Covalently bound azide on a (small) organic molecule or a (large) biomolecular structure has proven an important handle for bioconjugation. Azides are readily introduced, small, and stable, yet undergo smooth ligation with a range of reactive probes under mild conditions. In particular, the potential of azides to undergo metal-free reactions with strained unsaturated systems has inspired the development of an increasing number of reactive probes, which are comprehensively summarized here. For each individual probe, the synthetic preparation is described, together with reaction kinetics and the full range of applications, from materials science to glycoprofiling. Finally, a qualitative and quantitative comparison of azido-reactive probes is provided.
Azide (N3) is a popular functionality in organic chemistry. Azides possess unique chemical reactivity that has been heavily employed in organic synthesis.1 Most often, however, an azide plays the role of a masked form of an amine and is reduced at a final stage of a synthesis.2 The fact that azides are not retained in final products is not surprising, because azide functionalities are absent from natural products,3 with the possible exception of (S)-2-amino-3-azidopropanoic acid, which can be obtained from Salmonella grown in the presence of azide.4 In nature, the concentration of inorganic azide (N3−) is minute, which is fortunate, because inorganic azide displays a toxicity comparable to that of cyanide,5 with an LD50 in rats of 27 mg kg−1.6 Organic azides (RN3), in contrast, are uncharged and nontoxic compounds. Four azide-containing drugs—azidamfenicol and azidocillin (antibiotics), zidometacin (anti-inflammatory), and 3′-azidothymidine (anti-HIV)—have been clinically approved. In addition, organic azides display unique reactivity and selectivity and have therefore become a powerful tool in the field of bioconjugation. Moreover, introduction of an azide into a substrate can easily be achieved, either by chemical or biological modification. Chemically, the main strategies for the introduction of an azide group into an organic molecule involve nucleophilic substitution or diazo transfer. Both procedures can be readily applied to the majority of chemical structures; diazo transfer can also be used for protein modification as recently demonstrated by van Hest et al.7 Apart from that, synthetic azide-containing peptides can also be linked to proteins by adaptation of known ligation methods.8 Biologically, azides can be engineered into a protein by growing auxotrophic E. coli in methionine-deficient, azidohomoalanine-rich medium9 or by genetic encoding of azidophenylalanine in E. coli with the aid of a heterologous suppressor tRNA-aminoacyl tRNA synthetase pair with altered specificity.10 Other procedures have also been developed but lie outside the scope of this review. Moreover, introduction of an azide group onto a substrate generally has only minimal effect on its structure. As a consequence of the inertness and small size of the azide moiety, its facile introduction and, most importantly, its unique reactivity, azides have found wide application in bioorthogonal ligation.
There is a wealth of strategies for the chemoselective modification of biomolecules, such as proteins.8 However, it was not until 2000 that bioorthogonal chemistry was developed,11 based on the fundamental insight by Bertozzi and co-workers that a mutually selective reaction between two abiotic groups can be applied for in vivo ligation.12 In a seminal paper, it was shown that the long-known Staudinger reduction can also be applied for the formation of an amide bond, through rearrangement of the aza-ylide intermediate in a process known as Staudinger ligation (Scheme 1). Key to the Staudinger ligation is the intramolecular trapping of a nucleophilic aza-ylide (1) by an electrophilic ester, leading to a covalent amide bond (compound 2) via a putative pentacoordinate phosphine intermediate.13
The power of the Staudinger ligation for in vitro application was elegantly demonstrated in the cell-surface labeling of Jurkat and HeLa cells.11 The cells were incubated for three days with N-azidoacetylmannosamine (ManNAz), to introduce azides on the surfaces of the cells through metabolic conversion into N-azidoacetylsialic acid (SiaNAz) and incorporation into cell-surface glycoproteins. Afterwards, the cells were treated (1 h) with a biotin-labeled Staudinger probe, yielding fluorescent cells as determined by flow cytometry. The Staudinger ligation has also been applied for site-specific introduction of phosphoramides as phosphate isosteres in proteins14 and for in vivo experiments, such as for cell-surface remodeling,15 live-cell imaging16 and the visualization of O-linked glycosylation in mice.17 Further applications are not outlined here because the Staudinger ligation has been reviewed earlier.18
Although it has been shown that neither the Staudinger ligation nor the covalent phosphine adduct influences cell viability, for some applications it is desirable to avoid the presence of the phosphine oxide structure in the Staudinger adduct after ligation. Therefore, Bertozzi19 and others20 also developed a so-called traceless Staudinger ligation, involving the formation of an amide linkage (5, Scheme 2) from the intermediate aza-ylide 3, now with concomitant release of the phosphine oxide 4. The fact that a native amide bond is formed without a residual covalently attached phosphine oxide makes the traceless Staudinger ligation an attractive bioorthogonal alternative for chemical peptide coupling.
Although the Staudinger ligation was the first truly bioorthogonal ligation method, some specific disadvantages are apparent. First of all, because phosphines are prone to oxidation by air, shelf-stability is limited and oxidation may take place before ligation. Secondly, the Staudinger ligation is relatively slow (k=2.5⋅10−3M−1 s−1).13 Improvement of the kinetics and water solubility has proved troublesome,21 so alternative methods to improve bioorthogonal ligation have been developed.
The discovery by Meldal22 and Sharpless23 that the 1,3-dipolar cycloaddition of azides and acetylenes can be catalyzed by sources of copper(I) has been of tremendous value for the field of bioorthogonal ligations. Copper was found to catalyze the [3+2] cycloaddition of azides and acetylenes in aqueous solvents at physiological temperatures and the reaction is compatible with biomolecules such as sugars, proteins, and nucleic acids. The resulting 1,4-disubstituted triazole adducts are formed with exclusive regioselectivity and in high yields, which has led to the widespread application of this so-called “click” reaction, which has been reviewed extensively.24 Despite the power and versatility of click chemistry, the requirement for the toxic copper severely limits both its application in cellular systems (let alone living organisms) and also the scope of click materials, due to the potential for residual traces of copper. Nevertheless, the advent of click chemistry and its widespread use has been an important inspiration for the further development of truly bioorthogonal reactions based on copper-free approaches.
Cycloaddition with Cyclooctyne
In 1953, Blomquist and Liu25 published the first synthesis of (pure) cyclooctyne (7, Scheme 3), through oxidative decomposition of cyclooctane-1,2-dione dihydrazone (6), which was obtained through acyloin condensation of diethyl suberate (Scheme 3).26 In the same paper, the suspected ring strain in cyclooctyne, due to deformation of the triple bond, is already apparent from the reported explosive reaction of cyclooctyne and phenyl azide, leading to an uncharacterized viscous liquid. The structural identity of the resulting liquid was later established by Wittig and Krebs as 1-phenyl-4,5,6,7,8,9-hexahydro-1H-cycloocta[d][1,2,3]triazole (8), the product of a 1,3-dipolar cycloaddition.27 Wittig and Krebs reported that smaller cycloalkyne rings were too unstable for isolation, but their transient formation could be indirectly ascertained by trapping of the cycloheptyne or cyclohexyne generated in situ with phenyl azide and isolation of the resulting triazole products. In subsequent years, other procedures for the preparation of cyclooctynes have been developed and a variety of 1,3-dipolar cycloaddition and Diels–Alder reactions have been reported, as summarized recently.28
In 2004, Bertozzi et al. envisioned that the fast reaction between a strained cycloalkyne and an azide might also be applicable as a bioorthogonal ligation method.21 To this end, a functionalized cyclooctyne was prepared from cycloheptene in four synthetic steps and in an overall yield of 52 % (Scheme 4). Firstly, cycloheptene was converted into 9 by a known procedure29 involving an initial reaction with dibromomethane carbene, followed by electrophilic ring-opening in the presence of AgClO4 and trapping of the resulting E-allylic cation with methyl 4-(hydroxymethyl)benzoate. Subsequent elimination of the bromide and ester saponification yielded the carboxylic acid 11. In order to improve water solubility, a cyclooctyne lacking the lipophilic phenyl group was also prepared, with compound 12 being obtained in 10 % overall yield from cycloheptene.30 Both probes are significantly less prone to oxidation than their Staudinger ligation counterparts, but are nevertheless not fully shelf-stable, because prolonged storage of compounds 11 and 12 at −20 °C led to decomposition. However, one strong feature of the developed 3-alkoxycyclooctynes is that the rates of their cycloaddition with benzyl azide were found to be comparable to those of the Staudinger ligation (k=2.4×10−3M−1 s−1 for 11 and k=1.3×10−3M−1 s−1 for 12).31
To demonstrate the bioapplicability of strain-promoted alkyne azide cycloaddition (SPAAC), the glycoprotein GlyCAM-Ig was first expressed in the presence of peracetylated ManNAz, leading to the azide-labeled protein through metabolic transformation of Ac4ManNAz into SiaNAz and subsequent incorporation in the glycan part of the protein.21 Azide incorporation was visualized by staining with the biotin-labeled cyclooctyne 13 (Scheme 4), followed by Western blot detection. To show that not only glycoproteins but also living cells can be biotin-labeled with the cyclooctyne 13, Jurkat cells were incubated with Ac4ManNAz to introduce azide-containing sialic acid on cell-surface glycans through metabolic incorporation. After three days, the azide-modified cells were treated with 13 at several concentrations and with different densities of cell-surface azides and varying reaction times. It was found that more efficient labeling could be achieved with higher concentrations of 13, higher azide densities, and longer reaction times. Importantly, no negative effects on cell viability were observed. A side-by-side comparison of the cyclooctyne approach with the previously reported Staudinger ligation showed similar results, although Staudinger ligation gave labeling that was approximately twice as efficient.
An overview of the use of 3-alkoxycycloalkyne conjugates is provided in Table 1. Two recent applications deserve some comments. Lallana et al. (entry 7) observed a strong depolymerizing effect of CuI when CuAAC was applied on nanostructures based on azide-containing polymeric chitosan-PEG conjugates.32 Therefore, cyclooctynes were linked to fluorescein isothiocyanate (FITC) and an anti-BSA antibody. The FITC-cyclooctyne conjugate was ligated to chitosan-PEG-N3 and mixed with unfunctionalized chitosan-PEG to provide fluorescent nanoparticles. The nanoparticles were in turn decorated with cyclooctyne conjugated to anti-BSA rabbit IgG, as judged by dynamic light scattering (DLS) and staining of BSA-agarose beads.
Table 1. Use of 3-alkoxycyclooctyne conjugates and azides for bioorthogonal labeling.
[a] The cyclooctyne-lipid conjugate is mixed with living cells; detection was achieved with an azide-containing probe.
metabolic conversion of Ac4ManNAz into SiaNAz and detection in glycoprotein GlyCAM-Ig
Neef and Schultz33 recently followed an approach in which a cyclooctyne-fatty acid conjugate was first incorporated in the membranes of fibroblasts, followed by detection with 3-azido-7-(diethylamino)coumarin, a molecule that becomes fluorescent upon triazole formation. As a consequence, fatty acid labeling inside the cells could be monitored by fluorescence detection without any need for cell fixation, as was required for CuI-catalyzed [3+2] cycloaddition.
Because the stabilities of the 3-alkoxycyclooctyne derivatives are rather low, an alternative was developed with the alkyl-linked cyclooctyne conjugate 20 (Scheme 5).30 Compound 20 was prepared through α-alkylation of cyclooctanone (yielding 16), followed by triflate formation to afford 18, basic elimination, and finally hydrolysis of the ester moiety to afford 20. Indeed, the alkyl-linked cyclooctyne 20 was found to be more stable than cyclooctyne 11, although the reactivity of 20 towards azides (k=1.2×10−3M−1 s−1) was also two times lower.
Despite the non-optimal kinetics of the 3-alkylcyclooctyne 20, its improved stability and easy synthetic preparation have facilitated several biological applications, as summarized in Table 2. A microbial lipoic acid ligase (LplA), for example (entry 2), was redirected to attach azide-containing fatty acids—ideally 8-azidooctanoic acid—onto an engineered LplA acceptor peptide (LAP). Cyclooctynes conjugated to cyanine-3 (Cy3), Alexa Fluor 568, or biotin were subsequently used to visualize HEK cells containing fusion proteins either of LAP with cyan fluorescent protein (in turn fused to a transmembrane receptor) or of LAP with the low-density lipoprotein receptor.34
Table 2. Use of 3-alkylcyclooctyne conjugates and azides for bioorthogonal labeling.
[a] A cyclooctyne-biotin conjugate without the lipophilic phenyl ring was also applied (not shown).
metabolic conversion of detection in Jurkat cell- Ac4ManNAz to SiaNAz and surface glycoproteins
Wolfbeis et al. prepared synthetic peptides containing both propargylglycine and an N-terminal cyclooctyne moiety. Subsequent labeling with SPAAC followed by CuAAC with different fluorophores afforded peptide substrates for MMP-2, a diagnostic tumor indicator. Cleavage of the peptide by MMP-2 could be effectively monitored by Förster Resonance Energy Transfer (FRET).35 Alternatively, the cyclooctyne-labeled peptide could be loaded onto azido-functionalized silica nanoparticles (predoped with triazolylcoumarin), followed by CuI-catalyzed introduction of a fluorescent label to obtain dually labeled nanoparticles for FRET detection. In another contribution, the same authors demonstrated36 that a 3-alkylcyclooctyne conjugated to a red fluorophore can be useful for cell-surface staining of Chinese hamster ovary (CHO) cells.
Most recently, de Koster et al. used the cyclooctyne 20 for enrichment of azide-containing peptides from complex protein mixtures.37 To this end, the resin-immobilized cyclooctyne 22 (Scheme 6), containing a cleavable disulfide linker, was synthesized. Addition of the azido-trapping resin 22 to a mixture of peptides with or without azides, followed by disulfide reduction with triscarbethoxyethylphosphine (TCEP) and capping of the free thiols with iodoacetamide (yielding the peptides 23), greatly facilitated more accurate analysis of the azide-containing peptides. Photoactive yellow protein (PYP; Table 2, entry 6), for example, was expressed in auxotrophic bacteria (vide supra) in the presence of azidohomoalanine (AHA) to introduce six azidohomoalanine residues.38 After trypsin digestion, crude mass spectral analysis identified only four of the six expected AHA-containing segments. However, when the crude peptide mixture was first treated with the resin, filtered, and cleaved, all six different segments could be observed. The resin 22 could also be used to determine the rate of protein synthesis in E. coli.39 Finally, more accurate analysis was also feasible in the case of cytochrome c peptides, chemically cross-linked with an azide-containing cross-linker (entry 7).40
Because the alkyl-linked cyclooctyne had been found to be more stable but less reactive than earlier cyclooctynes, a fluoride substituent was introduced adjacent to the acetylene, on the assumption that lowering of the LUMO of the acetylene by a neighboring electron-withdrawing group would increase the reactivity towards azides (Scheme 5). To this end, cyclooctanone was fluorinated with Selectfluor and the resulting 2-fluorocyclooctanone (15) was subjected to the same sequence of reactions as for compound 20, yielding 21 in 15 % overall yield. The fluorinated compound 21 did indeed display a four-times enhanced reactivity towards benzyl azide (k=4.3×10−3M−1 s−1).30
Bertozzi et al. applied the fluorinated cyclooctyne 21 in the same cell-surface glycoprofiling experiments as for the previously reported cyclooctynes (Table 2, entry 1). Surprisingly, the Staudinger ligation still outperformed 21 in the biological assay, despite the better kinetics of the cyclooctyne, possibly because Staudinger ligation is more efficient with azides bearing electron-withdrawing or resonance-stabilizing groups.30 Nevertheless, cyclooctyne 21 has found one application in the cross-linking of azido-terminated star polymers prepared by ATRP (entry 8). Treatment of star polymers with a dimeric derivative of 21 led to time-dependent formation of a gel that could subsequently be photodegraded by irradiation with 350 nm light for two days.41
The fact that monofluorination of the cyclooctyne system enhanced the reactivity was a stimulus for Bertozzi et al. to synthesize a difluorinated cyclooctyne (DIFO).42 Cyclooctane-1,5-diol was therefore monoallylated (Scheme 7), followed by oxidation of the other alcohol moiety to give the ketone 24. Subsequent monofluorination of ketone 24 with Selectfluor, via the silyl enol ether intermediate, afforded both the cis- (compound 25) and trans-fluorinated (compound 26) diastereomers in high yield (ratio 2:1). The minor trans isomer was resubjected to the sequence of silyl enol ether formation-fluorination, leading to the difluoroketone 27. Apparently, similar treatment of the cis isomer proceeded unsatisfactorily, because it was recycled by epimerization with KHMDS, leading to a cis/trans ratio of 1.2:1. The allyl group in the difluoroketone 27 was next oxidized to a carboxylic acid, and this compound was finally converted into the DIFO 28 by triflation and elimination (11 % yield for two steps). The poor overall yield of the route (1.2 % over ten steps) is, however, nicely compensated for by a rate constant of 7.6×10−2M−1 s−1 for the reaction between the DIFO 28 and BnN3, about 17 times faster than the fastest previously reported Cu-free ligation. The high reactivity of the DIFO was also reflected in the staining of Jurkat cells bearing SiaNAz with the biotinylated DIFO (Table 3, entry 1), which proceeded with a 20-times higher efficiency than with other cyclooctynes. Again, no cellular toxicity was observed. A DIFO derivative fluorescently labeled with Alexa Fluor 488 was synthesized and applied for time-lapse imaging of labeled cell-surface glycans in CHO cells (entry 3). Moreover, in combination with another Alexa Fluor 568 conjugate, dynamic multicolor imaging of glycan internalization and trafficking by two consecutive rounds of incubation with Ac4ManNAz, followed by labeling with different fluorophores, indicated that glycan labeling did not perturb normal glycan trafficking to lysosomes.42 Finally, in vivo labeling of splenocyte glycans in mice was demonstrated by daily injection (one week) with DIFO-FLAG (entry 6).
Table 3. Use of the first-generation DIFO and azides for bioorthogonal labeling.
incorporation of azido-homoalanine in dihydrofolate receptor
In follow-up papers, in vivo labeling was also successfully demonstrated in zebrafish43 and Caenorhabditis elegans.44 Zebrafish embryos (entry 4) were incubated in media containing Ac4GalNAz at 3 to 72 h after fertilization and then treated with a DIFO-647 probe, to obtain a fluorescent signal. Alternatively, embryos were incubated with Ac4GalNAz from 3 h after fertilization and GalNAz incorporation was followed by labeling with DIFO-647 at time intervals of 12 h. Finally, the time dependency of sugar synthesis in certain regions was demonstrated by three-color detection with conjugates of the DIFO to Alexa Fluor 488, 555, and 647. Similar research has been reported43 for the incorporation of ManNAz, GalNaz, and GlcNAz into mucin O-glycans in the nematode C. elegans (entry 5, Table 3).44
A conceptually novel approach, reported by Zou and Yin,45 involved the identification of catalytically active adenylation domains of nonribosomal peptide synthetases (NRPSs) by fusion of the NRPS adenylation domain to the peptidyl carrier domain (PCP) and M13 phage. After pantethenylation, an azido-labeled 3-amino-5-hydroxybenzoate derivative (Table 3, entry 2) was enzymatically conjugated to PCP by the adenylation domain, as confirmed by Western blot, followed by cycloaddition of the biotin-conjugated DIFO. The conjugate was immobilized on a streptavidin-coated 96-well plate before addition of E. coli to give a tenfold increase in phage enrichment for phages displaying the adenylation-PCP domain.
Finally, Turro et al.41 elegantly demonstrated that the gelation times of azido-terminated star polymers can be simply tuned through the choice of cyclooctyne probe: the bis-DIFO functionalized ethylene-1,2-diamine (Table 3, entry 7) gave significantly faster cross-linking than a dimer of cyclooctyne 28 (Table 2, entry 8).
The great potential of the DIFO 28 warranted an alternative synthetic route to such compounds. To avoid the cumbersome elimination step to the alkyne, Bertozzi et al. developed a synthesis with the ether linkage replaced by an alkyl linker.46 The synthesis commenced with Selectfluor-mediated difluorination of cyclooctane-1,3-dione (29, Scheme 8),47 followed by a single Wittig olefination and hydrogenation to yield the intermediate 31. Conversion into the DIFO 32 was achieved by formation of the vinyl triflate, elimination, and ester saponification. Notably, elimination of triflate now proceeded in excellent yield (87 %), and the total synthesis of the DIFO 32 was completed in eight steps with an overall yield of 27 %.
The alternative DIFO derivative 36, lacking the phenyl ring in the linker, was also prepared in order to reduce nonspecific protein and cell binding as a result of the lipophilic character of DIFO. Although temporary protection of the acid moiety as an OBO orthoester (compound 34) was required, the synthesis was essentially similar to the previous one, leading to the DIFO 36 in ten steps and in a 21 % overall yield.
Kinetic experiments with the DIFOs 32 and 36, termed second-generation DIFOs, and benzyl azide revealed reaction rate constants of 4.2×10−2M−1 s−1 for 32 and 5.2×10−2M−1 s−1 for 36, values slightly lower than, but similar to, that obtained with the first-generation DIFO. The relative rate constants were also reflected in the labeling of Jurkat cells, which showed three-times higher fluorescence for the DIFO 28 than for the DIFOs 32 and 36. As would be expected, slightly less background labeling was observed in flow cytometry experiments with the DIFO 36, lacking the hydrophobic phenyl group, than with the DIFO 32. Live-cell imaging was also performed with CHO cells, with no notable difference with respect to earlier cyclooctyne probes.
For the preparation of microtissues (Table 4, entry 2), azide-containing Jurkat cells were conjugated to DIFO-containing single-stranded DNA. Multicellular structures were devised by selective hybridization of cells functionalized with complementary DNA sequences, eventually leading to microtissues with defined cell composition and stoichiometry.48
Table 4. Use of the second-generation DIFOs and azides for bioorthogonal labeling.
metabolic conversion of Ac4ManNAz into SiaNAz and detection in Jurkat and CHO cell-surface glycoproteins
Burkart et al. used the second-generation DIFO 32 to probe the protein interactions between communication-mediating (COM) domains of nonribosomal peptide synthetase (NRPS).49 To ensure smooth processing of biosynthetic intermediates, individual catalytic protein domains must properly interact by intramodular communication. A panel of pantetheine cycloalkynes and azides (Table 4, entry 3) was loaded onto PCP domains of NRPS modules to trap the transient protein–protein interaction during NRP biosynthesis. Cyclooctyne–azide conjugation was faster and more selective than CuI-catalyzed [3+2] conjugation, which also showed conjugation between non-interacting proteins.
Second-generation DIFOs have also found application in the patterning of microtissues.50 A doubly functionalized DIFO-labeled peptide was prepared (entry 4) and subsequently treated with poly(ethylene glycol) tetraazide to form hydrogels. Through the application of a peptide sequence with a metalloproteinase cleavage site, it was found that encapsulated cells could spread and migrate through the material.
Despite the great potential of DIFOs, Boons et al. envisioned that the introduction of fused aryl rings on the cyclooctyne system should enhance the reaction kinetics as a result of increased ring strain and conjugation.51 As shown in Scheme 9, the synthesis of the dibenzocyclooctynol (DIBO) 39 was achieved in five steps, starting with treatment of phenylacetaldehyde with Me3SiI for a week at −20 °C to form the tetracyclic system 37, which was transformed into the dibenzocyclooctenol 38 upon treatment with nBuLi. Subsequent conversion of the alkene into an alkyne proceeded effectively through bromination/double elimination, but required TBDMS protection of the alcohol prior to bromination, yielding 39 in an overall yield of 10 %.
In kinetic experiments, cycloaddition of the DIBO with benzyl azide in methanol proceeded with a rate constant of 0.17 M−1 s−1, twice as high as that for the DIFO 28.The DIBO system was also applied in the labeling of Jurkat cells and CHO cells. Both cell types were cultured in the presence of Ac4ManNAz and incubated for different timespans with a biotin-labeled dibenzocyclooctynol. An increase in fluorescence was observed when longer incubation times or higher concentrations of the biotin-labeled probe were applied, but background labeling was negligible.
In a follow-up paper with Popik et al., a photo-triggerable cyclopropenone system that formed the alkyne functionality upon UV irradiation at 350 nm was developed (Scheme 10).52 1,2-Bis(3-butoxyphenyl)ethane (41) was prepared through a Wittig reaction between 3-methoxybenzyltriphenylphosphonium chloride and 3-methoxybenzaldehyde53 with subsequent hydrogenation (51 % for two steps). Compound 41 was then converted into the photo-activated system 42 by double Friedel–Crafts alkylation with tetrachlorocyclopropene and in situ hydrolysis. Conveniently, the major product isolated had lost one butyl group by hydrolysis, thereby providing a handle for subsequent functionalization.
Whereas the cyclopropenone 42 showed no reaction with azides, irradiation at 350 nm resulted in the rapid formation (1 min) of a cycloalkyne that reacted with benzyl azide with a rate constant of 7.63×10−2M−1 s−1, as determined by a newly developed UV-based procedure.54 The biological applicability of the system was demonstrated by irradiation of mixtures of biotin-labeled 42 with azide-functionalized Jurkat and CHO cells, followed by fluorescent staining. Some background labeling was observed, probably due to the hydrophobicity of the DIBO, but chemical reactivity of the cyclopropenone 42 was excluded.
Van Hest et al. have applied DIBO conjugates for the functionalization of coatings by simultaneous thermal deposition of azido-functionalized RAFT polymers of methacrylate onto an alkyne-modified glass plate and cross-linking of the polymers with the tetraethylene glycol bis-DIBO derivative 43 (Scheme 11). The surplus free azides in the RAFT polymer were subsequently used for fluorescent labeling of the surface by treatment with four different fluorescently labeled probes: a simple alkyne, the Staudinger probe, dibenzocyclooctynol, and an oxanorbornadiene. Both CuI-catalyzed [3+2] cycloaddition and Staudinger ligation showed significantly lower functionalization than the dibenzocyclooctynol and the oxanorbornadiene, and the fastest labeling of the coating was observed with dibenzocyclooctynol.
As already described by Wittig and Krebs, rings smaller than cyclooctyne are too unstable for isolation and the formation of such cycloalkynes can only be deduced by in situ trapping with, for example, azide.27 In line with such an approach, Moses et al.,55 Larock et al.,56 Reddy et al.,57 and Feringa et al.58 reported that benzyne (45, Scheme 12) formed in situ rapidly reacts with azides to form the benzotriazole cycloadducts 46. Benzyne can be formed from, for example, trimethylsilylphenyl O-triflate (44). However, because protic solvents are not compatible with benzyne and chemical activation is necessary prior to reaction, bioorthogonal application of benzyne–azide cycloaddition appears futile.
The cyclooctynes described above have, without exception, negligible water solubility. In addition, background labeling due to hydrophobic interactions with cell membranes has been observed in biological experiments, and nonspecific binding to serum proteins cannot be excluded. Bertozzi et al. therefore set out to develop a more hydrophilic cyclooctyne through the introduction of nitrogen, to break the hydrophobic character of the ring.59 In addition, two methoxy groups were introduced to enhance polarity and water solubility. The synthesis started (Scheme 13) with methyl 4,6-O-benzylidene-α-D-glucopyranoside, which was converted in two steps into the bromide 47. The dialkene 48 was prepared from 47 through zinc reduction, followed by reductive amination and amine protection. After ring-closing metathesis (second-generation Grubbs catalyst), the alcohol was oxidized and the double bond was hydrogenated to provide the intermediate 49. Conversion into the alkyne 50 was effected through fragmentation of an intermediate selenadiazole, leading to the dimethoxyazacyclooctyne (DIMAC) 50. The DIMAC 50 showed a kinetic constant of 3.0×10−3M−1 s−1, slightly greater than that of the parent cyclooctyne 11 but significantly lower than that of the DIFO derivative 28. Nevertheless, the DIMAC 50 was found to be suitable for detection of metabolically incorporated ManNAz in Jurkat cells. As would be expected, background labeling was significantly lower than in the case of the cyclooctyne 12.
In a similar strategy, van Delft et al. envisioned that the more hydrophilic character of the DIMAC 50 and the more reactive dibenzocyclooctyne moiety of compound 39 could be combined in a single molecule: the dibenzoazacyclooctyne (DIBAC) 53 (Scheme 14).60 To this end, the precursor 51 was obtained by Sonogashira coupling of 2-ethynylaniline and 2-iodobenzyl alcohol, followed by Boc protection, reduction of the double bond, and oxidation of the alcohol. The resulting aldehyde 51 could be converted into the key intermediate 52 through a quantitative simultaneous deprotection and reductive amination. In the next step the introduction of the linker was required, because the alkyne was not stable in the presence of the free amine. Subsequent bromination, elimination, and hydrolysis of the ester yielded the DIBAC 53. Kinetic experiments with benzyl azide showed rate constants twice those of the dibenzocyclooctyne 39 in reactions with BnN3 in CD3OD (k=0.31 M−1 s−1) and with 2-azidopropanoic acid in D2O (k=0.36 M−1 s−1).
The DIBAC 53 was used for the efficient PEGylation of azide-modified CalB and HRP. To this end, 53 was attached to H2N-PEG2000-OMe by EDC coupling, and it was found that fast and complete conversion was achieved with only five equivalents of 53 within 3 h for both enzymes. In line with the rate constants, the DIBAC produced more efficient PEGylation of the enzymes than a similar PEG-conjugate derived from DIBO 39 and also proved significantly better than PEGylation by CuI-catalyzed [3+2] cycloaddition, which needed a large excess of reagents and showed only low conversion.61
The concept of cycloaddition of azides to strained unsaturated double bonds was also investigated for oxanorbornadienes by Rutjes et al.62 It was envisioned that ring-strained alkenes could undergo cycloaddition with azides, and that this could be further enhanced by introduction of electron-withdrawing groups, as in a report by Ju et al., who discovered that the spontaneous reactions of electron-deficient alkynes with azides63 also proceed in good yields under aqueous conditions.64 Thus, in a straightforward process, the oxanorbornadiene 54 was obtained through a Diels–Alder reaction between ethyl trifluorobutynoate and furan (Scheme 15).62 Treatment of 54 with an alkyl azide gave the triazoline intermediate 55, which underwent a retro-Diels–Alder reaction, leading to the triazole product 56 and furan. It was found that the oxanorbornadiene 54 is partly susceptible to undesired Michael addition as well as cycloaddition of azide on the other double bond. Introduction of a methyl substituent, by initial use of 3-methylfuran, effectively eliminated the latter disadvantage.65
Although the rate constant of cycloaddition of 54 with azide is relatively modest (k=8.5×10−4M−1 s−1), and slightly lower for the methyl-substituted derivative (k=4.2×10−4M−1 s−1), the advantages of the oxanorbornadiene system are its easy preparation and high water solubility. It was shown that the oxanorbornadiene 57 (Scheme 16) was useful for the labeling of tumor cells. To this end, 57 was modified with diethylene triamine pentaacetic acid (DTPA) and labeled with In3+ to form 58. The azide-containing RGD-peptide 59, known to bind to the αvβ3-integrin, was also synthesized. Because in vivo conjugation was too slow, the conjugates 60 and 61 were preformed in vitro and then injected into mice, showing a specific uptake in tumor tissue.66
An overview of the synthesis of all currently available chemical probes for metal-free ligation to azides has been provided. In addition, the use of these probes in a wide range of biological and/or material applications has been summarized. From these overviews it has become clear that there are no fundamental differences between the probes with regard to toxicity issues and reactivity towards azides in [3+2] cycloadditions, except in the case of the Staudinger ligation. A functionally more relevant difference between the Staudinger ligation and strain-promoted cycloaddition lies in the oxygen sensitivities of phosphine-based probes. For this reason, in research labs with the relevant synthetic expertise, a gradual shift from the Staudinger ligation towards the strain-promoted probes can be observed in recent years. For labs without synthetic chemists available, however, the accessibility of a particular probe is still an important issue in guiding the choice of which probe to apply. In this respect, it must be noted that whereas several of the Staudinger probes have long been commercially available,67 probes for strain-promoted reactions with azides did not become accessible until the beginning of 2010.68
If commercial availability is not an issue, one important aspect that governs the choice of a particular probe is its lipophilicity. As discussed above, several of the strained systems currently in use involve hydrocarbon scaffolds with limited solubility in water. Solubility issues apart, the bioavailability of a highly hydrophobic probe may be decreased due to sequestration by membranes or nonspecific binding to serum protein. In that respect, the DIMAC system (Table 5, entry 8) and the oxanorbornadienes (entry 10) clearly stand out as indicated by the provided calculated log P values.69 Nevertheless, neither the DIMAC nor the oxanorbornadienes have yet found significant application. The attractiveness of the hydrophilic DIMAC is limited mostly by the long synthetic route leading to it (eleven steps). Apart from that, the reaction kinetics for cycloadditions to azides are mediocre in relation to previously reported probes. The oxanorbornadienes, in contrast, are extremely simple to synthesize, requiring only two steps. In their case, however, cycloadditions with azides are too slow for in vivo applications. The early generation cyclooctynes (entries 2 and 3, Table 5) are relatively easy to prepare (4–5 steps, 12–52 % overall yields) and have provided the basis for current developments in copper-free ligation chemistries, but have clearly been superseded by more reactive cyclooctynes. In particular, the development of the DIFOs (entry 4, Table 5) heralded an era of novel opportunities for strain-promoted cycloadditions with azides, with reaction rates over ten times greater than those of earlier cyclooctynes. Moreover, the troublesome synthesis of DIFOs (ten steps, 1.2 % overall yield) was recently overcome by development of the so-called “second-generation” DIFOs (entry 5, Table 5), which can be prepared in acceptable overall yields (about 25 %). A disadvantage is that the synthetic route is still long and the reactivity slightly compromised in relation to the first-generation DIFO. The introduction of the dibenzocyclooctynes marked another big step forward in strain-promoted cycloaddition, because the DIBO (Table 5, entry 6) and the DIBAC (entry 9) currently show the highest reaction rate constants for reactions with azides, amounting to 0.17 and 0.31 M−1 s−1, respectively. That apart, synthetic access to both probes is good. The synthesis of the DIBO requires only five steps and proceeds with an overall yield of 10 %. The route to the DIBAC is more lengthy (nine steps), but many steps occur with (near) quantitative yields, giving an excellent overall yield of 41 % from cheaply available starting materials. Finally, the cyclopropane-based dibenzocyclooctyne precursor (Table 5, entry 7) deserves mention not so much because of its synthetic accessibility, but due to the unique property that a dibenzocyclooctyne with reaction kinetics only slightly slower than those of the DIBO is generated in situ on brief UV irradiation.
Table 5. Overview of synthesis of probes and their reactivities in [3+2] cycloadditions with azides.
No. of synthetic steps
k[a] [×10−3M−1 s−1]
[a] Second-order reaction rate constant determined with benzyl azide as model, unless otherwise indicated. [b] Of the corresponding carboxylic acid. [c] 2-Azidopropanoic acid used as azide to determine reaction kinetics.
In conclusion, the past decade has seen tremendous developments in the field of bioorthogonal ligations. Interest in the ligation of biomolecules, originally initiated by the versatile Staudinger ligation, exploded with the report on copper(I)-catalyzed azide alkyne cycloadditions. Despite its elegance, however, it has become clear that the potential for application of CuAAC in (bio)materials and living systems is very limited, due to the toxicity of copper. For in vitro and in vivo applications in particular, the development of copper-free—or preferably metal-free—ligation chemistries has therefore been instrumental. The majority of papers dealing with copper-free procedures have focused on the chemistry of azides with phosphines or strained alkynes, the topic of this review. Surprisingly, cycloadditions of strained cycloalkynes have been almost exclusively limited to azides, despite the large number of Huisgen reactions of alkynes with other dipoles such as nitrile oxides, diazo compounds, azomethines, and nitrones. Only recently have two independent papers reported on cycloadditions of dibenzocyclooctyne and dibenzoazacyclooctyne with nitrones, reactions that proceed up to 300 times more rapidly than with azides.70 Other bioorthogonal reaction pairs not involving strained alkynes have also recently been developed, in particular thiol-ene reactions,71 alkene–azamethine cycloaddition,72 and the cycloaddition of tetrazine with trans-cyclooctene.73 However, these topics lie outside the scope of this review and readers with interest in these topics are referred to other sources.18, 71, 74
It is clear that bioorthogonal chemistry has a great future. The challenge to synthetic chemists is to find reaction partners with ideal characteristics in terms of synthetic accessibility, polarity, reaction kinetics, and, of course, biocompatibility. Biologists, biochemists, and material scientists are finding their way to these novel chemistries for unique and unprecedented applications. It will be exciting to see where bioorthogonal chemistry will lead us in forthcoming years.
Marjoke Debets received her MSc in chemistry at the Radboud University Nijmegen in 2008, during which she did a research internship in the Synthetic Organic Chemistry group of Prof. Floris Rutjes on the synthesis of thrombin substrates. Another research internship was performed in the Synthetic Organic Chemistry group of Prof. Erik Sorensen (Princeton University, Princeton, USA). Since 2008 she has been pursuing her PhD at the Radboud University Nijmegen in the research groups of Prof. Floris Rutjes and Prof. Jan van Hest on the development and application of new bioorthogonal ligation methods.
Floris Rutjes obtained his PhD in chemistry at the University of Amsterdam (1993) under the supervision of Prof. Nico Speckamp. He then conducted post-doctoral research in the group of Dr. K. C. Nicolaou at The Scripps Research Institute (La Jolla, USA). In 1995, he was appointed assistant professor at the University of Amsterdam and became a full professor in synthetic organic chemistry at the Radboud University Nijmegen in 1999. His research interests comprise bio- and transition metal catalysis, total synthesis, the development of diagnostic tools, and synthesis in microreactors. He has received several awards including the Gold Medal of the Royal Netherlands Chemical Society (KNCV, 2002) and the AstraZeneca award research in organic chemistry (2003). He is co-founder of the companies Chiralix and FutureChemistry and in 2008 was entitled the “Most entrepreneurial scientist of the Netherlands”.
Christianus W. J. van der Doelen received his BSc in molecular life sciences from the Radboud University in 2009. As part of his MSc studies he performed a research internship in the Synthetic Organic Chemistry workgroup as part of a running collaboration with Synthon BV (Nijmegen, the Netherlands). Currently, he is attending management courses for his MSc.
Floris van Delft obtained his PhD in chemistry at the University of Leiden (1996, cum laude) under the supervision of the late Prof. J. H. van Boom. He then conducted post-doctoral research in the group of Dr. K. C. Nicolaou at The Scripps Research Institute (La Jolla, USA). In 1998, he became assistant professor in bioorganic chemistry at the University of Amsterdam. In 1999, he joined the Radboud University Nijmegen as an assistant professor. In 2000, he received a Vernieuwingsimpuls for a project on synthetic modification of aminoglycosides as novel antibiotics or antivirals. Other research interests involve the preparation of homogeneous glycopeptides and glycoproteins (or isosteres) and the development of new bioorthogonal chemistry tools.