A flow platform for degradation-free CuAAC bioconjugation

The Cu-catalyzed azide-alkyne cycloaddition (CuAAC) reaction is a cornerstone method for the ligation of biomolecules. However, undesired Cu-mediated oxidation and Cu-contamination in bioconjugates limits biomedical utility. Here, we report a generic CuAAC flow platform for the rapid, robust, and broad-spectrum formation of discrete triazole bioconjugates. This process leverages an engineering problem to chemical advantage: solvent-mediated Cu pipe erosion generates ppm levels of Cu in situ under laminar flow conditions. This is sufficient to catalyze the CuAAC reaction of small molecule alkynes and azides, fluorophores, marketed drug molecules, peptides, DNA, and therapeutic oligonucleotides. This flow approach, not replicated in batch, operates at ambient temperature and pressure, requires short residence times, avoids oxidation of sensitive functional groups, and produces products with very low ppm Cu contamination.

A significant limitation of the CuAAC reaction conducted under batch conditions is the need for a Cu catalyst; this can be problematic in a number of applications 5,6 . Cu-mediated oxidative damage of sensitive functional groups can result in product mixtures, which may complicate purification or lead to issues with bioassays due to the need for deconvolution of data or unknown pharmacology (Fig. 1b). In biomolecule tagging CuAAC modification of azide/alkyne biomolecules requires (super)stoichiometric loadings of Cu catalyst due to the presence of a number of Cu-chelating sites (e.g., N/S sites of peptides 7 , N7 of purines in nucleic acids 8 ), which can result in catalyst inhibition and the need for higher concentrations of Cu in the reaction (Fig. 1b) 6 . In addition, oxidative damage of biomolecules is a significant issue associated with current CuAAC-based bioconjugation strategies, severely limiting development [9][10][11] . These issues have inspired the development of a series of alternative Cufree click approaches such as strain-promoted azide-alkyne cycloadditions (SPAAC) 12 and inverse electron demand Diels-Alder (IEDDA) approaches using tetrazines 13 . Despite their moderate to fast kinetics 14 , these processes have their own issues; for example, lacking the chemo-and regiospecificity afforded by the CuAAC reaction due to the reactive (electrophilic) nature of the requisite cyclic alkynes/alkenes [15][16][17][18][19][20][21] , which are susceptible to side reactions with nucleophilic residues (e.g., thiol residues in glutathione). Furthermore, the installation of these large lipophilic groups has a significant impact on the overall physicochemical properties of the bioconjugate (Fig. 1c) 22 .
Whilst efforts have been made to overcome the oxidation and Cu contamination issues of the CuAAC reaction by the development of bespoke ligands, conducting these reactions under anaerobic conditions, the addition of oxidation inhibitors, and Cu scavengers, these issues extend from the requirement for high [Cu] to overcome slow catalytic turnover as a result of the numerous Lewis-basic groups typically found in proteins and nucleic acids 23 .
Here we describe the development of a rapid flow-assisted CuAAC reaction that overcomes these problems (Fig. 1d). Our operationally simple strategy couples solvent-induced erosion of a copper tube with the formation of a highly active CuAAC catalyst under laminar flow conditions. This enables the facile formation (t R ca. 1-10 min) of discrete ligation products and bioconjugates not possible using conventional batch conditions. Significantly, the level of Cu present in products is well below the reported mammalian cellular toxicity thresholds (e.g., <20 μM for DNA) 5,11,24 with no associated oxidative damage observed on a series of representative labile biomolecules, including peptides and DNA strands.
Whilst elemental Cu is an effective catalyst for flow CuAAC, elevated temperature and pressures are required, likely in order to solubilize some Cu in the eluent. However, H 2 O/organic mixtures are extremely effective and biocompatible solvent mixtures for CuAAC-mediated bioconjugation. In addition, the surface of Cu pipes is typically covered in a protective oxide layer, which are generally poorly soluble in organic solvents but more soluble in H 2 O. Indeed, erosion of Cu tubes with H 2 O is a well-studied engineering phenomenon, with Cu leaching a known problem in flow chemistry 31 .
Based on this, we hypothesized that an aqueous/organic mixture  26 . This hypothesis was found to be valid. Three benchmark CuAAC reactions, using three alkynes (1a-1c), with known differences in reactivity with benzyl azide (2) were evaluated in a laminar flow system comprising a pump and copper reactor at ambient temperature and pressure (Fig. 2a). The reaction does not proceed in pure MeCN or pure H 2 O and [Cu] in the eluent (10 mL collected, 1 mL/min under ambient conditions) was extremely low (<20 ppm). However, the addition of small amounts of H 2 O to the bulk MeCN resulted in the formation of triazole 3a-c, which peaked at 5:1 solvent mixture. Control experiments with an unused Cu reactor (Fig. 2b, red bars) vs. a reactor used for CuAAC reactions (Fig. 2b, blue bars) demonstrated greater erosion in the used reactor, consistent with a more exposed surface due to repeated chemistry; however the solvent composition/erosion trend was comparable, peaking at 1:1 H 2 O/ MeCN. The addition of small percentages of H 2 O to the carrier solvent (MeCN) enabled the CuAAC reaction of equistoichiometric ynamine 1a and BnN 3 (2) effectively at 5:1 MeCN:H 2 O (t R = 10 min; Fig. 2a). Whilst ynamine 1a exhibits faster batchreaction kinetics based on a pK a modifying Cu-ligation 42 , the mixed solvent system was also effective at enabling the CuAAC reaction of more standard alkynes 1b and 1c at the same flow rate. Analysis of the eluent by ICP-MS revealed that [Cu] was 14 ppm, which is well below the limit required for use in in vivo applications 5,11 . Importantly, control experiments identified a flow phenomenon. Attempting the CuAAC reaction of alkynes 1a-1c in flask experiments at 14 ppm Cu was unsuccessful for 1b and 1c and only moderately successful for the more reactive ynamine 1a (53% yield after 72 h), whereas the flow system results in quantitative conversion in 10 min (Fig. 2c). Residence times were also shortened significantly to ca. 1 min for more reactive substrates.
Scope of the flow platform. The scope of the flow CuAAC process was both broad and reproducible using three different alkyne classes (1a-1c) across a series of azide substrates (3-20; Fig. 3). Triazole products derived from simple azides, azido fluorophores, and azide possessing specific functions for downstream applications, were all isolated in high yield after a single pass. Importantly, ICP-MS analysis of the products again found the residual [Cu] was <20 ppm (see Supporting Information for full details).
We also examined the compatibility of the flow process with regards to established CuAAC chemoselectivity profiles (Fig. 3). Diyne 18, containing aliphatic alkyne and aromatic ynamine sites, underwent sequential CuAAC ligation, firstly with the coumarin azide 19 at the ynamine site followed by ligation with the nucleobase azide 20 at the aliphatic alkyne site; complete chemoselectivity was observed throughout. This demonstrates that established reactivity profiles 43 are replicated in the flow format and that our system enhances not only overall reaction kinetics but does so at very low [Cu].
The biomedical utility of the CuAAC reaction lies primarily in the ligation of bio-relevant molecules. We assessed the flow CuAAC process as a method for the ligation of representative alkyne-derivatives of nucleic acids and peptides, which have known susceptibility to form oxidized byproducts in the presence of a Cu catalyst (Fig. 4) 8,10 . Installation of a fluorinated residue onto a marketed PARP inhibitor 44 , and a common fluorophore onto a series of peptides and DNA strands containing oxidizable functionality produced triazole products with minimal formation of side-products. These include CuAAC ligations with oligodeoxyribonucleotides (ODNs) and the core ApoliproteinE (ApoE) peptide sequence (27) 45 , which has demonstrated utility as a delivery vehicle across the blood brain barrier 46 . Residues with known oxidative susceptibility (27a-e) under conventional CuAAC batch conditions were installed on the N-terminus to report any potential degradation Bioconjugation. Finally, we explored applying our flow-based CuAAC ligation approach to prepare therapeutic bioconjugates. Phosphoramidate morpholino oligonucleotides (PMOs) are a class of oligonucleotides with established therapeutic importance [47][48][49] . An essential requirement for in vivo efficacy of this class of biologics is the need conjugate a cell penetrating peptide sequence onto one of the termini to enable effective delivery to the central nervous system. The bioconjugate triazole 28 was prepared from precursors derived from a PMO azide with known in vivo efficacy as a splice-switching oligonucleotide for the treatment of Spinal Muscular Atrophy (SMA) and a peptide fragment derived from a portion of the ApoE protein 50 . Under flow conditions, the ApoE-PMO bioconjugate (28) was formed in 60% yield after 15 passes (1 mL/min; total t R = 30 min). No reaction was observed after 24 h under equivalent batch conditions, with only only 26% yield of 28 obtained in batch after 48 h using 100 equiv Cu.

Discussion
In summary, we have developed a rapid and operationally simple flow-based platform for the CuAAC reaction that operates at ambient temperature and pressure. Solvent-induced erosion of a  Cu pipe provides catalytically competent Cu to promote the CuAAC reaction of a range of both small molecules and biomolecules without oxidative damage to labile functional groups and with trace Cu contamination. We have demonstrated the dependency of the system on the composition of the medium and that the observed effect is unique to the flow conditions with comparable isolated experiments of low efficiency. We expect that these findings will significantly increase the utility of flow-assisted CuAAC across a series of academic and industrial applications.

Methods
General methods. See Supplementary Methods for further details supporting  experiments, Supplementary Tables 1-11 for additional data, and Supplementary  Figures 1-125 for spectra.
General procedure for the flow CuAAC process. Alkyne (0.2 mmol) and azide (0.2 mmol) were dissolved in 10 mL of MeCN/H 2 O (5/1). The CuAAC reactions were carried out in a commercial chemical flow reactor equipped with a 10 mL copper reactor (easy-Scholar from Vapourtec). The reaction mixture was flowed through a copper tube (diameter = 1 mm, volume = 10 mL, surface area = 400 cm 2 ) at a flow rate of 1 mL/min at rt (25°C, t R = 10 min). The reaction mixture was then collected and concentrated in vacuo to afford the crude product. Where necessary, purification was achieved by flash silica column chromatography (for small molecule products) or preparative HPLC (for peptide/DNA-based products).

Data availability
All data generated or analyzed during this study are included in this published article (and its supplementary information files). These data are also available from the author upon request.