Ligand‐Directed Chemistry on Glycoside Hydrolases – A Proof of Concept Study

Selective covalent labelling of enzymes using small molecule probes has advanced the scopes of protein profiling. The covalent bond formation to a specific target is the key step of activity‐based protein profiling (ABPP), a method which has become an indispensable tool for measuring enzyme activity in complex matrices. With respect to carbohydrate processing enzymes, strategies for ABPP so far involve labelling the active site of the enzyme, which results in permanent loss of activity. Here, we report in a proof of concept study the use of ligand‐directed chemistry (LDC) for labelling glycoside hydrolases near – but not in – the active site. During the labelling process, the competitive inhibitor is cleaved from the probe, departs the active site and the enzyme maintains its catalytic activity. To this end, we designed a building block synthetic concept for small molecule probes containing iminosugar‐based reversible inhibitors for labelling of two model β‐glucosidases. The results indicate that the LDC approach can be adaptable for covalent proximity labelling of glycoside hydrolases.


Introduction
[3][4] Thus, detecting which cell organelles hold active or inactive enzymes is vital for understanding and treating such diseases.In this regard, activity-based protein profiling (ABPP) has become an indispensable tool for the investigation of enzyme activity in complex matrices such as live cells and tissues.Providing information about the effective activity rather than abundance, ABPP reveals insights into enzyme-substrate interactionsinformation that is essential for diagnostic procedures as well as drug discovery. [5][11][12][13][14][15][16] In a number of approaches, Overkleeft and coworkers labelled different glycoside hydrolase families selectively.Besides others, they were successful in selective labelling human lysosomal acid β-glucocerebrosidase (GBA) in the presence of the other human retaining β-glucosidases (GBA3 and LPH). [17,18]These strategies for ABPP of glycoside hydrolases involve mechanism-based semi-or irreversible inhibitors forming a covalent bond in the active site of the target enzyme.Consequently, the activity of the labelled enzyme is permanently lost.
To maintain the activity of the enzyme after labelling an alternative approach has been reported by Hamachi and Tsukiji. [19]Ligand-directed chemistry (LDC) combines a reversible inhibitor and a cleavable electrophile for covalent labelling near -but not in -the active site.During the labelling process the reversible inhibitor is cleaved from the probe and, hence is able to depart from the active site thereby leaving the covalently modified -but still active -enzyme (see Figure 1).[21][22][23][24][25] Scenarios where the enzyme activity is needed after labelling are described in the literature [23] such as 19 F-NMR biosensing for in cell kinetic measurements, live-cell FRET imaging or turn-on fluorescent biosensors.
For LDC the small molecule probe consists of three modules, as shown in Figure 1.(A) A reversible inhibitor specific for the enzyme of interest is used as ligand.(B) For covalent labelling, the probe is equipped with a reactive cleavable site within the linker region between the ligand and (C) a reporter tag.The tag is used for detection of the labelled enzyme. [26]This probe design establishes protein selectivity through the ligand whereas the actual labelling process occurs in proximity to the active site.By release of the cleaved off reversible inhibitor the enzyme is covalently labelled and still maintains its catalytic activity.
We herein report the proof of concept for using LDC to label glycoside hydrolases.

Probe design and synthesis
Applying the LDC probe format, we designed and synthesized small molecule iminosugar-based probes.As models two commercially available enzymes, β-glucosidases from almonds (Prunus dulcis, PdGH1) and Thermotoga maritima (TmGH1) were chosen.
The modules for our probes are as follows (Figure1): (A) As ligand we use the iminosugar 1-deoxynojirimycin (DNJ), a known potent competitive inhibitor for the enzyme class under consideration. [27](B) As the cleavable reactive site, we first introduced a benzoate ester moiety (compounds 1 and 2, synthesis SI Scheme S1), similar to the phenyl esters introduced for protein labelling of marinopyrrole A. [28] Aromatic esters are convenient to handle and are relatively stable, which should minimize unwanted unselective labelling.We anticipated such esters to react with amines of lysine side chains which are located in close proximity after binding of the probe to the active site.However, no labelling was observed with probes 1 and 2 (SI Figure S6) indicating that a more reactive moiety as part B is needed.We therefore introduced a sulfonate ester as the electrophilic species since this has been shown previously to effect covalent labelling of histidines, tyrosines, glutamic and aspartic acids as well as cysteines. [19,29](C) To detect the labelled enzyme we use a two-step approach to avoid problems with solubility as well as possible steric hindrance between the bulky fluorophore and the target enzyme (Figure 1).To this end, the LDC probe was functionalized with a terminal azide that can undergo click chemistry with an alkyne-fluorophore in a second step, enabling visualization after the LDC labelling process.
The linker between the ligand and the cleavable electrophile determines the site selectivity of labelling. [19,30,31]By optimizing the linker length, the reactive moiety is positioned in close proximity to the amino acid residue to be labelled.This so-called proximity effect is vital for covalent labelling of the target enzyme.Therefore, we developed a building block synthetic concept which enables to tailor each module of the small molecule probe for the respective target enzyme (Scheme 1).To test the influence of the linker between the ligand and the cleavable electrophile, we synthesized probes of different linker lengths 1-6 (Figure 2).
Starting from 3-(chlorosulfonyl)benzoic acid 7, sulfonate ester 8 is formed by esterifying with 6-azidohexan-1-ol under standard conditions, installing the terminal reporter tag (C, Figure 1) in the first step of the synthesis.To insert the alkyl linker, classical mixed-anhydride conditions were employed, where different lengths of amino alcohols can be introduced.This allows for tailoring the distance between the ligand (A) and the cleavable electrophile (B).In this case, C-6 and C-11 amino Scheme 1. Synthesis of ligand-directed sulfonate ester (LDSE) probes 3-6. [32,33]  alcohols were used giving compounds 9 and 10, respectively.Dess-Martin oxidation followed by reductive amination of aldehydes 11 and 12 with 1-deoxynojirimycin (DNJ) 14 or 1,5dideoxy-1,5-imino-d-xylitol (DIX) 16 gave the desired probes 3-6 in yields of 40 to 60 %, respectively (Scheme 1).Incorporation of the iminosugar ligand in the last step of the synthesis enables using the same building block BÀ C for targeting different enzymes by exchanging the ligand (A).

Biologic Evaluation -Proof of Concept
Evaluation of sulfonate ester probes DNJ-C 6 À LDSE-N 3 3, DNJ-C 11 À LDSE-N 3 4 with the two model enzymes gave varying results, confirming the importance of the linker length.First, we assessed the initial inherent binding affinities of the probes prior to any covalent reaction by determining their inhibitory properties.All probes, benzoate esters 1 and 2 as well as sulfonate esters 3-6 proved to be competitive inhibitors as seen in the Lineweaver Burk plots (Table 1, details see SI, Fig S3-4).For DNJ-sulfonate ester probes 3 and 4 K i -values in the low to sub-micromolar range are seen with higher affinity towards TmGH1 compared to PdGH1.
For the labelling studies, the enzymes were incubated with respective LDC probes 3 and 4 for 20 hours at room temperature in a ratio of enzyme to probe 1 : 10, corresponding to 100 μM of the respective probe.Formation of a covalent bond should take place through displacement of the sulfonate, followed by introduction of a fluorophore via a click-chemistry protocol (Figure 1, details see experimental section).Viewing of SDS-PAGE gels under UV irradiation revealed fluorescent protein bands from the respective enzymes in the case of PdGH1 with DNJ-C 11 À LDSE-N 3 probe 4 and of TmGH1 with both probes 3 and 4 (Figure 3).The time-and concentration-dependence of labelling for each enzyme-probe combination were assessed by SDS PAGE analysis as shown in SI Figs S7-S14.In order to investigate the importance of the linker length, a concentration study for the labelling process has been performed.The most efficient labelling was seen with probe 4 and PdGH1, for which we observed bright fluorescent protein bands at concentrations as low as 50 μM and after incubating for 1 h, while 300 μM of probe 3 was required with an incubation time of 20 h (SI Figure S9-10, S13-S14).Somewhat similar results were seen for TmGH1 for which fluorescent bands were visible after incubating with 50 μM of probe 4 for 5 hours or with 50 μM of probe 3 after 20 h (SI Figure S7-S8, S11-S12).
To verify that the labelling process is based on the ligandenzyme interaction we performed a competitive labelling experiment.This was realised by pre-incubating the respective enzymes with conduritol B epoxide (CBE), [37,38] a known irreversible inhibitor, or isofagomine (IFG), [39,40] a more potent reversible inhibitor for glucoside hydrolases (K i -values: TmGH1 0.021 μM and PdGH1 0.41 μM), followed by incubation with 100 μM of the actual LDC probes (details see experimental  section).Pre-incubating TmGH1 with both IFG as well as CBE successfully prevents labelling of the enzyme with DNJ-C 6 À LDSE-N 3 probe 3 (Figure 4).These results verify that the labelling process is indeed largely dependent upon the ligand-enzyme interaction in the case of TmGH1 with DNJ-C 6 À LDSE-N 3 probe 3.However, preincubation of TmGH1 with CBE or IFG could not prevent labelling by DNJ-C 11 À LDSE-N 3 probe 4.This could be caused due to the longer linker of probe 4 compared to probe 3 and therefore reduced selectivity (SI Figure S18).
Finally, we tested the residual activity of the labelled enzyme.We incubated TmGH1 with the most potent candidates DNJ-LDSE probes 3 and 4 for 20 h, removed remaining small molecules by washing, and measured Michaelis-Menten param-eters for hydrolysis of p-nitrophenyl β-d-glucopyranoside.As shown in Figure 5 and Table 2, we observed significant activity of the enzyme (90 % and 70 % remaining activity in the best and worst case respectively) after labelling with similar kinetic parameters.These results highlight the applicability of liganddirected chemistry for labelling carbohydrate processing enzymes where preserving activity is important.

Conclusions
Herein, we designed a building block synthetic concept for iminosugar-based small molecule probes for covalent proximity labelling of glycoside hydrolases.In a proof of concept study synthesised probes 1-6 were biologically evaluated assessing the ligand-directed chemistry approach with two model βglucosidases.Key to our approach was use of the sulfonate ester as the reactive, cleavable site for covalent labelling.Moreover, our results confirm that the linker length between the ligand and the electrophile plays a significant role in the labelling process.The DNJ-C 6 À LDSE-N 3 probe 3 labels TmGH1 at low concentrations, for labelling PdGH1 higher concentrations are necessary.Hence, the length of the spacer can be used as another selectivity factor, when designing these probes for labelling specific enzymes in future applications.
Furthermore, we showed that the labelling process is driven by interaction of the ligand with the active site.This confirmed the fact that if the active site is blocked by a more potent reversible inhibitor (IFG) or an irreversible inhibitor (CBE), no labelling of TmGH1 with DNJ-C 6 À LDSE-N 3 probe 3 was observed.The results reveal that the choice of ligand and the linker length are crucial for labelling.Therefore, these modules must be tailored individually for a respective enzyme to optimize the probe.
Importantly, we demonstrated that catalytic activity is maintained after the labelling process.Taken together, our results show the applicability of the LDC approach for labelling carbohydrate processing enzymes where preserving activity is important.
Having achieved this proof of concept we are now ready to transfer the method of LDC labelling onto several glycoside hydrolases, such as human lysosomal enzymes.In principle, the LDC labelling can be transferable to α-and β-glycosidases, as well as retaining and inverting or exo-and endo-glycoside hydrolases with a respective probe design tailored for the enzymes of interest.

General methods
Optical rotations were measured at 20 °C on a Schmidt Haensch variPol C polarimeter at 589 nm with a path length of 5 cm.NMR spectra were recorded on a Bruker Ultrashield spectrometer at 300.36 MHz (1H) and 75.53 MHz (13 C), respectively.CDCl 3 was employed for protected compounds and CD 3 OD for unprotected inhibitors.Chemical shifts are listed in δ employing residual, non-   (1 L)) were employed, followed by heating on a hotplate.
Purification of the desired products by column chromatography was performed with the stated solvent systems on silica gel 60 (Acros Organics or Macherey-Nagel).Selected products were purified by flash chromatography using a Biotage® Selekt system and prepacked silica gel 20 columns.

General Procedure A: Mitsunobu reaction
A 10 % solution of the respective starting material (1.0 equiv.) in THF, Ph 3 P (1.0-1.1 equiv.),diisopropyl azodicarboxylate (DIAD, 1.0-1.1 equiv.)and the respective alcohol (1.0-1.1 equiv.) was stirred until complete conversion of the reactants was detected by TLC.Subsequently, the reaction mixture was diluted with CH 2 Cl 2 and washed consecutively with aqueous HCl (2 N) and saturated NaHCO 3 .After drying over Na 2 SO 4 , the filtrate was concentrated in vacuo to provide the corresponding crude product.

General Procedure B: Reductive amination employing NaBH 3 CN
A 20 % solution of the respective aldehyde (1.0 equiv.)and iminosugar (1.0-1.2 equiv.) in MeOH (containing a catalytic amount of AcOH) was stirred for 15 min before NaBH 3 CN (1.5-3.0 equiv.) was added.After complete conversion of the starting materials was detected by TLC, the reaction mixture was concentrated under reduced pressure to provide the corresponding crude title compound.

Biology General methods and materials
Solvents and components of buffering systems used were all commercially available and were used without further purification.The substrate p-nitrophenyl-β-d-gluco-pyranoside (pNP-β-Glc) and its hydrolysis product p-nitrophenol were purchased from Merck KGaA (EMD Millipore Corp. USA).Enzymes used in this proof of concept study were purchased and used without further purification.The β-glucosidase from Thermotoga maritima (TmGH1, LOT: 151102a) is commercially available from Megazyme (Headquarter in Bray, Ireland since 1996) as ammonium sulphate suspension with a specified activity of 460 U/ml.The β-glucosidase from sweet almonds (Prunus dulcis, PdGH1, LOT: BCCC7765) and α-mannosidase from Canavalia ensiformis (CeGH38, LOT: 016H9555) were purchased from Sigma Aldrich as lyophilized powder and a specified lot activity of 4.5 U/mg and 19 U/mg, respectively.Enzymes yielded more than one single band (at pH = 8.0, TRIS-HCl 50 mM) in sodium dodecyl sulphate (SDS) polyacrylamide gel electrophoresis (PAGE) with different molecular weights of approx.51 kDa for TmGH1 and 62 kDa for PdGH1 compared to 54 kDa and 135 kDa as specified by the supplier, respectively.Protein bands of 51 kDa for TmGH1 and 62 kDa for PdGH1 have been found to exhibit hydrolytic activity towards pNP-β-Glc and therefore were identified as the target enzymes.

Protein quantification (Bradford assay)
Protein quantification was done following the standard Bradford assay setup using the ROTI® Quant reagent from Carl Roth GmbH and Co. KG following instructions of the supplier.The calibration curve for protein quantification was determined in 96-well plates measuring the absorbance of the Coomassie Brilliant Blue Dye-G250 at 595 nm using calibration standards of BSA in the range between 0 to 100 μg/mL protein (details see SI Figure S19).

Labelling experiments
The enzyme (0.46 mg/mL) was incubated with the respective LDC probe (1-6) in a total volume of 200 μL including a final DMSO concentration of 2v%.Standard incubation conditions are found to be 100 μM of the ligand-directed chemistry (LDC) probe for 20 hours at room temperature, corresponding to an approximate molar ratio of 1 : 10 enzyme to probe.Consecutive removal of residual probe and small molecules by washing three times with 500 μL TRIS-HCl 50 mM pH = 8.0 and thereby exchanging the buffering system is performed by the use of Vivaspin® 500 centrifugal concentrator spin-columns from SARTORIUS (MWCO 10.000 PES membranes).Concentration of the protein solution yields approximately 20-25 μL of 4 mg/mL protein in TRIS-HCl 50 mM pH = 8.0.This solution is used for the second step, the introduction of the fluorescent tag via click chemistry.

Click chemistry protocol
Visualization of azide-functionalized enzyme was achieved by attaching the fluorescent dye AlexaFluorTM 594 alkyne (invitrogen, ThermoFisher Scientific) using the Click iT® TM Protein Reaction Buffer Kit (invitrogen, ThermoFisher Scientific).The fluorophore was introduced via copper-catalyzed click chemistry following the instructions of the manufacturer.The procedure was optimized to low volume reactions with the need for only 5 μL of the chemically modified enzyme (4 mg/ mL, in TRIS-HCl 50 mM pH = 8.0).The volume ratios of the reagents and order of the click iT® TM protocol are maintained (details see SI Figure S20).After the addition of the last reagent the reaction was stirred for 20 min at 1000 rpm at room temperature.SDS-PAGE analysis was performed without an additional wash protocol.

Enzyme kinetics
The hydrolytic activity of β-glucosidases from Thermotoga maritima (TmGH1) and sweet almonds (Prunus dulcis, PdGH1) were assayed spectrophotometrically.The release of p-nitrophenol at the expense of pNP-β-Glc was measured at 405 nm over 2-5 minutes using 96-well plates from SARSTEDT on a Spark® Multimode Microplate Reader (TECAN Group AG, Switzerland).Kinetic parameters K m , v max , and k cat were determined using constant amounts of enzyme together with 12 different substrate concentrations.Results and enzyme specific conditions are given in the supporting information Table S1 and Figure S2.Determination of inhibition constants (K i -values) and IC 50 values was performed by measuring residual hydrolytic activities after pre-incubation of β-glucosidases with the inhibitors at 7 different inhibitor concentrations and one control sample without the inhibitor, reflecting 100 % of activity at the specified conditions.The reaction was started by addition of three different substrate concentrations ([S] < K m , [S] = K m , [S] > K m ).Details on inhibitor and substrate concentrations are given in the SI Table S2.Initial rates for Michaelis Menten kinetics and inhibition kinetics were calculated with Excel.Lineweaver-Burk plots were constructed to validate the use of competitive or mixed type inhibition models and to assess the fit of the data (see SI Figures S3-S4).The data were then fit to the Michaelis Menten model or a competitive/mixed type inhibition model using non-linear regression analysis with Grafit 7.0.3.(Erithacus Software, details see SI equation 1-3).For the IC 50 values the series of initial rates obtained at 7 different inhibitor concentrations and one control sample at [S] = K m were analysed using non-linear regression with Grafit 7.0.3(Erithacus Software, details in SI).

Figure 1 .
Figure 1.Schematic illustration of the LDC strategy resulting in chemical modification of the enzyme nearby -but not in -the active site, thereby preserving its activity.(A) Ligand, (B) linker with reactive cleavable site, (C) reporter tag, (Nu) nucleophilic amino acid.

Table 1 .
K i -values of ligand-directed probes
[41]03deuterated solvent or residual H 2 O (CD 3 OD) as the internal standard.[41]CDCl3 : 7.26 ppm ( 1 H), 77.16 ppm ( 13 C); CD 3 OD: 4.87 ppm ( 1 H), 49.0 ppm ( 13 C).Signals were unambiguously assigned by COSY (correlation spectroscopy) and HSQC (heteronuclear single-quantum correlation spectroscopy) analysis.The signals of the aromatic groups are located in the expected regions and are not listed explicitly.MALDI-TOF was performed on a Micromass TofSpec 2E Time-of-Flight mass spectrometer.All reactions were monitored by thin-layer chromatography (TLC) performed on pre-coated aluminium plates silica gel 60 F254 and detected with UV light (254 nm).