New Irreversible α‐l‐Iduronidase Inhibitors and Activity‐Based Probes

Abstract Cyclophellitol aziridines are potent irreversible inhibitors of retaining glycosidases and versatile intermediates in the synthesis of activity‐based glycosidase probes (ABPs). Direct 3‐amino‐2‐(trifluoromethyl)quinazolin‐4(3H)‐one‐mediated aziridination of l‐ido‐configured cyclohexene has enabled the synthesis of new covalent inhibitors and ABPs of α‐l‐iduronidase, deficiency of which underlies the lysosomal storage disorder mucopolysaccharidosis type I (MPS I). The iduronidase ABPs react covalently and irreversibly in an activity‐based manner with human recombinant α‐l‐iduronidase (rIDUA, Aldurazyme®). The structures of IDUA when complexed with the inhibitors in a non‐covalent transition state mimicking form and a covalent enzyme‐bound form provide insights into its conformational itinerary. Inhibitors 1–3 adopt a half‐chair conformation in solution (4H3 and 3H4), as predicted by DFT calculations, which is different from the conformation of the Michaelis complex observed by crystallographic studies. Consequently, 1–3 may need to overcome an energy barrier in order to switch from the 4H3 conformation to the transition state (2, 5B) binding conformation before reacting and adopting a covalent 5S1 conformation. rIDUA can be labeled with fluorescent Cy5 ABP 2, which allows monitoring of the delivery of therapeutic recombinant enzyme to lysosomes, as is intended in enzyme replacement therapy for the treatment of MPS I patients.


Materials
Recombinant human iduronidase (rIDUA) was obtained from Genzyme (Aldurazyme). Normal human dermal fibroblasts (NHDF) were obtained from Lonza. Human patient fibroblasts (MPS-I and ML-II) were obtained with consent from donors from the Academic Medical Center in Amsterdam, the Netherlands. 4-methylumbelliferyl α-L-iduronide (4-MU-α-L-IdoA) was purchased from Glycosynth. Pierce TM Bicinchoninic acid (BCA) protein assay kit and Pierce™ Polyacrylamide Spin Desalting Columns 7K MWCO was acquired from Thermo Fisher Scientific. All other chemicals were obtained from commercial sources.

Apparent IC50 values of inhibitors and ABPs on recombinant human iduronidase and cell lysates
Inhibitors and ABPs were diluted at various concentrations in 12.5 µL assay buffer and incubated with 12.5 µL rIDUA (5.8 ng, or 70 fmol) or fibroblast lysates (50 µg protein), both diluted in assay buffer, at 37°C for 1 h (rIDUA) or 4 h (fibroblasts lysates), at 1% (v/v) DMSO concentration. This was followed by iduronidase activity assay (30 min incubation time) described in the previous section. The detected 4-MU fluorescence at each concentration for each compound were normalized to the fluorescence from the control sample without inhibitor, and data were fitted with [inhibitor] vs response -variable slope (four parameters) function in Graphpad Prism 7.0 software to obtain apparent IC50 values.

Labelling of recombinant human iduronidase at different pH
Assay buffers with pH 2.5 -8.0 were used to dilute rIDUA and ABP 2. The experiment was performed with 10 ng rIDUA at 10 µM ABP 2 (assay concentration) in 16 µL volume and 1 h incubation time, followed by denaturation, SDS-PAGE, and fluorescence scan.

Reversibility of labelling of ABP 2 on recombinant human iduronidase
5 µL rIDUA stock (0.58 µg/µL in PBS) was diluted with 20 µL assay buffer without Triton X-100, and incubated with either 25 µL ABP 2 (150 µM and 1 % (v/v) DMSO) or 25 µL assay buffer containing 0.1% (w/v) DMSO for 1 h at 37°C. Thereafter, 45 µL of sample was passed through a desalting column, and diluted with 216 µL assay buffer (with Triton X-100) to a final enzyme concentration of 1 ng/µL. For assessing the reversibility of labelling/inactivation of ABP 2 on rIDUA, both samples were kept at 4 °C until subjecting to iduronidase activity assay using 4-MU-α-IdoA at the following time points after desalting: 0.5, 28, and 100 h. Iduronidase activity assay was performed with 1 µL sample in triplicates, and assay buffer was used as blank ( Figure S2).

Kinetic parameters of ABP 2
The kinetic parameters of ABP 2 for rIDUA was determined by a SDS-PAGE-based assay, in which the intensity of fluorescent signal from ABP 2-labelled rIDUA on the wet slab gel is proportional to the extent of rIDUA inactivation, when compared to the signal from a control group with saturate labelled rIDUA. The experiment was performed by labelling rIDUA with ABP 2 for different time periods and at different ABP 2 concentrations (which were >> enzyme concentration), and the kinetic parameters kinact (pseudo first-order inactivation rate constant) and KI (inhibition constant) were obtained using non-linear curve fitting. Briefly, rIDUA was diluted in a series of 1.5 mL Eppdendorf tubes to 10 ng (120 fmol) per 14 µL, and labelled with 2 µL ABP 2 (diluted in DMSO and assay buffer to various concentrations and [DMSO] fixed at 8% (v/v)) for 30, 60, 90, 120, or 150 min. Concentrations in assay were 5 µM, 10 µM, 20 µM, 30 µM, 40 µM, 50 µM, or 60 µM for ABP 2, 7.5 nM for rIDUA, and 1 % (v/v) for DMSO. Reaction was terminated by incubating samples with sample buffer at 98°C for 5 min. Samples were subjected to SDS-PAGE and fluorescence scan, and Cy5 fluorescence from the ABP 2-labelled rIDUA was quantified using ImageQuant software (GE Healthcare). The incubation condition for maximum labelling (complete inactivation) on rIDUA for ABP 2 was determined to be 60 µM and 150 min, and a control sample with this condition was loaded on every gel to allow normalization for samples in the same gel. After normalization, the average values at each ABP 2 concentration from n =3 sets of experiment were plotted in a time vs % labelling (inactivation) graph, and the plotted data were fitted with one-phase exponential association function to obtain the rate constant k at each ABP 2 concentration. Finally, the obtained k values were plotted S9 against ABP 2 concentrations, and the data were fitted with a Michaelis-Menten equation to obtain kinact and KI values for ABP 2 on rIDUA. All non-linear curve-fitting was performed using Graphpad Prism 7.0 software.

ABP 2 labelling in fibroblast lysates
Fibroblast lysates (40 µg protein) were diluted in assay buffer and incubated with ABP 2 at final concentrations of 1, 2, 5, or 10 µM (1% (v/v) DMSO) at 37°C for 4 h. Samples were denatured and subjected to SDS-PAGE, and wet slab gel was scanned for Cy5 fluorescence and then stained with Coomassie Brilliant Blue solution for assessing total protein amounts in each lane of sample.

Fibroblasts uptake of ABP 2-labelled recombinant human iduronidase
rIDUA was labelled with ABP 2 at 75 µM for 1 h at 37 °C, and unbound ABP 2 in solution was removed followed the described procedures in an earlier section. The labelled rIDUA were diluted in assay buffer without Triton X-100, to a concentration of 10 ng/µL. For the uptake experiment, human normal and patient fibroblasts were sub-cultured 1 day before treatment in 12-well plates (1 mL culture medium per well) with or without glass coverslips. Cells were then pre-treated with 4 mM mannose-6-phosphate for 1 h, followed by treating with ABP 2-labelled rIDUA (100 ng/mL culture medium) for 20 h. A control group was included for each cell type, treated only with water and assay buffer. Confocal microscopy analysis was carried out following a previously described procedure 3 , where samples were fixed with 4% formaldehyde, permeabilized with 0.1% (w/v) saponin and 2% (w/v) BSA, and immuno-stained for the lysosomal membrane protein LAMP-1 using mouse anti human LAMP1 (Southern Biotech) as primary antibody and donkey anti mouse Alexa488 (Molecular Probes) as secondary antibody. The coverslips were mounted to microscopy slides using ProLong™ Diamond Antifade Mountant with DAPI (Thermo Fisher), and scanned using a Leica SP8 confocal microscope for DAPI, Alexa488, and Cy5 fluorescence with a 40x oil-immersed objective. Pictures at each fluorescence channel were captured at 1024 x 1024 resolution, with n =3 frame averages.

LC-MS/MS identification of rIDUA active site peptide
A total of 10 µg rIDUA was diluted in assay buffer and incubated with either 75 µM ABP 3 or DMSO (negative control) for 1 h at 37°C in 100 µL volume ([DMSO] = 0.5 % (v/v)). The samples were then added with 100 µg BSA (100 µL), and followed by chloroform/methanol precipitation and reduction/alkylation procedures described previously. 4 Consequently, the samples were dissolved in 2 % (w/v) SDS and diluted with 50 mM Tris-HCL (pH 7.8) to a final SDS concentration of 0.005 % (w/v). The samples were then concentrated with size exclusion columns (Amicon 10k) to a volume of 74 µL, and digested O/N at 25 ºC with 1.1 µg Chymotrypsin (Promega) in the presence of 10 mM CaCl2. Digested peptides were pulled-down using 50 µL of Streptavidin paramagnetic beads (MyOne T1, ThermoFisher) in 1 mL of pull-down buffer (50 mM Tris-HCL (pH 7.5), 150 mM NaCl, 0.5 % (w/v) SDS) for 1 h at RT under vigorous shaking. The beads were washed stringently following previously described procedures 4 and eluted with 100 µL of elution buffer (25 % (v/v) Acetonitrile, 5 % (v/v) formic acid, 70 S10 % H2O, 10 µM biotin) for 30 min at 37°C. Afterwards, acetonitrile in the supernatant were evaporated using a Speedvac at 45°C, and this was followed by desalting using StageTips. The eluate were evaporated and reconstituted in 20 µL of LC-MS sample solution (95:3:0.1, H2O:acetonitrile:formic acid) for LC-MS/MS analysis. Peptide samples were analysed with a two hour gradient of 5% to 25% acetonitrile on nano-LC, hyphenated to an LTQ Orbitrap and identified by manual search for the theoretical m/z values of the active site peptide and its MS/MS fragments labelled with ABP 3

Recombinant expression and purification of IDUA in seeds of Arabidopsis thaliana
Recombinant human -L-iduronidase (abbreviated to raIDUA to distinguish from rIDUA obtained from Genzyme) was produced in seeds of Arabidopsis thaliana cgl (complex glycan deficient) line A4.7 5 in which the seeds (T3 generation) accumulated raIDUA to 7.2 ± 0.6 % total soluble protein (9.8 μg/mg dry seeds). raIDUA was purified to homogeneity from the T3 seeds using concanavalin A-sepharose and anti-IDUA affinity chromatography as described previously 6,7 . In human IDUA, there are six N-linked glycosylation sites. The oligosaccharide structures at each site of rIDUA secreted from a Chinese hamster ovary (CHO) cell line have been determined by mass spectrometry 8 . The raIDUA expressed in the seeds of the cgl mutant of Arabidopsis has much reduced complexity in these N-linked glycans, the majority of which are non-matured, high mannose N-glycans 5,9 . Crystallization of raIDUA raIDUA was further purified by size exclusion chromatography using an S200 10/300 column (GE Healthcare) equilibrated in 20 mM Tris, pH 7.0, 500 mM NaCl, and 0.02 % sodium azide. The fractions containing raIDUA were buffer exchanged using a PD10 desalt column (GE Healthcare) into 20 mM dimethylglutaric acid, pH 6.0, 0.2 M NaCl, 5% (v/v) glycerol, and 5% (v/v) ethanol and concentrated to 10 mg/ml for crystallization. Crystallization was performed in a 24 well plate using hanging-drop vapour diffusion. The rhomboid-plate shaped crystals grew at room temperature from 0.1 M HEPES, pH 7.5, 0.26 M sodium potassium tartrate, 20% (w/v) polyethylene glycol 3350 (optimized from the crystallization condition reported by Bie et al. 10 ). Crystals were soaked in mother liquor containing a minute amount of solid 1 for 24 hours or 3 for 45 minutes, before being harvested. Crystals were cryo-protected in a solution containing the mother liquor plus 30% glycerol prior to vitrification in liquid nitrogen.
Data collection and processing for raIDUA crystals X-ray diffraction data were collected at Diamond Light Source (DLS) on beamlines I03 and I04; the data processing and refinement statistics can be found in Table S2. Diffraction data were processed either using the FastDP pipeline 11 (which utilises XDS 12 with Aimless 13 ) or Xia2 14 (also with XDS 12 with Aimless 13 ). Molecular replacement was performed using Phaser 15 with Protein Data Bank (PDB) entry 4JXO as the search model. Refinement was performed using REFMAC5 16 and manual model building was done using Coot 17 . Structures were validated using PDB_REDO 18 . Models for the fragments of 1 and 3 were built in JSME 19 and the libraries generated with PRODRG 20 . S11

Geometry optimization
By using the conformer distribution search option included in the Spartan 14 program 21 , exclusively the 4 H3 conformation of the structure was found. Only notable variations of the geometry were found at the C5-C7 bond, including multiple rotamers which were significant higher in terms of energy.
For α-L-idoA-aziridine the 4 H3 conformation was also found as lowest energy conformer, but in this case the 3 H4 was only 1.4 kcal/mol higher in terms of energy.
All calculations were performed with DFT as level of theory in combination with the B3LYP hybrid functional. A conformer distribution search option included in the Spartan 14 program 21 , in gas-phase with the use of 6-31G(d) as basis set, was used as starting point for the geometry optimization. All generated structures were further optimized with Gaussian 09 22 at 6-311G(d,p). Optimization was done in gas-phase and subsequently corrections for solvent effects were done by the use of a polarizable continuum model using water as solvent parameter.
The free Gibbs energy of the computed conformations was calculated using Equation (1)

= ∆ + ∆
The used free energies include unscaled zero-point vibrational energies. Visualisation of the conformations of interest was done with CYLview. 23 S13

NMR calculations
Based on the optimized lowest energy structure the spin-spin coupling constants were calculated according to the work of Rablen and Bally 24 with the use of 6-311g(d,p) u+1s as basis set and PCM(H2O) as solvent model and a scaling factor of 0.92. The calculated total nuclear spin-spin coupling terms were used as calculated spin-spin coupling constants. Table S3. Experimental coupling constants of α-L-idose configured cyclophellitol aziridine 11 compared to DFT calculated coupling constants.

General Experimental Details
All reagents were of a commercial grade and were used as received unless stated otherwise. Dichloromethane (DCM), tetrahydrofuran (THF) and N,N-dimethylformamide (DMF) were stored over 4 Å molecular sieves, which were dried in vacuo before use. Triethylamine was dried over KOH and distilled before using. All reactions were performed under an argon atmosphere unless stated otherwise. Solvents used for flash column chromatography were of pro analysis quality. Reactions were monitored by analytical thin-layer chromatography (TLC) using Merck aluminum sheets pre-coated with silica gel 60 with detection by UV absorption (254 nm) and by spraying with a solution of (NH4)6Mo7O24·H2O (25 g/L) and (NH4)4Ce(SO4)4·H2O (10 g/L) in 10% sulfuric acid followed by charring at ~150 ˚C or by spraying with an aqueous solution of KMnO4 (7%) and K2CO3 (2%) followed by charring at ~150 °C. Column chromatography was performed manually using either Baker or Screening Device silica gel 60 (0.04 -0.063 mm) or a Biotage Isolera™ flash purification system using silica gel cartridges (Screening devices SiliaSep HP, particle size 15-40 µm, 60A) in the indicated solvents. 1 H NMR and 13 C NMR spectra were recorded on Bruker DMX-600 (600/150 MHz) and Bruker AV-400 (400/100 MHz) spectrometer in the given solvent. Chemical shifts are given in ppm relative to the chloroform residual solvent peak or tetramethylsilane (TMS) as internal standard. Coupling constants are given in Hz. All given 13
(4R)-3-[(2E)-but-2-enoyl]-4-isopropyl-1,3-oxazolidin-2-one (2.20 g, 11.2 mmol) was dissolved in anhydrous DCM (20 mL). Then, a solution of 1.0 M dibutylboryltrifluoromethanesulfonate (DBBT) in anhydrous DCM (11.2 mL, 11.2 mmol) was added at -78°C. The resulting dark yellow mixture was removed from the cold bath to dissolve any frozen triflate and cooled again back to -78°C. Et3N (recently distilled, 1.79 mL, 12.8 mmol) was then added, causing the dark green color to fade. The solution was stirred for 1 h at -78°C and then at 0°C for 15 min and the solution turned yellow. While the reaction mixture was being cooled back down to -78°C, a solution of 2,3-di-O-benzyl-4,5-dideoxy-L-threo-pent-4-enose (3.31 g, 11.2 mmol) in anhydrous DCM (20 mL) was added to the reaction mixture via a cannula and argon atmosphere. The bath temperature was raised to -20 °C over 1 h and then the reaction mixture was stirred at this temperature overnight. The resulting yellow solution then warmed to -5°C and quenched with pH 7 phosphate buffer solution (25 mL). A 30% H2O2 solution (30 mL) was then added dropwise while maintaining the internal temperature below 5°C. Addition of the peroxide was continued until the internal temperature remained constant. The mixture was stirred for additional 30 min while slowly warming to room temperature. The reaction was then poured over aqueous saturated NaHCO3 solution (100 mL) and the aqueous layer extracted with DCM (3 x 100 mL). The combined organic layers were dried over MgSO4, filtered, and concentrated under reduced pressure. The crude product was purified by silica column chromatography (from pentane to pentane/EtOAc, 8:2) to afford the desired product 6 as colorless oil (3.32 g, 6.73 mmol, 60%).
To a solution of oxazolidinone 6 (2.55 g, 5.2 mmol) in THF (50 mL) at 0°C, H2O (2.5 mL) and LiBH4 (2 M solution in THF, 12.9 mL, 25.8 mmol) were added and stirred 1 h at 0°C. The reaction mixture was then warmed to room temperature and stirred for 1 h. The reaction was quenched with 2 N NaOH (aq) (50 mL) and diluted with Et2O (50 mL). After stirring for five minutes, the reaction mixture was poured over Et2O (50 mL) and the separated organic phase was washed with aqueous saturated NaHCO3 solution (20 mL) and brine (100 mL). The organic phase was dried over MgSO4, filtered, and concentrated under reduced pressure to give the crude alcohol. The crude was purified by silica column chromatography (pentane/EtOAc from 9:1 to 1:1) to afford the desired product 7 as a colourless oil (1.89 g, 5.13 mmol, 99%). Catalytic amount of TBAI (109 mg, 0.29 mmol), BnBr (1.75 mL, 14.7 mmol) and sodium hydride (60% dispersion in mineral oil, 587 mg, 14.7 mmol) were added to a solution of 8 in DMF (120 mL) at 0 ºC and the reaction mixture was stirred at room temperature 24 h. The reaction mixture was cooled to 0 °C and quenched by the addition of MeOH. The mixture was further diluted with water and subsequently extracted with Et2O. The organic phase was washed with brine, dried over MgSO4, filtered, and concentrated in vacuo. The crude was purified by silica column chromatography (from pentane to pentane/EtOAc 9:1) to afford the desired product 9 (1.36 g, 89%) as colourless oil. A suspension of 2-trifluoromethyl-3-aminoquinazolin-4-one (2.55 g, 11.4 mmol) in anhydrous DCM (70 mL) was added dropwise over a period of 60 min to a stirred suspension of PhI(OAc)2 (3.59 g, 11.1 mmol) in DCM (50 mL) at -25 ºC. The resultant mixture was stirred for additional 30 min and then a solution of the olefin 9 (2.9 g, 5.57 mmol) in DCM (20 mL) was added dropwise at a flowrate of 70 mL/h. The reaction mixture was then allowed to warm to oom temperature and stirred overnight. Then, the reaction mixture was diluted with brine (100 mL) and extracted with DCM (2 x 100 mL). The collected organic layers were dried over Na2SO4, filtered and the solvent was removed under reduced pressure and the crude was purified by silica column chromatography (from pentane to pentane/EtOaAC 8:2) to afford the desired aziridine 10 as a brown oil (1.79 g, 43% Ammonia (60 mL) was condensed at -60 °C. Lithium (0.67 g, 96 mmol) was added, and the mixture was stirred until the lithium was completely dissolved. A solution of aziridine 10 (1.79 g, 2.39 mmol) in THF (30 mL) was added. The reaction mixture was stirred for 1 h at -60 °C and subsequently quenched with milliQ-H2O (20 mL). The solution was allowed to achieve room temperature and stirred until all ammonia was evaporated. Next, the solution was concentrated in vacuo, redissolved in milliQ-H2O, filtered to remove the quinazolinone impurities and the filtrate was neutralized with Amberlite IR-120 H+. Product bound to the resin was washed with water and subsequently eluted with NH4OH solution (1 M) and evaporated under reduced pressure to give the fully deprotected aziridine 11 as a beige foam (389 mg, 93%). 8-Azido-octanoic acid (170 mg, 0.92 mmol, 1.3 equiv.) and EEDQ (227 mg, 0.92 mmol, 1.3 equiv.) were dissolved in anhydrous DMF (1.0 mL) and stirred at room temperature for 2 h. This pre-activated mixed solution (500 µL, 0.65 equiv.) was added to a solution of aziridine 11 (124 mg, 0.71 mmol, 1.0 equiv.) in DMF (5 mL) at 0 °C and stirred for 30 minutes after which the remaining portion of the preactivated solution (500 µL, 0.65 equiv.) was added. The resulting mixture was stirred at 0 °C for 3h. The reaction was quenched by 2 mL MeOH and the mixture was concentrated in vacuo. Then the crude product was purified by silica column chromatography (from DCM to DCM/MeOH 9:1) giving 13 as a colorless oil (56. Intermediate 12 (25 mg, 0.076 mmol), catalytic amount of TEMPO (0.08 mg, 0.493 μmol), and NaBr (3.12 mg, 0.030 mmol) were dissolved in water (3.0 mL) at 0 °C. A 13% sodium hypochlorite solution (0.079 ml, 0.167 mmol) was added dropwise in 10 portions to the mixture, at the same time adjusted to pH=10.5 by adding an aqueous solution of NaOH (0.5 M). The reaction mixture was stirred at 0 ºC for 2 h. After which, starting material remained and more TEMPO (0.04 mg, 0.25 mmol), NaBr (1.6 mg, 0.015 mmol) was added and subsequent sodium hypochlorite (40 μL, 0.084 mmol) was added in 5 portions adjusting the pH to 10.5. After stirring 1 h at 0 ºC the oxidation was finished, the reaction was quenched by adding EtOH (1mL) and the pH was adjusted to 7 by adding an aqueous solution of HCl solution (0.5 M). The reaction mixture was lyophilized and purified by semi-preparative reversed HPLC (linear gradient: 19%→25% B in A, 12min, solufons used A: 50mM NH4HCO3 in H2O, B: acetonitrile), the fractions were concentrated and lyophilized to afford desired product 1 as a white powder (3.6 mg, 14%). 1  Intermediate 13 (23 mg, 0.067 mmol, 1eq.), TEMPO (0.1 mg, 0.44 umol, 0.0065eq.), and NaBr (2.8 mg, 0.03 mmol, 0.4 eq.) were dissolved in water (5.0 mL) at 0 °C. A 13% sodium hypochlorite solution (0.07 ml, 0.148 mmol, 2.2eq.) was added dropwise in 10 portions to the mixture, while at the same time the pH was adjusted to pH=10.5 by adding aqueous 0.5 M NaOH. The reaction was conducted at 0 ºC in an ice bath and checked by LC/MS within the elution system of 10% NH4OAc. After 3 h product with the expected mass was detected and the reaction was quenched by adding 96% EtOH (1mL) and the pH was adjusted to 7 by adding aqueous 0.5 M HCl. The reaction mixture was freeze-dried and the crude product was purified by semi-preparative reversed HPLC (linear gradient: 14%→19% B in A, 12min, solutions used A: 50mM NH4HCO3 in H2O, B: acetonitrile), the fractions were concentrated and lyophilized to white powder product (2 mg, 5.6 µmol, 8.4%). 1 H NMR (500 MHz, CD3OD):

General Click Procedure for ABPs 2 and 3.
A solution of CuSO4·5H2O and sodium ascorbate (0.1M in MilliQ water) was prepared and degassed before using. Azido cyclophellitol aziridine 1 (1 equiv.) was dissolved in DMF (2 mL) and CuSO4·5H2O (1 M, 0.45 equiv.) and sodium ascorbate (1 M, 0.48 equiv.) were added to the solution under argon atmosphere. Then, the corresponding alkyne (1 equiv.) was added and the reaction mixture was stirred at room temperature overnight. Then, the reaction mixture was concentrated under reduced pressure and purified by semi-preparative reversed HPLC (linear gradient: %→% B in A, 12min, solufons used A: 50mM NH4HCO3 in H2O, B: acetonitrile), the fractions were concentrated and lyophilized to afford the desired product.