“Bind, cleave and leave”: multiple turnover catalysis of RNA cleavage by bulge–loop inducing supramolecular conjugates

Abstract Antisense sequence-specific knockdown of pathogenic RNA offers opportunities to find new solutions for therapeutic treatments. However, to gain a desired therapeutic effect, the multiple turnover catalysis is critical to inactivate many copies of emerging RNA sequences, which is difficult to achieve without sacrificing the sequence-specificity of cleavage. Here, engineering two or three catalytic peptides into the bulge–loop inducing molecular framework of antisense oligonucleotides achieved catalytic turnover of targeted RNA. Different supramolecular configurations revealed that cleavage of the RNA backbone upon sequence-specific hybridization with the catalyst accelerated with increase in the number of catalytic guanidinium groups, with almost complete demolition of target RNA in 24 h. Multiple sequence-specific cuts at different locations within and around the bulge–loop facilitated release of the catalyst for subsequent attacks of at least 10 further RNA substrate copies, such that delivery of only a few catalytic molecules could be sufficient to maintain knockdown of typical RNA copy numbers. We have developed fluorescent assay and kinetic simulation tools to characterise how the limited availability of different targets and catalysts had restrained catalytic reaction progress considerably, and to inform how to accelerate the catalytic destruction of shorter linear and larger RNAs even further.

14. Figure Table S1. Sequences and millimolar extinction coefficients values for peptides, unconjugated oligonucleotide and labelled linear target used for this study.

Figure S3. MALDI-ToF mass spectra Acetyl-[LR]3G-CO2H
Supplementary Figure S3. MALDI-TOF spectra of peptide Acetyl-[LR]3G-CO2H Spectra were recorded using a Bruker Daltonics Ultraflex TOF/TOF mass spectrometer. Figure S4. 1  6. Figure S5. General synthetic route of "bis" and "triple" conjugation Supplementary Figure S5. General synthetic route of conjugation. (A) Synthetic route for the production of "bis" conjugates, demonstrated by the example of BC5-L-ββ conjugate synthesis, when the β configuration was used for aminohexyl linker attachment at C1' position of abasic nucleotides. A similar synthetic route was used for the synthesis of BC5-L-αα, BC5-L-βα and BC5-L-αβ conjugates, when the  configuration was also used for aminohexyl linker attachment at the C1' position of abasic nucleotides. (B) Synthetic route for the "triple" conjugate BC5-L-βββ. Antisense oligonucleotide containing two or three internal abasic nucleotide either in  or β configuration was conjugated to catalytic peptide via aminohexyl linker. To avoid peptide self-conjugation and cyclisation during the amide coupling reaction, peptide was acetylated at the N-terminus. 7 9. Figure S8. 1 H NMR spectra of the bulge-loop inducing "bis" and "triple" conjugates Supplementary Figure S8. 1 H NMR spectra (400 MHz, Bruker Avance IIþ 400) of "bis" and "triple" conjugates indicating the prominent chemical shift of protons from oligonucleotides, peptide, aminohexyl linker and acetyl protecting group. In each spectrum, the breakdown of proton assignment for each region as well as integral intensity has been indicated. H3' region has not been assigned because water suppression prohibits full assignment.

5.
10. Figure S9. 1 H NMR comparison of "single", "bis" and "triple" conjugates Supplementary Figure S9. 1 H NMR spectra (400 MHz, Bruker Avance IIþ 400) of BC5-L-β "single" conjugate (A) BC5-L-ββ "bis" conjugate (B) and BC5-L-βββ "triple" conjugate (C). The prominent chemical shifts corresponding to oligonucleotides, peptide, aminohexyl linker and acetyl protecting group can be clearly seen. Careful integration of the oligonucleotide aromatic region as well as H1' sugar ring protons confirmed 1:1, 2:1 and 3:1 stoichiometric ratio of peptide to oligonucleotide in "single", "bis" and "triple" conjugates. Representative examples of the predicted complexes between the conjugate recognition motif and various 26-nt RNA regions from 5'-part of the tRNA Phe . Secondary structures and corresponding thermodynamics parameters (ΔG, ΔH, ΔS and Tm) were calculated using DINAMelt Server (www.unafold.org). The conjugate recognition sequences (TGGTGCGAATT-dR-dR-GATCGAACACAGGAC and TGGTGCGAATT-dR-dR-dR-GATC-GAACACAGGAC) were screened against every possible 26-nt section of the 5'-part of tRNA Phe by "sliding" it along this sequence with 1-nt incremental step to fully cover the region between 5'-terminal 1 G residue and 50 C nucleotide located in the beginning of TΨC-arm, as indicated by red trace and blue arrows. Labelling is shown in red for RNA and in blue for conjugate. 14. Figure

Model of target and catalytic conjugate interactions to explain atypical kinetics observed
Atypical kinetics of F-Q-RNA substrate cleavage by catalytic conjugates BC5-L-, BC5-L-β and BC5-Lββ observed under multiple turnover conditions (see Figure 7 and Table 2) could not be explained by Michaelis-Menten enzyme catalysis. The plotted correlation between the measured initial reaction velocity against target substrate concentration showed significant deviation from any hyperbolic curve consistent with Michaelis-Menten reaction model. New kinetic principles might need to be elucidated with a novel catalyst. We first consider whether side interactions of the substrate, catalyst and product can explain these behaviours within an otherwise Michaelis-Menten model, where the catalytic complexation of available F-Q-RNA substrate to the conjugate is considered to be in competition with weaker self-associations (Supplementary Figure S19), which both inactivate the catalytic conjugate and sequester substrate into unavailable complexes.
Indeed, when substrate depletion was considered, by subtraction of the observed increase in product, the decline in substrate appeared to be considerable during the progress of the reactions, at least for substrate excesses greater than 2-fold (Supplementary Figure 20 (A-C)). Such an extent of substrate depletion would be expected to exert a greater effect on the decline in velocity during the course of the reactions, which was not apparent in the progress curves in their relatively small curvatures (Figure 7 (A-C)) and non-linearity factors (Figure 7 (D-F)). This observation was explained by the possibility of self-association of the target substrate and conjugates into stable hairpin structures via intramolecular folding (Supplementary Figure S19). Such an explanation considers below the equilibrium dissociation constants (K) of each dominant interaction.

Model
The binding of free peptidyl-oligonucleotide catalytic conjugate (C) to free target substrate (S) and the formation of a cleavable substrate-catalyst complex (CS) may be considered, as a first approximation, in terms of a single-site competition, characterised by their respective self-association (see Supplementary Figure S19) and dissociation equilibria, which are dynamic as catalysis reactions progress. In this model (see Figure 8 in paper), the target substrate associates into a complex ( ), unavailable for binding to the catalyst, and the catalyst associates into inactive complex ( ), unable to attack the target substrate.
In order to determine whether such a model, operating alongside Michaelis-Menten kinetics, can simulate the observed velocities, we first considered the fractions (denoted as [concentrations] in square brackets) involved in each self-interaction separately, in isolation from each other, and then in competition according to well-established principles described, for example, in (2, 3).
Substituting into (7) from (5) The target substrate and catalytic conjugate were purified into respective water solutions and lyophilised, where the removal of water into ice crystals before sublimation will concentrate and promote their self-complexations (Supplementary Figure S19). With diminished water, and thereby ionic and hydrogen bonding, hydrophobic bonding within complexes will dominate. When reconstituted back into water and low ionic strength reaction mixtures, the positively-charged peptide in the structure of the catalytic conjugate is also likely to interact with the negatively-charged oligonucleotide (4) to reinforce self-complexations further.
Inactivated complexes of catalytic conjugates are likely to be activated, especially by binding of available substrate, initially to smaller exposed regions, with increase in activation facilitated by rise in substrate concentration. Competitive activation of catalytic conjugate by the substrate available from self-complexation can similarly consider the common conjugate available after their interaction at equilibrium, as above from (2, 3).

Calculation of fractions expected from this model
The expected target substrate complexed [ ] was calculated from input of its equilibrium dissociation and the known total substrate concentration [ ] initially in isolation (eq. 4), when the available substrate [ ] was obtained by difference (eq. 6). The complexed conjugate was also calculated initially in isolation from an approximated equilibrium dissociation (eq. 2), when the activation of conjugate could be similarly calculated from an initially approximated equilibrium dissociation in competition with the initially-available substrate [ ] to provide the activated concentration of conjugate available [ ] for each reaction. Initially approximated values were refined within the non-linear numerical fitting of the model below to simulate the observed data sets of every reaction sample for each conjugate catalyst.
The full binding of substrate and weaker competitive partial binding of substrate to the conjugate (Figure 8 in paper) were calculated from their input equilibrium dissociations and (respectively), when partial binding of two substrate molecules to a conjugate ("double substrate" complex with ) at higher substrate concentrations was also considered. As product formation was observed at each time point sampled, the corresponding depletion of substrate and the effect on the complexation of substrate were calculated to provide the dynamically-available substrate as the reaction progressed. The product bound to the available conjugate in competition with available substrate was also calculated for each time point from its dissociation , as its observed equilibrium concentration dynamically changed as the reaction progressed.
Formation of the cleavable substrate complex [ ] by full binding of the substrate to the different catalytic conjugates has dissociation characteristics reported here and earlier (5). Remaining dissociation characteristics were approximated initially as those expected from the dominant binding. The above calculations were completed alongside Michaelis-Menten kinetics for initial velocities 0 and initial available substrate [ ] 0 including the active conjugate [ ] (in the = [ ] × term): Non-linear numerical fitting (via 'Solver' add-in for Excel) was used to vary initial approximations of , and to refine approximated equilibrium dissociation characteristics ( , , ) by iteratively calculating the resulting expected velocities, until minimised against the observed velocities (i.e. minimum sum of squared differences between expected and observed velocities). All of the estimates of fractions of complexation expected were thereby calculated for each reaction time course, which are presented below on a log scale to discern also smaller-scale changes for three conjugate catalysts (Supplementary Figure S21). Substrate available (dark blue trace, Figure S21). Despite consumption of substrate into product, in this model, the fraction of "Substrate available" will not decline to the same extent and remains relatively steady. Such steady levels particularly at higher initial substrate concentrations were related by the replacement of "Substrate available" from complexed substrate as total substrate was depleted. This buffering of "Substrate available" manifests as slightly increasing fractional levels of "Substrate available" as substrate was consumed at lower initial substrate concentrations.
Active conjugate (purple trace, Figure S21). Of the total conjugate (5 μM in all cases), only a minor fraction (~10%) was available at small excesses of substrate and fractional levels declined during the course of the reaction, as activating substrate (at small excesses) declined with less buffering from complexed substrate. As the initial substrate concentration is increased, progressively greater proportions of the total conjugate will be activated in this part of the model. However, activated conjugate will still become inactivated at higher substrate concentrations (see below in Figures S22  and S23, not shown in the above Figure S21). Substrate and product-occupied conjugate. Cleavable substrate dominated the occupancy of the conjugate (grey trace Figure S21) due to formation of the "full-size", perfect-match complementary complex. As substrate is cleaved, the occupancy of the conjugate by product will rise but, given ~10× greater equilibrium dissociation constant, there will be less than 10% occupancy (Figure S22 (A-C) and orange-red trace, Figure S21). Products of cleaved substrate will have similar dissociation to partial binding of substrate in this model, and cleaved product leaves part of the conjugate unoccupied for the partial binding of substrate (S) (Figure S21). Partial binding of substrate ("Single half S occupancy", green trace, Figure S21) will compete with and follow the occupancy of product. However, growing with large excesses of substrate, some of the conjugate will become doubly-occupied by a pair of substrate molecules (see Figure 8 in paper), each partially bound with the other respective unbound portions of the substrates dangling. However, such "double substrate" occupancy of the conjugate (Figure S22 (D-F)) will decline in this model in competition with occupancy of the conjugate by product (Figure S21 (A -C), which increases during catalysis and will leave only a partial binding site for substrate. The rate of decline in double substrate occupancy was greater at lower substrate concentrations than higher substrate concentrations, when "double substrate" occupancy is retained for longer periods of reaction, and remained elevated at the higher substrate concentrations, where the conjugate was also prone to inactivation (Figure S22 (D-F)).
The two dangling ends of the unbound substrate molecules and their two or more dangling positivelycharged peptides in the structure of the catalytic conjugate are likely to promote agglomeration and inactivation of the conjugate, which would increase with the greater extent and longer duration of "double substrate" occupancy at higher substrate concentrations (Figure 8). The lesser rate of decline of 'double substrate' occupancy with increasing substrate concentrations was therefore used as the key determinant of inactivation in this model, where a power function of the more parabolic nature of agglomeration was estimated as best fit to the rate of "double substrate" occupancy with available substrate concentration.

Conjugate activation and inactivation model
The above characterisation of activation and inactivation interactions was used as a model to simulate the active conjugate concentration and the inactivated concentration of conjugate, as they vary with the dynamic changes in the available substrate concentration. The underlying equilibrium dissociation characteristics were refined numerically to produce the best fit (i.e., minimum sum of squared residuals) between the expected velocities calculated from this model and the velocities observed experimentally. The resultant sum of this activation and inactivation gave the active conjugate concentration ( Figure  S23 (A-C)). The model defined by the refined equilibrium dissociation characteristics gave the closest fit to the observed initial reaction velocities from non-linear fitting (Figure S23 (D-F) continuous purple curve). Residuals were raised slightly for 'single' conjugates at low substrate excesses (Figure S23, bottom row), suggestive of another lesser factor here, which had not been accounted for within the modelled complexity.

Competitive and uncompetitive inhibition
Michaelis-Menten kinetics were considered alone with competitive and uncompetitive inhibitions by product and substrate using Equation 12 (1): but there was no possible fit to the observed velocities, particularly when initial velocities were estimated by non-linear fitting. The hyperbolic nature of Michaelis-Menten kinetics alone was unsuited to the peculiar skewed bell-shaped or parabolic kinetics observed, where combination with conjugate activation and inactivation produced a much better fit ( Figure S23).
Consideration of product inhibition proved more fruitful when combined within the conjugate activation and inactivation model. Product inhibition was considered by the increases in product concentrations during the progress of each of the catalytic reactions, where velocities were approximated from the initial velocity 0 and non-linearity factor  as described by using Equation 13 (1): The velocities estimated for the sampled points during reaction progress were then compared to expected velocities calculated from using Equation 14 (1): where the active conjugate [ ] was used together with the changes in available substrate [ ] and product [ ] concentrations during the time (t) course of the reactions ( Figure S24). Although the available substrate declined less during the course of the reactions than the total substrate, at lower substrate concentrations, there was a greater decline in available substrate as product was produced.  (Table S2).
At low substrate concentrations, the reaction velocities increased during reaction progress for the conjugate catalysts with a single peptide (BC5-α, BC5-L-β), but velocities declined for the "bis" conjugate with two catalytic peptides (BC5-L-ββ). Although the velocities expected from Michaelis-Menten kinetics were in the same region, they did not show increases during reaction progress ( Figure  S24 red and pink traces). However, at higher substrate concentrations, available substrate buffering maintained less change in available substrate during the time course ( Figure S21). For these higher substrate concentrations, the velocities expected from the model of complex interactions around Michaelis-Menten kinetics, with product inhibition included in the model, approached those estimated from the reaction progress curves with relatively small residuals ( Figure S24).
Attempts were made to determine any competitive product inhibition using non-linear fitting by minimising residuals (sum of squared differences between estimated and expected velocities), but competitive inhibition was consistently insignificant with large values (>>100,000). However, uncompetitive product inhibition had consistently significant values in the region of 18-21 μM, a little less than the values 20-23 μM (Table S2). Similar values may be expected as the same oligonucleotide binding sequence is used in all cases, with changes in the way the catalytic peptide is attached and number of peptides being the main variable. Table S2. Kinetic characteristics input and output from the model described. The equilibrium dissociation constants for the cleavable substrate complex was input from the reciprocal of the equilibrium association constant estimated experimentally. The binding affinities of cleaved product and half substrate were assumed to be similar and input at the same value. All other kinetic characteristics were estimated by non-linear fitting. The principal difference in kinetics arising from the structural differences in the conjugate catalysts was in or the catalytic turnover . As may be expected, the conjugates with a pair of catalytic peptides had much higher reaction turnover (0.8 per hour), whereas, for single catalytic peptides, the  anomer produced the highest turnover (0.6 per hour), double that of the elongated β anomer (0.29 per hour).