Nanomole‐scale photochemical thiol‐ene chemistry for high‐throughput late‐stage diversification of peptide macrocycles

The photochemical thiol‐ene reaction is an efficient method for rapid and chemoselective formation of thioether linkages under mild conditions. It has found widespread use in small‐molecule synthesis as well as peptide and protein chemistry. While high‐throughput experimentation provides an invaluable tool for drug discovery, the considerable potential of the thiol‐ene reaction remains unexplored in this context. Herein, we report the development of nanomole‐scale photochemical thiol‐ene chemistry, performed using an automated approach in 1536‐well plates in a cost‐efficient custom reactor. Through careful reaction design and selection of reactants, this chemistry is applied to the lateral diversification of peptidic macrocycles to yield high purity crude mixtures. Selection of the photoinitiator 2,2‐dimethoxy‐2‐phenylacetophenone yields volatile breakdown products, which facilitates removal in vacuo. We demonstrate the use of this approach in late‐stage diversification of peptidic macrocycles with 96 examples averaging 95% conversion to the desired product.

or reductive release followed by cyclization with bis-electrophilic reagents to yield dithioether macrocycles (Figure 1a). [10]eviously, the use of solution-phase acylation of macrocycles containing diamino acids for diversification in high-throughput has been reported, enabling the synthesis of targeted libraries of small peptidic macrocycles to furnish binders against challenging targets. [4]wever, this approach is limited in scope regarding other two-electron reactive groups such as amines and carboxylic acids that are found in peptides.Additionally, the use of serine/threonine ligation was recently reported for similar applications, though this of course necessitates either a serine or threonine residue with a free amine. [11]rther, both of the aforementioned methodologies introduce additional polar groups, which may hamper membrane permeability of any lead compounds produced.As such, we sought to use chemoselective radicalmediated processes in the form of the photochemical thiol-ene reaction.
The thiol-ene reaction is of increasing interest in peptide and protein chemistry, having been applied to both conjugations and cyclizations, [8,12] as well as the synthesis of chemical probes. [13,14]Thiol-ene chemistry is tolerant to a wide variety of functional groups that may otherwise be reactive in other two-electron or redox chemistry, and therefore has been applied to a wide range of peptidic structures.Furthermore, thiol-ene has been described as a 'click-reaction', with high yields, few side-products and aqueous compatibility, all ideal for application to peptidic substrates. [15]These characteristics make thiol-ene chemistry an attractive approach for late-stage diversification.
However, high-throughput photochemistry has shown significant challenges in the past [16] including the development of a suitable and affordable photoreactor.In this work, we describe the development of nanoscale thiol-ene chemistry in high-throughput experimentation (HTE) format compatible with microtiter plates and automated liquid handling (Figure 1b,c).Following adaptation of the reaction to 100 μL volumes in 96-well plates, the reaction was further downscaled to single-digit μL volumes, enabling diversification of peptidic macrocycles at nanomole-scale.This methodology can be applied either to small numbers of examples in microlitre volumes or may be applied in a HTE context in 1536-well plates.To the best of our knowledge, this serves as the first reported example of the use of automated photochemical thiol-ene chemistry in 1536-well plates at microlitre volumes.The reaction proceeds in high-to-complete conversions with minimal side-products, while the volatility of the photosensitiser degradation products and other additives gives products with high crude purity suitable for direct further application.

| Materials
Reagents and solvents used were of analytical grade and were each used as obtained from commercial suppliers without further purification.1536-well cyclic olefin copolymer (COC) plates were obtained from Beckman Coulter (cat.#001-6969).

| Cyclisation of dithiol peptides using bis-electrophiles
Crude peptides were dissolved to 1 mM concentration in freshly prepared and degassed NH 4 HCO 3 buffer/MeCN solution (60 mM, pH 8.0) and the desired bis-electrophilic linker (2.0 equiv.assuming 50% yield from SPPS) was added.The reaction was agitated for 1 h or until observed as complete by LC-MS analysis.The reaction was then quenched by addition of β-ME (8.0 equiv.)and the macrocyclic product purified by preparative RP-HPLC.

| Optimised procedure for high-throughput thiol-ene 'click'
Peptide stock solutions (40 nmol, 1 μL from 40 mM solution in DMSO) were dispensed using a LabCyte Echo 650 acoustic dispenser into 1536-well plates, centrifuged and concentrated via RVC to give dried peptide pellets.2,2-dimethoxy-2-phenylacetophenone (DPAP, 40 nmol, 1 equiv.),TES (200 nmol, 5 equiv.)and thiol (400 nmol, 10 equiv.)were dispensed as a 4 μL of a single solution of 10 mM DPAP, 50 mM TES and 100 mM thiol in degassed DMSO using a CERTUS Flex automated bulk dispenser.The plates were centrifuged and sealed under argon atmosphere and then irradiated using a custom LED array and nail curing lamp setup for 1 h, followed by RVC to yield dried peptide thioether pellets.
For analysis, peptides were dissolved in DMSO (4 μL), of which 2 μL was diluted 10-fold in water: MeCN for LC-MS injection.

| RESULTS AND DISCUSSION
First, we set out to examine the compatibility of the thiol-ene reaction with high-throughput approaches involving small (μL) volumes and absence of stirring, as well as verifying an experimental setup suitable for performing the reaction in microwell plates.The absence of stirring is a particularly important caveat to be investigated for the thiolene reaction as it proceeds via diffusion limited reaction kinetics.The model reaction between allyl acetate and 3-mercaptopropionic acid (Mpa) was selected for this purpose.Reactions were performed in 7 wells each of 100 μL volumes in DMSO-d 6 for direct 1 H NMR analysis in a UV-certified glass 96-well plate with the photoinitiator/ photosensitiser pair DPAP and 4-methoxyacetophenone (MAP).Irradiation was achieved using 4 Â 9 W bulbs (top and side irradiation from a cost-efficient and low-energy, commercial nail curing system) together with a 7 W LED array designed and built in-house (Figure 2a; see Figures S1 and S2 for technical detail).The use of the nail curing lamp provided additional irradiation while also serving as a protective reactor housing to prevent exposure of the user to UV light. 1 H NMR analysis showed reaction completion in 3 h, even in the absence of MAP and with as little as 20 mol% DPAP (Figure 2b).Lowering the reaction time to 1 h did not provide complete conversion.In comparison to larger scale thiol-ene peptide functionalisations, this is likely as a result of slower reaction due to the absence of stirring/agitation.In comparison to other high-throughput photochemistry setups, the use of our in-house built reactor cost less than 100 USD, providing a highly accessible setup (Table S1).
With these promising results in hand, we moved to apply similar conditions to model peptide examples.A model peptide consisting of a Gly, allylglycine (Agl) and Tyr residue, capped with Mpa (Figure 2c) was synthesised by Fmoc-SPPS on thiol resin following procedures recently reported by Bognar et al. [10] Briefly, a disulphide-linked resin was prepared which bears a mercaptoethylamine (Mea) moiety with a free amine to which the first amino acid (AA) was coupled.Standard Fmoc-SPPS allowed chain elongation, following which TFA treatment removed side-chain protecting groups.In a separate step, the peptide was cleaved from the resin by reductive release to furnish the linear peptide.Cyclization was achieved through use of either oxidising conditions to yield the disulphide macrocycle, or biselectrophilic linkers in aqueous buffer (pH 8) to yield the dithioether macrocycle.
[18] To ensure efficient reaction in smaller volumes, we utilised stoichiometric DPAP.The plate was covered with a transparent seal, centrifuged and irradiated for 3 h.In the case of the linker-cyclized model peptide, complete conversion was achieved, however, LC-MS analysis of the disulphide-containing macrocycle revealed the presence of two compounds of mass corresponding to the target product, possibly due to occurrence of disulphide exchange with the external thiol during the reaction, as well as a compound with mass corresponding to the reduced disulphide (Figure S4).Of the peaks corresponding to the product mass, use of solvent with 0.1% TFA to prevent thiolate formation only gave a minor change in relative intensities, while use of three equivalents of TFA gave no further improvement (data not shown).As the thioether macrocycles have a number of advantages including improved stability and greater potential for diversity, along with showing better compatibility, we chose to focus on macrocycles obtained from such biselectrophilic cyclizations.However, this methodology should also be generally applicable to linear compounds or head-to-tail cyclized peptide macrocycles.
It was hypothesised that side-products arising from the breakdown of DPAP upon UV-initiation, namely benzaldehyde and benzaldehyde dimethyl acetal, may be removed in vacuo.Indeed, a drastic decrease in the intensity of HPLC peaks corresponding to DPAP breakdown was observed, giving an improved purity profile.In the case of a number of low molecular weight thiols, these could also be removed in vacuo.In addition, we performed the experiment at varying peptide concentrations (Figure 2d).At 10 mM complete hydrothiolation of the macrocycle was achieved, with lowered conversion (81%) at 5 mM.However, at 2.5 mM, conversion dropped to 46%, although it is possible that higher thiol equivalents may allow reaction at lower concentrations due to more efficient radical propagation.
We next moved to demonstrate the reaction on a number of scope peptides bearing canonical AAs of differing functionality.For this purpose, four peptides bearing either a Lys, Leu, Trp or Glu residue with a Tyr and Agl were synthesised by Fmoc-SPPS on thiol resin as before, cyclized using 2,6-bis(bromomethyl)pyridine and purified by RP-HPLC (Figure 3).These macrocycles were subjected to the conditions previously developed for the model peptide macrocycle and analysed by LC-MS.For these macrocycles, quantitative reaction was achieved.However, oxidation of thioethers was observed in the form of two +16 peaks and a +32 peak in the LC-MS trace, while otherwise demonstrating good reaction profiles.While thiol-ene chemistry has been used extensively on peptide and protein substrates, oxidation of thioethers is not often observed. [8]Hoppmann et al. previously found that oxidation occurred at the sulphur atom in thiol-ene cyclisation via aromatic photoswitchable linkers, hypothesising that this was due to the proximal photoactive group. [19,20]Additionally, oxidation proximal to aromatic groups in thiyl radical reactions initiated by blue light for small-molecule methodologies were recently reported. [21]With this combined experimental and literature evidence, it was hypothesised that the oxidation was occurring at the thioethers forming the macrocycle backbone.Indeed, such products were generally not found for those synthesised using aliphatic linkers.This therefore points to an important consideration in applying this methodology to substrates containing thioether moieties close to aromatic groups.We next sought to further optimise our protocol to prevent the formation of these oxidation by-products.First, inclusion of one equivalent of triethylsilane (TES) only gave slight decrease in oxidised products, with a ratio of 7:2:1 for product, singly oxidised products and doubly oxidised product (Figure 2d, entry 5).Use of DMF as solvent for addition of initiator, thiol and TES did not significantly impact the oxidation (entry 6).In attempt to prevent the formation of such oxidation products alternative reaction conditions were prevented oxidation of the thioether moieties (Figure 3b).Of course, carrying out this methodology in a glove box setup would negate this step and avoid sulfone formation.In cases where substrates are not sensitive to oxidation, or where a small extent of oxidation is acceptable, the reaction may be performed without inert handling and can tolerate extended reaction times allowing use of lower thiol equivalents (Figure 2d, entries 1-3).To examine the chemical orthogonality, these optimised conditions were applied to the remaining examples (Figure 3c), which furnished the desired products with good reaction profile (Figure 3d).It was noted that use of high excess of thiol can result in a minor dithiolated double addition product, wherein two equivalents of thiol were added across the alkene.It is important to emphasise, however, that the use of these conditions are only due to the presence of the aromatic thioether linkage.More simple peptide substrates that are not prone to oxidation can be functionalised using milder conditions and without inert atmosphere (Figure 2d, entries 1-3).
In an initial investigation of the thiol scope, application of these conditions to functionalisation with thiophenol or a benzylic thiol gave very low conversion.It was hypothesised that this was due to their increased propensity to dimerise via disulphide formation compared to aliphatic thiols.When present in significantly higher concentration than the alkene substrate, this allows dimerization to become the predominant pathway.
To avoid the build-up of dimers in the reaction mixture, incorporation of to side-products, [22] and it cannot be removed in vacuo.Indeed, TES has previously been utilised to supress disulphide formation during thiol-ene reaction of peptide substrates. [23]In the case of the benzylic thiol example, reaction in presence of TES gave quantitative conversion.However, the thiophenol example still gave no conversion for either thiophenol, 4-methoxythiophenol or 4-hydroxythiophenol.
][26][27] Solvent was removed by RVC, and dispensing of thiol, DPAP and TES in degassed DMSO was achieved using a Certus FLEX automated bulk dispenser.
The removal of the DMSO from which peptide stocks were dispensed allows for use of entirely degassed solvent in the reaction, a step which may be omitted if substrates are not sensitive to oxidation.
Despite this process using air compression, oxidation as a result of this approach is not significant compared to previous experiments using pipetting.Irradiation was performed as for previous experiments and volatiles were removed via RVC (Figure 4a).Of the scope peptides, it is unsurprising that peptide a rendered the lowest conversion, since amines have been shown to slow the kinetics of thiol-ene chemistry, often as a result of their basic nature. [28]This may also in part explain the slightly lower conversions observed for macrocycles based on the pyridine linker, which additionally is prone to oxidation.Still, conversions for these examples remained excellent.The different alkenes were well tolerated, with differences only apparent for the Cys examples (row 3, columns i-l).
The Agl, Alloc-Lys and β-allylglycine examples all proceeded to similar high conversions despite Cys being present in lower equivalents.The N-allylglycine example, however, gave lower conversion, likely as a result of lower stability of the carbon-centred radical intermediate formed following attack by the thiyl radical, enabling the reverse addition to occur to a greater extent. [12,29]Despite this, quantitative conversion was observed for six of the eight thiols for this peptide example.

| CONCLUSION
We have reported the development of facile high-throughput photochemical thiol-ene methodology which is compatible with microtiter plates in an accessible and affordable photoreactor.We applied ADE and automated bulk dispensing to perform nanoscale reactions in 1536-well plates in single-digit microlitre volumes.This approach provides a method for rapid and efficient parallel evaluation of photochemical thiol-ene reaction conditions of different substrates without the requirement for large amounts of potentially precious materials.
Additionally, this methodology can be performed without the need for state-of-the-art instrumentation such as the Echo system and dispensers, as demonstrated in earlier experiments performed via pipette addition of reactants.
The protocol was subsequently applied as an efficient approach for diversification or late-stage functionalisation of peptidic alkenecontaining macrocycles.The volatile nature of the photoinitiator products and thiols allows for removal of a significant proportion of reactants in vacuo.As a result, the crude reaction mixtures demonstrated high purity as well as excellent conversions, with potential for direct application without further purification in drug-screening campaigns or other biological assays.Whilst aromatic thioethers are prone to oxidation under the reported conditions, we have shown that, if desired, this can be prevented through careful use of inert atmosphere and sealing of plates in a simple plastic glove bag and degassed solvents.
This also provides further understanding of the oxidation of thioethers during photochemical reactions, which is not currently well understood.A variety of canonical amino acids, alkene-containing amino acids and thiol partners are well tolerated, with 96 successful examples presented.Disulphide-based macrocycles represent an important limitation, although this is unsurprising for a thiyl-radical based method.In comparison to previously reported methodologies for late-stage diversification at this scale, this approach is compatible with presence of various two-electron reactive groups such as amines and carboxylic acids.
This work provides a novel and highly efficient approach to highthroughput peptide conjugation, with diverse applications for drug discovery.Upscaling of the methodology could facilitate access to even greater numbers of compounds.This may be applied to optimisation of peptide structures or structure-activity relationship campaigns.Ongoing work will investigate the application of this methodology for late-stage diversification peptides to improve binding properties.
U R E 1 (a) Diversification and macrocyclization using biselectrophilic linkers.(b) Additional thiol-ene mediated diversification of alkene containing macrocycles.(c) High-throughput elements incorporated in this work.

F
I G U R E 2 (a) UV reactor for photoinitiation of thiol-ene in microwell plates.(b) Proton nuclear magnetic resonance ( 1 H NMR) based validation of successful thiol-ene reaction in 96-well plates via monitoring of alkene consumption.Consult Figure S3 for additional spectra.(c) Model alkene macrocycle functionalisation reaction.(d) Optimisation of thiol-ene functionalisation of alkene macrocycles in 1536-well plates.investigatedthat would avoid oxidation by taking advantage of its slower kinetics when compared to those of the desired thiol-ene reaction.Higher thiol equivalents give faster reaction and will therefore allow shorter reaction times.Indeed, reaction of 10 equivalents of Mpa with the alkene scaffold for 1 h yielded <10% total oxidation products ($3% each; entry 9).Use of five equivalents of thiol gave incomplete conversion (entry 8).Performing the reaction in degassed solvent lessened the oxidation to a total of <2% (entry 10), and further, sealing the plate in an argon environment in a simple plastic glove bag (Aldrich ® AtmosBag) after reactant distribution (entry 12)

F
I G U R E 3 (a) Reaction of scope peptide with oxidation-prone aromatic linker.(b) LC-MS traces for different conditions applied to prevent oxidation.(c) Scope peptide structures baring different 2-electron reactive groups.(d) LC-MS traces for scope peptides.'*' Denotes minor dithiolated double addition product.the volatile reducing agent TES into the reaction mixture was investigated.Whilst tris(2-carboxyethyl)phosphine (TCEP) is often used for disulphide reduction, the reasons for avoiding this additive are twofold; TCEP has been shown to desulfurize thiols in UV conditions which could lead U R E 4 (a) Workflow for diversification of alkene macrocycles in high throughput.(b) Structures of peptides and thiols investigated and heatmap of conversions in high throughput diversification.'*' Denotes Cys not fully dissolved at desired concentration.'**' Denotes values represent integral of desired product compared to starting material or other side-products.See Data S1 for UV chromatograms for all compounds in the respective heatmap.fashion.For this, peptides were distributed into 1536-well plates compatible acoustic dispensing as either source or destination plates from 40 mM stock solutions in DMSO by ADE using an Echo 650 acoustic liquid handler (Labcyte).Precedent for use of acoustic dispensing in HTE has recently been established, and has proven an

For
this experiment, three sets of peptides were utilised (Figure4b); the scope peptides previously employed for optimisation and compatibility investigations (a-d), the initial model sequence cyclized with four different linkers (e-h), and a set of peptides containing different alkene-containing amino acids (i-l).For potential application in direct biological assay of the products, good reaction profile when performed on crude substrates may be required, meaning, that the quenched bis-electrophilic linker may be present in the reaction mix.With this in mind, the set of peptides with varying linkers (2,6-bis(bromomethyl)pyridine, 1,3-bis(bromomethyl)benzene, 1,3-dichloroacetone and divinylsulfone), were used in their crude form.For the set of varied alkene peptides, these were synthesised bearing an Ala and Trp residue, as well as either Agl, Alloc-Lys, β-allylglycine or N-allylglycine (Figure4b).In addition to the 12 peptides used, eight thiols were evaluated to give a total of 96 examples; four aliphatic thiols and four aromatic thiols were included.LC-MS analysis of the reaction mixtures was used to calculate the extent of conversion to the desired product.This was calculated as the % desired product of the sum of all peptidic material.Of the 96 examples, 81 (85%) showed >90% conversion, with an average of almost 95% conversion across the entire set.In the case of the cysteine examples (row 3), dissolution of the thiol component at the desired concentration was not possible and is therefore likely the cause for the observed decrease in conversion, with starting material accounting for the remaining material in these examples.Use of up to 20% water did not provide significant improvement in solubility.Discounting these Cys examples, 96% of the other entries showed conversions of >90%.The benzyl thiols (rows 4-7) in particular showed excellent reaction profile under the applied conditions.Cyclohexyl thiol (row 8), being a relatively bulky thiol, gave slightly reduced conversions, although only one example was below 90% conversion to the desired product.