N‑Azidoacetylglucosamine (GlcNAz)-Derived Sugar Oxazolines as a New Class of Substrates for Chemoenzymatic Site-Specific Antibody Bioconjugation

Abstract

Endoglycosidase-catalyzed Fc glycan remodeling using glycan oxazolines as substrates has become a general method to produce homogeneous antibody glycoforms and site-specific antibody–drug conjugates (ADCs). While endoglycosidases generally tolerate structural modifications at nonreducing terminal sugar moieties, the reducing-terminal GlcNAc oxazoline has been regarded as an essential and largely immutable structural motif for enzyme recognition. In this work, we sought to better understand substrate recognition by endoglycosidases and to develop simpler and more efficient strategies for synthesizing site-specific ADCs. We have performed site-selective modifications on the GlcNAc oxazoline moiety concurrent with structure–activity relationship (SAR) studies. We found that while the endoglycosidase Endo-S2 does not tolerate modifications at the C-3 or C-6 positions, the enzyme can tolerate certain modifications at the methyl group of the oxazoline portion, allowing the introduction of azide and halogen atoms at this site. In contrast to the conventional method of introducing tags to the nonreducing terminal glycans through ether bonds, which requires tedious protection–deprotection steps, this new strategy enables the one-step introduction of a tag (e.g., azide) to the reducing-terminal glucosamine moiety while leaving free hydroxyl groups on other positions intact. This approach significantly enhances the efficiency of the ADC preparation. These findings open a new avenue to antibody tagging and bioconjugation with a class of much simpler disaccharide substrates. In addition, we found that a second sugar moiety β-1,4-linked to the GlcNAc oxazoline also played an important role, with the mannose moiety being the most efficient for enzymatic transglycosylation. A structural modeling analysis indicated that there was a cavity in the enzyme pocket that permits certain modifications at the methyl group. This new method was successfully used to produce site-specific ADCs. Cell-based assays showed that the resulting ADCs exhibited potent cell killing of cancer cells that overexpressed the corresponding antigen.

Introduction

Antibody–drug conjugates (ADCs) have emerged as an important class of anticancer therapeutics that combine the specificity of antibodies with the high potency of cytotoxic drugs. − Over a dozen new ADCs were approved by the U.S. Food and Drug Administration (FDA) for the treatment of cancers within the past 10 years, and hundreds are at various stages of preclinical and clinical development. First-generation ADCs have been generated through nonspecific conjugation, which usually results in heterogeneous mixtures with varied drug/antibody ratios (DARs) and complicated pharmacological properties. Recent studies have demonstrated that site-specific ADCs are superior to randomly conjugated ADCs, with improved pharmacokinetics, in vivo stability, and/or favorable safety profiles. ,

Different strategies have been attempted to generate homogeneous ADCs, including antibody engineering, enzymatic manipulations, and chemoselective reactions. One promising method is glycosite-specific bioconjugation, which takes advantage of the conserved Fc N-glycan at Asn-297 as a handle for selective chemoenzymatic manipulations. − This method allows for conjugation of payloads specifically to the Fc domain and does not require protein engineering. In particular, the enzymatic Fc glycan remodeling strategy using azide- and other functionalized sugar oxazolines as donor substrates, as reported by our group and others, has become an efficient approach for synthesizing site-specific antibody–drug and antibody–ligand conjugates (Figure a). − Specifically, we have demonstrated that selectively azide-tagged disaccharide oxazolines can act as donor substrates of IgG-specific endoglycosidases from Streptococcus pyogenes (e.g., Endo-S and Endo-S2) for a one-pot two-step or one-pot single-step antibody–drug conjugation to synthesize site-specific antibody-drug conjugates. , However, for previous studies, all the sugar oxazoline donor substrates used have been derived from oligosaccharides carrying an unmodified N-acetylglucosamine (GlcNAc) at the reducing terminus, and all modification/functionalization has been carried out at the nonreducing sugar moieties. It is not clear whether these enzymes can tolerate modifications directly on the 5-membered oxazoline ring or on the oxazoline-fused sugar moiety. To address this issue with the goal of developing simple and effective antibody conjugation methods, we synthesized a series of disaccharide derivatives with selective modifications at the oxazoline and its fused sugar moiety and evaluated their enzymatic activities. We observed that Endo-S2 could not tolerate most modifications at the methyl group of the oxazoline heterocycle or at the 3-, 4-, and/or 6-hydroxyl group of the oxazoline-fused sugar moiety. Nevertheless, we found that an azide substitution on the oxazoline methyl group was well tolerated, and the resulting N-azidoacetylglucosamine (GlcNAz)-derived disaccharide oxazolines were excellent substrates of Endo-S2 for antibody Fc glycan remodeling. In addition, while plain GlcNAc-oxazoline was not a substrate, an attachment of a monosaccharide moiety (mannose, glucose, or galactose) at the 4-position of the GlcNAc oxazoline could rescue substrate activity. Molecular modeling analysis indicated that the additional sugar moiety made essential contacts with the enzyme, with the mannose moiety making the most favorable contributions. Since the GlcNAz derived sugar oxazolines can be prepared easily from readily available starting materials, this discovery, in combination with subsequent click reactions, provides a new and facile method for producing site-specific antibody–drug conjugates (Figure b).

1.

Comparison of the endoglycosidase-catalyzed Fc glycan remodeling methods for ADC synthesis. (A) Previous work: azide-tagged glycan oxazolines as donor substrates in Endo-S/S2 (WT/mutants) catalyzed IgG functionalization over the past decade. (B) This work: Using easily accessible GlcNAz-derived oxazolines as donor substrates for Endo-S2 catalyzed site-specific IgG functionalization.

Results and Discussion

Synthesis of Disaccharide Oxazolines with Selective Modifications on the Oxazoline Ring and the Fused Sugar Moiety

Our previous studies have shown that Manβ1,4-GlcNAc oxazoline (1) and its derivatives with modifications at the mannose moiety could serve as excellent substrates of wild-type Endo-S2 for Fc glycan remodeling. , To investigate if Endo-S2 can tolerate any modifications on the oxazoline ring and the fused sugar moiety, we sought to modify the 3/6-OH with either an azido group or to introduce a tag at the oxazoline methyl group as the first set of substrate analogues (compounds 2, 3, and 4 in Scheme A). The synthesis of 2 started with the removal of the acetyl protecting groups in 5, followed by benzylidenation to obtain 6. Compound 6 was converted to the 6-azide-derivative (7) in 4 steps in 58% overall yield. Glycosylation of 7 and the glycosyl donor (8) using Crich et al.’s β-mannosylation method gave the disaccharide derivative (9) in 61% yield (Scheme B). The N-phthaloyl group in 9 was then changed to an N-acetyl group by treatment with ethylenediamine followed by N-acetylation to afford 10 in 78% yield. Chemoselective hydrolysis of benzylidene acetals under acidic conditions, followed by selective removal of the PMB and Bn groups in the presence of azide using oxidative conditions (NaBrO₃/Na₂S₂O₄), provided the free disaccharide (11) in 60% yield in two steps. Finally, treatment of 11 with 2-chloro-1,3-dimethylimidazolinium chloride (DMC)/TEA in water yielded the disaccharide oxazoline (2) (Scheme B).

1. Synthesis of GlcNAc Modified Disaccharide Oxazoline Derivatives .

^a (a) NaOMe, MeOH; benzaldehyde dimethyl acetal, TsOH; (b) BnBr, NaH, DMF; CSA, MeOH; TsCl, Et₃N, CH₂Cl₂; NaN₃, DMF; (c) 8, TTBP, BSP, CH₂Cl₂, Tf₂O, 4 Å MS, −60 °C, 3 h; (d) NH₂CH₂CH₂NH₂, n-BuOH, 90 °C, 4 h; Ac₂O, Py, rt, overnight; (e) CSA, MeOH; NaBrO₃, Na₂S₂O₄, EtOAc/H₂O; (f) DMC, water, Et₃N; (g) MeI, NaH, DMF; Et₃SiH, BF₃·Et₂O, CH₂Cl₂; (h) H₂, Pd/C, Pd­(OH)₂; (i) Ac₂O, acetonitrile/water, NaHCO₃; (j) N₃–CH₂COONHS ester, acetonitrile/water, NaHCO₃.

Synthesis of the C-3 modified derivative is depicted in Scheme C. At the beginning, we were attempting to introduce a 2-azidoethyl group to the C-3 position; however, treatment of compound 6 with NaH and 2-azidoethyl tosylate did not yield the desired product. Then, we switched to introducing a methyl group, which was achieved by treatment of 6 with NaH and MeI in high yield. Next, the regioselective opening of the benzylidene group with Et₃SiH/BH₃·Et₂O afforded 12 with a free OH at the C-4 position in 61% yield over two steps. Glycosylation of 12 and the mannosyl donor (8) gave disaccharide 13 in 69% yield. Conversion of the N-Phth group into the NHAc group following the same procedure as described in the synthesis of 10 furnished disaccharide 14 in 85% yield over two steps. Finally, global deprotection followed by oxazoline formation gave disaccharide oxazoline 3 in high yield (Scheme C).

Synthesis of the Manβ1,4-GlcNAz oxazoline (4) is demonstrated in Scheme D. Glycosylation of the monosaccharide acceptor (16) and the mannosyl donor (8) afforded the known disaccharides (17) in 66% yield. Hydrogenation of 17 followed by the installation of an N-acetyl and N-azidoacetyl group furnished the disaccharide Manβ1,4-GlcNAc (18) and Manβ1,4-GlcNAz (19), respectively. Finally, one-step oxazoline formation by treatment of the free disaccharides (18 and 19) with DMC/TEA in water gave the Manβ1,4-GlcNAc oxazoline (1) , and the Manβ1,4-GlcNAz oxazoline (4), respectively, in high yields (Scheme D).

Evaluation of the Activity of the Disaccharide Oxazoline Derivatives (1–4) as Substrates of Endo-S2

With the ManGlcNAc oxazoline derivatives (1–4) in hand, we evaluated their activity in Endo-S2-catalyzed transglycosylation reactions with trastuzumab as the acceptor substrate. As expected, the nonmodified Manβ1,4GlcNAc oxazoline (1) was an excellent substrate for Endo-S2 catalyzed glycosylation, and a quantitative conversion was achieved within 10 min when a catalytic amount of enzyme (1:500, w/w, enzyme/antibody) and 20 equiv of oxazoline (1) (per reaction site) were used under the described conditions (Figure ). However, under the same conditions, the C-6 and C3-GlcNAc modified oxazoline derivatives (2, 3) exhibited a dramatically decreased activity toward Endo-S2, giving less than 15% yield at 1 h under the same conditions. These results suggest that Endo-S2 is sensitive to the modification of the C-3 and C-6 positions of the oxazoline-fused sugar ring moiety. On the other hand, we found that the Manβ1,4GlcNAz oxazoline (4) could act as an excellent substrate for Endo-S2 and gave a >95% conversion yield within 10 min and a quantitative conversion within 30 min under the same conditions (Figure ).

2.

Evaluation of the Manβ1,4GlcNAc oxazoline derivatives with selective modification of the sugar moiety fused with the oxazoline ring. (A) Schematic presentation of the Endo-S2 catalyzed reactions with the Manβ1,4GlcNAc oxazoline derivatives (1–4); (B) side-by-side comparison of the enzymatic transglycosylation efficiency. Reaction conditions: trastuzumab, 500 μg (25 mg/mL); oxazoline, 20 equiv/site; Endo-S2 WT, 1/500 (w/w); 100 mM phosphate buffer, pH 7.0, 30 °C.

Synthesis and Evaluation of Glcβ1,4GlcNAz Oxazoline (24) and LacNAz Oxazoline (27)

Inspired by the promising Fc glycan remodeling efficiency of Manβ1,4GlcNAz oxazoline (4), we sought to evaluate the scope of GlcNAz disaccharide oxazolines as enzyme substrates by replacing the nonreducing terminal mannose with other monosaccharides, such as glucose or galactose. Accordingly, we synthesized Glcβ1,4GlcNAz oxazoline (24) and LacNAz oxazoline (27) (Scheme ). Glycosylation of acceptor 16 and glucosyl donor 20 was achieved using TMSOTf as the catalyst to give the disaccharide derivative (21) in 85% yield. Deacetylation, followed by simultaneous reduction of the azido group and debenzylation via catalytic hydrogenation over Pd/C in compound 21, afforded the Glcβ1,4-glucosamine (22) in 90% yield over two steps. Installation of the azidoacetyl group to 22 followed by oxazoline formation gave Glcβ1,4-GlcNAz oxazoline (24) in excellent yield (Scheme A). LacNAz oxazoline (27) was synthesized through a simple 2-step conversion from commercially available lactosamine (25) in excellent yield (Scheme B). This facile two-step synthesis of the azide-modified disaccharide oxazoline substrates significantly improves the efficiency of ADC preparation compared to our previously developed disaccharide oxazoline carrying an azido tag on the nonreducing terminal sugar moieties, which required a multiple step synthesis (a total of 15 steps for the synthesis of the 6-N₃-PEG₃-LacNAc oxazoline).

2. Synthesis of Glcβ1,4-GlcNAz-ox (24) and LacNAz-ox (27)­ .

^a (a) TMSOTf, CH₂Cl₂, −10 °C to r.t., 1 h, 85%; (b) MeONa, MeOH, r.t., 6 h; (c) MeOH, HCl, H₂, Pd/C, r.t., 4 h, 90% over two steps; (d) N₃–CH₂COONHS, acetonitrile/water, NaHCO₃, 0 °C to r.t., 2 h, 82% for 23, 75% for 26; (e) DMC, water, Et₃N, 0 °C, 2 h, 86% for 24, 84% for 27.

We then tested the activity of the synthetic Glcβ1,4-GlcNAz-ox (24), LacNAz-ox (27), and Manβ1,4-GlcNAz-ox (4) for enzymatic glycan remodeling. We found that the sugar moiety attached at the C-4 of the oxazoline-fused ring affected the substrate activity, with the mannose substitution contributing most favorably, followed by the glucose and galactose moieties. For example, under the conditions tested when 0.5% enzyme (w/w) was used, a complete conversion was achieved for Manβ1,4-GlcNAz-ox (4) within 10 min, while a 98% conversion and a 75% conversion were achieved at 30 min for Glcβ1,4-GlcNAz-ox (24) and LacNAz-ox (27), respectively (Figure , reaction condition a). By increasing the enzyme usage to 1% (w/w), complete conversion was also achieved for LacNAz-ox (27) (Figure , reaction condition b). To demonstrate the scope of glycosite-specific antibody conjugation, we performed glycan remodeling on another IgG antibody, brentuximab, a CD30-specific monoclonal antibody. EndoS2-mediated transglycosylation with Manβ1,4-GlcNAz-ox (4) was completed within 10 min under the same reaction conditions used for the preparation of 28 (Figure , reaction condition a), demonstrating similar efficiency to that observed for trastuzumab. No hydrolysis of glycan-remodeled brentuximab was observed after an extended reaction time of 3 h. The corresponding data are provided in the Supporting Information (Figure S1). We also evaluated the transglycosylation activity of the Endo-S enzyme on the two disaccharide oxazoline substrates (24 and 27). Our results showed that Endo-S exhibited significantly less activity than Endo-S2. In comparison, under the same transglycosylation conditions, i.e., 30 mg/mL antibody concentration in 1× PB (pH 7.0), 20 equiv/site oxazoline, and 1/200 (w/w) enzyme usage, Endo-S gave 35% and 70% tranglycosylation yield after 3 h for LacNAz-ox (27) and Glcβ1,4-GlcNAz-ox (24), respectively, while Endo-S2 gave >95% yield (Figure S2).

3.

Evaluation of the three GlcNAz-derived disaccharide oxazolines (4, 24, and 27) at the second sugar moiety. (A) Schematic presentation of the Endo-S2 catalyzed reactions with the three disaccharide oxazolines (Manβ1,4GlcNAz oxazoline (4), Glcβ1,4GlcNAz oxazoline (24), and Galβ1,4GlcNAz oxazoline (27)); (B) side-by-side comparison of the enzymatic transglycosylation efficiency. Reaction conditions: (a) trastuzumab, 500 μg (30 mg/mL); oxazoline, 20 equiv/site; Endo-S2 WT, 1/200 (w/w); 100 mM phosphate buffer, pH 7.0, 30 °C. (b) trastuzumab, 500 μg (30 mg/mL); oxazoline, 20 equiv/site; Endo-S2 WT, 1/100 (w/w); 100 mM phosphate buffer, pH 7.0, 30 °C.

Synthesis and Evaluation of Disaccharide Oxazolines with Additional Substituents at the Methyl Group of the Oxazoline Ring

To further investigate to what extent Endo-S2 can tolerate modifications at the oxazoline methyl group for enzymatic recognition, we selected lactosamine, a readily available disaccharide, as the starting material to synthesize various LacNAc oxazoline derivatives with different substitutions at the methyl group (Scheme ). Amidation of lactosamine with free acids or their activated form, either the acid-NHS ester or the acyl chloride, yielded LacNAc derivatives 31a–31h. Then, one-step oxazoline formation by treatment of 31a–31h with 2-chloro-1,3-dimethylimidazolinium chloride (DMC) in the presence of TEA gave the corresponding sugar oxazolines (32a–32h).

3. Synthesis of LacNAc Oxazoline Derivatives with Modifications at the N-Acetyl Group.

Enzymatic evaluation of the derivatives indicated that both the size and the nature of the substituents at the methyl group affected the enzymatic recognition and substrate activity. For example, substitution with an azide, fluorine, or additional methyl group (30, 33b, and 33e) did not significantly affect the substrate activity (Scheme ). However, substitution at the oxazoline methyl group with a chlorine (33c) or hydroxyl group (33d) decreased the substrate activity. Interestingly, although Endo-S2 could tolerate modification on the methyl group by adding an azido group or one methylene group (27/33e vs 32a), compounds carrying an azido group through a longer linker (32g and 32h) were found to be poor substrates for Endo-S2 (<5%, even with a large amount of enzyme and a prolonged reaction time). These results suggest that Endo-S2 is sensitive to the size of the moiety at the methyl group on the oxazoline ring, but that azide and fluorine are well-tolerated, as the resulting fluorinated and azide-substituted disaccharide oxazolines acted as an excellent substrate for enzymatic transfer, making it an attractive method to introduce an azide and fluorine group site-specifically into the antibody product, respectively.

4. One-Pot IgG Glycoengineering with Oxazoline Modified LacNAc Derivatives .

^a Transglycosylation conditions: trastuzumab, 500 μg; oxazoline, 20 equiv/site; Endo-S2 WT (w/w), 0.5% or 2%; 100 mM phosphate buffer, pH 7.0, 25 μL; 25 °C, 3 h.

Structural Insights into the Recognition of GlcNAz Oxazolines by Endo-S2

Structure–activity relationship studies demonstrated that the mannose-containing disaccharide oxazoline had the highest activity for enzymatic transfer, followed by the glucose-containing disaccharide oxazoline, with the galactose-containing disaccharide oxazoline being the least active among the three. In addition, our comparative evaluation indicated that Endo-S2 could accommodate certain substitutions, such as an azide or a fluorine group, on the methyl group of the oxazoline ring but could not tolerate larger substituents. Thus, several factors, including the size, electronegativity, and potential of hydrogen-bonding capacity of the substituents, can play an important role in affecting enzymatic recognition. To elucidate the structural basis of the difference in activity when mannose-, glucose-, or galactose-GlcNAz oxazolines were used as substrates, we modeled the potential interactions between the disaccharide substrates and the enzyme, Endo-S2. Because the second monosaccharide moieties of the disaccharide oxazolines are epimers that differ only in the configurations of the C2 and C4 hydroxyl groups, we focused on interactions involving O2 and O4. In the crystal structure of Endo-S2 with a complex-type glycan, the axial O2 of Man(−2) hydrogen bonds to the side chains of E288 and Y339, while the equatorial O4 forms hydrogen bonds with the D108 and H109 side chains (Figure A). We observed the same interactions in our model of the mannose-GlcNAz disaccharide bound by Endo-S2 (Figure B). In our model of the glucose-GlcNAz–Endo-S2 complex, while the equatorial O4 remains in a similar position to interact with residues D108 and H109, the now equatorial O2 is at too great a distance to interact with Y339, while interactions with E288 remain. The equatorial O2 is positioned closer to N295, but the modeled distance suggests that this is a potentially weak interaction (Figure C). The loss of the hydrogen bond to Y339 compensated only by what appears to be a weak polar interaction with N295, as compared to that of mannose, could explain, at least in part, why the activity is reduced when mannose is replaced with a glucose moiety. Finally, the O2 of galactose is equatorial, while the O4 is axial. Thus, the weaker interactions involving O2 are similar to those of glucose. Additionally, the now axial O4 is likely unable to interact with D108 and is much further from the side chain of H109 (Figure D). This suggests that the interactions between Endo-S2 residues and galactose-GlcNAz are weaker than those between both the mannose and glucose counterparts, reflecting the decrease in activity. Although our method of alignment and modeling does not account for any conformation changes that may be induced in either the glycan substrate or enzyme due to the presence of the oxazoline, azide, or reduced sugar size as compared to the reported crystal structure, the suggested interactions support the observed differences in the activity of different disaccharide oxazoline substrates in the enzymatic Fc glycan remodeling.

4.

(A) Interactions between Endo-S2 residues and the O2 and O4 of Man(−2) according to a crystal structure (PDB ID: 6MDS). (B–D) Interactions between Endo-S2 residues and the O2 and O4 of mannose (B), glucose (C), or galactose (D) GlcNAz disaccharide oxazolines.

To elucidate the structural basis of the difference in activity with different substitutions on the methyl group, we also modeled the potential interactions between the methyl-substituted substrates and the enzyme. The modeling analysis demonstrated that N₃, F, Cl, and OH modifications exhibit similar fitting and binding contacts within the enzyme’s active site, as seen in Figure , where N₃, F, Cl, and OH groups align well with the enzyme’s structural architecture. However, the key difference in their activity may be attributed to the electronegativity of the substituent groups and their effect on the local electrostatic environment within the pocket (Figure S3). The fluorine and chlorine atoms, being highly electronegative, may influence the binding dynamics differently compared with the less electronegative hydroxyl group, potentially altering enzyme–substrate interactions and catalytic efficiency. On the other hand, modifications with larger groups, such as (CH₂)₃N₃, and (OCH₂CH₂)₂N₃, are unable to be accommodated within the glycan-binding pocket (Figure S4). These larger substituents likely introduce steric clashes that prevent optimal binding, suggesting that the enzyme’s active site has a limited capacity for larger substituent groups. Thus, the enzyme’s spatial constraints and the electrostatic interactions between the functional groups and the binding pocket appear to dictate the observed differences in activity and the accommodation of substituent sizes.

5.

Interactions between Endo-S2 residues and the methyl modifications of the GlcNAc oxazolines. Interactions between Endo-S2 residues and N₃ modification (A), F modification (B), Cl modification (C), or OH modification (D).

Synthesis and Cancer Cell Killing Activity of Antibody–Drug Conjugates Using the Fc Glycan Remodeling Method

We next evaluated the preparation of ADCs through click chemistry using the azido-tagged antibodies obtained (29 and 30). The conjugation between 29/30 and DBCO-PEG5-VC-PAB-MMAE was carried out in 30% DMSO at r.t. with a final concentration of the antibody at 2 mg/mL and 20 mol equiv of the payload per click handle being used (Scheme ). The reaction was monitored by LC-ESI-MS (Figure ), which indicated the completion of conjugation within 6 h to give ADCs 34 and 35.

5. Synthesis of Homogeneous ADCs through Click Reaction.

6.

LC-ESI-MS analysis of the antibody–drug conjugates (34, 35, 36, and 37). Deconvoluted mass of Fc domains was released by IdeS treatment. Asterisked peaks indicate the ion fragments derived from the Fc domain.

To generate an ADC with increased drug loading, we clicked branched payload (DBCO-PEG5-Bis-VC-PAB-MMAE (S5), Scheme S1 for synthesis of S5) to the azido-tagged antibodies (30 and 29) to obtain ADCs 36 and 37 with DAR 4.

Further characterization of the four ADCs was performed by using size exclusion chromatography (SEC) and hydrophobic interaction chromatography (HIC). The SEC analysis, performed with agarose-dextran particle-based columns, revealed that all four ADCs exhibited less than 2% aggregate content in PBS buffer (Figure S5A). HIC analysis showed the high homogeneity of these ADCs (Figure S5B).

We evaluated the in vitro cytotoxicity of the obtained ADCs with SK-BR-3 (high HER-2 expressing) and T-47D (low HER-2 expressing) cells. For the SK-BR-3 cell line, our novel ADCs 34 and 35 demonstrated dose-dependent killing of the SK-BR-3 cell (Figure A) with IC₅₀ values of 17.2 ng/mL and 17.0 ng/mL, respectively, which is comparable with ADCs obtained using glycan oxazolines carrying azido groups in the nonreducing terminus of the glycan (13.9–25.6 ng/mL). , The in vitro cytotoxicity of the obtained ADCs 36 and 37 with SK-BR-3 and T-47D cells was also performed. For the SK-BR-3 cell line, 36 and 37 demonstrated a dose-dependent killing (Figure B) with IC₅₀ at 7.7 and 7.5 ng/mL, respectively, showing better cell killing than the ADCs 34 and 35 with lower drug loading (DAR 2).

7.

Cytotoxicity assays of the antibody–drug conjugates with the SK-BR-3 (HER2 overexpression) and the T-47D (HER2 low expression) cancer cell lines. All assays were performed in triplicate.

Conclusions

We developed a facile and efficient chemoenzymatic strategy for site-selective antibody modification and conjugation using N-azidoacetylglucosamine (GlcNAz)-derived sugar oxazolines as enzymatic substrates. Comprehensive structure–activity relationship (SAR) studies reveal that Endo-S2 tolerates specific modifications at the oxazoline ring methyl position, including azide substitution and fluorination, without compromising enzymatic activity. Molecular modeling suggests that a pre-existing cavity within the enzyme active site accommodates these modifications, providing a structural basis for the observed substrate tolerance. The GlcNAz-derived disaccharide substrates are simpler and much easier to synthesize than the previously reported azide-disaccharide substrates. Thus, the discovery of the GlcNAz-derived disaccharide oxazolines as a class of new enzyme substrates enables simpler and more versatile approaches for constructing site-specific antibody–drug conjugates for further therapeutic discovery and development.

Experimental Section

General Procedures

Chemicals, reagents, and solvents were purchased from Sigma-Aldrich and/or TCI and used as received, unless otherwise specified. Monoclonal antibody trastuzumab was purchased from RefDrug Inc. (Hillsborough, NJ). Payload MMAE was purchased from MedChemExpress (Monmouth Junction, NJ). Thin-layer chromatography (TLC) was performed on silica gel 60-F254 on glass plates (Merck) and stained with p-anisaldehyde. Flash chromatography was performed on an Isolera One system with ZIP KP-Sil columns (Biotage). NMR spectra were recorded on a 600 MHz spectrometer (Bruker, Tokyo, Japan) with CDCl₃ or D₂O as the solvent. MALDI-TOF analysis was performed using a Bruker UltrafleXtreme (UTX) mass spectrometer in positive reflectron mode with DHB (ACN/H₂O 1/1) as the matrix. Analytical reverse-phase HPLC was performed on a Waters Alliance e2695 HPLC system equipped with a dual absorbance 2489 UV/vis detector. Separations were performed by using a C18 column (YMCTriart C18, 4.6 × 250 mm, 5 μm). ESI-MS spectra were obtained by using a Waters SQ Detector 2 single quadrupole mass spectrometer. Preparative RP-HPLC was performed on a Waters 600 HPLC system equipped with a dual absorbance UV detector using a C18 column (Waters-Symmetry Prep C18, 19 × 300 mm, 7 μm) at a flow-rate of 10 mL/min or a C18 column (Waters XBridge, Prep Shield, 10 × 250 mm, 5 μm) at a flow-rate of 4 mL/min. LC-ESI-MS analysis was performed on an Ultimate 3000 HPLC system coupled to an Exactive Plus Orbitrap mass spectrometer (Thermo Fischer Scientific) with a C4 (whole antibody, gradient, 5–95% aq MeCN containing 0.1% FA for 6 min, 0.4 mL/min) or C8 (IdeS digestion, gradient, 25–35% aq MeCN containing 0.1% FA for 6 min, 0.4 mL/min) column. Deconvolution data were transformed with the MagTran software.

Synthesis of Glcβ1,4-GlcNAz Oxazoline (24)

A mixture of acceptor 16 (1.0 g, 2.01 mmol), donor 20 (1.24 g, 2.52 mmol), and 4 Å molecular sieves (2 g) in anhydrous CH₂Cl₂ (30 mL) was stirred under a N₂ atmosphere for 30 min and then cooled to −10 °C. TMSOTf (38 μL, 0.21 mmol) was added to the mixture. The reaction mixture was stirred at −10 °C for 30 min, then warmed to rt, and further stirred for 30 min, at the end of which time TLC (silica, 1:2 EtOAc–hexanes) showed it was complete. The reaction mixture was quenched with Et₃N (30 μL) and filtered. The filtrate was evaporated in vacuo to give a residue, which was purified by silica gel column chromatography (silica, EtOAc–hexanes, 1:5 → 1:2) to give compound 21 (1.44 g, 85% yield) as a colorless syrup. MeONa (7 mg, 0.13 mmol) was added in one portion to a solution of 21 (1.00 g, 1.24 mmol) in MeOH (15 mL) at room temperature. The reaction mixture was stirred at room temperature overnight. TLC (silica, 3:1 EtOAc–hexanes) showed that the reaction was complete. Then acidic ion-exchange resin (Amberlite IR-120 (H+), Alfa Aesar) was added to neutralize the reaction mixture. The mixture was filtered. To the filtrate was added Pd/C (15 mg) in one portion. The reaction mixture was then stirred under an atmosphere of hydrogen (1 atm) at room temperature for 4 h. The reaction mixture was filtered through a pad of Celite using MeOH (5 mL) as the eluent, and the resulting filtrate was concentrated to a volume of 5 mL. Et₂O (50 mL) was added to the solution to precipitate out the crude product, which was further dissolved in 5 mL of MeOH and precipitated again by adding Et₂O (50 mL). The precipitate was dried to give compound 22 (0.42 g, 90% over two steps) as a white powder. NaHCO₃ (1 M aq. solution, 0.33 mL, 0.33 mmol) was added to a cold solution (0 °C) of Glcβ1,4-glucosamine·HCl (22, 50 mg, 0.13 mmol) in 1 mL of water. The reaction mixture was stirred at 0 °C for 5 min before azidoacetic acid NHS ester (32 mg, 0.16 mmol) in 0.5 mL of MeCN was added. The resulting mixture was further stirred for 2 h until MALDI analysis showed completion. The reaction mixture was purified by RP-HPLC to give Glcβ1,4-GlcNAz (23, 46 mg, 82%) as a white powder after lyophilization. To a solution of Glcβ1,4-GlcNAz (12 mg, 42 μmol) in water (1.2 mL) was added triethylamine (39 μL, 0.28 mmol) and 2-chloro-1,3- dimethylimidazolinium chloride (DMC) (24 mg, 0.14 mmol) at 0 °C. The reaction mixture was stirred on an ice bath for 2 h. The glycan oxazoline product was purified by gel filtration on a Bio-Gel P-2 column, eluting with 0.1% triethylamine. The carbohydrate containing fractions were pooled and lyophilized to give Glcβ1,4-GlcNAz oxazoline (24, 9.9 mg, 86%) as a white powder. ¹H NMR (600 MHz, D₂O) δ 6.24 (d, J = 7.3 Hz, 1H, H-1), 4.52 (d, J = 7.9 Hz, 1H, H-1’), 4.49 (dd, J = 3.0, 1.9 Hz, 1H, H-3), 4.35–4.30 (m, 1H, H-2), 4.23 (ddd, J = 50.9, 16.5, 1.8 Hz, 2H, N₃–CH ₂), 3.95 (dd, J = 12.4, 2.2 Hz, 1H, H-6’), 3.85 (dd, J = 12.4, 2.5 Hz, 1H, H-6), 3.80–3.74 (m, 2H, H-6’, H-4), 3.72 (dd, J = 12.4, 6.3 Hz, 1H, H-6), 3.55–3.46 (m, 3H, H-3′, H-5, H-5′), 3.45–3.39 (m, 1H, H-4’), 3.31 (dd, J = 9.5, 8.0 Hz, 1H, H-2’). ¹³C NMR (150 MHz, D₂O) δ: 165.0, 103.6, 100.4, 77.7, 75.4, 75.1, 72.6, 70.7, 69.0, 68.3, 64.5, 61.2, 60.1, 45.6. HRMS (ESI-Orbitrap) m/z [M + H]⁺ Calcd for C₁₄H₂₃N₄O₁₀ ⁺: 407.1409, found: 407.1378.

Synthesis of LacNAz Oxazoline (27)

NaHCO₃ (1 M aq. solution, 0.33 mL, 0.33 mmol) was added to a cold solution (0 °C) of lactosamine·HCl (25, 50 mg, 0.13 mmol) in 1 mL water. The reaction mixture was stirred at 0 °C for 5 min before azidoacetic acid NHS ester (32 mg, 0.16 mmol) in 0.5 mL of MeCN was added. The resulting mixture was further stirred for 2 h until MALDI analysis showed completion. The reaction mixture was purified by RP-HPLC to give LacNAz (26, 42 mg, 75%) as a white powder after lyophilization. To a solution of LacNAz (18 mg, 42 μmol) in water (1.8 mL) were added triethylamine (59 μL, 0.42 mmol) and 2-chloro-1,3- dimethylimidazolinium chloride (DMC) (36 mg, 0.21 mmol) at 0 °C. The reaction mixture was stirred on an ice bath for 2 h. The glycan oxazoline product was purified by gel filtration on a Bio-Gel P-2 column eluting with 0.1% triethylamine. The carbohydrate containing fractions were pooled and lyophilized to give LacNAz oxazoline (27, 14.5 mg, 84%) as a white powder. ¹H NMR (600 MHz, D₂O) δ 6.17 (d, J = 7.3 Hz, 1H, H-1), 4.44 (dd, J = 3.1, 1.9 Hz, 1H, H-3), 4.39 (d, J = 7.8 Hz, 1H, H-1’), 4.28–4.24 (m, 1H, H-2), 4.17 (ddd, J = 42.0, 16.5, 1.8 Hz, 2H, N₃–CH ₂), 3.87 (d, J = 3.0 Hz, 1H, H-4’), 3.83–3.62 (m, 6H, H-4, H-5′, H-6, H-6’), 3.59 (dd, J = 10.0, 3.4 Hz, 1H, H-3′), 3.51–3.42 (m, 2H, H-2’, H-5). ¹³C NMR (150 MHz, D₂O) δ 165.1 (N = C), 104.2­(C-1’), 100.4 (C-1), 77.5 (C-4), 74.7 (C-5′), 72.0 (C-3′), 70.8 (C-5), 70.3 (C-2’), 68.5 (C-3), 68.1 (C-4’), 64.4 (C-2), 61.1, 60.6 (C-6, 6’), 45.6 (N₃-CH₂). HRMS (ESI-Orbitrap) m/z [M + H]⁺ Calcd for C₁₄H₂₃N₄O₁₀ ⁺: 407.1409, found: 407.1382.

One-Step Functionalization of Wild-Type Trastuzumab with Glcβ1,4-GlcNAz Oxazoline (29)

To a solution of wild-type trastuzumab (2.0 mg, 25 mg/mL final concentration) and Glcβ1,4-GlcNAz oxazoline (24, 0.22 mg, 20 equiv per reaction site) in 100 mM phosphate buffer (pH = 7.0) was added wild-type Endo-S2 (10 μg, 0.5% of the antibody, w/w). The mixture was incubated at 30 °C. LC-ESI-MS monitoring indicated complete transglycosylation within 1 h. The functionalized trastuzumab was purified by protein A affinity chromatography to give 29 (1.78 mg as measured using Nanodrop). LC-ESI-MS: calcd for the whole antibody, M = 146,678 Da; found (m/z), 146,678 (deconvolution data); after IdeS digestion, LC-ESI-MS: calcd for the Fc part, M = 24,543 Da; found (m/z), 24,542 (deconvolution data).

One-Step Functionalization of Wild-Type Trastuzumab with LacNAz Oxazoline (30)

To a solution of wild-type Trastuzumab (2.0 mg, 25 mg/mL final concentration) and LacNAz oxazoline (27, 0.22 mg, 20 equiv per reaction site) in 100 mM phosphate buffer (pH = 7.0) was added wild-type Endo-S2 (20 μg, 1% of the antibody, w/w). The mixture was incubated at 30 °C. LC-ESI-MS monitoring indicated complete glycosylation within 2 h. The functionalized Trastuzumab were purified by protein A affinity chromatography to give 30 (1.70 mg as measured using Nanodrop). LC-ESI-MS: calcd for the whole antibody Trastuzumab-GNF-LacNAz (30), M = 146 678 Da; found (m/z), 146,677 (deconvolution data); after IdeS digestion, LC-ESI-MS: calcd for the Fc part, M = 24 543 Da; found (m/z), 24,541 (deconvolution data).

Preparation of Antibody–Drug Conjugates 34 and 35 with DAR 2

A solution of azide-tagged antibody 30 or 29 (1 mg, 6.8 nmol) and the DBCO-PEG5-VC-PAB-MMAE (230 μg, 20 equiv) in a phosphate buffer (50 mM, pH 7.2) containing 30% dimethyl sulfoxide (DMSO) (final volume, 0.5 mL) was incubated at room temperature. The reaction mixture was shielded from light and gently vortexed. The reaction was monitored by LC-ESI-MS analysis. After 6 h, the click reaction was complete as indicated by LC-ESI-MS. The mixture was then diluted with phosphate buffer (5 mL, 50 mM, pH 7.2) and filtered using a 0.22 μm syringe filter to remove most of the unreacted hydrophobic payload. The filtrate was purified by protein A chromatography to give ADC 34 (0.84 mg, 82%) and 35 (0.87 mg, 85%). LC-ESI-MS: calcd for the whole ADC 34, M = 150,082 Da; found (m/z), 150,083 (deconvolution data); after IdeS digestion, LC-ESI-MS: calcd for the Fc part, M = 26,244 Da; found (m/z), 26,243 (deconvolution data). LC-ESI-MS: calculated for the whole ADC 35, M = 150,082 Da; found (m/z), 150,085 (deconvolution data); after IdeS digestion, LC-ESI-MS: calcd for the Fc part, M = 26,244 Da; found (m/z), 26,244 (deconvolution data).

Preparation of Antibody–Drug Conjugates 36 and 37 with DAR 4

A solution of azide-tagged antibody 30 or 29 (1 mg, 6.8 nmol) and the DBCO-PEG5-bis-VC-PAB-MMAE (S5, 430 μg, 20 equiv) in a phosphate buffer (50 mM, pH 7.2) containing 50% dimethyl sulfoxide (DMSO) (final volume, 1 mL) was incubated at room temperature. The reaction mixture was shielded from light and gently vortexed. The reaction was monitored by LC-ESI-MS analysis. After 18 h, the click reaction was complete as indicated by LC-ESI-MS. The mixture was then diluted with phosphate buffer (5 mL, 50 mM, pH 7.2) and filtered using a 0.22 μm syringe filter to remove most of the unreacted hydrophobic payload. The filtrate was purified by protein A chromatography to give ADC 36 (0.73 mg, 70%) and 37 (0.79 mg, 76%). LC-ESI-MS: calcd for the whole ADC 36, M = 153,108 Da; found (m/z), 153,109 (deconvolution data); after IdeS digestion, LC-ESI-MS: calcd for the Fc part, M = 27,757 Da; found (m/z), 27,757 (deconvolution data). LC-ESI-MS: calculated for the whole ADC 37, M = 153,108 Da; found (m/z), 153,110 (deconvolution data); after IdeS digestion, LC-ESI-MS: calcd for the Fc part, M = 27,757 Da; found (m/z), 27,757 (deconvolution data).

Cell Lines and Culture Conditions

SK-BR-3 cells (ATCC) were maintained in McCoy’s 5a Medium (ATCC) containing 10% fetal bovine serum (FBS, not heated), 100 U/mL penicillin, and 100 μg/mL streptomycin in T-75 flasks (CELLTREAT). T-47D cells (ATCC) were maintained in RPMI-1640 Medium (ATCC) containing 10% fetal bovine serum (FBS, preheated), 4 mg/L insulin, 100 U/mL penicillin, and 100 μg/mL streptomycin in T-75 flasks (CELLTREAT).

Cytotoxicity Assay

SK-BR-3 and T-47D (ATCC) cells were planted into 96-well plates (cell number: 10,000cells per well), and the plates were incubated for 24 h at 37 °C with 5% CO₂. The ADC samples were diluted by 3-fold serial dilution with the corresponding medium from 5000 to 0.085 ng/mL (11 concentrations) and then added to the wells in triplicate (150 μL per well) for every single concentration. The cells were cultured at 37 °C with 5% CO₂ for 3 days before the removal of the medium and addition of Cell Counting Kit-8 (Sigma). The absorbance of formazan released by viable cells was measured at 450 nm using a spectrophotometer after incubation at 37 °C with 5% CO₂ for 2–3 h, and the background absorption was deducted by 550 nm absorbance. Finally, the cell viability curve and IC₅₀ values were calculated using GraphPad Prism software.

Molecular Modeling

We constructed Man-GlcNAz, Glc-GlcNAz, and Gal-GlcNAz oxazolines in PyMOL (Schrodinger, LLC. The PyMOL Molecular Graphics System, Version 1.8. 2015) and aligned them with the GlcNAc(−1) and Man(−2) of the crystal structure of Endo-S2 bound by a complex-type glycan (PDB ID: 6MDS).

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.5c01772.

• Figures S1–S5; Scheme S1; general procedures for the synthesis of disaccharide substrates and enzymatic antibody Fc glycan remodeling; LC-MS analysis of antibody conjugates; ¹H and ¹³C NMR spectra (PDF)

The authors declare the following competing financial interest(s): L.X.W and G.Z. are inventors on a provisional patent application filed with the United States Patent and Trademark Office by University of Maryland College Park relevant to the work in this manuscript. All other authors have no competing interests.