A Sensitive Gel-based Method Combining Distinct Cyclophellitol-based Probes for the Identification of Acid/Base Residues in Human Retaining β-Glucosidases*

Background: Cyclophellitol-derived activity-based probes label the nucleophile of retaining β-glucosidases by exploiting their catalytic mechanism. Results: Combining two distinct cyclophellitol-derived probes with azide-mediated activity-rescue of mutated retaining β-glucosidases enables acid/base residue identification. Conclusion: A sensitive, new gel-based method for catalytic residue identification is developed. Significance: Cyclophellitol-based probes are proposed as useful tools for the identification of catalytic residues in retaining glycosidases. Retaining β-exoglucosidases operate by a mechanism in which the key amino acids driving the glycosidic bond hydrolysis act as catalytic acid/base and nucleophile. Recently we designed two distinct classes of fluorescent cyclophellitol-type activity-based probes (ABPs) that exploit this mechanism to covalently modify the nucleophile of retaining β-glucosidases. Whereas β-epoxide ABPs require a protonated acid/base for irreversible inhibition of retaining β-glucosidases, β-aziridine ABPs do not. Here we describe a novel sensitive method to identify both catalytic residues of retaining β-glucosidases by the combined use of cyclophellitol β-epoxide- and β-aziridine ABPs. In this approach putative catalytic residues are first substituted to noncarboxylic amino acids such as glycine or glutamine through site-directed mutagenesis. Next, the acid/base and nucleophile can be identified via classical sodium azide-mediated rescue of mutants thereof. Selective labeling with fluorescent β-aziridine but not β-epoxide ABPs identifies the acid/base residue in mutagenized enzyme, as only the β-aziridine ABP can bind in its absence. The Absence of the nucleophile abolishes any ABP labeling. We validated the method by using the retaining β-glucosidase GBA (CAZy glycosylhydrolase family GH30) and then applied it to non-homologous (putative) retaining β-glucosidases categorized in GH1 and GH116: GBA2, GBA3, and LPH. The described method is highly sensitive, requiring only femtomoles (nanograms) of ABP-labeled enzymes.

Retaining ␤-glucosidases hydrolyze glucosidic substrates through a two-step double-displacement mechanism driven by a discriminate nucleophile and acid/base residue (Fig. 1, a and  b) (18). In the first step the nucleophile attacks the anomeric carbon of the glucose moiety, whereas the acid/base residue protonates the interglucosidic oxygen, resulting in aglycone release and concomitant nucleophile glucosylation (28). In the second step enzyme deglucosylation is promoted by activation of an incoming water nucleophile by the deprotonated acid/ base, completing the hydrolysis event with overall retention of configuration at the anomeric carbon of the released glucose molecule.
Recently we reported the development of two different activity-based probe (ABP) classes for retaining ␤-glucosidases (29,30). The first class concerns cyclophellitol (␤-epoxide) derivatives functionalized with a fluorescent BODIPY moiety grafted to the C8 position (MDW933, hereafter named ABP 1; see Fig.  1c) (29). They bind irreversibly to the nucleophile Glu-340 of GBA in an acid/base-dependent manner; the acid/base residue protonates the ␤-epoxide, which is concomitantly attacked by the nucleophile; thereby the oxirane opens and forms a stable ester bond with the nucleophile residue ( Fig. 1d) (29,30). Aziridine analogs of cyclophellitol functionalized with a BODIPY group grafted to the ␤-aziridine nitrogen comprise the other ABP class (MDW1044, hereafter named ABP 2; see Fig. 1c) (30). These bind nucleophiles of various ␤-glucosidases through a similar mechanism but that can occur independently of the acid/base status (Fig. 1e) (30).
An elegant method has been developed in the past to identify the nucleophile and acid/base residues of retaining glycosidases, namely by using (sodium) azide as an exogenous anion FIGURE 1. Background. a, protein sequence homology of all retaining ␤-glucosidases expressed in man, categorized in glycosylhydrolase families GH1, -30, and -116. b, double displacement mechanism of retaining ␤-glucosidases, with distinct acid/base (A/B, top) and nucleophile (N, bottom). c, structure formulas of cyclophellitol-based ABP 1 (MDW933 (29)) and aziridine containing ABP 2 (MDW1044 (30)) and their nucleophile binding mechanisms (d and e, respectively). Azide-mediated rescue of hydrolysis in case of absent acid/base (f) or nucleophile (g). that can function as a substitute for the acid/base or nucleophile residue, allowing the rescue of substrate hydrolysis by enzyme molecules with a mutated catalytic residue (Fig. 1, f and g) (31)(32)(33). In this approach, definite identification of the acid/base and nucleophile can be achieved by determining the azidoproduct stereochemistry via NMR. In retaining ␤-glucosidases, sodium azide-mediated rescue results in retention of stereochemistry in the case of acid/base absence (␤-azidoglucoside; Fig. 1f) and inversion of stereochemistry when the nucleophile is lacking (␣-azidoglucoside; Fig. 1g) (31)(32)(33). Success of this method hinges on the ability to isolate and identify the nature of the glucosazide adduct formed, which essentially means that the approach works best when executed on isolated, soluble mutant glycosidases.
In a series of related studies, Withers et al. (28, 34 -37) have employed various 2-deoxy-2-fluoroglycosides to identify the nucleophile of retaining ␤-glycosidases by trapping the nucleophile. Covalent and (semi)permanent modification of the nucleophilic residues in retaining ␤-glucosidases occurs due to the substitution of the C2-hydroxyl by a fluorine, which markedly destabilizes the positive charge developed during the oxocarbenium ion-like transition states flanking the covalent glycosyl-enzyme intermediate, thereby effectively reducing the catalytic rates involved in the double-displacement mechanism by up to 10 7 -fold. Identification of the thus covalently and irreversibly modified nucleophile in this setting is achieved by mass spectrometry analysis of trypsin-digested protein, and the method thus relies on labeling of pure, recombinant enzyme with 2-deoxy-2-fluoroglycosides.
We here report a new and complementary method for the identification of the acid/base and nucleophile residues in the known human retaining ␤-glucosidases. Our approach entails a combination of site-directed mutagenesis, subsequent identification of putative catalytic residues by sodium azide-mediated rescue of enzymatic activity, and finally discrimination between the acid/base and nucleophile by detection of ABP 2-labeled ␤-glucosidases on slab-gels. The ABP labeling of retaining ␤-glucosidases is characterized by high sensitivity and selectivity, allowing visualization of femtomoles, i.e. nanograms, of enzymes within complex mixtures such as cell extracts.
As model we first used GBA (CAZy family GH30) with known nucleophile residue Glu-340 and acid/base residue Glu-235 to validate our method (39). Next, we studied GBA3 (CAZy GH1) for which crystal data predicted Glu-165 and Glu-373 as acid/base and nucleophile, respectively (40). In addition, LPH (GH1), with known catalytic residues in its two pockets (41), was investigated. Finally we studied the enzyme GBA2. Recently, the acid/base and nucleophile of a distant GBA2 homologue in Sulfolobus solfataricus (GH116) have been identified via azide rescue and covalent labeling of the nucleophile with 2-deoxy-2-fluoroglucoside (42). Based on homology alignment, the acid/base and nucleophile in human GBA2 were tentatively appointed previously (42) and by our own work, and these are in line with the residues predicted from the GBA2homologue in S. solfataricus. We also demonstrate through application of our ABP methodology that none of the proteins klotho, ␤klotho, and KLPH (GH1) possesses ␤-glucosidase activity.
Protein Expression and Isolation-Confluent COS-7 cells were transfected with empty pcDNA3.1 vector (Mock) or the vector with the described insert, in conjunction with FuGENE (Roche Applied Science) and harvested after 72 h by scraping in 25 mM potassium phosphate buffer (pH 6.5, supplemented with 0.1% (v/v) Triton X-100 and protease inhibitor mixture (Roche Applied Science)). After determination of the protein concentration (BCA kit, Pierce), lysates were aliquoted and frozen at Ϫ80°C.
Sodium Azide-mediated Rescue of Enzymatic Activity-Samples were incubated with ABP (with 10 g of total protein) or artificial substrate (with 100 g of total protein) in the presence of 0 -5 M sodium azide (solubilized in 10 mM Tris-HCl, adjusted to the optimal pH for each enzyme). As elegantly described earlier (see Ref 33), the presence of (sodium) azide can rescue the enzymatic activity in the absence of the nucleophile or that of the acid/base residue. After substrate reactions, tubes were centrifuged (3000 rpm for 3 min at room temperature) after which the supernatant was collected and vigorously mixed, and liberated chromogenic/fluorogenic product was measured. After ABP labeling reactions, proteins were precipitated with 10% (w/v) trichloroacetic acid for 10 min on ice and centrifuged at 10,000 rpm for 10 min at 4°C, and the resulting pellet was washed 3 times with cold 50% (v/v) acetone in nanopure water before further use for His-tag pulldown or directly solubilized in 1ϫ Laemmli buffer and applied to SDS-PAGE. Rescue of ABP labeling by the addition of acetic acid (10 M) or formic acid (15 M) followed exactly the same procedure.

Elucidation of the Acid/Base and Nucleophile in GBA-
The nucleophile (Glu-340) and acid/base (Glu-235) of GBA have been identified unequivocally (Fig. 2a) (29,39), and we therefore used this enzyme to validate our envisioned method where nucleophile-modifying cyclophellitol ABPs can distinguish the location of the acid/base and nucleophile in retaining ␤-glucosidases. Accordingly, GBA isoforms with substituted nucleophile and/or acid/base residues were generated. COS-7 cells were transfected with constructs expressing C-terminally myc/His-tagged human GBA or variants with E235G, E235Q, E340G, E340Q, or E235G/E340G substitutions. All mutagenized GBA proteins were deficient in enzymatic activity toward the artificial substrates 4MU-, 4NP-, and 2,4DNP-␤-D-Glc (Fig. 2b). Next, we attempted to rescue the enzymatic activity of the aforementioned GBA mutants by co-incubating a fixed concentration of artificial substrate together with a range of sodium azide concentrations. In this context, azide acts as exogenous anion and is at certain concentrations able to substitute the absent acid/base or nucleophilic residue (Fig. 1, f and  g) (31,33,42). The use of sodium azide has become an important staple in the laboratory toolkit available for catalytic residue discovery and has been thoroughly reviewed by Ly and Withers (32), and its use is elegantly described in manuscripts such as Bravman et al. (33).
Titration of the various GBA mutants with sodium azide did not result in an enhanced hydrolysis rate (i.e. rescue) of the acid/base and nucleophile mutants in combination with fluorogenic 4MU-␤-D-Glc (Fig. 2c). Partial, although significant rescue was observed with 4NP-␤-D-Glc except for the double E235G/E340G GBA mutant, which lacks both the acid/base and nucleophile (Fig. 2b). The most pronounced rescue was achieved with 2,4DNP-␤-D-Glc (Fig. 2e). This can readily be explained by considering that 4-methylumbelliferone (4MU) and 4-nitrophenol (4NP) are inferior leaving groups to 2,4-dinitrophenol (2,4DNP), and all require activation by protonation by the acid/base in the glucosylation step (31)(32)(33). At high concentrations, sodium azide severely impaired activity of GBA toward all artificial substrates tested (Fig. 2, d-f).
A final step the classic approach to determine the definite location of the acid/base and nucleophile is determination of the stereochemistry of the reaction products, formed during the presence of sodium azide in combination with the various mutants (31)(32)(33)42). This methodology hinges on the feasibility of isolation/purification of, first, the mutagenized glycosidase and, second, of the various azido products. We propose the use of specific, fluorescent cyclophellitol-type ABPs. The ABPs described are highly specific, only bind to the nucleophilic residue, and enable the sensitive detection of attomoles of ABP-labeled retaining ␤-glucosidases on slab gels (29,30). In addition, mutagenized proteins can be labeled directly within complex lysates and living cells (29,30) and can be, when required, purified before gel electrophoresis. To illustrate this ABP application, complete lysates of COS-7 cells expressing recombinant wild-type or mutant GBA were incubated with ␤-epoxide 1 or ␤-aziridine 2 and thereafter purified via their His tag with TALON beads. Thereafter, the labeled GBA was visualized by SDS-PAGE and subsequent fluorescence scanning (Fig. 3a).
The use of 10 g of total protein as the source for ABP labeling and subsequent purification step enabled evident detection of ϳ12 ng (i.e. 200 fmol) of recombinant and ABP-labeled GBA molecules on gel by using a Typhoon Variable Mode Imager (Amersham Biosciences) with medium sensitivity settings. In contrast, the limit of detection of the GBA loading control via Western blotting for the myc epitope required at least 100 g of total protein to be loaded on the gel (Fig. 3a).
␤-Epoxide 1 requires a priori a protonated acid/base and thus did not label E235G and E235Q GBA. Sodium azide completely restored the ability of E235G and E235Q GBA variants  DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51 to label with ABP 1 by acting as an external acid/base in the form of hydrazoic acid (HN 3 ; Fig. 3b). Labeling of acid/base mutants E235G and E235Q with ABP 1 reached its maximal intensity when using sodium azide (Fig. 3b) at concentrations where no rescue in catalytic activity could be identified when using 2,4DNP-␤-D-Glc (Fig. 2, d-f). The combination of sodium azide with latter substrate did not reach wild-type activity levels, even at optimal conditions. Complete restoration of ABP 1 labeling was also obtained when acetic acid (HAc) or formic acid (HFo) were added instead of sodium azide (Fig. 3c).

Cyclophellitol-based Probing of ␤-glucosidases
␤-Aziridine 2 reacts with the nucleophile Glu-340 regardless of the acid/base status in GBA molecules and independently of the presence of sodium azide (Fig. 3, a and b) (30). Substitutions of nucleophile Glu-340 in GBA to Glu-340 or Gln-340 inherently prevented labeling by both ABP classes (Fig. 3b). Having established that ABPs 1 and 2 indeed can be employed to sensitively differentiate the acid/base from nucleophilic residue as compared with the classic sodium azide rescue approach, we next applied the developed method to the other known human retaining ␤-glucosidases.
Elucidation of the Acid/Base and Nucleophile in GBA3 from CaZY Family GH1-We next analyzed the applicability of the ABP-labeling approach for the cytosolic retaining ␤-glucosidase GBA3, which is categorized in CAZy family GH1 and which is non-homologous to GBA (GH30, see also Fig. 1a). Crystallography studies of GBA3 strongly indicate that residues Glu-165 and Glu-373 function as its acid/base and nucleophile, respectively (Fig. 4a) (40). Mutated GBA3 containing E165G and/or E373G displayed no activity toward 4MU-, 4NP-, and 2,4DNP-␤-D-Glc (Fig. 4b). Titration with sodium azide resulted in a minor but significant activity rescue for E165G, whereas the E373G mutant did not hydrolyze 4NP-␤-D-Glc, indicating their involvement in substrate catalysis (Fig. 4c). A more evident rescue of activity was observed by using 2,4DNP-␤-D-Glc as substrate (Fig. 4d).
Substitution of Glu-373 prevented GBA3 labeling by both ABP 1 and 2, whereas substitution of Glu-165 allowed labeling by ␤-aziridine ABP 2 but not by ␤-epoxide 1, indicating Glu-165 functions as acid/base. This role was further confirmed by rescue of ␤-epoxide 1 labeling with sodium azide (Fig. 4e). Our results confirm the prior indications that indeed Glu-373 is the nucleophile and Glu-165 the acid/base residue in GBA3. More importantly, the results illustrate the developed method can be applied to retaining ␤-glucosidases non-homologous to GBA (GH30), categorized in other CaZY families such as GH1.
Labeling of G1065 LPH with ␤-epoxide 1 only occurred in the presence of sodium azide, whereas labeling of this mutant with ␤-aziridine 2 was independent of the presence of sodium azide, confirming Glu-1065 acts as an acid/base residue (Fig.  5e). The Gly-1273 LPH variant did not bind ABP 1 or ABP 2 in the absence or presence of sodium azide (Fig. 5e), whereas its enzymatic activity was rescued (2,4DNP-␤-D-Glc; Fig. 5c), indicating that Glu-1273 is indeed the nucleophile of pocket III.
Pocket IV of LPH appears not targeted by ABPs 1 or 2 as mutagenesis of its nucleophile and acid/base residue (Glu-1538 and Glu-1749, respectively) did not influence ABP labeling (Fig.  5e) nor did it result in a significantly reduced ␤-glucosidase activity (Fig. 5b). It seems, therefore, that substitutions in pocket IV have little impact on pocket III but, intriguingly, not vice versa. Wild-type LPH is irreversibly inhibited by ABPs 1 and 2 in its hydrolysis of 2,4DNP-␤-D-Glc, and relatively low IC 50 values were observed (1.3-1.4 and 0.19 -0.21 M for ABP 1 and 2, respectively). In contrast, ABP 1 and ABP 2 are not potent irreversible inhibitors of lactose hydrolysis mediated by ␤-galactosidase pocket IV, again indicating that the activity-based probes bind exclusively to the ␤-glucosidase pocket III of LPH.
Klotho, ␤Klotho, and KLPH-Sequence analysis indicates that KLPH possesses one pocket, and klotho and ␤klotho have two pockets. Moreover, each is predicted to lack either an acid/base or nucleophile residue (Fig. 6a). The absence of ␤-glucosidase activity toward 4MU-, 4NP-, and 2,4DNP-␤-D-Glc was confirmed by no demonstrable ABP labeling (Fig. 6b). Consistently, even after incubations of 18 h with an excess of either ABP (1 mM) and the presence of sodium azide, no labeling was observed.
The designation of Glu-527 as the nucleophile and Asp-677 as the acid/base in human GBA2 overlaps with the location of these residues identified in SSO1353 of S. solfataricus (see Fig.  8a) and at their suggested locations in human GBA2 as extrapolated from the data from S. solfataricus (42).

DISCUSSION
Retaining ␤-glucosidases fulfill important functions in the human body as is perhaps illustrated best by the deleterious consequences of GBA deficiency causing Gaucher disease and the emerging implication of GBA2 in cancer and neuronal pathologies. Better understanding of human ␤-glucosidases requires identification of their crucial catalytic amino acids for catalysis, the nucleophile, and acid/base residues.
Elegant methodologies have been developed in the past to identify these key residues, such as the use of sodium azide as an external acid/base and/or nucleophile, allowing catalytic activity rescue of enzyme molecules with a mutant catalytic residue and NMR analysis of the resultant product (31)(32)(33), or trapping the nucleophile with a 2-deoxy-2-fluoroglycoside and identification via mass-spectrometric detection. These methodologies hinge on the ability to isolate the wild-type or mutagenized enzyme before analysis, a feat usually reserved for at least partly soluble enzymes, and the feasibility to purify sufficient 2-deoxy-2-fluoroglycoside-labeled peptide or azido-glycoside from the treated samples.
We have described here a complementary method that combines the previously reported sodium azide-mediated rescue of enzyme activity to identify the catalytic residues with ABP-labeling in complex lysates to directly discriminate between the acid/base and the nucleophile. ABP labeling can occur directly with high specificity within living cells (data not shown) or in total lysate (29,30), after which, if required, the recombinant retaining ␤-glucosidases can be purified via for instance an implemented His tag.
We capitalize on the unique features of epoxide-containing ABP 1 and aziridine-containing ABP 2, which fundamentally differ in their ability to label retaining ␤-glucosidases. Both ABPs bind covalently to the nucleophile, but ␤-epoxide 1 FIGURE 7. Mapping of catalytic residues in GBA2. a, schematic depicting GBA2 putative catalytic residues (shades of gray), with possible acid/base D667 (left, gray) and nucleophile Glu-527 (black). b, sequence alignment of GBA2 homologues of various species, with highly conserved aspartic (D) and glutamic acids (E). DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51 requires a functional acid/base, whereas ␤-aziridine 2 does not. Upon labeling of the nucleophile with the fluorescent ABP, the protein can be detected by highly sensitive fluorescence scanning of SDS-PAGE slab-gels.

Cyclophellitol-based Probing of ␤-glucosidases
The high sensitivity of the presented method warrants discussion. Emitted fluorescence from ABP-labeled proteins on slab-gels can be detected in the attomole range (29), and here 50 femtomoles (i.e. 3 ng) of recombinant GBA, labeled with either ABPs, were clearly visualized. This is in sharp contrast to the at least 10-fold higher protein input minimally required for detection by Western blot and in the enzymatic assays. Therefore, relatively small amounts of recombinant enzymes expressed by mammalian cells after transfection are sufficient for accurate detection with fluorescent ABPs (29,30).
As we demonstrated, ␤-aziridine 2 reacts with the nucleophile of GBA, GBA2, GBA3, and the ␤-glucosidase pocket III of LPH to form a stable covalent adduct through a mechanism independent of the presence of a functional acid/base residue. In contrast, labeling by ␤-epoxide 1 of GBA, GBA3, and the ␤-glucosidase pocket III of LPH is prohibited by mutation of the acid/base, coinciding with loss of enzymatic activity that is the result of such mutations. Both enzyme activity and ␤-epoxide 1 labeling in these mutants can be restored by the addition of sodium azide as external acid/base. Binding of ABP 1 to the nucleophile in absence of the acid/base is completely rescued by employing several orders of magnitude less azide than required for substrate hydrolysis rescue.
Combining the outcome of fluorescent labeling with ABP 1 and 2 of wild-type and site-directed mutagenized ␤-glucosidases with the sodium azide-mediated rescue of their enzymatic activity and labeling allowed us to identify the acid/base and nucleophile in all human ␤-glucosidases studied: GBA (Glu-235/Glu-340), GBA3 (Glu-165/Glu-373), LPH (pocket III: Glu-1065/Glu-1273), and GBA2 (Asp-677/Glu-527). The cata-lytic residues designated in human GBA2 overlap with the recently identified acid/base and nucleophile in the GBA2-homologue SSO1353 from S. solfataricus (42) and their suggested locations in human GBA2. We noted that the proteins klotho, ␤klotho, and KLPH, although evolved from ␤-glucosidases given the close homology to members of GH1, are not labeled by either ABP, consistent with their lack of ␤-glucosidase activity, due to the absence of either a nucleophile or acid/base residue in their proposed structure.
In conclusion, our study demonstrates the versatility of combined use of cyclophellitol-based ABPs 1 and 2 in the highly sensitive identification of the acid/base and nucleophile residues via direct visualization on slab-gels. Rescue of enzyme activity with sodium azide in combination with selective ABP labeling of the nucleophile with ␤-aziridine 2 in the absence of the acid/base residue may serve as a general approach to discriminate between the catalytic residues of retaining ␤-glucosidases using fluorescence scanning of gels as a read-out. ␤-Epoxide 1 labeling in the presence or absence of sodium azide may be used to validate the results obtained with ␤-aziridine 2 in enzymes that tolerate labeling by this ABP. Broadening of this approach beyond ␤-glucosidases to other classes of retaining glycosidases should be feasible through variation of the cyclophellitol core structure (30).