azide groups is reduced. The H 2 N-SMC-N 3 -conjugate rapidly formed at an initial rate (v max = 2.72-3.03 nMÁs À1 ) comparable to reduction of the monovalent N3. Perhaps remarkably, the fully reduced (H 2 N-SMC) 2 was observed already in the early reaction phase although the rate of formation was slower than that of N3 (v max = 0.15-0.31 vs 1.80-1.98 nMÁs À1 ; Figure 3B and Table S1). This provides unambiguous evidence that strand replacement (of the dmTCEP-probes) occurs.

| Templated photoreduction
Ruthenium complexes such as Ru(bpy) 2 (mcbpy) ( Figure 1B) are suitable catalysts of azide reduction. As the ruthenium complex requires excitation and the presence of an electron donor to be active, the setup of the templated reaction was slightly modified. Samples were irradiated with a collimated LED lamp (1 W, λ = 455 nm) and 5 mM sodium ascorbate was added to the reaction buffer.
As already reported by others, [16,26,38,39] the Ru(II)-catalyzed reduction was faster than the Staudinger reduction (v max = up to 4.44 nMÁs À1 vs 1.98 nMÁs À1 ) and, in contrast to the phosphine reactions, proceeded well also at later reaction phases until 90% conversion within 20 to 30 minutes ( Figure 4A and Table S1). Conversely, in the absence of RNA template, only 2% to 3% product was formed after 90 minutes. A gap of a single unpaired nucleotide (12 + N3) produced the most efficient reaction (t 1/2 = 4.9 minutes), resulting in a relative speed increase over background of close to 900-fold. The reason for the slightly longer half-lives of Ru(II)-catalyzed reactions compared to the dmTCEP system, was the occurrence of a short lag-phase of 1 to 2 minutes in the beginning of the reaction. This was possibly due to the relatively low intensity of the LED light source used and thus time required to build up the reductively active Ru(I) species. In a separate experiment, we also verified the specificity of the templated reaction. When a scrambled azido-SMC probe scrN3 was used instead of N3, the same yield as in the nontemplated reaction was obtained ( Figure S6A). This confirmed that the catalytic effect of the RNA on the reaction rate is sequencedependent. 3.4 | Comparison of PNA and 2 0 -OMe-RNA PNA has frequently been used as a biostable recognition unit in nucleic acid-templated chemistry. However, the hydrophobicity of PNA conjugates can be problematic ( Figure S4). Most nucleic acid tools are based on oligonucleotide scaffolds. Phosphorothioate structures and 2 0 -O-methoxy ribonucleotides are frequently introduced into oligonucleotide tools to prevent degradation by nucleases. In a few studiesaddressing inhibition or localization of miRNA or mRNA was PNA compared side-by-side with metabolically stable oligonucleotide tools. [40][41][42][43][44] A comparison of different backbones in templated chemistry was, to our best knowledge, lacking and we, therefore, set out to explore the RNA-triggered photocatalytic reduction of azidopeptides on 2 0 -O-methoxy oligoribonucleotides scaffolds.
To maintain comparability between PNA-and 2 0 -OMe-RNA probes in templated reactions, they should possess the same affinity to the target RNA. To that end, the duplex melting temperatures of different 2 0 -OMe-RNA sequences were evaluated ( Table 2, see also Figure S9). An initial test with the same 9-mer sequences as used for the PNA, showed T m values more than 10 C lower (39.1 C and 33.7 C vs 49.8 C and 46.2 C).
By increasing the length to 12-mers, almost matching melting temperatures (50.1 C and 45.0 C) could be obtained.
The 2 0 -OMe-and PNA-based Ru(II) probes were compared in templated reactions at different stoichiometric ratios of RNA target (0.01-1.0 eq.). This should reveal potential catalytic differences as well as determine the maximum achievable turnover of each system. To facilitate turnover in template, the reactions were measured at 37 C.
When the reactions were performed at a template concentration that matched the concentration of the azidopeptide, both PNA and 2 0 -OMe probe performed identically, reaching their maximum conversion (about 90%) after 15 minutes ( Figure 5). Interestingly, at 37 C, the lag phase of the ruthenium complex was a bit shorter (t 1/2 = 3 minutes; Figure S6B). Also, the acceleration of the reaction over background by the template increased slightly, to around 1100-fold. When the amount of template was decreased, catalytic differences between the PNA-and 2 0 -OMe-RNA-based probes became apparent ( Figure 5, Figure S7). In the presence of 0.3 eq. template (300 nM), the PNA probes achieved nearly the same yield as with 1 eq. RNA (albeit slightly later; t 1/ 2 = 7.1 minutes), while the maximum conversion for the 2 0 -OMe-RNA probes dropped to 56%. With as little as 0.01 eq. (10 nM) of template RNA, the PNA-system still reached 15% conversion. At this concentration, the templated reaction of 2 0 -OMe-RNA probes provided only slightly more product than the non-templated background reaction, yielding only about 5% product after 60 minutes.
The experiments point to significant differences between RNAtemplated photoreduction of azidopeptides on PNA and 2 0 -OMe-RNA probes. First, the yields obtained with PNA probes are higher. Second, the reaction with PNA is faster, especially when photoreduction occurs at low template to azidopeptide ratios. One factor that allows the PNA conjugates to achieve better conversion than the 2 0 -OMe congeners is most likely a different stability of the Ru(II)-complexes.
The UPLC measurements revealed that significant amounts of catalyst degrade during irradiation, and this effect is more pronounced for the 2 0 -OMe-RNA-based probes ( Figure S8A). PAGE analysis of irradiated and nonirradiated samples ( Figure S8C) showed no fragmentation of the conjugate, so a photocatalytic cleavage of the phosphodiester bonds, as has been reported for other Ru complexes, [45,46] can be ruled out. Instead it is likely an oxidative process (a mass difference of +16 could be detected for one of the degradation products; Figure S8B). Ruthenium polypyridyl complexes have been shown to cause oxidative damage to nucleic acids, particularly guanine, [47,48] of which the longer 2 0 -OMe-RNA probe has one more than the PNA. It is also conceivable that the latter generally has lower susceptibility to oxidative damage, as PNA has been found able to suppress photooxidation of a hybridized DNA. [49] However, the differential stability of the Ru(II)-complexes cannot account for the differences in reaction rates at early phases because the amount of intact Ru(II)-strand remains identical within the first 10 minutes of the reaction ( Figure S7A). While the initial rates on one equivalent template are similar, reactions on PNA occur at twice the speed when the photoreduction is performed under turnover conditions. This indicates a different strand exchange behavior. To achieve a similar T m , 2 0 -OMe-RNA probes are longer than PNA probes. The dissociation kinetics of the longer 2 0 -OMe-RNA probes might be slower, since the incremental affinity loss induced upon end fraying (caused by the dissociation of a base pair at the terminus) is smaller than with the shorter PNA probes, which provide higher stability per base pair. It may also be that 2 0 -OMe-RNA oligonucleotides have generally different hybridization kinetics than PNA. In fact, comparatively low k off values have been reported for other 2 0 -ribose modified oligonucleotides. [50] Regardless of the causes, the comparison between PNA and RNA probes points to the special properties of PNA that indicate a particular suitability for template-controlled reactions. In addition, PNA generally has a low tolerance for base mismatches. Though not subject of this study, we refer to previous work of our group in which we have demonstrated the high sequence fidelity provided by templated reaction of short PNA probes. [51][52][53][54][55][56][57][58] 4 | CONCLUSIONS In summary, we have developed a system for an RNA-induced reductive activation of azide-modified Smac mimetic peptides. The comparative study revealed that the photoredox catalyst Ru(bpy)(mcbpy) 2 provided higher rates and higher reaction yields than the phosphine based reducing agent dmTCEP. Using azidopeptide conjugates and RNA template in equimolar ratios at 37 C, up to 90% yield of acti- in vivo. [27,38] Although PNA is more difficult to deliver into cells than negatively charged 2 0 -OMe-RNA, this limitation may be overcome with physical transfection techniques, like electroporation. An alternative could be positively charged guanidinium-PNA, which other groups have successfully used for probe delivery. [28,59,39]

ACKNOWLEDGMENTS
The authors thank Prof Shaomeng Wang for providing the XIAP L-BIR2-BIR3 plasmid. This work was funded by the European

CONFLICT OF INTEREST
The authors declare no competing financial interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study (UPLC-MS traces, binding data, kinetics and melting curves) are available in the supplementary material of this article. Raw data are available upon request from the authors.