Cell Surface Engineering by a Modified Staudinger Reaction

See allHide authors and affiliations

Science  17 Mar 2000:
Vol. 287, Issue 5460, pp. 2007-2010
DOI: 10.1126/science.287.5460.2007


Selective chemical reactions enacted within a cellular environment can be powerful tools for elucidating biological processes or engineering novel interactions. A chemical transformation that permits the selective formation of covalent adducts among richly functionalized biopolymers within a cellular context is presented. A ligation modeled after the Staudinger reaction forms an amide bond by coupling of an azide and a specifically engineered triarylphosphine. Both reactive partners are abiotic and chemically orthogonal to native cellular components. Azides installed within cell surface glycoconjugates by metabolism of a synthetic azidosugar were reacted with a biotinylated triarylphosphine to produce stable cell-surface adducts. The tremendous selectivity of the transformation should permit its execution within a cell's interior, offering new possibilities for probing intracellular interactions.

Chemoselective ligation reactions designed to modify only one cellular component among all others have provided insight into cellular processes (1). The goal in developing such transformations is to equal the tremendous selectivity of noncovalent recognition events, such as antibody-antigen binding, that direct many normal biological processes and are now powerful experimental tools. In order to achieve this, the two participating functional groups must have finely tuned reactivity so that interference with coexisting functionality is avoided. Ideally, the reactive partners would be abiotic, form a stable adduct under physiological conditions, and recognize only each other while ignoring their cellular surroundings. The demands on selectivity imposed by cells preclude the use of most conventional covalent reactions, and thus far only two have proven utility in a biological environment.

One chemoselective ligation reaction, that between a ketone and an aminooxy or hydrazide group, has enabled us to engineer the composition of cell surfaces (2). We introduced ketones onto cells through unnatural sialic acid biosynthesis. Human cells metabolize the unnatural precursor N-levulinoylmannosamine (compound 2, Fig. 1), a ketone-bearing analog of the native sugarN-acetylmannosamine (compound 1, Fig. 1), to the corresponding keto–sialic acid residues on cell surface glycoconjugates. Chemically orthogonal to native cell surface components, the ketone can then react selectively with externally delivered aminooxy or hydrazide reagents to form stable covalent adducts. Applications of this reaction include the chemical construction of new glycosylation patterns on cells (3), new approaches to tumor cell targeting (4), and novel receptors for facilitating viral-mediated gene transfer (5).

Figure 1

(top). Metabolic delivery of chemically orthogonal functional groups, ketones and azides, to cell surfaces by unnatural sialic acid biosynthesis. Compound 1: N-acetylmannosamine; compound2: N-levulinoylmannosamine; compound3: N-azidoacetylmannosamine.

Although useful for cell surface chemistry, ketone ligation reactions have limited intracellular use owing to competition with endogenous keto-metabolites. Tsien and co-workers reported a second chemoselective ligation reaction that circumvents this problem—condensation of a cysteine-rich hexapeptide motif with a bis-dithioarsolane (6). This enabled the targeting of a synthetic fluorescent dye to a single protein within the environs of a living cell. In order to augment existing chemical approaches to the study and manipulation of cellular components, the identification of new cell-compatible chemoselective ligation reactions is of fundamental importance. We have therefore focused on refining traditional chemical transformations in accordance with cellular demands.

The Staudinger reaction occurs between a phosphine and an azide to produce an aza-ylide (Fig. 2A) (7, 8). In the presence of water, this intermediate hydrolyzes spontaneously to yield a primary amine and the corresponding phosphine oxide. The phospine and the azide react with each other rapidly in water at room temperature in high yield. Both are abiotic and essentially unreactive toward biomolecules inside or on the surfaces of cells. Thus, in its classical form, the Staudinger reaction meets many of the criteria required of a chemoselective ligation in a cellular environment. Where it falls short is that the initial covalent adduct, the aza-ylide, is not stable in water. Our solution to this problem was to design a phosphine that would enable rearrangement of the unstable aza-ylide to a stable covalent adduct. We envisioned that an appropriately situated electrophilic trap, such as a methyl ester, within the phosphine structure would capture the nucleophilic aza-ylide by intramolecular cyclization (Fig. 2B). This process, in turn, would ultimately produce a stable amide bond rather than the products of aza-ylide hydrolysis. We tested this hypothesis in a model reaction (Fig. 2C) between a simple phosphine and methyl 2-azidogalactose in aqueous tetrahydrofuran (THF); only the ligation product was observed with no evidence of aza-ylide hydrolysis.

Figure 2

(bottom). Classical and modified Staudinger reactions. (A) The classical Staudinger reaction of phosphines and azides. Hydrolysis of the aza-ylide produces an amine and a phosphine oxide. (B) A modified Staudinger reaction that produces a stable covalent adduct by amide bond formation, even in the presence of water as solvent. (C) A model reaction that produces a single amide-linked product. The limited water solubility of the phosphine necessitated an organic cosolvent (THF).

The cell surface is a far more demanding environment, and to test the modified Staudinger reaction in this context we required (i) a method of installing azides on cells and (ii) a water-soluble phosphine reagent. The azide was selected rather than the phosphine for cell surface display because of its small size and the synthetic accessibility of azidosugars as metabolic precursors. On the basis of our earlier work with N-levulinoylmannosamine, we predicted that N-azidoacetylmannosamine (compound 3,Fig. 1) (9) would be well tolerated by the sialic acid biosynthetic machinery. Biotinylated phosphine 5(10) (Fig. 3A) was designed for water solubility, by virtue of the tetraethyleneglycol linker, and for detection of the ligated cell-surface product. The synthesis of compound 5 was performed as shown in Fig. 3A; a versatile intermediate, compound 4, bears a carboxylic acid to which any biological probe or biopolymer can be appended. The proposed reaction of 5 with cell surface azido-sialic acid is depicted in Fig. 3B.

Figure 3

Reaction of phosphines and azides on cell surfaces. (A) Synthesis of a water-soluble biotinlyated phosphine for quantifying the reaction with cell surface azides [synthetic procedures for compound 4 are provided in (10)]. (B) Reaction of biotinylated phosphine5 with cell surface azido sialic acid generated by metabolism of acetylated 3.

Jurkat cells were incubated with N- azidoacetylmannosamine, in acetylated form (11), at a concentration of 20 μM for 3 days. The cells were washed and then reacted with compound 5[1 mM in phosphate-buffered saline (PBS), pH 7.4)] for 1 hour. After staining with fluorescein isothiocyanate (FITC)–avidin, the cells were analyzed by flow cytometry (Fig. 4A). Jurkat cells treated with acetylated 3 showed a marked increase in fluorescence that indicated the accumulation of biotin moieties on the cell surface, whereas untreated cells showed only a background level of fluorescence after exposure to phosphine5. The fluorescence signal was reduced by the addition of tunicamycin during incubation of Jurkat cells with the azidosugar, in agreement with previous observations that most sialic acids on Jurkat cells reside within N-linked glycans (3). The background fluorescence was identical to that observed with Jurkat cells that were not exposed to any reagents (12) and thus represents autofluorescence of cells and not nonspecific uptake of the biotin probe or FITC-avidin.

Figure 4

(A) Analysis of cell surface reaction by flow cytometry. Jurkat cells (1.25 × 105 cells per milliliter) were cultured in the presence or absence (control) of acetylated 3 (20 μM for 3 days). The cells were washed twice with 1 ml of buffer (0.1% fetal bovine serum in PBS, pH 7.4) and diluted to a volume of 240 μl. Samples were added to 60 μl of a solution of 5 (5 mM in PBS, pH 7.4) and incubated at room temperature for 1 hour. The cells were washed and resuspended in 100 μl of buffer, then added to 100 μl of FITC-avidin staining solution (1:250 dilution in PBS). After a 10-min incubation in the dark at 4°C, the cells were washed with 1 ml of buffer and the FITC-avidin staining was repeated. The cells were washed twice with buffer, then diluted to a volume of 300 μl for flow cytometry analysis. Similar results were obtained in two replicate experiments. (B) Progress of reaction over time. Assays were performed as in (A) with 40 μM acetylated 3 and varying the duration of the reaction with compound 5. (C) pH profile of reaction. Assays were performed as in (A) with 40 μM acetylated3 and varying the pH of the buffer used during incubation with compound 5.

HeLa cells responded similarly to incubation with acetylated3 followed by reaction with compound 5. Notably, HeLa cells that were cultured for an additional 3 days after the modified Staudinger reaction showed no change in growth rate. Thus, neither metabolism of azidosugars, reaction with phosphine5, nor the covalent attachment of phosphine oxide adducts to the cell surface appears to affect cell viability.

Using biotinylated beads of known biotin density, we were able to correlate the fluorescence intensities observed by flow cytometry with the number of dye molecules on a particle or cell (2). On this basis, we determined that Jurkat cells treated with 40 μM acetylated 3 for 3 days, followed by reaction with 1 mM compound 5 for 1 hour, accumulated ∼850,000 biotin moieties on the cell surface. This value places a lower limit on the number of azides present on the cell surface, as some azides may be concealed within the glycocalyx and therefore not accessible to the phosphine reagent. Furthermore, the cell surface reaction may not proceed in quantitative yield as observed with the model reaction (Fig. 2C). Higher densities of cell surface biotin moieties could be achieved by extending the reaction time as shown in Fig. 4B. Increasing the concentrations of the azidosugar or phosphine probe also elevated the level of cell surface modification. For example, Jurkat cells treated with 40 μM acetylated 3 for 3 days, followed by reaction with 2 mM compound 5 for 3 hours, accumulated ∼4.5 million biotin moieties on the cell surface. We observed a dependence of the cell surface reaction yield on pH; reaction at pH 6.5 produced 75% of the fluorescence signal observed at pH 7.4 (Fig. 4C). This is consistent with previous observations that protonation of aza-ylides facilitates their hydrolysis, a competing side reaction of the modified Staudinger process (8).

We considered one alternative explanation for the azide-dependent localization of biotin on cells. Phosphine 5 might have reduced cell surface azides to the corresponding amines by the classical Staudinger reaction, simultaneously producing phosphine oxide6 (Fig. 5A). Compound6, in turn, might nonspecifically acylate cell surface amines. If so, the reaction would lose the critical element of selectivity that we sought for biological applications. To address this possibility, we independently synthesized compound 6 and investigated its reactivity with cells. Two populations of Jurkat cells were pretreated with the azidosugar to engender cell surface azides. One population was then further reacted with a water-soluble trisulfonated triphenylphosphine to intentionally reduce the azides. In both cases, no cell surface biotinylation was observed. This result contrasts markedly with the extensive biotinylation of azidosugar-treated cells reacted with phosphine 5 (Fig. 5A). We conclude that the chemoselective ligation reaction proceeds as designed without complications arising from nonspecific amine acylation.

Figure 5

Specificity of the modified Staudinger reaction. (A) Cell surface biotinylation does not proceed by classical Staudinger azide reduction followed by nonspecific acylation. Jurkat cells were cultured in the presence of acetylated 3as described in Fig. 4. Cell surface azides were either reduced intentionally with a trisulfonated triphenylphosphine or left unreduced. Phosphine oxide 6, the product of the classical Staudinger reaction, was prepared independently and incubated with the cells (1 mM for 1 hour). Analysis by flow cytometry was performed as inFig. 4. (B) Triarylphosphines do not reduce disulfide bonds at the cell surface. Jurkat cells were incubated with a 1 mM solution of triarylphosphine 4 or TCEP for 1 hour at room temperature. The cells were centrifuged (2 min, 3000 g), washed with PBS, and diluted to a volume of 240 μl. Samples were combined with 60 μl of a solution of iodoacetylbiotin (5 mM in PBS). After incubation in the dark at room temperature for 1.5 hours, the cells were washed with buffer, stained with FITC-avidin, and analyzed by flow cytometry. In both (A) and (B), error bars represent the standard deviation of two replicate experiments.

To satisfy the requirement of chemical orthogonality, both participants in the reaction may not engage functional groups endogenous to cells. Triarylphosphines are mild reducing agents, which raises the the possibility of disulfide bond reduction as an undesirable side reaction. We addressed this issue by incubating Jurkat cells with triarylphosphine 4, an intermediate in the synthesis of 5, and quantifying the appearance of free sulfhydryl groups on the cell surface with iodoacetylbiotin and FITC-avidin (Fig. 5B). After 1 hour in the presence of 1 mM4, no detectable increase in free sulfhydryl groups was observed relative to cells exposed to iodoacetylbiotin alone. In a positive control experiment, we incubated Jurkat cells with the trialkylphosphine TCEP (1 mM, 1 hour), a commercial disulfide bond reducing agent. A marked increase in free cell-surface sulfhydryl groups was observed in this case (Fig. 5B). We conclude that triarylphosphines such as 4 and 5 are essentially unreactive toward disulfide bonds under these conditions, rendering ligation with azides the predominant pathway for reactivity.

In a side-by-side comparison with our previously reported cell surface ketone reaction (2), the cell surface Staudinger process was superior in several respects. Using the same reagent concentrations, azidosugar metabolism followed by phosphine reaction produced twofold higher fluorescence than ketosugar metabolism followed by hydrazide reaction. This may reflect either a faster reaction at the cell surface or more efficient metabolism of azidosugar 3 as compared with ketosugar 2 (Fig. 1). The azide has a major advantage over the ketone in that its reactivity is unique in a cellular context owing to its abiotic nature. Ketones, by contrast, abound inside cells in the form of metabolites such as pyruvic acid and oxaloacetate. The modified Staudinger reaction is chemically orthogonal to ketone ligations and should allow tandem modification of cell surfaces with the two chemistries.

The susceptibility of azides to reduction during the metabolic process warrants some consideration in light of the reducing potential of the cell's interior. Monothiols such as glutathione can reduce alkyl azides at alkaline pH, but the rates of such reactions under physiological conditions are insignificant on the time scale of our experiments (13). Correspondingly, metabolic studies of the azido drug AZT (azidothymidine) showed 90% recovery of the azide, either in its administered form or metabolized to the glucuronidated compound, without significant reduction (14).

The delivery of azides to cell surfaces through other carbohydrate biosynthetic pathways could significantly expand applications of cell surface engineering. Azides and phosphines are abiotic structures both inside and outside cells, which raises the exciting possibility that their ligation could proceed in the intracellular environment. Given existing powerful methods for incorporating unnatural building blocks into other biopolymers, one need not be restricted to cell surface oligosaccharides as hosts for these chemical handles (15, 16). Azido–amino acids, for example, could be introduced into proteins and later targeted with phosphine probes. The introduction of the reactive partners into transiently associated biopolymers might allow their covalent trapping within a cell and, as a result, the identification of previously unobservable interactions.

  • * To whom correspondence should be addressed. E-mail: bertozzi{at}


View Abstract

Navigate This Article