Automated Glycan Assembly of 19F‐labeled Glycan Probes Enables High‐Throughput NMR Studies of Protein–Glycan Interactions

Abstract Protein–glycan interactions mediate important biological processes, including pathogen host invasion and cellular communication. Herein, we showcase an expedite approach that integrates automated glycan assembly (AGA) of 19F‐labeled probes and high‐throughput NMR methods, enabling the study of protein–glycan interactions. Synthetic Lewis type 2 antigens were screened against seven glycan binding proteins (GBPs), including DC‐SIGN and BambL, respectively involved in HIV‐1 and lung infections in immunocompromised patients, confirming the preference for fucosylated glycans (Lex, H type 2, Ley). Previously unknown glycan–lectin weak interactions were detected, and thermodynamic data were obtained. Enzymatic reactions were monitored in real‐time, delivering kinetic parameters. These results demonstrate the utility of AGA combined with 19F NMR for the discovery and characterization of glycan–protein interactions, opening up new perspectives for 19F‐labeled complex glycans.


Introduction
Glycans are ah ighly diverse class of biomolecules involved in several processes such as cellular communication and recognition and play important structural and modulatory roles. [1]Pathogens invade the host by mimicking or exploiting host glycans present on endothelial cells.T his process is often mediated by lectins,aclass of glycan-binding proteins (GBPs) expressed by both pathogens and hosts.Ty pically,m ammalian glycans have low affinity for mammalian receptors,w hile showing higher affinity for bacterial lectins. [2]Profiling glycan-lectin interactions is ac rucial step towards the understanding of the biological functions of glycans.Still, the extreme complexity and diversity of glycans pose as evere bottleneck to the characterization of these generally weak and promiscuous interactions.
Synthetic glycans are valuable probes to dissect glycanprotein interactions.H owever, lengthy synthetic protocols hampered their systematic and widespread use in glycobiology.A utomated glycan assembly (AGA) enables fast access to complex and well-defined glycans. [3,4] ith AGA, glycans are typically assembled in an overnight run, permitting the production of broad collections of glycans for systematic screenings. [5][8] Most of these strategies rely on immobilized glycans (e.g.m icroarray technology) [6][7][8][9][10] or require large amounts of samples and analysis time (ITC, [11] SPR, [12] or X-ray crystallography [13] ).In contrast, NMR allows for the detection of protein-glycan interactions in solution in af ast and reliable manner,p roviding information on the binding mode in ah omogeneous assay format in absence of immobilization protocols. [14,15]MR active labels are commonly introduced to simplify NMR analysis.[15,17] Among all, the 19 Fnucleus stands out due to its unique properties such as:i )high sensitivity to local chemical environment, ii)short acquisition times,i ii)simple spectra, iv) broad chemical shift range,a nd v) absence in biological systems (no background signal).[18,19] Even though 19 FNMR has enabled the description of peptide (mis)folding, real-time in vivo events, [18][19][20][21][22] protein-ligand interactions,and high-throughput ligand screening, [23,24] the use of fluorinated glycans to investigate protein binding [25] and enzymatic reactions [26][27][28] is just at the beginning.T he labor-intensive multistep synthesis of 19 F-labeled glycans represents the main bottleneck and has limited these studies to small collections of short and relatively simple glycans.[14,[29][30][31][32] Still, 19 F-labeled glycans have the potential to dissect protein-glycan interactions. [33,34] erein, we present ah igh-throughput NMR-based approach for the screening and characterization of proteinglycan interactions using 19 F-labeled glycans.A GA enabled quick access to ac ollection of 19 F-labeled Lewis type 2 complex glycans.Lewis type 2antigens are involved in several physiological and pathological processes,i ncluding cancer, where they act as cell adhesion or recognition mediators.[35,36] Subtle differences in the fucosylation pattern strongly impact their interaction with proteins and ultimately can lead to host immune system elusion.[37][38][39][40] The 19 F-labeled glycan probes (hereafter F-glycans) were screened against mammalian and bacterial lectins as well as enzymes.A mong mammalian lectins,weselected Langerin [41] and the dendritic cell specific ICAM-3 grabbing non-integrin (DC-SIGN) [42] both of which are known to bind high-mannose N-glycans.D C-SIGN also selectively recognizes specific fucosylated glycans, [43] playing acrucial role in the biology of viral pathogens (e.g.HIV).In addition, we screened soluble lectins produced by some opportunistic pathogens responsible for lung infections,s uch as Pseudomonas (LecA and LecB) [44] and Burkholderia (BambL) [45] species.F inally,w es elected two different sialyltransferases and screened their interactions with Lewis antigens,g iven the importance and widespread occurrence of terminal sialylation in Lewis antigens.[46,47] Thel abeled glycan probes in combination with 19 FNMR proved to be valuable for detecting binding events in real-time,identifying new weak protein-glycan interactions,a nd determining affinities (K d )aswell as kinetics of enzymatic reactions.

Automated Synthesis of F-Glycans
Recently,a ne legant procedure to access ac ollection of Lewis type 2a ntigens by AGAw as reported. [48]We envisioned as imilar approach to produce as et of 19 F-labeled analogs to screen protein binding in asimple 19 FNMR assay.Since the position of the 19 Freporter is thought to be crucial to obtain valuable information, F-glycans (F-Lac, F-nLac 4 , F-Le x , F-H type 2,a nd F-Le y )w ere designed with the 19 Freporter in the lactose inner core subunit (Figure 1A).This position is distal from the binding site (i.e.non reducing end) to minimize the effect of the fluorine atom during the binding event. [49,50]We hypothesize that labeling of the inner core glucose unit should maintain sensitivity to the binding event due to overall changes in the correlation time of the glycan in the bound state,r eporting changes in the 19 FNMR signal. [20] 19 -labeled analogs of Lewis type 2a ntigens were assembled on as olid support (functionalized Merrifield resin, L1) Figure 1.Integrated approach for the preparationo f 19 F-labeled Lewis type 2g lycans by AGA and screening against lectins and enzymes.A) BBs 1-5,i ncludingBB1 bearing the 19 Freporter,were employed for the AGA of acollection 19 F-labeled Lewis type 2antigen analogs represented following the Symbol Nomenclature ForGlycans (SNFG).[16] B) The F-glycansw ere screeneda gainst proteins, including mammalian and bacterial lectins, as well as enzymes.The enzymes were screened in the absence of donor (i.e.CMP-Neu5Ac) to probe binding to the substrate.The binding strength was defined depending on the changes observed in the NMR after addition of the protein (right panel).Strong binding (blue) is defined as adecrease in peak intensity higher than À25 %orachemical shift perturbation (CSP) bigger than 0.01 ppm in the 19 FNMR.Weak/ medium binding (light blue) is defined as ad ecrease in peak intensity higher than À25 %i nthe CPMG-filtered 19 FNMR.No binding (white) is defined as adecrease in peak intensity lower than À25 %i nC PMG-filtered 19 FNMR.
using building blocks (BBs) 1-5 (Figure 1A).TheB Bs are equipped with at hioether or ad ibutylphosphate reactive leaving group.O rthogonal cleavage of the 9-fluorenylmethoxycarbonyl (Fmoc) and levulinoyl (Lev) temporary protecting groups permits regioselective chain elongation.Benzyl (Bn), benzoyl (Bz), and N-trichloroacetyl (TCA) groups protect the remaining functionalities.b-Stereoselectivity during glycosylation with BBs 1-4 is ensured by anchimeric assistance of the protecting groups at C-2, while a-stereoselectivity with BB 5 was verified in previous studies. [48]BB 1 is labeled with the 19 Fr eporter at the C-3 position. [51]Each oligosaccharide was assembled in an overnight run following previously reported conditions for unlabeled analogs (see SI). [48] Post-AGA manipulations included solid-phase methanolysis, [51] photocleavage [52] from the solid support, and hydrogenolysis (see SI).As ingle final purification step afforded the target F-glycans in overall yields of 5% to 16 %o ver 7to1 5steps. 19

FNMR Screening of F-Glycan Library
A 19 Fand CPMG NMR screening was performed to probe the interactions of five F-glycans (F-Lac, F-n-Lac 4 , F-Le x , F-H type 2,a nd F-Le y )w ith mammalian (Langerin, DC-SIGN) and bacterial (LecA, LecB,B ambL) lectins and enzymes (a(2,3)-sialyltransferase from Pasteurella multocida (Pma23ST) [53] and a(2,6)-sialyltransferase from Photobacterium damsela (Pda26ST) [54] )( Figure 1B).Upon protein binding,t he molecular tumbling rate of the glycan is drastically affected resulting in ad ecrease of the 19 Fsignal intensity. [20]Monitoring 19 Fchemical shift perturbation (CSP) or change in peak intensity upon addition of protein allowed us to qualitatively evaluate the strength of the interaction.A decrease in peak intensity or aC SP in 19 FNMR indicates strong binding.Application of aCPMG-based spin echo filter allows us to detect weak binders.Asaresult, bacterial (LecA, LecB,and BambL) and mammalian (DC-SIGN ECD) lectins preferred fucosylated glycans (Figures S2A, S2B,S 2C,a nd S2E).No binding to F-glycans was observed in presence of Langerin ECD (Figure S2D), in agreement with previous reports. [55]In contrast, the enzymes showed much weaker interactions and aslight preference for shorter non-branched glycans (Figure S3).

Reporter Position on F-Glycans Does Not Affect Binding to Mammalian and Bacterial Lectins
DC-SIGN recognizes cellular ligands and pathogens that express Lewis antigens.Inparticular,Le x and Le y present on Schistosoma mansoni [56] and Helicobacter pylori [43] or endothelial cells, [57] respectively,a re known binding partners for DC-SIGN. [58]Thestrong preference of DC-SIGN for fucosylated ligands has also been elucidated with the crystal structure of the carbohydrate-binding site of DC-SIGN bound to Le x . [59]Theq ualitative CPMG NMR screening of mammalian lectins confirmed the interaction of DC-SIGN with fucosylated glycans F-Le x , F-H type 2,a nd F-Le y (Fig- ure 2A), as indicated by changes in the NMR peak intensity of the reporter molecule.T his effect is maximized with aprotein-to-ligand ratio of 2:1( Figure S4A).
First, we explored the role of the 19 Freporter in F-glycan binding to DC-SIGN.W ep erformed protein-observed 15 NHSQC NMR and recorded an HSQC NMR spectrum of DC-SIGN CRD in the presence of F-Le x and Le x .B oth ligands promoted similar changes in the backbone of DC-SIGN CRD (Figure 2B and S4B).Next, we investigated the effect of the reportersp osition on the ability to reveal binding events.W ec onjugated aC F 3 moiety to the remote end of the aminopentyl linker on Htype 2 (CF 3 -H type 2), far from the carbohydrate-binding site,and tested the new ligand in 19 Fa nd CPMG NMR.Remarkably,i ts binding was observed with both mammalian (DC-SIGN,F igure 2A)a nd bacterial lectins (BambL, Figure S5).These results indicate that the positioning of the 19 Fr eporter on the Glc unit does not affect the binding of F-glycans with proteins.F urthermore,the 19 Freporter can be remote to the glycan binding site to avoid any interference with the binding event, while preserving excellent sensitivity.H owever, the functionalization of the amino linker with aC F 3 moiety prevents any further conjugation of the glycan (e.g.t op rotein, surface, liposome).
We further investigated the interactions of DC-SIGN CRD with F-Le y and F-H type 2 in 15 NH SQC NMR (Figure 3A and S6A).Even though Le y is known for its interaction with DC-SIGN,s tructural data are lacking. [57]oth ligands promoted CSPs of the residues located in the carbohydrate-binding and remote sites of DC-SIGN CRD.Binding to F-Le y promoted larger changes in DC-SIGN CRD than F-H type 2 or the monosaccharide positive control dmannose (Figure 3B and S6B).This result proved that the avidity effect plays ac rucial role in the interactions between DC-SIGN and Lewis type 2a ntigens,a ss imilarly noted for high-mannose structures. [60][63] Cumulatively,w eb elieve these probes are valuable tools for the description of the interaction mechanisms between DC-SIGN and fucosylated blood antigens.

Binding Affinity of F-Glycans to Bacterial Lectins
Bacterial lectins show ar emarkably high affinity for fucosylated blood group antigens. [35,64] hei nteraction of BambL from Burkholderia ambifaria with Htype 2 has been thoroughly investigated and two binding sites were identified in acrystal structure of the complex (Figure 4A). [35]We set on to verify this interaction for F-glycans in 19 Fa nd proteinobserved NMR.
First, we performed 19 FNMR screening and titration experiments with fucosylated F-glycans. 19FNMR experiments allowed us to confirm the interaction and obtain affinity constants for F-H type 2 (K d = 9 AE 2 mm,F igure 4B and 3C)a nd F-Le y (K d = 14 AE 2 mm,F igure S7A).Given that BambL has two binding sites available for glycan binding,we applied one-and two-binding site models to derive the affinities for both sites.Both models resulted in matching K d values,i na greement with values reported by ITC. [35]Even though we did not observe ad ifference in the affinities between the two sites in 19 FNMR, we showed that 19 FNMR can be applied reliably to derive affinities while considerably reducing the amount of ligand needed for ITC.
We verified the interaction of F-H type 2 (Figure 4D)and F-Le y (Figure S7B) with 15 N-labeled BambL in proteinobserved 15 NTROSY NMR.Changes in protein backbone similar to the one obtained with a-Me-l-fucose indicate that the a-l-fucose branch was mainly responsible for the binding (Figures S7C and 4D).To derive affinities,w et itrated both ligands and followed the changes in peak intensities and CSPs for the peaks in slow (F-H type 2: K d = 12 AE 8 mm,F igure 4F and F-Le y : K d = 17 AE 3 mm,Figure S7D), and fast (F-H type 2: K d = 94 AE 33 mm,F igure 4G and F-Le y : K d = 245 AE 29 mm, Figure S7E) exchange regimes,r espectively.H owever,p rotein-observed NMR is not well suitable for the determination of K d for ligands with high affinities and thus,ithampered the accurate derivation of the K d . [65]This underscores the advantage of the 19 FNMR ligand-observed approach.
In addition to the known strong interactions of LecB and BambL with fucosylated glycans, [66] CPMG NMR screening revealed weak interactions between LecA and fucosylated Fglycans.T oc onfirm this observation, we performed 19 FR 2filtered, protein-observed 19 F( PrOF) and 15 NTROSY NMR experiments.F-H type 2 showed af aster relaxation in presence of protein, indicating aweak interaction with LecA (Figure S8B).Protein-observed NMR experiments with 5fluorotryptophan (5FW,F igure S8C) and 15 N-labeled LecA (Figure S8D and S8E) confirmed that this interaction takes place in the canonical carbohydrate-binding site of LecA, as indicated by perturbation of W42 and CSPs promoted in as imilar manner to d-galactose,r espectively.T ot he best of our knowledge this is the first report of such weak binding detected using ab iophysical method. [67,68] hese results demonstrate that F-glycans serve as probes for the affinity determination and discovery of new interactions using low amounts of protein and ligand.

Enzyme Binding and Real-Time Kinetics with F-Glycans
The 19 FNMR assay allowed us to monitor the binding of F-glycans (F-Lac and F-nLac 4 )t oe nzymes.T wo sialyltransferases (Pma23ST [53] and Pda26ST [54] )w ere screened in the absence of donor (i.e.C MP-Neu5Ac) and revealed weak binding to the glycan substrate (Figure 1B and S3).This is particularly relevant because binding sites of transferases usually have av ery low affinity for the acceptors,m aking these interactions difficult to detect.Shorter non-branched glycans (F-Lac and F-nLac 4 )s howed stronger binding than longer branched ones.F-Le x did not show any binding with Pma23ST or Pda26ST,matching its known poor reactivity as acceptor (Figure S3). [69]In contrast, Pda26ST showed weak binding to F-H type 2,inagreement with previously reported enzymatic activity (Figure S3). [54]This simple assay could be envisioned as screening platform to identify acceptor substrates for known enzymes and for the discovery of new glycosyltransferases. [70,71] Theh igh sensitivity of the 19 Fr eporter to subtle modifications in its chemical environment offers avaluable tool for real-time monitoring of enzymatic reactions.T he possibility to place the 19 Freporter on ac arbohydrate unit in proximity to the functionalization site is crucial for detecting achemical shift perturbation.We selected two enzymes (b-galactosidase [72] and Pma23ST [53] )a nd we monitored their activity on amodel substrate, F-Lac.Glycosidic bond cleavage,mediated by b-galactosidase,was followed by 19 FNMR.Cleavage of the terminal b-galactose induced ac hemical shift perturbation and real-time 19 FNMR tracking allowed for derivation of the K M of the enzymatic reaction (Figure 5A).Next, glycosidic bond formation promoted by Pma23ST [53] was monitored in real-time.N-Acetyl-neuraminic acid (Neu5Ac) is transferred from an activated cytidine monophosphate donor (CMP-Neu5Ac) to the C-3 OH of the terminal galactose unit of F-Lac to yield F-sLac.T he electron-withdrawing nature of Neu5Ac induced ac hemical shift perturbation of 0.2 ppm on the 19 F-labeled acceptor,allowing to track in real-time the enzymatic sialylation process (Figure 5B).When the 19 Fr eporter was positioned remotely to the reactive site of the acceptor (> 3sugar units away, F-nLac 4 ), no chemical shift perturbation was noticed, despite the success of the enzymatic transformation (Figure S10).Thus,i nc ontrast to what is observed for protein binding,t he position of the 19 Fr eporter is key for monitoring enzymatic reactions.

Conclusion
AGAe nabled the fast assembly of 19 F-labeled Lewis type 2a ntigens for the high-throughput screening of protein binding.Mammalian and bacterial lectins as well as enzymes were analyzed. 19FNMR screening of Fglycans permitted aq uick qualitative evaluation as well as ar eliable quantification of lectin binding (K d ).Thea ssay does not require labeled proteins or complex 2D NMR experiments.All NMR experiments can be performed in an extremely small scale (few nmol of glycan and protein per experiment).Enzymatic reactions,i ncluding sialylation, were monitored in real-time, demonstrating that 19 F-labeled glycans hold agreat potential as molecular probes to uncover enzymatic processes and for high-throughput screening. [27]Protocols for the selective 19 Flabeling of monosaccharides are available; [73][74][75] the implementation of these novel BBs in AGAwill fuel the production of new classes of glycan probes.Given the high dispersion of 19 FNMR signals,libraries of F-glycans with diverse chemical shifts can be designed to increase the high throughput of this approach. [76]Theability of 19 Fglycan probes to reveal binding or enzymatic transformation in solution and in real-time could open the way to in cell NMR applications,o ften hampered by high background signals. [14,77,78] OFNMR using F-glycans.A) 19 FNMR of F-Lac incubated with b-galactosidase. 19FNMR real-time tracking of product formation (black arrows) upon incubation of F-Lac with b-galactosidase (right).Kinetic data were derived plotting the product formation rate as afunctiono fthe substrate concentration.The best fit of the experimental data provides a K M value of 86.5 AE 10.5 mm according to the Henry-Michaelis-Menten equation( left).B) 19 FNMR of F-Lac incubated with Pma23ST in presence of CMP-Neu5Ac.The formation of F-sLac (black arrows) can be followed by 19 FNMR in real-time.Product formation was confirmed by HPLC (Figure S9).

Figure 2 .
Figure 2. Mammalian lectin (DC-SIGN) binding to F-glycansa nd study on the reporter position.A) CPMG NMR screening of F-glycansa lone (gray) and in presence of DC-SIGN ECD (blue).DC-SIGN ECD binds to F-Le x , F-H type 2,and F-Le y as shown by adecrease in peak intensity in presence of protein (orange lines, left panel).CPMG NMR spectra of CF 3 -H type 2 alone (gray) and in presence of DC-SIGN ECD (blue; right panel).B) Cartoon of assigned domains of DC-SIGN CRD (unassigned resonances in dashed line) and CSP plot of assigned resonances in presence of F-Le x and Le x showing that F-Le x -perturbed resonances similarlyt ounlabeled Le x .

Figure 3 .
Figure 3. Mammalian lectin (DC-SIGN) binding to F-Le y .A) HSQC NMR (left) shows the interaction of F-Le y with 15 N-labeled DC-SIGN CRD and the perturbed residues were mapped on astructure of DC-SIGN CRD (blue).Surface diagram of the crystal structure of DC-SIGN CRD (PDB:1sl4;r ight).F-Le y targets the carbohydratebinding site of DC-SIGN CRD based on changes in resonances (e.g.321Leu, 365Asn and 368Lys, gray).B) Cartoon of assigned domains of DC-SIGN CRD (unassigned resonances in dashed line) and CSP plot showing that F-Le y -perturbed resonances similarly to d-mannose (red, positive control).The magnitudeo fF-Le y -promoted CSPs is higher compared to d-mannose.CSPs exceedingthe threshold (dashed line at 0.005 ppm) and intensities decreasing by more than 50 %were used for mappingthe binding site of F-Le y on astructure of DC-SIGN CRD.

Figure 4 .
Figure 4. Bacterial lectin (BambL)b inding to F-glycans.A) Surface diagram of the crystal structure of BambL in complex with H-F type 2 (PDB:3 zzv).Sites 1a nd 2c orrespond to the carbohydrate-bindingsites within amonomer and between two monomers, respectively.B) 19 FNMR screening of F-glycansalone (gray) and in presence of BambL (blue).BambL binds F-Le x ,F-Le y ,and F-H type 2 strongly as shown by CSP in presence of protein (orange line).The 19 FNMR titration spectra shows F-H type 2 undergoingslow exchange on the chemical shift timescale upon increase of BambL concentration.C) The K d of F-H type 2 was calculated from the changes in peak intensity and fitted to one-and two-site models resulting in a K d of 9 AE 2 mm.D) TROSY NMR verified F-H type 2 binding to 15 N-labeled BambL.Given that BambL has two binding sites, peaks showing aslow (30, 7, and 33), intermediate and fast exchange (5, 17, and 62) on the chemical shift timescale have been observed upon titration of F-H type 2.One-site model for slow (E) and fast exchange (F) peaks was applied to derive the K d values of 12 AE 8 mm and 94 AE 33 mm,r espectively.G)CSP plot showing the resonances perturbed in presence of a-Me-l-fucose and F-H type 2.

Figure 5 .
Figure 5. Real-time enzyme kinetics by19 FNMR using F-glycans.A)19 FNMR of F-Lac incubated with b-galactosidase.19FNMR real-time tracking of product formation (black arrows) upon incubation of F-Lac with b-galactosidase (right).Kinetic data were derived plotting the product formation rate as afunctiono fthe substrate concentration.The best fit of the experimental data provides a K M value of 86.5 AE 10.5 mm according to the Henry-Michaelis-Menten equation( left).B)19 FNMR of F-Lac incubated with Pma23ST in presence of CMP-Neu5Ac.The formation of F-sLac (black arrows) can be followed by19 FNMR in real-time.Product formation was confirmed by HPLC (FigureS9).