Click Chemistry in Complex Mixtures: Bioorthogonal Bioconjugation

Abstract

The selective chemical modification of biological molecules drives a good portion of modern drug development and fundamental biological research. While a few early examples of reactions that engage amine and thiol groups on proteins helped establish the value of such processes, the development of reactions that avoid most biological molecules so as to achieve selectivity in desired bond-forming events has revolutionized the field. We provide an update on recent developments in bioorthogonal chemistry that highlights key advances in reaction rates, biocompatibility, and applications. While not exhaustive, we hope this summary allows the reader to appreciate the rich continuing development of good chemistry that operates in the biological setting.

A. Introduction: Bioorthogonal Click Chemistry

Chemical biology involves the creation of non-biological molecules that exert an effect on, or reveal new information about, biological systems. Central to this field is the property of selectivity: ultimately, one wishes for molecules with perfectly selective biological function; in practice, one starts with as much chemical selectivity as possible and tests and refines from there. Therefore, the ability to make chemical modifications that enable the direct detection of, or interaction with, biomolecules in their native cellular environments is at the heart of the chemical biology enterprise.

Genetically encoded reporters, such as GFP and tetracysteine motifs, have been used to excellent effect for protein tagging, but other molecules such as glycans, lipids, metabolites and myriad post-translational modifications are not often amenable to this type of labeling. Monoclonal antibodies usually provide sufficient target specificity, but are laborious to generate and are often unable to enter cells and tissues. Covalent chemical modification has therefore emerged as an alternative strategy. Bioorthogonal reactant pairs, which are most suitable for such applications, are molecular groups with the following properties: (1) they are mutually reactive but do not cross-react or interact in noticeable ways with biological functionalities or reactions in a cell, (2) they and their products are stable and non-toxic in physiological settings, (3) ideally, their reaction is highly specific and fast (Sletten and Bertozzi, 2009). Rate is an often underappreciated factor by the casual user of bioorthogonal chemical technology: very high rate constants are required for labeling cellular processes that occur on fast time scales or with low abundance structures in (or on) the cell.

Bioorthogonal chemical reactions have emerged as highly specific tools that can be used for investigating the dynamics and function of biomolecules in living systems (Jewett and Bertozzi, 2010; Lang and Chin, 2014; Lim and Lin, 2010b; Patterson et al., 2014; Prescher and Bertozzi, 2005; Sletten and Bertozzi, 2009). Click chemistry, inspired by nature’s use of simple and powerful connecting reactions, describes the most specific bioorthogonal reactions that are wide in scope, easy to perform, and usually employ readily available reagents that are insensitive to oxygen and water (Kolb et al., 2001)(Hawker and Wooley, 2005; Kolb and Sharpless, 2003; Wu et al., 2004). In vivo bioorthogonal chemistry and click chemistry therefore overlap quite a bit, reflecting the same underlying chemical principles applied in somewhat different ways toward the discovery or development of molecular function and information.

To meet stringent requirements of rate, selectivity, and biocompatibility, the development of bioorthogonal reactions proceeds through several steps. First, of course, is the identification or invention of a highly specific ligation process that works well in water. Potential problems associated with reactant/product stabilities and reaction biocompatibility must be anticipated and addressed. The reaction is first optimized “in the flask,” where the fundamental scope, limitations, and mechanistic modifications are explored. Then the reaction is tested in a variety of biological environments, escalating in complexity from aqueous media to biomolecule solutions to cultured cells. The most optimized transformations are then tested and employed in living organisms and animals (Sletten and Bertozzi, 2011b). The reactions highlighted in the following section are at different stages of development towards the ultimate goal of in vivo application. Second-order rate constants for bioorthogonal reactions reported to date span ten orders of magnitude, with the fastest labeling reactions reaching rates up to 10⁵ M⁻¹s⁻¹.

This perspective provides a critical review of emerging bioconjugation strategies with comments on their general utility and challenges. We recommend several excellent published reviews for more comprehensive accounts of the underlying chemistries (Lang and Chin, 2014; Patterson et al., 2014; Prescher and Bertozzi, 2005; Sletten and Bertozzi, 2009). We also provide a few examples of emerging areas of applications such as triggered release and prodrug activation.

B. Bioorthogonal Conjugation Strategies and Applications

Typically, bioorthogonal reactions are exothermic by virtue of their highly energetic reactants or strongly stabilized products, and most involve the formation of carbon-heteroatom (mostly N, O and S) bonds. The reactions are typically fusion processes that leave no byproducts or release nitrogen, or are condensation processes that produce water as the only byproduct. Most applications involve a two-step approach that requires introduction of bioorthogonal reporters by chemical, metabolic, or genetic means, followed by modification of the biomolecule with the bioorthogonal labeling reaction (Figure 1). From a practical point of view, a bioorthogonal process finds useful application if the reporter group is stable inside the cell and can be introduced by systemic distribution, metabolically, or through a promiscuous enzymatic pathway (Sletten and Bertozzi, 2009). Most bioorthogonal reactions reported in the literature are chemoselective with respect to many, but not all, biological functionalities, and have been used to label protein in vitro or at the cell surface. Only a very few have been used in much more complex systems inside living cells or animals: the field, therefore, is really only just getting started.

Figure 1.

Strategies for introducing and conjugating functionality to biomolecule targets. (A) The two-step bioorthogonal labeling strategy, in the first step exogenous functionality is introduced either genetically, metabolically or chemically, and the second step involves highly specific bioorthogonal reaction. (B) Site-specific bioconjugation based on native functionalities present in proteins.

1. Site-specific bioconjugation based on native functionalities

Intrinsic bioconjugation reactions selectively target side chains and termini of the 20 naturally occurring proteogenic amino acids (Ban et al., 2013). Classical reactions such as thiol-maleimide additions or amide-activated ester acylations are widely used for derivatizing proteins in vitro, but are typically not sufficiently specific or rapid to modify a particular biomolecule in a higher complexity cellular environment. Carbodiimide-based methods for coupling of carboxylates are much less popular, probably for reasons of positional selectivity or the potential for undesired protein crosslinking, although there are certainly exceptions (Schlick et al., 2005). As applications of bioconjugates become more complex, the development of reliable chemoselective methods for introducing functional molecules into the natural amino acid residues of peptides or proteins with high site-selectivity has become important. The addressable functional groups can be introduced as unnatural amino acids (Wang et al., 2006), metabolic precursors, or installed by bioconjugation reactions. In the last category, the traditional processes, which usually proceed at low coupling rates or with poor chemoselectivity, have started to yield to new, highly selective organic transformation and/or mechanistic modifications that improve ligation kinetics and the stability of the resulting conjugates.

a. Emerging approaches for lysine modifications

Lysine, arginine, histidine, and N-terminal amines contribute to proteins positive charge at neutral pH and are often solvent exposed, making them amenable to chemical modification via acylation or alkylation (Table 1, entries 1–4). While primary amines readily undergo reactions with activated esters, anhydrides, carbonates, isothiocyanates and a range of other acylating and alkylating agents at slightly alkaline pH, these reactions are not site-specific nor protein specific. For example, N-hydroxysuccinimide (NHS) esters of a variety of types can cross-react with the side-chain hydroxyl groups of tyrosine, serine, and threonine if there is a nearby histidine to take up the NHS ester as an acyl imidazole, which in turn reacts with the –OH group resulting in an ester bond (Hermanson, 2013).

Table 1.

Recent developments in the bioconjugation of amino groups in proteins.

Entry	Residue	Reagent(s)	Product	Reference
1	Lys	X = H, SO3Na		Standard reagents; see (Hermanson, 2013)
2	Lys
3	Lys
4	Lys	typical for X = S: pH = 9.0, 4°C, 24 h
5	Lys	100 μM, pH 8.0, RT, 20 min 500 μM, 1.5 mM NalO4 pH 6.5, RT, 15 min		(Hooker et al., 2006)
6	Lys	[Cp*Ir(bipy)(H2O)]SO4 (20 μM) pH 7.4, 37 °C		(McFarland and Francis, 2005)
7	Lys	R = ligand		(Tanaka et al., 2008; Tanaka et al., 2014)
8	Lys, N-term.			(Cal et al., 2012)
9	Lys, N-term.			(Diethelm et al., 2014)

A variety of promising alternative approaches have recently appeared, taking advantage of the special reactivity of aniline groups, imine/iminium intermediates, and diazonium centers. Thus, a two-stage methodology from Francis and coworkers connects the primary amine of lysine to an aniline center by virtue of a reasonably selective reaction with isatoic anhydride, followed by a highly selective oxidative coupling mediated by NaIO₄ (Table 1, entry 5) (Hooker et al., 2006). Aniline modifications to proteins can also be introduced at N-termini by acylation (Behrens et al., 2011), or can be genetically introduced using amber codon suppression, (Mehl et al., 2003), and can undergo oxidative couplings with aminophenols, mediated by NaIO₄ (Behrens et al., 2011) or more recently by ferricyanides (Obermeyer et al., 2014b).

The venerable reductive amination reaction has never been much use under biological conditions, since it requires acidic conditions to preserve the cyanoborohydride (or equivalent) reductant. However, the use of an iridium complex has been found to make this unnecessary, opening this pathway for the decoration of lysine amines (Table 1, entry 6). (Gildersleeve et al., 2008; Jentoft and Dearborn, 1979). The acidic conditions required for NaBH₃CN reduction of the aldiminium intermediate can be avoided with iridium-catalyzed transfer hydrogenation using [Cp*Ir-(bipy)(H₂O)]SO₄. Iminium intermediates can also be intercepted intramolecularly with an adjacent diene group, resulting in intramolecular tandem 6π electrocyclization followed by autooxidation and hydrolysis to yield zwitterionic pyridinium products (Table 1, entry 7) (Tanaka et al., 2008). This strategy has been employed for peptide labeling in conjugation to a cyclic RGDyK motif for visualizing α_vβ₃-integrins displayed on cell surfaces (Tanaka et al., 2014), but seems also to have promise in more complex situations if undesired conjugate addition can be avoided. Furthermore, an adjacent boronic acid center can stabilize iminoboronate intermediates derived from lysine or N-terminal amines; such bonds can be reversed upon addition of fructose, dopamine, or glutathione (Table 1, entry 8) (Cal et al., 2012).

Lastly in this category, diazonium salts are one of the most reactive types of compounds easily accessible in aqueous media. While increasingly exploited for reaction with tyrosine (see below), it has been something of a surprise that other biological nucleophiles have not been more widely explored for productive ligation reactions with diazonium salts. This lack of action has recently ended with a report from Carreira and coworkers describing the use of simple aminoterephthalate derivatives (Diethelm et al., 2014). Diatozation and reaction with protein amines is followed by intramolecular trapping of triazenyl intermediates with an adjacent ester group to give hydrolytically stable triazin-4(3H)-ones. While rate constants were not reported, the reaction appears to be roughly between the speed of amine-NHS ester and thiol-maleimide couplings, but with excellent chemoselectivity. One can expect modifications of this already-productive methodology in the near future.

b. N-terminal modifications

The N-terminal amine of peptides and proteins can be selectively modified by a variety of reactions (Table 2, entries 1–6). The α-amino group reacts with moderate to high selectively with ketenes (4:1 to >99:1 preference for α-amino group over ε-amino groups) under mild conditions (On-Yee Chan et al., 2012). N-Terminal serine or threonine residues can be oxidized with periodate (Geoghegan and Stroh, 1992), and N-terminal amines can undergo a variety of reactions with aldehydes, including pyridoxal phosphate mediated N-terminal transamination (Gilmore et al., 2006; Scheck et al., 2008; Witus et al., 2010), thiazolidine ring formation (Bernardes et al., 2013; Wade et al., 2001), and Pictet-Spengler cyclization (Sasaki et al., 2008). And, of course, native chemical ligation (NCL) and its variants exploit S-to-N acyl migration for rapid and selective N-terminal acylation (entry 7) (Dawson et al., 1994; Dawson and Kent, 2000). However, a few examples of recent variations on these themes have appeared.

Table 2.

Recent developments in the bioconjugation of N-terminal amines in proteins.

Entry	Residue	Reagent(s)	Product	Reference
1	Any	typical: pH 6.3, RT		(On-Yee Chan et al., 2012)
2	Ser, Thr	NaIO4		(Geoghegan and Stroh, 1992)
3	Pro	K3Fe(CN)6 pH 7.5, 30 min		(Obermeyer et al., 2014a)
4	Any			(Gilmore et al., 2006; Scheck et al., 2008; Witus et al., 2010)
5	Cys			(Bernardes et al., 2013; Wade et al., 2001)
6	Trp			(Sasaki et al., 2008)
7	Cys			(Dawson et al., 1994; Dawson and Kent, 2000; Esser-Kahn and Francis, 2008)
8	Cys	typical: pH 7.4		(Ren et al., 2009)
9	Ser			(Levine et al., 2014)

Thus, Rao and coworkers reported the development of a thiol-based ligation between 2-cyanobenzothiazole (CBT) and D-cysteine (Table 2, entry 8) (Ren et al., 2009; Van de Bittner et al., 2013). The reaction works well under physiological conditions with a second order rate constant of 9.1 M⁻¹s⁻¹ (Ren et al., 2009; Van de Bittner et al., 2013). It was subsequently applied to protein labeling (Ren et al., 2009) and imaging protease activity in living cells (Liang et al., 2010), and most recently for the self assembly of a fluorescent molecule in apoptotic cells for imaging chemotherapeutic efficacy in vivo (Ye et al., 2014). An intriguing NCL variation uses salicylaldehyde as an adapter to marry N-terminal serine with a a PEG derivative (entry 9) (Levine et al., 2014).

c. Chemoselective connections to arginine, histidine, tryptophan, and tyrosine sidechains

Site specific conjugation to these amino acids is less common, and seems even today to be dominated by new applications of old chemistry – a characteristic of many click and bioconjugation protocols. For example, strategies for the addressing of arginine residues are rare, as the strongly basic guanidino group (pK_a > 12.0) is almost always protonated under physiological conditions. Among the oldest known reactions at arginine is condensation with α,β-dicarbonyl compounds, but this is usually much too slow for efficient bioconjugation: bovine serum albumin requires 14 days for room-temperature Maillard reactions with 100 mM methylglyoxal at pH 7.4 (Oya et al., 1999). In spite of its low rate, this process has been applied to the labeling of arginine residues of lysozymes (Table 3, entry 1) (Gauthier and Klok, 2011). While the long-known reactivity of histidine with vinylsulfones continues to receive productive attention (del Castillo et al., 2014), histidine residues are most commonly exploited in a chemoselective sense for the connection of proteins to complexed metal ions (Co³⁺, Ni²⁺, Cu²⁺), metal surfaces, and metal nanoparticles; such binding interactions are beyond the scope of this review.

Table 3.

Recent developments in bioconjugation at Arg, Met, Trp and Tyr residues.

Entry	Residue	Reagent(s)	Product	Reference
1	Arg			(Gauthier and Klok, 2011)
2	Met			(Kramer and Deming, 2012)
3	Trp	Rh2(OAc)4, t-BuNHOH pH 6.0		(Antos and Francis, 2004; Antos et al., 2009)
4	Tyr	Pd(OAc)2 (40 μM) TPPTS (0.48 mM) pH 8.5–9		(Chen et al., 2009; Tilley and Francis, 2006)
5	Tyr	pH 9, 4 °C, 15 min Na2S2O4 (100 mM) pH 7.2, r.t., 2 h NalO4		(Hooker et al., 2004)
6	Tyr	pH 9, 4 °C, 15 min pH 6, r.t., 2 h		(Gavrilyuk et al., 2013; Schlick et al., 2005)
7	Tyr	pH 6.5, r.t., 18 h		(Joshi et al., 2004; McFarland et al., 2008; Romanini and Francis, 2007)
8	Tyr			(Ban et al., 2010; Ban et al., 2013; Jessica et al., 2013)
9	Tyr	Ni(OAc)2, Gly-Gly-His MMPP, pH 7		(Kodadek et al., 2005; Meunier et al., 2004)
10	Tyr	[Ru(bpy)3]Cl2, (NH4)2S2O8		(Sato and Nakamura, 2013)
11	Tyr	(NH4)2Ce(NO3)6 pH 6, 1 h, r.t.		(Sato and Nakamura, 2013; Seim et al., 2011)
12	Tyr	IPyBF4, H2O; Pd(OAc)2, TPPTS K3PO4, ArBF3, H2O		(Vilaró et al., 2008)
13	Tyr	Tyrosinase, O2		(Long and Hedstrom, 2012)

More exciting are the renewed appreciation of the mild nucleophilicity of methionine (thioether), tryptophan (indole), and tyrosine (phenol) sidechains as potential sites for selective bond formation. Kramer and Deming have returned to simple alkylation as a method to address methionine, finding that the resulting sulfonium ions are often quite stable to aqueous base, heat, and other extreme conditions (Table 3, entry 2) (Kramer and Deming, 2012, 2013). So far, this chemistry has not found wide application, being restricted in the reported cases to polypeptides containing many methionine residues, but it is promising. The Francis lab has developed a tryptophan modification using rhodium carbenoids in aqueous solution (Antos and Francis, 2004) over a broad pH range (entry 3) (Antos et al., 2009). This rhodium catalysis has been applied to site-specific modification of aromatic side chains (Trp, Tyr and Phe) using metallopeptides capable of molecular recognition (Antos et al., 2009; Chen et al., 2011).

Tyrosine contains an ionisable phenolic side chain (pKa 9.7–10.1) that is often solvent exposed (Koide and Sidhu, 2009), and can be site-specifically modified by a number of reactions (Jones et al., 2014). For obvious reasons of selectivity, the direct alkylation of the phenolic oxygen is quite rare, but it has been recently managed by Pd-catalyzed alkylation (Jones et al., 2014; Tilley and Francis, 2006) or prenylation^58,59 via Claisen rearrangement of the O-prenylated precursor, mediated by prenyltransferase enzymes (Table 3, entry 4) (McIntosh et al., 2013; Rudolf and Poulter, 2013). The aromatic ring is a more distinctive nucleophile, however, and diazonium salts have long been known to react with phenols to make azo dyes. Variations on this reaction have also been used to label tyrosine aromatic rings with moderate rate and high selectivity (Table 3, entries 5–6) (Bruckman et al., 2008; Hooker et al., 2004; Jones et al., 2012; Schlick et al., 2005). The resulting azo intermediate can be processed in a number of ways, such as by oxidation to o-iminoquinone for hetero-Diels-Alder reaction (Hooker et al., 2004), or simply kept as a connector to another functional group that can be addressed by a second bioorthogonal reaction (such as oxime formation) (Gavrilyuk et al., 2013; Schlick et al., 2005). Similarly, a three component Mannich reaction between tyrosine and imines formed from aldehydes and electron rich anilines has also been described (entry 7) (Joshi et al., 2004), although it is relatively slow and requires a large excess of regents (McFarland et al., 2008; Romanini and Francis, 2007). The aqueous ene-type reaction of tyrosine with cyclic diazodicarboxamides seems more generally promising (Ban et al., 2010; Ban et al., 2013; Jessica et al., 2013), which reacts selectively at the o-position of the phenol side chain, and was shown to be thermally and hydrolytically stable (entry 8).

Quite a number of tyrosine functionalization processes rely on one-electron oxidation as an initial activating step, selectivity arising from the fact that phenol is the most easily oxidized common sidechain. Thus, nickel(II)-catalyzed oxidative coupling of two phenols can be used for protein crosslinking (Kodadek et al., 2005), and has been followed by bioconjugation on tyrosine residues present on the capsid proteins of virus-like particles (entry 9) (Meunier et al., 2004). Similarly, ligand directed single electron transfer from a photocatalyst ([Ru(bpy)₃]²⁺) bound to a protein ligand results in the efficient formation of tyrosyl radicals, which can be trapped by agents such as N′-acetyl-N,N-dimethyl-1,4-phenylenediamine (entry 10) (Sato and Nakamura, 2013). Oxidative conjugation of tyrosines, including O-alkylation, with various peptides containing phenylenediamine and anisidene moieties has been accomplished with cerium(IV) ammonium molybdate (entry 11) (Seim et al., 2011). A milder oxidant, potassium ferricyanide, has also been shown to be capable of coupling anilines with o-aminophenols on proteins, but without contaminating thiol oxidation (Obermeyer et al., 2014b). Electrophilic iodination of tyrosines with ¹²⁵I and ¹³¹I is also useful (Schumacher and Tsomides, 2001), since the products can undergo aqueous Suzuki-Miyaura cross coupling (Vilaró et al., 2008). Tyrosine can also be iodinated at the ortho position using bis(pyridine)iodinium tetrafluoroborate (IPy₂BF₄, entry 12) (Chalker et al., 2010; Espuña et al., 2006) for subsequent palladium catalyzed cross coupling.(Chalker et al., 2010; Chalker et al., 2009c; Espuña et al., 2006) Particularly attractive for potential in vivo applications is oxidation by tyrosinase to the corresponding o-quinones, followed by selective nucleophilic trapping (Long and Hedstrom, 2012).

d. Cysteine Modifications

The thiol group of cysteine side chains is one of the most widely used functional groups for the synthetic modification of peptides and proteins (Chalker et al., 2009a), since thiols are less abundant and more reactive than amines (Fodje and Al-Karadaghi, 2002). Cysteine residues are rarely found on highly solvent-accessible surfaces of a protein; most conjugation reactions therefore require pretreatment with a reducing agent such as dithiothreitol (DTT) to cleave an accessible disulfide or reduce an oxidatively passivated cysteine thiol (as in a sulfenic acid). Many of the electrophiles traditionally used with thiols (iodoacetamides, maleimides, vinyl sulfones, vinylpyridines, epoxides; Table 4, entries 1–4) (Chalker et al., 2009a; Stenzel, 2012) are subject to competing reactions with other nucleophilic amino acids (usually lysine and histidine) to an extent that can be highly variable and occasionally substantial (Baldwin and Kiick, 2011; Heukeshoven, 1980; Lewis and Shively, 1998)(Alley et al., 2008; Shen et al., 2012). Disulfide exchange is another common option for modifying cysteine (entry 5) (Pollack and Schultz, 1989), provided that there is thermodynamic preference for the disulfide on the protein. However, this method suffers from sluggish rates and limited control of product distributions.

Table 4.

Recent developments in bioconjugation at Cys residues.

Entry	Residue	Reagent(s)	Product	Reference
1	Cys			Standard reagents; see (Chalker et al., 2009a; Hermanson, 2013; Kim et al., 2008)
2	Cys
3	Cys
4	Cys
5	Cys
6	Cys			(Marculescu et al., 2014)
7	Cys			(Bryden et al., 2014)
8	Cys			(Koniev et al., 2014)
9	Cys			(Abbas et al., 2014)
10	Cys, Lys			(Hong et al., 2009a; Kislukhin et al., 2011; Kislukhin et al., 2012)
11	Cys	pH 5.5–8.0		(Toda et al., 2013)
12	Cys	pH 6–8 NaBH4		(Badescu et al., 2014)
13	Cys	AuCl/AgOTf CH3CN/H2O		(On-Yee Chan et al., 2013)
14	Cys			(Crich et al., 2006)
15	Cys			(Spokoyny et al., 2013)
16	Cys			(Hoppmann et al., 2012; Hoppmann et al., 2014)
17	Cys	(50 mM) pH 10–11		(Bernardes et al., 2008)
18	Cys	hν or AIBN		(Dondoni et al., 2009; Li et al., 2012)

Maleimide chemistry remains the subject of continual investigation and modification. Frequently regarded as a reliable connection (Bednar, 1990; Kim et al., 2008), the thiol-maleimide conjugate addition product has recently been shown by Baldwin and Kiick to be labile or exchangeable under physiologically-relevant conditions (Baldwin and Kiick, 2011). While uncontrolled or contaminating cleavage can often be a problem, engineered cleavability can be an advantage for applications that require the release of a cargo from a carrier protein, for example. The incorporation of a leaving group such as alkoxide into the maleimide changes the reaction mechanism with thiol from conjugate addition to addition-elimination reaction (Table 4, entry 6), creating a class of maleimides that offers reversible modification and up to three conjugation points. Introducing two leaving groups (entry 7) allows a disulfide linkage to be conveniently replaced with a bridging cyclic structure (Bryden et al., 2014). Cysteine thiol groups also undergo addition reactions with electron-deficient alkynes (Shiu et al., 2009) such as alkynones and alkynoate amides or esters, the adducts of which also have the capability to undergo exchange with additional thiol. Modifications to these electrophiles, such as 3-aryl propionitriles (entry 8) (Koniev et al., 2014) and allenamides (entry 9) (Abbas et al., 2014) are designed to retain high chemoselective cysteine tagging in complex aqueous media, while stabilizing the resulting adducts toward thiol exchange and hydrolysis.

Another alternative to maleimides are electron-deficient oxanorbornadienes (entry 10), which display thiol addition rates similar to those of maleimides, but have greater aqueous stability (Hong et al., 2009a; Kislukhin et al., 2011; Kislukhin et al., 2012). An additional fragmentation feature of this linkage is discussed at the end of this article. Julia-Kocienski like sulfone derivatives (Toda et al., 2013) based on methylsulfonyl-functionalized heteroaromatic derivatives (entry 11) have been used for rapid protein/peptide conjugation and their conjugates were found to be more stable than the maleimide-cysteine conjugates in human plasma. Similarly, latently reactive PEG-di-sulfone (Balan et al., 2007) and PEG-mono-sulfone (Badescu et al., 2014), have been used for chemoselective cysteine conjugation. The PEG mono-sulfone reagent undergoes facile elimination of toluene sulfinic acid under mild conditions to generate the reactive α,β-unsaturated PEG reagent for conjugation (entry 12). Retro-Michael de-conjugation can be prevented by treating the ketone with a mild reducing agent such as sodium borohydride. These reagents are highly selective for cysteine over lysine.

The range of thiol-reactive functional groups explored for bioconjugation continues to expand. Thiols undergo highly specific conjugation with allenes (in the presence of gold catalyst, entry 13) (On-Yee Chan et al., 2013) or allyl selenosulfate salts (entry 14) (Crich et al., 2006). Positional selectivity has been elegantly achieved by genetically encoding reactive azobenzene-electrophile groups onto proteins (entry 15). Site-specific linkage can occur with a nearby cysteine residue in proximity-enabled fashion (Hoppmann et al., 2012; Hoppmann et al., 2014). Such bridge formation is highly specific even in the presence of other thiols groups on the scaffold, and the light-driven isomerization of the azobenzene bridge from the trans to the cis conformation can be used to change protein conformation and activity. Among the most interesting electrophiles are perfluoroaromatic compounds, identified by Pentelute and coworkers to react cleanly with cysteine residues under mild conditions (entry 16). This strategy has been applied to site-selective modification of cysteine residues in unprotected peptides, serving as a molecular staple (Spokoyny et al., 2013) for the synthesis of hybrid macrocyclic peptides (Zou et al., 2014). Remarkably, S_NAr reactions can also be catalyzed by naturally occurring glutathione S-transferase for efficient ligation between a polypeptide and an N-terminal glutathione sequence (Zhang et al., 2013). In this way, unique chemical orthogonality can be achieved, modifying multiple cysteine sites with different chemical probes while avoiding protecting groups and additional synthetic procedures.

The cysteine residue can be rendered electrophilic as well. Oxidative elimination by treatment with O-mesitylenesulfonyl hydroxylamine (MSH) under basic conditions affords dehydroalanine (Bernardes et al., 2008), which can be modified by conjugate addition of thiols (entry 17) (Chalker et al., 2009a), or can undergo olefin cross-metathesis with ruthenium catalysts (Lin et al., 2008). Additionally, thiol-ene click reactions that are so useful in materials contexts have also been used for bioconjugation (entry 18), as in the glycosylation of a single cysteine moiety on glutathione (Dondoni et al., 2009) or the labeling of alkyne-bearing proteins expressed in E. coli (Li et al., 2013b; Li et al., 2012).

2. Bioorthogonal reactions based on extrinsic functionalities

a. Transimination reactions of ketones/aldehydes with hydrazides/alkoxyamines

Among the first functionalities to be explored as extrinsic bioorthogonal reactants were ketones and aldehydes (Rideout, 1986). Although they are not completely abiotic and are present in some intracellular metabolites, their synthetic accessibility and small size has made them amenable to incorporation into biomolecules. Popular routes include uptake and processing of unnatural intermediates of glycan biosynthesis (Jacobs et al., 2000; Luchansky et al., 2004; Mahal et al., 1997; Sadamoto et al., 2004), genetically-controlled incorporation into proteins in a residue-specific (Datta et al., 2002; Ngo and Tirrell, 2011; Tang et al., 2009) or site-specific manner (Carrico et al., 2007; Chin et al., 2003; Hudak et al., 2012; Hudak et al., 2011; Liu and Schultz, 2010; Wang et al., 2003a; Wu et al., 2009), and use in peptide tags that direct enzymatic ligation of aldehyde or ketone bearing small molecules (Chen et al., 2005; Rashidian et al., 2012). Aldehydes can also be introduced into proteins by periodate oxidation of N-terminal Ser/Thr residues (Geoghegan and Stroh, 1992), site-specific protein transaminations (Geoghegan et al., 1993; Gilmore et al., 2006; Witus et al., 2010; Witus et al., 2013), periodate oxidation of sialic acids displayed on proteins (Hage et al., 1997), and by addition of ketone containing small molecules to protein C-terminal thioesters generated by expressed protein ligation (Esser-Kahn and Francis, 2008).

Once incorporated, the carbonyl group can be addressed by a number of highly specific methods, as illustrated in Figure 2. The discovery by Dawson and coworkers of aniline catalysis of the click reaction with α-effect nucleophiles such as hydrazine and aminooxy groups at pH 4–6 has made this popular process even more effective (Figure 2A) (Gaertner et al., 1992; Jencks, 1959; Sander and Jencks, 1968), with the organocatalytic process proceeding through a highly reactive Schiff base (Cornish et al., 1996; Dirksen and Dawson, 2008; Dirksen et al., 2006a; Dirksen et al., 2006b; Jencks, 1959; Rashidian et al., 2013; Ulrich et al., 2014). To improve biocompatibility and increase reaction rate at pH 7, substituted anilines (Dirksen and Dawson, 2008; Dirksen et al., 2006b; Rashidian et al., 2013; Wendeler et al., 2013) such as water-soluble 5-methoxyanthranilic, 3,5-diaminobenzoic (Crisalli and Kool, 2013b), and 2-aminobenzenephosphonic acid (Crisalli and Kool, 2013a) have been found to accelerate the condensation reaction by up to 40-fold at pH 7 as compared to the aniline-catalyzed reaction. While the electronic and acid/base properties of the nucleophilic reactants strongly influence the rate at biological pH (Kool et al., 2014), it must be remembered that all of these ligations are reversible, and that hydrazones are especially prone to dissociation at low concentrations (Dirksen et al., 2006a; Dirksen et al., 2006b; Kalia and Raines, 2008), whereas oximes are more stable (Kalia and Raines, 2008). Hydrazone and oxime condensations are therefore best suited for in vitro studies or reactions at cell surfaces (Tuley et al., 2014), since the necessary reactant concentrations can be too high to achieve intracellularly and α-effect nucleophiles can undergo reactions with carbonyl-bearing metabolites such as pyruvate, oxaloacetate, sugars, and various cofactors. The Pictet-Spengler reaction between aldehydes and tryptamine nucleophiles generates hydrolysis-resistant oxacarbolines via an intermediate oxyiminium ion in acidic environments, followed by spontaneous cyclization (Figure 2B) (Agarwal et al., 2013a; Agarwal et al., 2013b; Stöckigt et al., 2011). A recent variant that joins glyoxyl-modified peptides with an aminobenzamidoxime proceeds at rates that are comparable to the hydrazino-Pictet Spengler reaction (3 M⁻¹s⁻¹ at pH 6.0, Figure 2C) (Kitov et al., 2014). Additionally, aldehydes have been shown to undergo oxidative condensation with aryldiamines in the presence of Cu(II) or Zn(II); this process was employed for the chemical modification of a arylenediamine-containing T4 lysozome (Figure 2D) (Ji et al., 2014).

Figure 2.

Polar condensation reactions of carbonyl compounds.

b. Azide-based bioorthogonal reactions

Azides are essentially absent from biological systems (Griffin, 1994) and are unreactive with biological functionality, and so have become one of the premiere bioorthogonal participants (Debets et al., 2010b). Azides are small, not very polar, and incapable of significant hydrogen bonding, making them unlikely to significantly change the properties of structures to which they are attached. They can be easily introduced into biomolecules such as glycans (Baskin et al., 2007; Chang et al., 2010; Laughlin et al., 2008), proteins (Beatty et al., 2010), lipids (Hang et al., 2007; Kho et al., 2004), and nucleic acids by biosynthetic pathways. Azide modified proteins have been used to label biomolecules in living systems and animals. Likewise, alkynes are abiotic and can be easily introduced into biological molecules (Grammel and Hang, 2013; Johnson et al., 2010; Ngo and Tirrell, 2011).

Although only about 15 years old (Saxon et al., 2000), the Staudinger-inspired reaction between azides and phosphines to yield amide linkages developed by Bertozzi and coworkers is now a classic among bioorthogonal processes and remains highly useful (Kiick et al., 2002; van Berkel et al., 2011). The byproduct phosphine oxide can remain attached to, or can be released from, the final product (Figure 3) (Nilsson et al., 2000; Saxon et al., 2000). While the kinetics are sluggish (k₂ approximately 3 × 10⁻³ M⁻¹ s⁻¹) (Lin et al., 2005) and the phosphine reagents prone to oxidation by air or metabolic enzymes, the Staudinger ligation is a popular choice for in vivo work due to its remarkable selectivity and compatibility with everything from cells (Agard et al., 2006; Chang et al., 2007; Saxon and Bertozzi, 2000) to animals (Prescher et al., 2004).

Figure 3.

Non-traceless and traceless Staudinger ligations.

Copper-catalyzed azide-alkyne cycloaddition (CuAAC, Figure 4) (Rostovtsev et al., 2002; Tornøe et al., 2002; Worrell et al., 2013) has emerged as a highly versatile ligation process for bioconjugation applications of all types, except in living cells and tissues where free copper ions are toxic. Such operations as the immobilization, labeling, or capture of proteins, nucleic acids, and polysaccharides (Moses and Moorhouse, 2007) benefit from the small size and unobtrusive nature of the terminal alkyne and azide moieties. Among many other applications, the CuAAC process has been used to label the outer (Banerjee et al., 2010; Banerjee et al., 2011; Hong et al., 2009b; Wang et al., 2003b) and inner (Abedin et al., 2009; Hovlid et al., 2014) surfaces of azide- and alkyne-functionalized virus particles, modify proteins in vitro and in vivo (Deiters et al., 2003; Liu and Schultz, 2010; Ngo and Tirrell, 2011), tag azide-modified lipids (Hang et al., 2011), and label azide- or alkyne-modified nucleic acids. The CuAAC reaction proceeds at considerably faster rates than the Staudinger ligation (more than 25-fold) and is usually 10–100 times faster than copper-free strain-promoted azide-alkyne cycloaddition (discussed below) (Jewett et al., 2010; Presolski et al., 2010), with second-order rate constants of 10–200 M⁻¹s⁻¹ in the presence of 20–500 μM Cu^I.

Figure 4.

Key features of the Cu^I-catalyzed azide-alkyne cycloaddition (CuAAC) reaction. (A,B) Overall process and mechanistic overview of the CuAAC catalytic cycle. (C) Accelerating ligands commonly used in the CuAAC reaction. (D) Strategies to accelerate rates by chelation of the azide component. (E) Cu-promoted generation of reactive oxygen species. (F) Dehydroascorbic acid mediated cross-linking of protein side chains.

With the proper selection of copper-binding ligand (Hong et al., 2009b; Presolski et al., 2010), CuAAC reactions proceed rapidly within richly functionalized biological environments such as cell lysates. Indeed, the labeling of groups incorporated onto the outer surface of live cells can be managed very efficiently with the best catalysts (Besanceney-Webler et al., 2011; Hong et al., 2010). Note, however, that CuAAC bioconjugations usually require greater amounts of Cu^I and accelerating ligand than the biomolecular azide and alkyne, which are usually present in nanomolar-micromolar concentrations.

Oxidative stress and biological damage associated with CuAAC has been attributed to Cu^I-promoted generation of reactive oxygen species (ROS) from O₂ (Brewer, 2010; Hong et al., 2010; Wang et al., 2003b). Catalytically active Cu^I is not stable under physiological conditions, and the oxidation of Cu^I to Cu^II, by either O₂ or H₂O₂ (via Fenton processes) facilitates the production of superoxide and hydroxyl radicals, respectively (Figure 4E) (Biaglow et al., 1997; Tabbi et al., 2001). The presence of ROS can affect the structural and functional integrity of biomolecules, causing degradation of amino acids and cleavage of peptide chains (Stadtman, 2006), and this has been observed under CuAAC conditions (Hong et al., 2009b; Kennedy et al., 2011; Kumar et al., 2011). Furthermore, the ascorbate reducing agent used in these reactions can also do damage by virtue of the electrophilic properties of its oxidized form (dehydroascorbate), which can react with lysine, arginine (Reihl et al., 2004), and cysteine (Kay et al., 2013) side chains leading to protein crosslinking (Figure 4F) (Corti et al., 2010). Additives such as excess Cu-binding ligand and aminoguanidine are used as ROS and dehydroascorbate scavengers, respectively (Hong et al., 2009b).

CuAAC is a ligand-accelerated process (Chan et al., 2004; Hong et al., 2009b), and a variety of ligands have been found to be effective under different conditions. In general, such ligands enhance cell compatibility by stabilizing the Cu^I oxidation state and increasing its catalytic efficiency. The most widely used ligands (Figure 4C) include sulfonated bathrophenanthroline (BPDS) (Lewis et al., 2004), tris(benzimidazole) (BimC₄A)₃ (Rodionov et al., 2007b), tris-triazole TBTA (Chan et al., 2004), water soluble THPTA (Hong et al., 2009b), BTTES (Soriano del Amo et al., 2010), BTTAA (Besanceney-Webler et al., 2011), and bis(L-histidine) (Kennedy et al., 2011). Chelation-assisted copper catalysis is a recent and complementary approach to ligand-based acceleration (Brotherton et al., 2009; Kuang et al., 2011; Kuang et al., 2010), designed to enhance the “weakest link” in the CuAAC mechanism: the association of azide with the metal center (Rodionov et al., 2007a). For example, a bidentate picolyl azide (shown in Figure 4D) was used in combination with BTTAA for site-specific conjugations of proteins on the surface of live cells with metabolically labeled RNA and protein molecules, providing 25-fold enhancement in specific labeling rate relative to conventional non-chelating azides at low copper concentrations (10–100μM) (Jiang et al., 2014; Uttamapinant et al., 2013; Uttamapinant et al., 2012). Azides bearing tighter binding ligands such as tetradentate bis(triazolyl)amino azide, have also exhibited outstanding reactivity with alkynes under dilute conditions in the presence of complex cellular media, and was used for the tracking of paclitaxel inside living cells (Bevilacqua et al., 2014).

The most effective way to improve the biocompatibility of azide-alkyne cycloaddition has been to eliminate the requirement of Cu^I catalysis and accelerate the reaction with alkynes that are activated by ring strain. The strain-promoted azide-alkyne cycloaddition (SPAAC, Figure 5A) (Agard et al., 2006; Agard et al., 2004) has been used effectively in a variety of in vivo contexts, including in mammalian cells (Agard et al., 2004; Baskin et al., 2007) and animals (Baskin et al., 2007; Laughlin et al., 2008). Significant effort has been focused on increasing the rate, biocompatibility and pharmacokinetic properties of the cycloooctyne reagents used in SPAAC. An optimal balance between reactivity and stability must be achieved for effective performance in vivo. The reactivity of the cyclooctyne can be modulated (Figure 5B) by appending electron withdrawing groups at the propargylic position (MOFO, DIFO) (Agard et al., 2006; Baskin et al., 2007), or by augmentation of strain energy through aryl ring (DIBO, DIBAC and BARAC) (Debets et al., 2010a; Jewett et al., 2010; Ning et al., 2008) or cyclopropyl (BCN) (Dommerholt et al., 2010) ring fusion (Sletten et al., 2014). The superior reaction rate of BARAC is counterbalanced by its instability toward hydrolysis in phosphate buffered saline (t_1/2= 24 h), and tendency for intramolecular rearrangement under acidic conditions (Chigrinova et al., 2013). Additionally, cyclooctynes have been reported to undergo nucleophilic addition with cellular nucleophiles such as glutathione (Beatty et al., 2010; Chang et al., 2010), homotrimerization (Sletten et al., 2010), and reaction with cysteine sulfenic acids (with trapping rates that are an order of magnitude faster than SPAAC and exceed the rates of previously sulfenic acid capture reactions by more than 100-fold) (Figure 5C,D) (Poole et al., 2014).

Figure 5.

Strain-promoted 1,3-dipolar cycloaddition reactions useful for bioconjugation. (A) Strain-promoted azide-alkyne cycloaddition (SPAAC). (B) Common cyclooctyne reagents and associated rate constants for reaction with benzyl azide. (C,D) Side reactions of strain-promoted click chemistry reagents or products. (E) Alternative strain-promoted 1,3-dipolar cycloaddition reactions.

The major constraint on the use of SPAAC in bioconjugation derives from the relatively large size and hydrophobic nature of the cyclooctyne component, which can affect their distribution and change the biological properties of the species to which they are attached. Despite the synthetic tractability of the dibenzocyclooctynes, and kinetic benefits of aza analogs thereof, the original difluorinated cyclooctynes remain useful reagents for copper-free click chemistry for this reason. DIFO is less sterically encumbered than the benzannulated variants and has been shown to be more effective for site-specific incorporation into proteins. In one case, the larger dibenzocyclooctyne could not be site-specifically incorporated into proteins through amber stop codon techniques (Plass et al., 2011). A different early limitation of SPAAC, the requirement for independent synthesis of cyclooctynes, has been largely eliminated for most users by the commercial availability of a variety of derivatives, including those with pendant and orthogonal reactive functionality such as NHS ester or maleimide. Among the many recent productive uses of SPAAC methodology are the derivatization of purified proteins in vitro, (Agard et al., 2004), glycans on cell surfaces for probing sialic acid biosynthesis (Beatty et al., 2010), and in vivo experiments with zebrafish (Dehnert et al., 2012), the C. elegans nematode (Laughlin and Bertozzi, 2009), and mice (Chang et al., 2010) following biological uptake of ManNAz, now also commercially available.

Alternative 1,3-dipoles such as nitrones (McKay et al., 2011; McKay et al., 2010; Ning et al., 2010), nitrile oxides (Sanders et al., 2011; Singh and Heaney, 2011), diazoalkanes (McGrath and Raines, 2012), and syndones (Wallace and Chin, 2014) have also been explored as means for improving the kinetics and biocompatibility of reactions with multiple-bond partners (primarily strained alkynes, Figure 5E). Strain-promoted alkyne-nitrone cycloaddition (SPANC) reactions demonstrate rate constants up to 60 M⁻¹s⁻¹ (McKay et al., 2012; McKay et al., 2010; Ning et al., 2010), and have been used for N-terminal peptide modification (Ning et al., 2010), direct protein labeling, and pre-targeted labeling of ligand-receptor interactions on cell surfaces (McKay et al., 2011). Cyclic nitrones display greater stability toward hydrolysis and faster kinetics than their acyclic counterparts, and nitrone reactivity is tunable to allow for simultaneous SPANC reactions for multiplex labeling (MacKenzie and Pezacki, 2014; MacKenzie et al., 2014). Nitrile oxides have also been explored as alternative 1,3-dipoles in reactions with cyclooctynes to yield isoxazoles, and the reaction has been applied for generating oxime containing nucleotides and peptides (Jawalekar et al., 2011) as well as carbohydrates (Sanders et al., 2011). Nitrile oxides must be generated in situ due to their more reactive nature. Additionally, syndones have been used as 1,3-dipoles in Cu^I catalyzed cycloadditions with terminal alkynes (Kolodych et al., 2013), and more recently in strain-promoted cycloadditions with bicyclononyne (rate constant = 0.054 M⁻¹s⁻¹, comparable with that of azides). This method was applied to the efficient fluorescent tagging of cyclooctyne-modified proteins (Wallace and Chin, 2014). Recently, the rate of this cycloaddition process has been improved 30-fold by using 4-chlorosyndones (k₂ = 1.6 M⁻¹s⁻¹) (Taran et al., 2014).

Azide modified proteins have also been found to undergo [3+2] cycloaddition with strained olefin probes such as oxanorbornadienes, which form 1,2,3-triazoles upon retro-Diels-Alder fragmentation (Figure 6A) (van Berkel et al., 2007). While relatively slow (rate constants in the 10⁻⁴ M⁻¹s⁻¹ range), the reaction has been used for functionalizing azide-modified peptides (van Berkel et al., 2007) and remains attractive due to the relatively easy synthetic accessibility of the oxanorbornadiene fragment. Nitrile oxides have also been used in 1,3-diplar cycloaddition with norbornenes leading to a regioisomeric mixture of 2-isoxazolines (Figure 6B). This approach has been implemented for the fluorescent modification of oligonucleotides (Gutsmiedl et al., 2009; Singh and Heaney, 2011). While in situ generated nitrile imines are prone to hydrolysis, their 1,3-dipolar cycloaddition is faster and has been applied to the fluorescent labeling of acrylamide containing proteins (Wang et al., 2014). This reaction, however, is strongly influenced by pH and chloride ion concentration.

Figure 6.

Strain-promoted cycloadditions of (A) azides and (B) nitriles oxides with strained alkenes; (C) quadricyclanes with a nickel complex.

An alternative strain-promoted reaction involving the compact quadricyclane structure was recently reported by Sletten and Bertozzi (Sletten and Bertozzi, 2011a). In its initial form, the desired [2+2+2] cycloaddition was found to occur most productively with the π-system of a Ni-bis(dithiolene) complex (Figure 6C, rate constant = 0.25 M⁻¹s⁻¹). Quadricyclanes are quite easy to make (photochemical reaction with norbornadienes), are stable in aqueous media at room temperature, and are not known to react with biomolecules. Using this reaction, quadricyclane modified bovine serum albumin (BSA) was efficiently labeled in the presence of cell lysates, but the reaction has yet to achieve wider use.

c. Photo-inducible click chemistry

The supreme chemoselectivity of bioorthogonal reactions makes them attractive for the selective probing of molecular events in biological environments. A key component of such investigations is time, since learning when a species appears or reacts in a cell is often as useful as knowing where it is. The temporal component can be controlled to some degree by triggering the expression of a protein or nucleic acid with genetic promotors, but such modulation of the chemistry is less common. For this reason, the recent development of photochemically-triggered bioorthogonal reactions is exciting, and is certainly in its infancy (Lim and Lin, 2011). A light-induced reaction between diaryl tetrazoles and alkenes was first developed for bioconjugation by Lin, et al. (Song et al., 2008b; Wang et al., 2007), involving the photochemical ejection of dinitrogen from the tetrazole followed by rapid trapping of the resulting nitrile imine by alkene (Figure 7A) (Wang et al., 2008; Wang et al., 2007). Similarly, azirines can be photoactivated as nitrile ylides (Figure 7B), a technique that has been used for PEG modification of lysozomes (Lim and Lin, 2010a).

Figure 7.

Photo-induced 1,3-dipolar cycloadditions of 2,5-diaryltetrazoles and alkenes.

Further improvements have been made by tailoring the tetrazole precursor. Electron-donating groups serve to increase the energy of the highest occupied molecular orbital of the nitrile-imine, leading to increased reactivity with terminal alkenes (Wang et al., 2009), and extending the tetrazole conjugation with unsaturated substituents shifts the photochemical event to longer wavelength light, minimizing biomolecule damage. Introducing strain in the alkene component results in further rate enhancement (rates up to 34000 M⁻¹s⁻¹) (Lim and Lin, 2010a; Yu and Lin, 2014). The power of the technique is just starting to be explored, with initial reports of the time- and positionally-controlled labeling of alkene-containing (homoallyl glycine) proteins overexpressed in E. coli cells (Lim and Lin, 2011; Song et al., 2008a). In addition, tetrazole-amino acids site-specifically incorporated into proteins expressed in E. coli (Wang et al., 2010) and mammalian cells have also been addressed (Song et al., 2010; Yu et al., 2012).

d. Tetrazine ligations: inverse-electron demand Diels-Alder reactions

In the quest for faster rates, and therefore lower usable concentrations, for bioorthogonal connecting reactions, a new standard has been set by the inverse-electron-demand Diels-Alder reaction between 1,2,4,5-tetrazines and strained alkenes (Table 5) (Devaraj and Weissleder, 2011). Although only recently developed, this process has received wide attention, including uses of trans-cyclooctene (TCO) (Blackman et al., 2008; Taylor et al., 2011), cyclooctyne (Chen et al., 2012), cyclopropene (Patterson et al., 2012; Yang et al., 2014a; Yang et al., 2012b), norbornene (Devaraj et al., 2008; Han et al., 2010; Vrabel et al., 2013), cyclobutene (Pipkorn et al., 2009), azetidine (Engelsma et al., 2014), and isonitrile (Stairs et al., 2013) dienophiles. Reaction rates with TCO are truly spectacular (in the 10⁶ M⁻¹s⁻¹ range); other dienophiles react with slower, but still useful, speed (Table 5) (Lang et al., 2012b).

Table 5.

Tetrazine-based inverse-electron demand Diels-Alder reactions with strained alkenes and alkynes.

Tetrazine	Dienophile	Product	Rate constant	Reference
			210 – 2,800,000 M−1s−1	(Blackman et al., 2008; Taylor et al., 2011)
			0.12 – 9.5 M−1s−1	(Devaraj et al., 2008; Han et al., 2010; Vrabel et al., 2013)
			not reported	(Pipkorn et al., 2009)
			3.3 – 41 M−1s−1	(Chen et al., 2012)
			0.03 – 13 M−1s−1	(Patterson et al., 2012; Yang et al., 2014a; Yang et al., 2012b)
			0.39 M−1s−1	(Engelsma et al., 2014)
			not reported	(Stairs et al., 2013)

As these examples illustrate, the tetrazine ligation reaction has undergone significant optimization to improve the reaction kinetics, product stability, and cell permeability of the tetrazine (Karver et al., 2011) and trans-cyclooctene (Taylor et al., 2011) derivatives. Increasing the strain in the trans-cyclooctene moiety through cyclopropyl fusion resulted in a 50-fold rate enhancement (Lang et al., 2012b; Seitchik et al., 2012). Additionally, π-conjugated tetrazines exhibit strong fluorescence upon cycloadditions with dienophiles such as cyclopropenes and trans-cycloctene (Wu et al., 2014). Applications of tetrazine-TCO ligations have included labeling of newly synthesized proteins (Lang et al., 2012a) and cancer cells (Devaraj et al., 2010; Devaraj et al., 2009), in vivo cancer imaging with ¹¹¹In (Rossin et al., 2010) and ¹⁸F radiolabeling (Keliher et al., 2011; Li et al., 2010; Reiner et al., 2011), cancer cell detection (Haun et al., 2010), fluorescent imaging of cytoskeletal proteins within living mammalian cells (Liu et al., 2012), and recently the methodology has been used with amino acids modified by tetrazine (Seitchik et al., 2012) and trans-cyclooctene (Lang et al., 2012b) for genetic incorporation into proteins. The favorable kinetics of tetrazine-trans-cyclooctene ligation will make it particularly useful for tracking fast biological processes and labeling low-abundance proteins.

The availability of the tetrazine component has been dramatically improved by new synthetic methods (Wu et al., 2014; Yang et al., 2012a; Yang et al., 2014a) and by the commercial availability of a few derivatives. The trans-cyclooctene component, like cyclooctynes used for SPAAC, suffers from some disadvantages of size and hydrophobicity. However, tetrazines react reasonably well with the much smaller moieties listed above (Table 5), making the components of the tetrazine ligation reaction not much more perturbing than azides and alkynes used in CuAAC, although probably a bit more cross-reactive with other functional groups in biology.

e. Bioorthogonal organometallic reactions

Although the application of transition metals in medicine has been investigated for centuries, their use to mediate chemoselective transformations is much more recent (Yang et al., 2014b). The nature of late transition metals makes them well suited to the manipulation of unsaturated and polarizable functional groups (olefins, alkynes, aryl iodides, arylboronic acids, etc.) and to the retention of reactivity in aqueous environments. The presence of oxygen and other biological oxidants might have been anticipated as a general impediment, but a variety of procedures are known (and can be anticipated) that are not sensitive to oxidation. Notable successes at present are catalytic cycles mediated by Pd^II/Pd^IV species, and the preservation of Cu^I species in reducing environments (although not without cost, as described above for CuAAC chemistry). Much more challenging for the practical application of organometallic reactions in biology, however, is the presence of relatively high (mid-micromolar) concentrations of thiols in cellular media, principally in the form of glutathione. General solutions to this problem remain to be developed.

In 2006, Meggers and Streu used a ruthenium catalyst to mediate allyl carbamate deprotection of a caged fluorophore inside living cells (Streu and Meggers, 2006). The lack of apparent toxicity associated with this procedure prompted the investigation of heterogeneous palladium chemistry in cells. Pd(0)-functionalized microspheres were shown to be able to enter cells and mediate allyl carbamate deprotections and Suzuki-Miyaura cross-couping in the cytoplasm (Sasmal et al., 2013; Yusop et al., 2011). Applications of palladium based applications in cell culture include copper-free Sonagashira coupling (Li et al., 2011), extracellular Suzuki coupling on the surface of E. coli cells (Spicer et al., 2012), and conjugation of thiol groups with allyl selenosulfate salts (Crich et al., 2006).

While Pd-catalyzed cross-coupling reactions have promise in the bioconjugation arena because of their excellent functional group tolerance and compatibility with water (Shaughnessy, 2006)(Nicolaou et al., 2005), some optimization is always required; examples are shown in Figure 8. Thus, an early report of a Mizoroki-Heck modification of a genetically engineered Ras protein bearing an p-iodophenylalanine with a biotin-alkene reagent proceeded in very low yield (Figure 8A) (Kodama et al., 2006). The reaction was facilitated by the water soluble Pd-TPPTS (triphenylphosphine-3,3′,3″-trisulfonate) catalyst, although the outcome and requirement for inert atmosphere highlights limitations displayed by many early examples of transition metal-catalyzed bioconjugation reactions. The same modified protein substrate was later derivatized more successfully with a propargyl-tethered biotin under aqueous Sonagashira coupling conditions (Figure 8B) (Kodama et al., 2007), followed by much better conditions for related Sonagashira reactions of homoparpargylglycine-containing substrates (Figure 8C) (Li et al., 2011). The same type of water-soluble Pd complex was also used for an efficient Suzuki-Miyaura coupling reactions (Spicer et al., 2012) (including the example in Figure 8D) (Chalker et al., 2009b), following the pioneering use of a standard organometallic catalyst (Pd⁰-dibenzylidene acetone, Pd-DBA) under moderately forcing conditions (pH 8.5, 70 °C) (Brustad et al., 2008; Lim and Lin, 2010b).

Figure 8.

Examples of bioorthogonal reactions mediated by organometallic complexes.

Since it uses the small bioorthogonal alkene group, olefin metathesis is of particular interest for potential use in biological systems. Ruthenium complexes currently dominate this application, due to an advantageous combination of reaction rate, accessibility, tailorability, lack of toxicity, and functional group tolerance (Lin et al., 2013; Lin et al., 2009, 2010). S-allylcysteine can be easily introduced into proteins by a variety of methods, including conjugate addition of allyl thiol to dehydroalanine, direct allylation of cysteine, desulfurization of allyl disulfide, or metabolic incorporation as a methionine surrogate in methionine auxotrophic E. coli (van Hest et al., 2000), or by reassignment of the amber stop codon (Ai et al., 2010). The resulting allyl sulphides (Lin et al., 2008) are effective substrates for cross metathesis using standard (Figure 8E) and new catalysts developed for aqueous-phase applications (Burtscher and Grela, 2009; Skowerski et al., 2012; Tomasek and Schatz, 2013).

f. Responsive strategies: ligation and release

While complete, or even representative, coverage of the concept of cleavable linkers is beyond the scope of this article, it is useful to mention that bioorthogonal click reactions have an important role to play in the creation of molecules and materials that create linkages, release active ingredients, or perform other functions in response to changes in biological location or cellular condition. Figure 9A shows an example of such a triggered bioorthogonal ligation process: the rapid hetero-Diels-Alder ligation of a quinolinone-based quinone methide, generated in situ, with vinyl thioethers (Li et al., 2013a). The process has been used quite successfully for site-specific protein labeling as well as imaging bioactive small molecules inside live cells. A particularly interesting recent example of controlled bond formation and release (Figure 9B) takes advantage of an oxidation event to create an electrophile suitable for capture as an oxime ether (Park et al., 2014). Reduction of the adduct then sets the stage for facile N-O bond cleavage, releasing the captured molecule in a manner that is not the reverse of the original bond-forming step. Such sequential diverse chemistry provides more opportunities for control of each step of the process. Another variant on this theme is provided by the aforementioned oxanorbornadiene linkages (Table 4, entry 10). These feature a built-in cleavage trigger, since their Michael adducts undergo retro-Diels-Alder fragmentation at rates predetermined over several orders of magnitude by their substitution pattern (Figure 9C) (Kislukhin et al., 2012). If the resulting furan and thiomaleate moieties do not compromise the function of the released functional molecules, this provides a well-controlled triggered cleavage pathway for drug release.

Figure 9.

Examples of responsive linkages enabled by bioorthogonal ligations.

The development of prodrugs, molecules activated in the body to reveal therapeutic function in order to limit off-target toxicity or enhance properties such as biodistribution, constitutes a rich area of application of bioorthogonal chemistry. An example is the use of abiotic bioorthgonal reactions to provoke drug activation (Figure 10A), in contrast to the traditional reliance on endogeneous processes such as hydrolysis or enzymatic cleavage (Figure 10B). In a relatively simple example, the reduction of an azide by phosphine (Brakel et al., 2008) has been used as the prodrug activator in vivo, producing the corresponding aromatic amine that releases the cytotoxic cargo by 1,6-benzylic fragmentation (Figure 10C). The Staudinger ligation was also used to trigger drug liberation by virtue of a similar process triggered by O-to-N acyl migration to the iminophosphorane intermediate in the Staudinger process (Figure 10D) (Azoulay et al., 2006). Although conceptually attractive, these approaches suffer from slow reaction kinetics and the oxidative sensitivity of the phosphine reagent. The installation by Robillard and coworkers of a carbamate drug linkage next to the double bond of a trans-cyclooctene dienophile created a clever solution to these problems for antibody-drug conjugates (Versteegen et al., 2013). Fast tetrazine ligation was followed by facile rearrangement to the corresponding pyridazine with ejection of the allylic carbamate leaving group, thus releasing the drug only upon receipt of the bioorthogonal reaction signal (Figure 10E).

Figure 10.

Prodrug activation strategies using bioorthogonal chemistry. (A) Overall strategy. (B) Examples of traditional endogenous prodrug activation mechanisms. (C,D) Phosphine-based activation examples. (E) Activation assisted by tetrazine ligation.

C. Conclusion

This examination of the recent literature of bioorthogonal click reactions has revealed continuing exciting activity in “traditional” ligations to protein termini and sidechains, as well as a wealth of new chemistry that has pushed the boundaries of rate and selectivity. However, new reactions have yet to find much application to other biological substrates; ligations to nucleic acids, lipids, and glycans are almost exclusively done with well-established methods, reflecting both the ubiquity of applications for peptide and protein derivatization, and the fact that proteins have more functional groups to address. But these chemically more challenging biomolecular substrates are just as important for applications in medicine, materials science, and fundamental chemical biology, and so define several aspects of the future growth of the field of bioorthogonal chemistry.

The overall dream shared by many investigators is the precise insertion of non-biological chemical reactions into ever-more-complex biological systems, up to and including higher organisms. In this quest, we choose to dance with the incredible molecular and functional complexity of evolution, without stepping on Nature’s toes. The music is sublime, the destination rewarding, and we have a few good moves. Finding and using new ones is an endeavor most worthwhile.