A metabolic inhibitor blocks cellular fucosylation and enables production of afucosylated antibodies

Significance

The monosaccharide fucose is a common constituent of glycoproteins from multicellular organisms. Fucosylation modulates the functions of various proteins, thereby exerting control over a range of physiological processes. For example, fucose present on conserved N-glycans of antibodies diminishes their affinity for a receptor found on white blood cells, reducing their ability to kill cancer cells. Using a mechanism-inspired approach, we create a conveniently prepared fucose mimetic that can potently block cellular fucosylation. This fucose mimetic can be used as an additive, enabling production of desirable potent afucosylated anticancer antibodies. Moreover, we show that this glycomimetic blocks fucosylation but is not incorporated by diverse cell lines, making this a convenient and simple strategy useful for diverse academic and industrial applications.

Abstract

The fucosylation of glycoproteins regulates diverse physiological processes. Inhibitors that can control cellular levels of protein fucosylation have consequently emerged as being of high interest. One area where inhibitors of fucosylation have gained significant attention is in the production of afucosylated antibodies, which exhibit superior antibody-dependent cell cytotoxicity as compared to their fucosylated counterparts. Here, we describe β-carbafucose, a fucose derivative in which the endocyclic ring oxygen is replaced by a methylene group, and show that it acts as a potent metabolic inhibitor within cells to antagonize protein fucosylation. β-carbafucose is assimilated by the fucose salvage pathway to form GDP-carbafucose which, due to its being unable to form the oxocarbenium ion-like transition states used by fucosyltransferases, is an incompetent substrate for these enzymes. β-carbafucose treatment of a CHO cell line used for high-level production of the therapeutic antibody Herceptin leads to dose-dependent reductions in core fucosylation without affecting cell growth or antibody production. Mass spectrometry analyses of the intact antibody and N-glycans show that β-carbafucose is not incorporated into the antibody N-glycans at detectable levels. We expect that β-carbafucose will serve as a useful research tool for the community and may find immediate application for the rapid production of afucosylated antibodies for therapeutic purposes.

Protein glycosylation is a common posttranslational modification that has diverse roles in controlling protein folding, stability, and activity (1). Many classes of glycosylation are known, and these are generated by the sequential attachment of monosaccharides to proteins. The resulting glycan structures often manifest compositional variation, known as microheterogeneity, that results from differences in the numbers, types, and connectivities of the monosaccharide residues that are present. At the outermost position of glycans are terminal monosaccharides, which typically show the greatest microheterogeneity and play a particularly important role in mediating biological processes. For example, L-fucose (Fuc, 1α, 1β) (Fig. 1C) is a common terminal monosaccharide that exerts a range of biological effects by controlling interactions with cell surface protein receptors (2). The best characterized example of a fucose-regulated protein interaction occurs when the fucose-containing tetrasaccharide sialyl Lewis X (sLeX), present on the surface of leukocytes, binds to the selectin receptors on the surface of endothelial cells that are expressed during inflammation (3–6). However, the biological roles for fucosylation in regulating various processes, including tuning of Notch signaling by modification of the epidermal growth factor repeats of Notch, continue to emerge (7). These and other roles of fucose-containing glycans in inflammatory diseases and cancers (2, 5, 8) have drawn significant attention, which in turn has stimulated interest in ways to manipulate the levels of fucose within glycans and its consequent presentation on proteins.

Fig. 1.

Antibody fucosylation and existing metabolic inhibitors of fucosylation. (A) ADCC (antibody-dependent cell cytotoxicity), CDC (complement-dependent cytotoxicity), and ADCP (antibody-dependent cell phagocytosis) rely on the binding of Fc domain and Fc receptor. Created with https://BioRender.com. (B) Structural representation of the IgG1 Fc domain and a representative complex N-glycan. (C) Structure of previously reported fucosylation inhibitors. (D) Schematic representation of the biosynthetic pathways leading to fucosylation. Enzymes responsible for biosynthetic steps are in italics.

The area where manipulation of protein fucosylation has gained most attention is in tuning the glycan structures of antibodies (9, 10). The composition and structure of fucose-containing glycans that decorate recombinant human IgG1 antibodies can profoundly affect their use in therapeutic strategies involving antibody-dependent cell cytotoxicity (ADCC), complement-dependent cytotoxicity (CDC), or antibody-dependent cell phagocytosis (ADCP) (9–11) (Fig. 1A). Fucose residues α-1,6 linked to the N-acetylglucosamine residue at the reducing end of the N-glycans at Asn297 in the CH2 domains of IgG1 antibodies markedly impair binding to the Fc gamma receptor IIIa (FcγRIII) (12–14) that is expressed on effector cells or the immune system (Fig. 1 A and B). The absence of fucose at this “core” position within the IgG1 N-glycan improves antibody binding by 20-fold to 100-fold (12–14). Accordingly, various approaches to decreasing core fucosylation on antibodies have been pursued. These include the genetic engineering of cell lines to disrupt fucosyltranferases (FuTs) that install fucose onto glycans (15, 16) or impairing the de novo fucose biosynthetic pathway (17, 18). Alternatively, approaches involving trimming fucose from the N-glycans of purified antibodies using recombinant fucosidases (19), or exploiting chemoenzymatic methods to install specific fucose-free glycans (20, 21), have been demonstrated. Although efficient, these strategies present some limitations such as a lack of control of the resulting glycan, reduced production yields of IgG, and complex implementation to industrial processes. As an alternative, various chemical approaches to block fucosylation have attracted attention owing to their potential ease of implementation. Among these, the most successful have been metabolic inhibitors of fucosylation (2, 22–24). Here, structural mimics of fucose are converted into analogues of the activated sugar nucleotide guanosine diphosphate (GDP)-fucose (GDP-Fuc) via the fucose salvage pathway (Fig. 1D). These unnatural GDP-Fuc analogues are often poor substrates for the fucosyltransferases (FuTs) and function in one of two possible ways: i) by directly inhibiting endogenous FuTs or ii) by antagonizing enzymes of the de novo fucose biosynthetic pathway, which leads to depletion of GDP-fucose within cells.

Among the various fucose analogues that have been assessed (22), the most potent metabolic inhibitors are 5-thio-L-fucose (5SFuc, 2) (25), 2-deoxy-2-fluoro-L-fucose (2FFuc, 3) and 6-deoxy-6-fluoro-L-fucose (6FFuc, 4) (26, 27), 6-alkynyl-fucose (6AlkFuc, 5) (28), and 7-alkynyl fucose (7AlkFuc, 6) (29), and 6,6,6-trifluoro-L-fucose (Fucostatin, 7) (3) (Fig. 1C). Unfortunately, despite their ability to block antibody fucosylation, these compounds have universally shown undesirable incorporation into cellular glycans. While incorporation in some cases is as low as 0.8% (2FFuc) to 5% (6AlkFuc) (30) for some analogues it extends beyond 50% (27, 29, 31). Even at very low levels, the incorporation of unnatural monosaccharides into glycans is problematic as their biological impact cannot be fully assessed, which raises concerns for the use of this strategy for therapeutic applications (30). Alternatives to monosaccharides have also been pursued including, for example, phosphonate (3) and elegant mechanism-based phosphate metabolic precursors (32) that covalently inactivate the enzyme GDP-mannose 4,6-dehydratase (GMDS) within the de novo fucose biosynthetic pathway (Fig. 1D). These later compounds, however, have modest stability in solution and adversely affect cell growth. Furthermore, their preparation requires involved synthetic methods that may limit their use in applications requiring larger quantities of materials. Despite these challenges, these reports from both academia and industry point to clear interest in the development of simple and effective fucosylation inhibitors that are not incorporated into glycans.

Here, we describe β-carbafucose (8β) as an efficient metabolic fucosylation inhibitor. We detail simple synthetic routes that enable convenient access to carbafucose and various analogues. We show that carbafucose is readily assimilated into the de novo biosynthetic pathway and acts in various cell lines to block fucosylation both at the cell surface and on secreted antibodies. We further demonstrate that carbafucose does not affect cell growth and, using both capillary electrophoresis (CE) and mass spectrometry (MS), that it is not incorporated into antibody glycans. Thus, carbafucose can be conveniently used to tune the extent of protein and, more specifically, antibody fucosylation. We expect that carbafucose will find common use as a research tool to block cellular fucosylation and enable fundamental studies into the physiological roles of fucosylation.

Results and Discussion

One of the characteristic features of glycosyltransferases is that they catalyze their reactions by stabilizing cationic transition states (33, 34). These oxocarbenium ion-like transition states are thought to adopt a half-chair conformation in which the positive charge is distributed between the anomeric carbon and the endocyclic oxygen (35) (Fig. 2A). Activated GDP-2FFuc was earlier shown to be an inhibitor of FuTs, presumably by virtue of the fluorine substituent inductively destabilizing such oxocarbenium ion-like transition states (26, 36). Nevertheless, incorporation of 2FFuc into glycans reveals that this destabilization is inadequate to completely prevent transfer to glycans (26, 27, 30). We therefore considered alternative approaches to destabilize these oxocarbenium ion-like transition states. We recognized that replacement of the endocyclic oxygen with a methylene unit would preclude charge delocalization, which would accordingly require the transfer reaction to proceed through a far less stable carbocation-like transition state, and therefore prevent transfer to glycans (Fig. 2B). We were encouraged in this regard by previous in vitro studies showing that GDP-carbafucose is not a competent substrate for FuT3/5 (37, 38). In those studies, the synthesis of a pseudoanomeric diastereomer of carbafucose (8α) that mimics α-L-fucose (1α) was described starting from L-fucose (37). However, given that the previously described 8α cannot undergo mutarotation to mimic β-L-fucose (1β), we realized this compound would be unable to serve as a metabolic precursor for the fucose salvage pathway (Fig. 1D). Additionally, while the synthesis of protected forms of the enantiomeric β-D-carbafucose (ent-8β) has been reported from D-galactose (39), and racemic β-D/L-carbafucose has been prepared in 13 steps from myo-inositol (40), neither would be useful for the present study. We therefore set out to develop an enantioselective synthesis of β-L-carbafucose (Fig. 2C, 8β), which we expected would possess the required functionality and absolute stereochemistry to act as a competent metabolic precursor for the enzymes of the fucose salvage pathway. In addition, we aimed to generate a small panel of carbafucose analogues (9-11) that incorporated either a 2-fluoro feature (9) as seen in the successful metabolic fucosylation antagonist 3, a 6,6,6-trifluoro group (10) as featured in Fucostatin (7), or the per-O-acetylated derivative of carbafucose (11) (Fig. 2C).

Fig. 2.

Mechanism by which carbafucose analogues act to block fucosylation. (A) Proposed mechanism for enzymatic transfer of GDP-fucose and its analogue GDP-2FFuc. (B) GDP-carbafucose hypothetical carbocation-like transition state is not formed. (C) Structures of the synthesized carbafucose 8β and its derivatives.

To prepare carbafucose (8β) and 2-deoxy-2-fluoro-carbafucose 9 (Scheme 1) we started with two separate proline-catalyzed α-halogenation-aldol reactions (41, 42) involving the protected hydroxyacetone derivative 12 and the L-ribose-derived aldehyde 13. Using N-chlorosuccinimide (NCS) as the halogenating agent we accessed the syn-chlorohydrin 14 in good yield on small scales (~50%), though the yield decreased (~20%) on scales larger than 2 mmol. Following the same protocol but using the fluorinating agent Selectfluor gave the corresponding syn-fluorohydrin 15 (42). From chlorohydrin 14, a reaction sequence involving olefination to generate the corresponding diene 16, followed by deprotection and ring closing metathesis using Grubb’s second generation catalyst (43, 44) afforded the alkenyl carbasugar 18. A similar sequence of reactions executed on the fluorohydrin 15 gave the fluorinated alkenyl carbasugar 19. Hydrogenation of 19 completed the synthesis of 2-deoxy-2-fluoro-carbafucose (9), which was produced as the major component of an 8:1 mixture of diastereomers (9 and 20), the stereochemistry of which was confirmed by analysis of 2D-NOESY spectra. To access carbafucose (8β), the crude chlorinated alkenylcarbasugar 18 was converted into epoxide 21 under basic conditions in refluxing water, which also effected epoxide opening to give the alkenylcarbasugar 22 in reasonable yield over 4 steps from diene 16 (14% overall, ~60% yield per step). Hydrogenation then gave carbafucose (8β) as the major component of a separable mixture of C5 diastereomers (8β:23 = 1.5:1). The stereochemistry of 8β was confirmed by analysis of 2D-NOESY spectra and NMR spectroscopic data recorded for 8β, which we found closely matched data reported for D/L-8β (39). Finally, acetylation of 8β afforded per-O-acetylated carbafucose 11 (not shown).

Scheme 1.

Synthetic routes for the preparation of candidate carbafucose metabolic inhibitors. (A) Synthesis of alkenyl carbasugars 18 and 19. (B) Synthesis of β-carbafucose (8β) and 2-deoxy-2-fluoro-carbafucose (9). (C) Synthesis of 6,6,6-trifluorocarbafucose 10.

To access the corresponding 6,6,6-trifluorocarbafucose (10) an alternative synthetic route was required (Scheme 1). Starting from the known bromoalkene 24, which has previously been prepared in four steps from bromobenzene (45), protecting group manipulations gave the bis-silyl ether 25. Lithium-halogen exchange followed by reaction with triisopropyl borate and subsequent oxidation then afforded the ketone 26 in good yield. The reaction of the ketone function in 26 with the Ruppert–Prakash reagent (CF₃TMS) (46) followed by hydrolysis of the resulting trimethylsilylether provided a mixture of the tertiary alcohols 27. While xanthate derivatives of these alcohols proved unstable, the corresponding oxalates were prepared in good yield and smoothly underwent radical deoxygenation (47) to afford the diastereomeric trifluoromethyl products 28 and 29 in ~1.2:1 ratio, which were separable by flash column chromatography after removal of TBS groups. Removal of the acetonide protecting group from the resulting diol gave the targeted 6,6,6-trifluorocarbafucose (10) the stereochemistry of which was assigned based on analysis of 2D-NOESY spectra.

We next set out to evaluate the ability of carbafucose 8β to inhibit fucosylation in cells. We chose Chinese Hamster Ovary K1 (CHO K1) cells as a model since these cells are commonly used to produce therapeutic antibodies (48). Moreover, CHO K1 cells have been shown to only express FuT8 (49), and their glycans accordingly almost exclusively exhibit core fucosylation (50, 51), making them an appropriate model to evaluate candidate inhibitors of protein fucosylation. To assess fucosylation in a quantitative manner, we developed an image-based multiwell 384-well microplate assay using Aleuria Aurentia Lectin (AAL) as a detection method (Fig. 3A). AAL has been shown to bind to a broad range of fucose linkages (52), and is commercially available as FITC-conjugate (FITC-AAL), enabling convenient and quantitative fluorescence imaging. Blockade of fucosylation, using candidate metabolic inhibitors, should reduce FITC-AAL binding to fucosylated glycoconjugates at the cell membrane. To perform this assay, we optimized FITC-AAL concentration, to obtain the best possible signal (SI Appendix, Fig. S1). CHO K1 cells were seeded in a 96 well-plate and, after fixation, treated with various concentrations of the lectin. We assessed the assay using commercial 2FFuc as a control (Fig. 3B). The integrated fluorescence intensity for each image was extracted and normalized for the number of DAPI-stained nuclei. This method allowed us to measure in-cell IC₅₀ values. We found that 2FFuc showed an IC₅₀ value of 152 µM ± 96 which is consistent with previous reports (3) (Fig. 3C). In addition, we used the number of cells per well to measure the cytotoxicity and confirmed that 2FFuc had no effect on cell growth. We next miniaturized the image-based assay, adapting it to 384-well microplate format, and demonstrated the robustness of this workflow through reliable use in semiautomated format (Z′ = 0.33, SI Appendix, Fig. S2). In parallel, because acetylated monosaccharide analogues are often used for efficient metabolic engineering of cells (53), we synthesized per-O-acetylated carbafucose (11). Finally, we assessed carbafucose derivatives 8β-11 in a series of dose–response experiments and, surprisingly, only the parent carbafucose 8β showed clear inhibition with an IC₅₀ value of 17 µM ± 8 (Fig. 3 B and C and SI Appendix, Fig. S3). Surprisingly, 11 was not superior to 8β (SI Appendix, Fig. S3), and it was found to be cytotoxic, perhaps because of cell line dependent off-target activity of such per-O-acetylated monosaccharides (54). These data suggest that, while showing promiscuity toward either C6 modifications or the removal of the endocyclic oxygen, the biosynthetic enzymes from the de novo pathway appear to not tolerate the presence of both unnatural features.

Fig. 3.

Cell-based fucosylation assay. (A) Schematic representation of the lectin-based assay to assess fucosylation in CHO cells. Created with https://BioRender.com. (B) Representative microscopy images of CHO K1 cells treated with vehicle, 2-deoxy-2-fluoro-L-fucose (100 µM), or carbafucose (100 µM). (C) Dose–response effect of treatment with dimethyl sulfoxide (DMSO), 2-deoxy-2-fluoro-L-fucose, and carbafucose on fucosylation of CHO K1 cells and corresponding IC₅₀ values. Image-based quantitative analysis was performed using integrated fluorescence. n = 3 independent biological replicates.

We next examined whether carbafucose 8β would exert effects in other cell types. We therefore assessed its effects in other cell lines including human fibroblasts, SK-N-SH, and U2OS cells, and found that fucosylation was blocked in all of these cell lines with a similar level of efficiency (SI Appendix, Fig. S4). We then set out to examine whether carbafucose 8β could block other types of fucosylation by extending the study to two cell lines that have been previously shown to present other fucosylation motifs in both N- and O-glycans, namely HepG2 and HL-60 cells. After screening the cells in a lectin-based assay using AAL, which broadly recognizes fucose, and Lotus Tetragonolobus (LTL), which recognizes α1,3-linked fucose, we measured a significant and dose-dependent decrease of both lectin binding in these two cell lines (SI Appendix, Fig. S5). To confirm that various forms of fucosylation were affected by treatment with carbafucose, we analyzed the fucosylation of these cells as well as SK-N-SH by MALDI-MS profiling of their glycans, with or without treatment with carbafucose. The data showed a clear decrease of all fucosylated species, suggesting that carbafucose enables the inhibition of fucosyltransferases other than FuT8 (SI Appendix, Figs. S6–S8).

Returning our focus to CHO K1 cells and having shown that treatment with carbafucose (8β) led to a total loss of cell surface fucosylation, we decided to test whether 8β was, as we expected, metabolized into its end product GDP-carbafucose. To that end, we treated CHO K1 cells with concentrations of 8β that led to a total inhibition of fucosylation (100 µM for 4 d). Cells were subsequently harvested and lysates analyzed by LC–MS (Fig. 4 A and B). Commercial GDP-fucose and synthetic GDP-carbafucose (37) (SI Appendix, Fig. S9) were used as standards to confirm the retention times of metabolites, and commercial GDP-glucose was used within all the samples as an internal standard for quantitation. In control cells treated with vehicle, a peak within the chromatograph having a retention time corresponding to that of GDP-fucose was detected. Notably, this peak was not observed when analyzing the lysates obtained from carbafucose-treated cells. However, we observed a peak having a greater retention time matching that of a GDP-carbafucose standard (Fig. 4B). Mass spectrometry (MS) analysis of both lysates confirmed that the detected species were GDP-fucose and GDP-carbafucose (Fig. 4C). Furthermore, quantification based on the internal standard showed the presence of GDP-fucose and the absence of GDP-carbafucose in control cells, and, conversely, the absence of GDP-fucose and the presence of GDP-carbafucose in cells treated with 8β (Fig. 4D). Altogether, these results confirm that cellular uptake of 8β is followed by its metabolic processing into GDP-carbafucose. Additionally, the fact GDP-fucose was not detected within cells treated with 8β indicates that accumulated GDP-carbafucose inhibits de novo biosynthesis of GDP-fucose as described in previous studies using fucose-based metabolic inhibitors (3, 26). These data suggest the principle mechanism of action of 8β is through decreasing GDP-fucose levels within cells, which is why 8β may serve to decrease the installation of fucose in different structural contexts.

Fig. 4.

Carbafucose is metabolized in GDP-carbafucose in CHO K1 cells. (A) Workflow for lysates analysis of CHO K1 cells after treatment with carbafucose 8b. Created with https://BioRender.com. (B) LC analysis of lysates from cells treated with vehicle (green), carbafucose (red), and corresponding standards (gray). (C) LC–MS analysis of lysates and identification of GDP-fucose and GDP-carbafucose. (D) Quantification of GDP-fucose and GDP-carbafucose in CHO K1 cells.

Following the validation of 8β as a cellular fucosylation inhibitor and having characterized its mechanism of action in CHO K1 cells, we decided to apply this metabolic inhibitor to the production of afucosylated antibodies. We produced the IgG1 antibody Herceptin (Trastuzumab®) in GCHO/2103-LC2/HC4 GPex cells. Herceptin is an FDA-approved therapeutic antibody used in the treatment of breast cancer (55). Previous studies demonstrate that afucosylated Herceptin shows an increased in vivo efficacy compared to its fucosylated counterpart (56) and we therefore envisioned that it would be a pertinent model for assessing the effects of carbafucose 8β. Herceptin was produced in the absence or presence of a range 5 to 150 µM 8β over 15 d (SI Appendix, Table S2). Cell growth and viability, as well as viable cell density, antibody production, and levels of metabolites were monitored over the production time and no significant differences were observed between the different doses of carbafucose (SI Appendix, Figs. S10–S18). We then harvested, isolated, and purified the IgG1 antibody expressed in the presence of 40 or 60 µM of carbafucose 8β or in the absence of treatment. LC–MS analysis showed a decrease of 292.08 Da in the mass of the intact purified antibody obtained from the carbafucose-treated cells relative to antibody from control cells (Fig. 5A). This mass difference corresponds to the mass of two fucose units (328 Da) minus two molecules of water (36 Da), confirming the loss of fucose on both N-glycans from the IgG Fc domain. To further investigate the effects on glycosylation, we produced small batches of Herceptin using similar conditions, but this time in the presence of a range of concentrations of 8β. Subsequent capillary electrophoresis analysis of N-glycans released from Herceptin isolated from these samples showed a clear dose-dependent effect on the antibody glycosylation (Fig. 5B). The amount of fucosylated species we detected decreased with increasing doses of 8β whereas, conversely, the levels of afucosylated species increased. By compiling the results, we obtained the relative fractions of fucosylated and afucosylated antibodies as a function of the concentration of carbafucose (Fig. 5C). Based on these data, we established that carbafucose 8β blocks core fucosylation in cells with an EC₅₀ value of 15 µM ± 8 and therefore shows that carbafucose enables the efficient production of afucosylated IgG1 antibodies.

Fig. 5.

Carbafucose enables production of afucosylated antibodies and is not incorporated into the resulting glycans. (A) MS analysis of intact Herceptin expressed in the presence or absence of carbafucose 8β. Created with BioRender.com (B) Capillary electrophoresis analysis of N-glycans released from Herceptin produced in the presence of various concentrations of 8β. (C) Relative fractions of fucosylated and afucosylated antibodies as a function of the concentration of 8β. (D) LC–MS/MS of glycopeptides obtained by trypsin digestion of Herceptin expressed in the presence or absence of 8β (40 or 60 µM). (E) Corresponding extracted ion chromatography (XIC).

As discussed earlier, a major limitation of previously described metabolic inhibitors of fucosylation is their low-level incorporation into the resulting glycan in place of fucose. Our set of data suggested that carbafucose 8β is not incorporated into glycans (Fig. 5A), however, we wanted to rigorously establish this in glycans in the context of antibody production. To that end, the purified IgG proteins were submitted to trypsin digestion, and the resulting glycopeptides were analyzed by LC–MS/MS. The identified fucosylated glycopeptides were quantified based on the relative peak areas of their respective extracted ion chromatograms (XIC) (Fig. 5 D and E). Analysis of the samples from cells treated with carbafucose 8β confirmed the lack of fucosylated species (<1%) and, despite extensive effort, we were unable to detect any carbafucosylated glycopeptides (Fig. 5E). In contrast, for the control samples treated with vehicle, we observed only fucosylated glycopeptide. We also noted a minor peak corresponding to the hypothetical carbafucosylated species (~0.6 %), however, given the control samples were never exposed to carbafucose, we conclude that this species is likely an isobaric peptide. Notably, when repeating these analyses in SK-N-SH, HepG2, and HL-60 cell lines, we were also unable to detect any transfer of carbafucose on the analyzed glycans, suggesting that the observed results are not limited to core fucosylation in CHO cells. Therefore, carbafucose 8β represents a marked improvement over existing metabolic inhibitors of fucosylation, being both more potent as well as not being incorporated into cellular glycans.

Conclusion

We describe carbafucose 8β as a rationally designed and mechanism-inspired metabolic inhibitor of fucosylation. The replacement of the endocyclic oxygen with a methylene group prevents the formation of the oxocarbenium ion transition state used by fucosyltransferases to enable fucose transfer from activated nucleotide sugar. The synthetic routes we detail, starting from readily available inexpensive precursors, enabled easy access to parent 8β, as well as carbasugar mimics of known carbohydrate-based fucosylation inhibitors. A robust semiautomated lectin-based assay permitted rapid assessment of the potency of this series of compounds and exhibits favorable assay characteristics that make it compatible with high-throughput screening. We speculate this could be a convenient way to screen for cell active fucosylation inhibitors. Using this assay, we established that 8β compared favorably with currently used metabolic inhibitors of fucosylation. Furthermore, we find that 8β acts in various cell lines to decrease levels of fucose at the cell surface in different contexts. Though we have not illustrated the effects of 8β on all forms of cell surface fucosylation, we believe that the generality of the effects we observe upon treatment of cells with 8β is most consistent with this compound efficiently blocking the biosynthesis of GDP-fucose. We expect that the resulting reduction in the levels of the natural donor should broadly antagonize the function of all cellular FuTs. Finally, we demonstrate that, when applied to the production of the FDA-approved therapeutic antibody Herceptin, carbafucose 8β enables the efficient production of afucosylated antibody, with no effects on cell growth nor antibody yield. Analytical studies confirm that 8β is a metabolic fucosylation inhibitor that is not incorporated into glycans. Given that therapeutic antibody-based drugs have demonstrably higher efficacy, the rapid production, and screening of the performance of antibodies acting by ADCC in both their fucosylated and afucosylated forms is a topic of high interest. Given the performance of carbafucose 8β in enabling production of afucosylated Herceptin, we envision this simple molecule could serve as a convenient and robust approach to rapidly access afucosylated antibodies. An added benefit is that by using variable levels of carbasugar 8β it should also be possible to conveniently tune the extent of antibody fucosylation for certain applications. Furthermore, we expect our findings to be useful beyond the field of antibody production and in fundamental research where modulating cellular fucosylation could be of interest in uncovering the roles of protein fucosylation in various contexts including more recently proposed roles for core fucosylation (55), to terminal fucosylation as seen in sLeX (3–6) through to fucosylation of epidermal growth factor repeats (7). Moreover, as previously shown with 2-deoxy-2-fluoro-fucose (27), carbafucose 8β is also likely to be useful in vivo study, perhaps even more so given its greater potency. Accordingly, carbafucose 8β could serve as a useful tool for fundamental studies of fucosylation in preclinical models. Finally, due to the mechanistic relatedness of glycosyltransferases, (34) we expect that it should be feasible to apply the concept of carbasugars as nonincorporated metabolic inhibitors to other forms of glycosylation.

Materials and Methods

Complete detailed methods are provided in SI Appendix, including details on general methods, cell culture, image-based lectin fucosylation assay, HPLC methods for analysis of nucleotide sugars, Herceptin production in CHO cells, LC–MS analysis of intact antibodies, in-solution proteolytic digestion of IgG proteins, glycopeptide analysis by LC–MS, glycopeptide identification and quantification, glycomic analysis of cell lines by MALDI-MS, complete methods for the synthesis of all compounds, and spectral characterization of all compounds.

Analysis of CHO K1 Cell Lysates: Nucleotide Sugars.

CHO K1 cells were seeded in a 10 cm dish according to the procedure described above. After overnight incubation at 37 °C and 5 % CO₂, the media were removed and replaced by fresh media containing 100 µM carbafucose 8β or vehicle (DMSO). Nucleotide sugars were extracted from cells based on previously established methods (1, 2). In brief, cells were washed with PBS, typsinized, and collected in a 1.5 mL microcentrifuge tube. Cells were then pelleted by centrifugation at 300 × g, and the supernatants were discarded. Cell pellets were snap-frozen in liquid nitrogen and stored at −20 °C until they were processed. GDP-glucose (10 μL of a 100 μM stock solution in water) was used as an internal standard and added to all cell samples before extraction. Then, 750 μL of 70% ice cold ethanol was added to cells after which cell pellets were subsequently vortexed on with a vortex mixer (Mandel; Benchmark Scientific Inc.) and dispersed via sonication with an ultrasonic water bath cleaner (VWR). After sonication, samples were placed in the freezer (−20 °C) for 20 min, and insoluble cell material was removed by centrifugation at 13,000 × g for 5 min; the soluble supernatants, which contained the nucleotide sugars, were placed into new 1.5 mL microcentrifuge tubes. The supernatants were immediately snap-frozen in liquid nitrogen and dried on a Savant SpeedVac concentrator (Thermo Scientific). Dry supernatants were stored in the freezer until solid-phase extraction.

Crude, dried cell extracts were redissolved in water (0.2 mL) and applied to Supelclean™ 250 mg ENVI-Carb 3 mL SPE-tubes (MilliporeSigma) conditioned prior to use by washing with 3 mL 80% acetonitrile (ACN) + 0.1% trifluoroacetic acid (TFA) followed by 6 mL water. All SPE procedures were carried out using positive pressure. After loading the SPE cartridges, they were sequentially washed with water (3 mL), 25% ACN (3 mL), and 50 mM triethylammonium acetate (TEAA, 3 mL) after which nucleotide sugars were eluted in 50 mM TEAA containing 25% ACN (2 × 0.75 mL). The nucleotide sugar-containing fractions were pooled, immediately snap-frozen in liquid nitrogen and concentrated using a SpeedVac. Samples were reconstituted in 100 μL water for HPLC–MS analysis. Nucleotide sugar standards were prepared for HPLC analysis as 10 μM solutions.

Synthetic Route to Carbafucose (8β).

Chloro ketone (14).

To a stirred solution of aldehyde 13 (626 mg, 2.44 mmol) in CH₂Cl₂ (6.1 mL) at 0 °C was added N-chlorosuccinimide (358 mg, 1.10 mmol) and (S)-Proline (224 mg, 0.800 mmol). The reaction mixture was stirred for 30 min at 0 °C. Next, ketone 12 (919 mg, 4.88 mmol) in DMSO (12.2 mL) was added followed by H₂O (0.2 mL). The reaction mixture was warmed to room temperature and stirred overnight. Brine and ether were added to the reaction mixture and the organic phase was washed five more times with brine. The combined aqueous phases were extracted with diethyl ether one last time. The combined organic layer was dried over anhydrous MgSO₄, filtered, concentrated, and the resulting crude product was purified by column chromatography (5% Et₂O in hexane) to afford 14 (257 mg, 22%) as a colorless viscous liquid. Data for 14: [α]²⁰_D = −25.3 (c 1.19, CHCl₃); ¹H NMR (600 MHz, CDCl₃) δ 5.66−5.58 (m, 4 H), 5.36−5.27 (m, wwww2 H), 4.83 (dd, J = 10.5, 3.0 Hz, 1 H), 4.41 (m, 1 H), 2.14 (s, 3 H), 2.08 (s, 3 H), 2.06 (m, 1 H), 1.97 (s, 3 H), 1.92–1.78 (m, 2 H), 1.24 (s, 18 H), 0.95 (d, J = 6.6 Hz, 3 H); ¹³C NMR (151 MHz, CDCl₃) δ 209.4, 137.7, 117.8, 78.5, 76.9, 71.4, 64.8, 25.7, 25.4, 18.2, 18.1, 17.8, 12.6, −4.8, −5.0; IR (neat) ν 3,522, 2,994, 1,721, 1,099, 838 cm⁻¹; HRMS (ESI-TOF) m/z: [M + NH₄]⁺ Calcd for C₂₃H₅₁NClO₄Si₂ 496.3040; Found 496.3045.

Chloro di-alkene (16).

To a stirred solution of Julia’s reagent (724 mg, 3.26 mmol) in THF (5.4 mL) at −78 °C was added lithium hexamethyldisilane (1.0 M in hexanes, 1.63 mL, 1.63 mmol). The reaction mixture was stirred for 30 min at −78 °C. Next, chlorohydrin 14 (260 mg, 0.542 mmol) in THF (2.7 mL) was added and the reaction mixture was stirred for 30 min at −78 °C. The reaction mixture was quenched with H₂O and extracted with EtOAc. The organic layer was dried over anhydrous MgSO₄, filtered, concentrated, and the resulting crude product was purified by column chromatography (6% Et₂O in hexane) to afford 16 (186 mg, 72%) as a colorless viscous liquid. Data for 16: [α]²⁰_D = –15.1 (c 0.96, CHCl₃); ¹H NMR (500 MHz, CDCl₃) δ 6.09 (ddd, J = 7.2, 10.3, 17.4 Hz, 1 H), 5.31 (dt, J = 1.3, 17.3 Hz, 1 H), 5.22 (dt, J = 1.0, 10.3 Hz, 1 H), 5.0 (m, 1 H), 4.54 (m, 1 H), 4.33 (d, J = 5.4 Hz, 1 H), 4.04 (d, J = 8.6 Hz, 1 H), 3.98 (dd, J = 5.6, 8.7 Hz, 1 H), 2.12 (d, J = 5.5 Hz, 1 H), 1.73 (s, 3 H), 1.14−1.04 (m, 21 H), 0.88 (s, 9 H), 0.10 (s, 3 H), 0.03 (s, 3 H); ¹³C NMR (126 MHz, CDCl₃) δ 145.1, 138.3, 117.2, 115.4, 77.5, 77.2, 70.4, 65.3, 25.9, 18.2, 18.2, 16.6, 12.6, −4.6, −5.0; IR (neat) ν 3,661, 2,957, 1,572, 1,462, 1,084, 836 cm⁻¹; HRMS (ESI-TOF) m/z: [M + H]⁺ Calcd for C₂₄H₅₀ClO₃Si₂ 477.2982; Found 477.2977.

Chloro di-alkene triol (S-3).

To a stirred solution of 16 (36.4 mg, 0.0763 mmol) in THF (2 mL) at 0 °C was added tetra-n-butylammonium fluoride (1.0 M in THF, 0.24 mL, 0.24 mmol). The ice bath was removed, and the reaction mixture was stirred for 3 h. The reaction mixture was then treated with saturated aq NH₄Cl solution and extracted with EtOAc. The aqueous layer was separated and further extracted with EtOAc. The combined organic phases were washed with brine, dried over anhydrous MgSO₄, filtered, concentrated, and the resulting crude was purified by column chromatography (50% EtOAc in petroleum ether) to afford S-3 (11 mg, 70%) as a white foam. Data for S-3: [α]²⁰_D = −11.5 (c 0.46, CHCl₃); ¹H NMR (400 MHz, CDCl₃) δ 5.87 (ddd, J = 17.3, 10.5, 6.8 Hz, 1 H), 5.53 (dt, J = 17.2, 1.2 Hz, 1 H), 5.42 (dt, J = 10.6, 1.1 Hz, 1 H), 5.15 (dt, J = 2.0, 1.0 Hz, 1 H), 5.06 (t, J = 1.8 Hz, 1 H), 4.31 (d, J = 5.3 Hz, 1 H), 3.61 (dd, J = 7.9, 5.3 Hz, 1 H), 3.49 (dd, J = 6.9, 4.3 Hz, 1 H), 3.22 (dd, J = 7.9, 4.3 Hz, 1 H), 2.34 (brs, 2 H), 1.82 (t, J = 1.2 Hz, 3 H); ¹³C NMR (151 MHz, CDCl₃) δ 142.7, 136.0, 129.6, 117.7, 112.8, 87.0, 83.9, 75.4, 74.0, 18.4; IR (cast film) ν 3,434, 3,405, 3,343, 3,077, 2,980, 2,917, 1,429, 1,381, 1,216, 1,110, 767 cm⁻¹; HRMS (ESI-TOF) m/z: [M + Na]⁺ Calcd for C₉H₁₅ClO₃Na 229.0602; Found 229.0613.

Alkenyl tetrol (22).

To a stirred solution of triol S-3 (37 mg, 0.18 mmol) in CH₂Cl₂ (8.9 mL) was added Grubbs II catalyst (15.3 mg, 0.018 mmol). The reaction mixture was stirred for 1 h at 40 °C. The reaction mixture was then filtered through a plug of celite and flushed with CH₂Cl₂. The filtrate was concentrated in vacuo. To the resulting crude taken in THF (0.90 mL) was added a 2 M aqueous solution of sodium hydroxide (0.18 mL). The reaction mixture was stirred at reflux for 12 h and concentrated. The resulting crude product was purified by column chromatography (20% CH₃OH in EtOAc) to afford 22 (5.8 mg, 20% over two steps) as a white foam. Data for 22: [α]²⁰_D = −61.5 (c 0.90, CH₃OH); ¹H NMR (400 MHz, CD₃OD) δ 5.40 (m, 1 H), 3.95 (d, J = 4.3 Hz, 1 H), 3.88 (m, 1 H), 3.57 (dd, J = 7.6, 10.5 Hz, 1 H), 3.38 (dd, J = 4.3, 10.5 Hz, 1 H), 1.82 (t, J = 1.7 Hz, 1 H); ¹³C NMR (151 MHz, CD₃OD) δ 136.3, 128.3, 74.1, 73.6, 72.8, 72.0, 20.8; IR (cast film) ν 3,724, 3,698, 2,349, 1,664, 1,436, 1,084 cm⁻¹; HRMS (ESI-TOF) m/z: [M + NH₄]⁺ Calcd for C₇H₁₆NO₄ 178.1074; Found 178.1075.

Carbafucose (8β).

To a stirred solution of 22 (20.0 mg, 0.124 mmol) in EtOH (2 mL) at rt was added Pd/C (2 mg). Then, the reaction mixture was equipped with a balloon filled with H₂ and stirred overnight. The reaction mixture was then filtered through a pad of celite and flushed with CH₃OH. The resulting solution was concentrated in vacuo to obtain a mixture of 8β and 23 which exhibits ~1:1.5 diastereomeric ratio in favor of 8β, as determined by ¹H NMR spectroscopy in CD₃OD at 27 °C This mixture was purified by column chromatography (20% CH₃OH in EtOAc) to afford 8β and 23 (10.2 mg combined, 50%) as a white foam. Data for 8β: [α]_D²⁰ = +22.3 (c 0.05, CH₃OH); ¹H NMR (600 MHz, CD₃OD) δ 3.68 (brs, 1 H), 3.49 (app t, J = 9.3 Hz, 1 H), 3.39−3.35 (m, 1 H), 3.27 (dd, J = 9.6, 3.0 Hz, 1 H), 1.65−1.60 (m, 1 H), 1.57−1.48 (m, 2 H), 1.02 (d, J = 6.8 Hz, 3 H); ¹³C NMR (151 MHz, CD₃OD): δ 76.5, 76.4, 74.9, 73.8, 35.8, 33.1, 17.8; IR (neat) ν 3,368, 2,958, 2,928, 2,857, 1,731, 1,668, 1,462, 1,261, 1,067, 1,022, 799 cm⁻¹; HRMS (ESI-TOF) m/z: [M + Na]⁺ Calcd for C₇H₁₄NaO₄ 185.0784; Found 185.0784.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

The authors confirm that the data supporting the findings of this study are available within the article and SI Appendix.