Imaging the glycome

Abstract

Molecular imaging enables visualization of specific molecules in vivo and without substantial perturbation to the target molecule's environment. Glycans are appealing targets for molecular imaging but are inaccessible with conventional approaches. Classic methods for monitoring glycans rely on molecular recognition with probe-bearing lectins or antibodies, but these techniques are not well suited to in vivo imaging. In an emerging strategy, glycans are imaged by metabolic labeling with chemical reporters and subsequent ligation to fluorescent probes. This technique has enabled visualization of glycans in living cells and in live organisms such as zebrafish. Molecular imaging with chemical reporters offers a new avenue for probing changes in the glycome that accompany development and disease.

Molecular imaging reveals a wealth of information about biomolecules in their native environments (1). Subcellular localization of proteins (2) and RNAs (3), protein-expression patterns (4), protein–protein interactions (5), and ion concentrations (6) can be assayed with minimal perturbation to the organism or process under study. Despite the power of molecular imaging, the varieties of biomolecules amenable to such detailed scrutiny are few. Proteins are perhaps most accessible to in vivo visualization, because they are easily manipulated with genetics to generate autofluorescent protein chimeras (7). Certain small molecules and ions, including calcium (8), copper (9), lead (10), zinc (11), mercury (12), cAMP (13), and hydrogen peroxide (14), may also be imaged by using cleverly designed small-molecule fluorophores or reengineered fluorescent proteins. However, extension of molecular imaging to other biomolecule classes (i.e., glycans and lipids) has proved challenging (15).

Glycans are particularly attractive targets for in vivo imaging. These biopolymers play key roles in numerous biological processes (16, 17). For example, cell-surface glycans participate in cell–cell interactions involved in embryonic development (18), leukocyte homing (19), and cancer cell metastasis (20, 21). The wide range of functions that glycans can fulfill reflects their structural variety (Fig. 1) (22). Vertebrate glycans are components of glycoproteins, glycolipids, and proteoglycans, and may be membrane-associated, intracellular, or secreted. Built from monosaccharide building blocks that are connected in both linear and branched geometries, the glycans possess structural diversity that can far exceed that of the linear biopolymers. Further, the totality of glycans that a cell produces, collectively termed the cell's “glycome,” is influenced by the cell's genome, transcriptome, and proteome, as well as environmental cues and nutrients (23). Thus, the glycome reports on the physiological state of the cell and, not surprisingly, changes in the glycome have been associated with disease. The ability to witness such changes in the context of living organisms would augment our understanding of systems biology and provide new clinical tools for disease diagnosis.

Fig. 1.

Examples of glycoconjugate structures found in vertebrates. The glycans can be long and linear, as in the glycosaminoglycan chondroitin sulfate, or as simple as a single monosaccharide, as in cytosolic and nuclear O-GlcNAc-modified proteins. Branched structures are typical for the N-glycans and O-glycans found on glycoproteins and in glycolipids.

Glycans are not directly encoded in the genome and are thus not amenable to imaging techniques that rely on genetic reporters (24). Rather, efforts to image glycans in vivo have focused on the use of affinity reagents and chemical tools (15, 24). Herein, we review the various strategies available to image glycans, with a focus on our own work: using the bioorthogonal chemical reporter strategy to visualize glycans both ex vivo and in living organisms.

Imaging Glycans with Lectins and Antibodies

Lectins are naturally occurring glycan-binding proteins (25) that have been widely used for the detection (26) and enrichment (27) of glycoconjugates. Lectins are able to recognize structures as varied as the monosaccharides sialic acid (28) and fucose (29), and higher-order structures such as the Tn (30) and sialyl-Tn tumor antigens (31) and the conserved core region of N-glycans (32). However, lectins typically have low affinity for their glycan epitope and require multivalency for high-avidity binding (33). Moreover, lectins are generally tissue-impermeable and often toxic (34, 35). For these reasons the utility of lectins for imaging in living systems is limited, although they have been widely used to visualize glycans ex vivo. The use of lectins to probe specific glycans on cultured cell lines is well precedented (36, 37). Lectins have enabled glycan visualization on tissue sections or whole-mount specimens at discrete time points in mouse (38), chick (39), and fly (40) embryogenesis, as well as in the mature mouse thymus (41), rat endothelial vasculature (42), and human kidney (43). Further, lectins have been used to facilitate screening of Caenorhabditis elegans mutants with glycosylation defects (44, 45).

Like lectins, antibodies generated against glycan structures enable the visualization of these molecules ex vivo but have limited use in vivo. Antibodies are able to recognize very specific glycan structures. Indeed, there are well-characterized commercially available monoclonal antibodies that bind distinct epitopes on heparan (46) and chondroitin sulfate (47), as well as sialyl Lewis x (48), sulfoadhesin (49), and O-linked N-acetylglucosamine (O-GlcNAc) (50), among many others. Theoretically, an antibody may be generated against any glycan epitope; however, the synthesis of even small glycan structures can be unwieldy (51), complicating the generation of a hapten for immunization. Like lectins, antibodies are also tissue-impermeant, and most antibodies generated against glycan epitopes are of the low-affinity IgM subtype. Nevertheless, applications to immunofluorescence microscopy of cultured cells or tissue sections abound. Antibodies have been used for glycan visualization on fixed cells (52), as well as sections from chick cornea (53), bovine mammary gland (54), rabbit appendix (55), and human thymus (56), among others.

Only in 1 report has a glycan-specific antibody been used for in vivo imaging. Licha et al. (57) succeeded in visualizing the peripheral lymph node endothelial glycan termed sulfoadhesin in mice by using the MECA-79 antibody. This sulfated glycan, which serves as a ligand for the leukocyte adhesion molecule L-selectin, is also a marker of pathological inflammation that occurs during diabetes, asthma, and arthritis (58). Although in the published work only lymph nodes were imaged, a similar approach may permit clinical imaging of chronic inflammatory diseases.

Lectin and antibody-based imaging methods provide a snapshot of the glycome at a particular point in time, but are difficult to implement in the context of dynamic studies. Further, with limited in vivo applicability, these reagents require removal of the cells or tissues of interest from their native environment before analysis. Thus, we have focused on developing complementary methods for glycan imaging that permit in vivo analysis of dynamic changes in the glycome.

Imaging Glycans with Bioorthogonal Chemical Reporters

We developed a 2-step approach to glycan imaging that employs only small-molecule reagents (Fig. 2A) (15). First, a chemically reactive moiety is incorporated into target glycans by metabolic labeling with an unnatural monosaccharide substrate. The chemically reactive group is termed a “chemical reporter.” In the second step, the reporter group is visualized by covalent reaction with an imaging probe. The requirements on the chemical nature of the reporter are demanding. The reporter must be small enough to be ignored by the cell's metabolic enzymes, and inert to the endogenous chemical functionality of the cell. This latter criterion is also referred to as the property of “bioorthogonality.”

Fig. 2.

The bioorthogonal chemical reporter strategy for imaging glycans. (A) Glycans can be metabolically labeled with unnatural sugars bearing a chemical reporter group. The chemical reporter group, typically an azide or terminal alkyne, can be detected in a second step by covalent reaction with a probe. (B) Bioorthogonal reactions used to visualize chemical reporters appended to unnatural sugars (R). Azides can be detected by Staudinger ligation with triaryl phosphines, resulting in the formation of an amide linkage between the reporter and the probe. Azides and terminal alkynes can be detected by reaction with each other via CuAAC, forming in a triazole linkage between the reporter and probe. The azide can also be detected by Cu-free click chemistry with strained cyclooctynes. This latter reaction avoids the use of a cytotoxic metal catalyst.

The first chemical reporter used to visualize glycans was the ketone group, which can be detected by oxime or hydrazone formation with aminooxy- or hydrazide-functionalized probes, respectively (59). However, these covalent reactions perform best at reduced pH (5.5–6.5) and are sluggish at pH 7.4, thus limiting their utility in vivo. Greater progress in glycan imaging has been made by using azides and alkynes as bioorthogonal chemical reporters. Both of these functional groups are small, biologically inert, and capable of reacting with other bioorthogonal functional groups at physiological pH. The azide can be detected by reaction with phosphines via the Staudinger ligation (60), linear alkynes via the Cu-catalyzed azide-alkyne cycloaddition (abbreviated CuAAC) (61, 62), and with a variety of cyclooctynes via the strain-promoted azide-alkyne cycloaddition (also termed “Cu-free click chemistry”) (Fig. 2B) (63–65). The alkyne can be detected with exogenously delivered azides via the CuAAC reaction (Fig. 2B). The azide and alkyne reactions described above are not equally compatible with living systems. The Cu catalyst required in the CuAAC reaction is cytotoxic, thus limiting this reaction to use with fixed cells or tissues.

Metabolic Labeling of Glycan Subtypes with Chemical Reporters.

Most glycan subtypes, with the notable exception of glycosaminoglycans and glycosylphosphatidylinositol anchors, have been imaged by metabolic labeling with azido or alkynyl monosaccharides (Fig. 3). The first of these were glycoconjugates bearing the terminal monosaccharide N-acetylneuraminic acid (Neu5Ac), a member of the sialic acid family. This sugar holds a privileged status in the field of glycobiology, having been identified as a determinant of viral infection, leukocyte-endothelial cell adhesion, neuronal development, immune cell activation, cancer metastasis, and many other normal and pathological processes (66, 67).

Fig. 3.

Azide- and alkyne-bearing monosaccharides used for metabolic labeling of glycans. Sialic acids may be labeled with N-azidoacetylneuraminic acid (SiaNAz), ManNAz, 9-azido N-acetylneuraminic acid, and alkynyl ManNAc. GalNAc-containing glycans and O-GlcNAc-labeled glycans may be labeled with azides by using GalNAz. O-GlcNAc-labeled glycans may also be labeled with GlcNAz. Fucose-containing glycans may be labeled with 6AzFuc or alkynyl fucose. Structures are shown in peracetylated form, which are typically used for metabolic labeling experiments. After cell entry by passive diffusion, the acetyl groups are cleaved by cytosolic esterases.

Sialic acid-containing glycans can be visualized by metabolic labeling with analogs of its biosynthetic precursor N-acetylmannosamine (ManNAc) or with derivatives of sialic acid (Fig. 3). The biosynthetic machinery will tolerate the addition of chemical reporters to the N-acyl group of either substrate class [e.g., Ac₄ManNAz (60), alkynyl ManNAc (68), and SiaNAz (69)] or at C-9 of the sialic acid (e.g., 9-azido Neu5Ac) (70). Indeed, when mammalian cell lines are incubated with Ac₄ManNAz, SiaNAz replaces 4–41% of natural sialic acids (71). Ac₄ManNAz has been used to visualize sialic acids in a diverse range of cell types (71), as well as living mice (72) and zebrafish (73). Alkynyl ManNAc has been used to visualize sialic acids on human cancer cell lines and for glycoproteomic analysis of labeled glycoproteins (68, 74).

Mucin-type O-linked glycans have also been studied with the chemical reporter strategy. These structures share a common core N-acetylgalactosamine (GalNAc) residue that links the glycan to serine or threonine residues within the underlying protein (Fig. 1). Metabolic labeling of the core GalNAc residue can be achieved by treating cells or animals with per-O-acetylated N-azidoacetylgalactosamine (Ac₄GalNAz) (75). This unnatural sugar is processed by the GalNAc salvage pathway to form the intermediate uridine diphospho (UDP)-GalNAz, which is recognized by a family of polypeptide GalNAc transferases in the Golgi compartment. Ac₄GalNAz has been used to label mucin-type O-linked glycans in numerous cell lines, live mice (76), and zebrafish (73).

O-GlcNAc-modified proteins (77), which occur in the cytosol and nucleus, have been labeled with bioorthogonal chemical reporters by using either N-acetylglucosamine (GlcNAc) or GalNAc analogs. Per-O-acetylated N-azidoacetylglucosamine (Ac₄GlcNAz) (78) is modified by the GlcNAc salvage pathway enzymes to form UDP-GlcNAz, which is used as a substrate by the cytosolic O-GlcNAc transferase. Alternatively, Ac₄GalNAz is metabolized to form UDP-GalNAz, which can be converted to UDP-GlcNAz in situ by an epimerase (79). Ac₄GlcNAz has been used for the identification of O-GlcNAc-modified proteins in cell lysates (78) and Ac₄GalNAz has been used to image nuclear O-GlcNAc-modified proteins in cultured cells (M. J. Hangauer and C.R.B., unpublished work).

Fucosylated glycans, like those bearing sialic acid, have been implicated in myriad normal and pathological processes (80). As such, they have been attractive and heavily pursued targets for imaging with chemical reporters (68, 81, 82). Fucosylated glycans have been metabolically labeled by using per-O-acetylated 6-azidofucose (6AzFuc) or per-O-acetylated 6-alkynylfucose (alkynyl fucose), both of which exploit the fucose salvage pathway. So far, these fucose reporters have been limited to use in cultured cells because of low levels of metabolic incorporation and cytotoxicity.

Imaging of Glycans on Cultured Cells with the Chemical Reporter Strategy.

Imaging of azide- and alkyne-labeled glycans on cultured cells has been accomplished with fluorophore-conjugated phosphines and alkynes. Chang et al. (83) were able to visualize cell-surface SiaNAz residues on live cells by Staudinger ligation with fluorescein-, rhodamine-, and Cy5.5-conjugated phosphine reagents. Among these reagents, the Cy5.5 derivative afforded the best sensitivity because of its intrinsically low nonspecific cell binding activity and concomitantly low background fluorescence.

The background fluorescence that is frequently associated with poor clearance of unreacted fluorophore can be minimized by using “smart” probes, which become fluorescent only after reaction with their target. Lemieux et al. (84) developed a smart phosphine probe based on a coumarin scaffold (Fig. 4A). Before reaction, the phosphine is in the reduced state and its lone pair of electrons quenches the coumarin's fluorescence; however, after Staudinger ligation with azides, the resulting phosphine oxide eliminates quenching to yield a fluorescent product. The reagent was still prone to some background fluorescence because the phosphine oxide is generated both by nonspecific oxidation and the Staudinger ligation. More recently, Hangauer et al. (85) described the use of a red-shifted fluorogenic phosphine reagent that overcomes the limitations of its predecessor and shows promise for in vivo imaging applications (Fig. 4B). Capitalizing on the Staudinger ligation's mechanism (86), we introduced a fluorescence quencher at the ester group that is cleaved during the course of the reaction. Consequently, background fluorescence derived from the unreacted probe was minimized and labeled glycans could be visualized on cells with high sensitivity.

Fig. 4.

Fluorogenic phosphine dyes activated by Staudinger ligation with azides. (A) Fluorogenic phosphine in which the phosphine lone-pair electrons quench the fluorescence of the coumarin fluorophore. Staudinger ligation or nonspecific oxidation sequesters the lone-pair electrons and generates a fluorescent product. (B) A fluorogenic phosphine that utilizes a FRET quencher, which is cleaved during the Staudinger reaction but not during nonspecific oxidation.

Wong and coworkers (68, 82) developed a smart probe to visualize 6AzFuc-, alkynyl fucose-, and alkynyl ManNAc-labeled glycans on fixed cells by using the CuAAC reaction (Fig. 5A). This fluorogenic reagent comprised a 1,8-napthalimide scaffold with a pendant azide or alkyne group positioned to quench the molecule's fluorescence. Reaction of the dye's azide or alkyne moiety with a complementary azido or alkynyl sugar resulted in a significant fluorescence enhancement. When using the azide-containing 1,8-napthalimide probe for fixed cell imaging, however, the authors observed high background fluorescence (68) and turned to a similar azide-bearing smart probe based on a coumarin scaffold (Fig. 5B) (87). Interestingly, after fixation and permeabilization of azide-labeled cells, these quenched fluorophores enabled the visualization of fucose residues inside the cell, likely those trafficking through the secretory pathway.

Fig. 5.

Fluorescent dyes activated by the CuAAC reaction based on (A) 1,8-napthalimide or (B) coumarin scaffolds. After the reaction, the triazole modifies the electronics of the fluorophore to yield a fluorescent product.

Imaging Glycans in Living Organisms.

Ultimately, the constituents of the glycome must be visualized in living organisms to capture glycans in action as they perform their myriad functions. In principle, the Staudinger ligation and strain-promoted reaction of azides and cyclooctynes are applicable to in vivo imaging because the reagents involved are both bioorthogonal and devoid of intrinsic toxicity. Indeed, the Staudinger ligation has been shown to proceed in live mice, enabling the covalent tagging of SiaNAz- or GalNAz-modified glycoproteins in various tissues (72, 76). However, the Staudinger ligation suffers from sluggish reaction kinetics so that fast (i.e., less than a few hours long) biological processes involving glycans cannot be studied in real time.

By contrast, the reaction of azides with cyclooctynes (Cu-free click chemistry) has the potential for much improved kinetics under physiological conditions. Baskin et al. (64) demonstrated that a cyclooctyne reagent bearing a gem-difluoro group adjacent to the alkyne, termed “DIFO” (Fig. 6A), reacts 30–60 times faster with azides than nonfluorinated cyclooctynes or Staudinger ligation reagents. The superior reactivity of DIFO coupled with its selectivity and biocompatibility enabled the analysis of glycan trafficking dynamics in live cells (64). Azide-labeled glycans were reacted with a DIFO–fluorophore conjugate for 1 minute and the movement of the glycans within the cell was tracked by fluorescence microscopy. Further, pulse–chase experiments were performed in which azido glycans produced at different time points were labeled with discrete fluorophores. Thus, temporally distinct glycan populations could be color-coded as “old” or “new.” The addition of fused phenyl rings to the cyclooctyne scaffold can also increase its reactivity with azides, a phenomenon exploited by Boons and coworkers in generating alternative reagents for visualizing cell-surface azido glycans (88).

Fig. 6.

Application of the bioorthogonal chemical reporter strategy for in vivo imaging of glycans. (A) Zebrafish embryos were treated with azidosugars during their development, resulting in metabolic labeling of glycans with azides. The azides were visualized by reaction with fluorescent DIFO reagents. (B) An example of a zebrafish embryo metabolically labeled with Ac₄GalNAz and reacted with Alexa Fluor 647-conjugated DIFO (DIFO-647) at 60 h postfertilization (hpf) followed by Alexa Fluor 488-conjugated DIFO (DIFO-488) at 63 hpf to detect newly synthesized glycans. (Left) Single z-plane brightfield image. (Left Center) z-projection of DIFO-647 fluorescence. (Right Center) z-projection of DIFO-488 fluorescence. (Right) z-projection of DIFO-647 and DIFO-488 fluorescence merge. (Scale bar, 100 μm.)

The strain-promoted cycloaddition of azides with DIFO reagents has recently been extended to visualization of glycans in developing zebrafish, the first example of glycan imaging in a living organism (Fig. 6A) (73). The zebrafish is well suited to optical imaging because it develops ex utero and is transparent during most of the process (89). Further, zebrafish embryogenesis, a popular model for vertebrate development, is well characterized morphologically (90). Zebrafish embryos were grown in media containing GalNAz or ManNAz, resulting in metabolic labeling of mucin-type O-linked glycans or sialylated glycans, respectively, with azides. Glycans produced at various time-points during development were imaged by simply bathing the embryos with DIFO–fluorophore conjugates. Multicolor labeling experiments allowed the discrimination of temporally distinct glycan populations and analysis of their trafficking patterns during development (Fig. 6B).

Future Prospects

Glycans have now been added to the list of biological targets amenable to in vivo interrogation with molecular imaging techniques. The bioorthogonal chemical reporter approach has been successfully implemented to image glycans in developing zebrafish, thus augmenting the arsenal of tools for probing the dynamic glycome. Similar applications in other model organisms such as Drosophila and C. elegans are on the horizon, and further extension to mammalian disease models and even human clinical settings are worthy of pursuit. These future goals will be accompanied by new challenges, such as designing reagents based on consideration of metabolic stability and pharmacokinetic properties in addition to selectivity and kinetics.

In addition, new chemistries will be required to expand the fraction of the glycome that is revealed by using chemical reporters. A single azidosugar labels only a portion of the glycome; multiple unnatural sugars will be required to achieve broader coverage. Distinguishing multiple sugars will require additional chemical reporters that can be visualized independently of azides and alkynes. Thus, a major challenge in the field involves identifying small, bioorthogonal functional groups that are also orthogonal to the reagents described above. A collection of such reagents would enable a more thorough analysis of how glycan patterns change during normal and pathological processes.