Imaging Cell-Surface Glycans in Animals with Bioorthogonal Chemistry
The myriad functions of cell-surface glycans have rendered them attractive targets for in vivo molecular imaging. Until very recently, methods for specifically visualizing this class of biomolecules were lacking. The advent of bioorthogonal chemistry, i.e., reactions among functional groups that do not interact with or interfere with biological systems, provided a platform for probing glycans on cells and in animals. Herein, the progression of bioorthogonal reaction development and use of these chemistries for glycan imaging are described.
KeywordsBioorthogonal Imaging In vivo Metabolic Zebrafish Azide Cyclooctyne Glycan Reporter
The cell-surface glycome comprises diverse glycans that contribute to virtually all aspects of a cell’s social life. Glycans contribute directly to cell–cell and cell–matrix interactions, affect the organization and turnover kinetics of cell-surface molecules, and modulate the activity of signaling receptors. Accordingly, they are central players in cell differentiation, proliferation, and migration and in the translation of signals from the tissue microenvironment to cellular responses. The cell-surface glycome can also be considered a complex data set that reports on the underlying cell’s physiology. Governed by numerous inputs – expression of glycosyltransferase genes, protein and lipid scaffold availability, nutrient status, and secretory pathway activity, for example – the composition of the cell-surface glycome changes in response to cellular changes in a manner that has diagnostic value. The unique glycomic signatures of cancer cells, embryonic stem cells, and activated leukocytes and endothelial cells, in particular, have attracted much attention as targets for diagnostic probes.
The ability to visualize changes in cell-surface glycosylation as a function of space and time, ideally in live organisms, is therefore an important goal toward advancing our understanding of glycobiology. While molecular imaging techniques targeting proteins have advanced considerably in recent decades (e.g., genetically encoded fluorescent proteins, bioluminescence methods, antibody-targeted diagnostics), it is only recently that the challenge of in vivo glycan imaging has begun to be met. Initial work focused on using glycan-binding proteins such as antibodies and lectins to direct a conjugated fluorescent probe to glycan structures of interest. These techniques were quite effective for profiling global glycosylation changes on cultured cells and tissue slices but have limited applicability for imaging in live organisms (Laughlin and Bertozzi 2009). Antibodies and lectins often have low binding affinities for their glycan epitopes, poor tissue penetrance, and, in the case of lectins, can be toxic to cells (Laughlin and Bertozzi 2009).
Initial inspiration for the metabolic labeling strategy came from the work of Werner Reutter and coworkers demonstrating that unnatural analogs of N-acetylmannosamine (ManNAc), a metabolic precursor to sialic acid, are converted to the corresponding unnatural sialic acid derivatives in cultured cells and in rodents (Kayser et al. 1992). This was an important discovery, since it showed that the sialic acid biosynthetic pathway could tolerate unnatural precursors and deliver them to a locale, the cell surface, accessible to myriad probe reagents.
The next conceptual challenge was to identify functional groups that could be integrated into such unnatural ManNAc analogs and that could undergo highly selective conjugation reactions with probes. For the chemical reactions to occur in vivo, the reagents would have to be bioorthogonal, meaning that they must neither interact nor interfere with surrounding biological functionalities. Moreover, their mutually selective reaction must proceed under physiological conditions: in water, at 37 °C and pH 7, and with no toxic by-products. Additionally, the reaction partner found on the metabolic precursor sugar, referred to as the “reporter ” group, must be small enough to enable processing by biosynthetic enzymes. Finally, the reaction must have inherently fast kinetics. This requirement stems from the relatively low concentrations of glycan-associated reporter group and circulating probe molecule that can be achieved in living systems. As well, observing dynamic changes in glycosylation that occur over short-time periods mandates the use of reactions that occur on a fast time scale.
The first reaction used for probing glycans in living systems was the condensation of carbonyls and α-effect nucleophiles. While the condensation between ketones or aldehydes and amines is typically sluggish, the corresponding reaction with α-effect nucleophiles such as aminooxy or hydrazide compounds is much faster (Fig. 1b). The nitrogen atoms of aminooxy and hydrazide groups have enhanced nucleophilicity due to their direct attachment to another heteroatom.
Typically, the ketone has been used as the metabolic label and the hydrazide for probe delivery. In an early example by Bertozzi and coworkers, a ketone-functionalized analog of ManNAc, N-levulinoylmannosamine (ManLev), was used as the metabolic precursor following Reutter and coworkers’ precedent (Mahal et al. 1997). The electrophilic ketone was chosen as the chemical reporter group because a hydrazide-based reporter would readily react with other biological ketones and aldehydes present in the cytosol (e.g., pyruvate and free monosaccharides). Although ketones and aldehydes are found within cells and in circulation, bringing into question the true bioorthogonality of this reaction, they are not otherwise present on cell surfaces where the probe ligation reaction occurs. Accordingly, it was demonstrated that treatment of cells with ManLev produced N-levulinoyl sialylated glycans on cell surfaces, enabling their visualization with hydrazide-functionalized imaging probes. While ManLev with free hydroxyl groups was administered to cells at millimolar concentrations to achieve metabolic labeling, micromolar concentrations could be employed using the peracetylated analog. Presumably, peracetylated ManLev diffused through membranes more efficiently than the hydrophilic free sugar and was subsequently deacetylated by cytosolic esterases prior to metabolism.
While useful for some applications, condensation reactions of carbonyls and α-effect nucleophiles were not particularly useful for imaging glycans in animals. These reactions require acidic pH (i.e., 5.5–6.5) for efficient labeling, which is impossible to achieve in vivo . As well, the hydrazide or oxime bond formed in the reaction is susceptible to hydrolysis. Recently, a carbonyl condensation reaction that produces a hydrolytically stable C-C bonded adduct was reported (Agarwal et al. 2013). But no matter what the stability of the product is, nor the kinetics that can be achieved at physiological pH, ketone and aldehyde reporter groups suffer from the fundamental problem that they are not unique in a biological setting. Thus, an important component of this in vivo glycan imaging approach is the development of bioorthogonal reactions among functionalities that have no counterpart whatsoever in animals. The first reaction that met this challenge was the Staudinger ligation.
The Staudinger ligation is highly a selective reaction between azide and phosphine reagents, resulting in the formation of an amide bond (Prescher et al. 2004). The development of this reaction was inspired by the classic Staudinger reduction of azides to the corresponding amines. In the Staudinger ligation, the nucleophilic phosphine first attacks the electrophilic azide (Prescher et al. 2004). Following extrusion of dinitrogen, an aza-ylide intermediate forms and rapidly reacts in intramolecular fashion with a nearby ester group to form a stable amide bond.
For glycan imaging, the azide has been used as the reporter group, while the phosphine was used to deliver the imaging probe. The azide has proven to be a quintessential reporter group due to its small size, chemical stability in biological settings, unique reactivity, and absence from biology (Sletten and Bertozzi 2009). Prior to its use in metabolic labeling, the azide was widely employed in synthetic chemistry as protected amine equivalent. Accordingly, there are many known methods for installing azides in small molecule substrates, particularly sugars where the azide has long been used in the construction of aminosugar analogs.
The Staudinger ligation reaction circumvented many of the limitations of ketone and aldehyde-based condensations. The ligation proceeds efficiently at neutral pH, and the generated amide bond is stable in vivo . These features, in combination with its exceptional selectivity and biocompatibility, enabled the Staudinger ligation to be the first bioorthogonal chemical reaction performed in live mice (Prescher et al. 2004). In this study, mice were injected with peracetylated N-azidoacetylmannosamine (ManNAz) once daily for several days and then subsequently administered a biotinylated phosphine probe. After a few hours, organs were harvested and the Staudinger ligation product was detected by flow cytometry analysis of splenocytes and by Western blotting of organ lysates.
Unfortunately, the Staudinger ligation did not prove to be very useful for in vivo glycan imaging , due primarily to its relatively slow kinetics (second-order rate constant k ≈ 10−3 M−1s−1). There are two strategies for accelerating reactions with inherently slow kinetics: elevate temperature and increase reagent concentrations. For in vivo imaging, temperature cannot be altered outside the physiological range. The concentration of metabolically azide-labeled glycans in vivo is low and also not subject to much external control. The only variable that can be experimentally altered, therefore, is probe concentration administered to the organism. However, for fluorescent probes, high concentrations are difficult to achieve due to solubility limits, and if injected at high levels into animals, unreacted probe molecules are difficult to wash away. The consequence is high background fluorescence that obscures any fluorophore chemically attached to glycans of interest. Attempts were made to design phosphine reagents that react more rapidly with azides . Unfortunately, phosphines with enhanced azide reactivity were also more susceptible to spontaneous air oxidation, a significant liability for biological imaging applications.
Other work focused on designing phosphine reagents that underwent an enhancement in fluorescence during the Staudinger ligation; in principle, such “fluorogenic” probes could be used at higher concentrations without the burden of washing away unreacted probe and associated background fluorescence. Two design concepts were explored for producing fluorogenic phosphine probes (Sletten and Bertozzi 2009). In the first, the mechanism for fluorescence turn-on relied on photo-induced electron transfer (PET) quenching by the phosphorous atom’s lone pair of electrons. During the Staudinger ligation, the phosphine is converted to a phosphine oxide that lacks a lone pair of electrons and therefore no longer quenches fluorescence. Unfortunately, air oxidation of the phosphine also gave rise to fluorescence turn-on, limiting its use for biological imaging studies.
A second fluorogenic probe was designed based on fluorescence resonance energy transfer (FRET) quenching. For this probe, a fluorophore was conjugated to the phosphine and a FRET quencher was attached through the ester linkage. Cleavage of the ester during the Staudinger ligation reaction released the quencher and activated fluorescence. This probe was used to image glycans on live cells; however, the slow reaction rate necessitated multi-hour incubations, which, although never directly tested, would likely undermine in vivo imaging due to competing liver metabolism of the phosphine probe. Thus, faster reaction rates were urgently sought to realize the goal of in vivo glycan imaging using azidosugar reporters.
Two years after the Staudinger ligation was described, the fast and highly chemoselective copper-catalyzed azide–alkyne cycloaddition (CuAAC) reaction was reported (Rostovtsev et al. 2002). Prior to this, the cycloaddition reaction between azides and alkynes to form triazoles was well known in the organic chemistry literature. This classic transformation is thermodynamically favorable (ΔG° ≈ −61 kcal mol−1), but under standard conditions it does not proceed readily due to a high kinetic barrier. In CuAAC, the copper (I) catalyst lowers the kinetic barrier by combining with the alkyne to form a reactive copper acetylide species. This intermediate reacts rapidly with azides to form the 1,4-triazole product. Sharpless and coworkers coined the term “click chemistry” to refer to such fast, selective, and reliable transformations, and CuAAC is often considered to be the quintessential click chemistry reaction.
The speed of CuAAC dwarfs that of the Staudinger ligation by over three orders of magnitude. However, copper(I) is highly toxic to cells at concentrations required for catalysis, largely precluding the use of CuAAC for imaging live cells and organisms. CuAAC was, however, used to image glycans on fixed cells and enabled visualization of a snapshot of glycans in the process of intracellular trafficking (Hsu et al. 2007). Recently, a copper ligand was reported that enables in vivo CuAAC labeling of glycans on cells and in developing zebrafish embryos (Soriano del Amo et al. 2010). While promising, this reaction still must be performed for a short duration, suggesting that toxicity during longer copper(I) exposures is an issue.
An alternative means for enhancing the azide–alkyne cycloaddition reaction kinetics has been identified that does not require a toxic metal catalyst: ring strain. Alkynes constrained in eight-membered rings, i.e., cyclooctyne s, possess considerable ring strain (~19 kcal/mol) due to bond angle distortion from the ideal of 180° down to about 155°. As a result, cyclooctynes are high-energy substrates for cycloaddition with azides, during which much of this strain energy is released (Fig. 1b). Unsubstituted cyclooctynes react spontaneously with azides to form triazole products, whereas their linear unstrained counterparts require elevated temperatures or pressures.
The reactivity of cyclooctynes in this “strain-promoted azide–alkyne cycloaddition (SPAAC),” also sometimes referred to as “copper-free click chemistry,” can be tuned over at least three orders of magnitude. Unsubstituted cyclooctynes typically react with azides with second-order rate constants on par with that of the Staudinger ligation. However, introduction of two fluorine atoms next to the alkyne, as in the difluorinated cyclooctyne “DIFO,” increases the second-order rate constant 60-fold (Sletten and Bertozzi 2009). This improvement was sufficient to enable glycan imaging after reacting azidosugar-treated cells with a DIFO–fluorophore conjugate for one minute.
Further enhancements in cyclooctyne reactivity were achieved by fusing the ring to two benzene rings, as in the dibenzocyclooctynes (DIBO), dibenzo-aza-cyclooctynes (DIBAC), and bisaryl-aza-cyclooctynones (BARAC) (Jewett et al. 2010). As well, a substantive kinetic enhancement was observed upon fusion of the cyclooctyne to a cyclopropyl ring, generating a bicyclononyne (BCN) structure. The increased reactivity of these various ring-fused cyclooctynes may be due to enhanced ring strain. Considerable computational work has now been performed to understand the relationship of cyclooctyne structure and azide cycloaddition reactivity, particularly by Houk, Goddard, and Alabugin (Sletten and Bertozzi 2009). DIBO, DIBAC, BARAC, and BNC react with azides 100–1,000 times faster than the parent unsubstituted cyclooctyne and can be used to image glycans on cultured cells at sub-micromolar concentrations. Nowadays, many of these cyclooctyne-probe conjugates are available from commercial sources.
While informative, these studies were limited to visualizing of glycans on the outermost layer of embryo cells. Attempts to image internal zebrafish glycans revealed that the reaction rate is not the only parameter that determines labeling efficiency. Visualization of internal zebrafish glycans requires tissue access by cyclooctyne probes, which appears to be limited, at least for the commercial fluorophores that have been tested. As well, efforts to use cyclooctyne probes to image azidosugar-labeled glycans in mice were initially hampered by physical properties of those probes (Chang et al. 2010). For example, while DIFO conjugates react efficiently with cell-surface azidosugars in culture cells, in mice such conjugates have limited bioavailability due to their tight and possibly covalent association with serum albumin. Thus, a major future goal is to optimize cyclooctyne-probe conjugates for pharmacokinetic properties including tissue distribution, metabolic stability, and bioavailability in serum.
Nonetheless, cyclooctyne–azide cycloaddition chemistry has been employed for glycan imaging in mice, with a focus on tumor detection. Kim and coworkers showed that cyclooctyne–nanoparticle conjugates can detect azido sialic acid residues metabolically introduced onto on tumor cells in live mice (Koo et al. 2012). With several cyclooctynes per nanoparticle, the resulting multivalency was thought to increase the nanoparticle’s reaction efficiency with cell-surface glycans. Further, each nanoparticle was outfitted with multiple fluorophores, generating a bright signal and enhancing detection sensitivity. Still, to achieve detectable labeling above background, the azidosugar substrate, ManNAz, had to be directly injected into the tumor rather than administered systemically; thus, it is unclear how much of the observed signal resulted from reaction of the nanoparticles with unprocessed ManNAz that was not associated with cell-surface glycans.
Although the kinetics and bioorthogonality of copper-free click chemistry are suitable for many imaging applications, there is a continued need for new bioorthogonal reactions with similar or improved properties. Having multiple bioorthogonal reactions that are chemically orthogonal to one another would permit the simultaneous imaging of different glycan types, and more generally different types of biomolecules, in vivo. This capability could be useful, for example, in monitoring the metabolic flux of different glycosylation pathways that are modulated during cancer progression.
Although several bioorthogonal reactions have been reported over the last decade, few possess a reaction partner that is capable of acting as a metabolic reporter . An exception is the recently reported inverse-electron-demand Diels–Alder (IED-DA) reaction between a substituted cyclopropene reporter and a tetrazine probe. In the late 2000s, Weissleder and coworkers discovered that strained alkenes can react with tetrazines >1,000-fold faster than the most rapid cyclooctyne–azide cycloaddition (Devaraj et al. 2008). While the kinetics of this reaction were extremely promising for bioorthogonal chemistry, the reaction partners (e.g., trans-cyclooctenes, norbornenes) were too large to serve as broadly applicable reporter groups. Later, the Prescher and Deveraj groups discovered that tetrazine also reacts rapidly with the much smaller cyclopropenes (Patterson et al. 2012; Yang et al. 2012). While the reaction rates were slower than those observed with norbornenes and tetrazines, they were comparable to fast cyclooctyne–azide cycloadditions. A possible complication was the known reactivity of cyclopropenes with thiols, but this unwanted side reaction was abolished by the addition of a vinylic methyl group. The reaction of tetrazines and methyl cyclopropenes was made even more promising for imaging applications by the development of fluorogenic tetrazine probes. Methyl cyclopropene-functionalized sialic acid analogs have now been shown to metabolically label cell-surface glycans, which can then be imaged by reaction with tetrazine–fluorophore probes, even in the presence of other bioorthogonal functionalities (Patterson et al. 2012). Together, azide–cyclooctyne and cyclopropene–tetrazine reactions may enable simultaneous visualization of different glycan structures.
Over the past decade, significant progress has been made in the development of bioorthogonal reactions for imaging glycans in model organisms. Following the observation that ketone-modified metabolic precursors could be processed by endogenous biosynthetic pathways and detected on live cells, significant work has gone into developing reactions with improved properties. Now, azide-functionalized sugars and cyclooctyne–probe as well as tetrazine–probe conjugates are all available from commercial sources, making in vivo glycan imaging accessible to the broader community of biologists. Early applications have focused on studies of glycan expression during zebrafish embryogenesis and tumor detection in murine xenograft models, but there is a rich future for expansion of the metabolic labeling/bioorthogonal chemistry approach into other systems. In vivo models of inflammation, wound healing, and microbial infection are all attractive venues for imaging the dynamics of cell-surface glycosylation associated with disease.