Mechanism-based Inhibition of Yeast α-Glucosidase and Human Pancreaticα-Amylase by a New Class of Inhibitors

2-Deoxy-2,2-difluoroglycosides are a new class of mechanism-based inhibitors of α-glycosidases, which function via the accumulation of a stable difluoroglycosyl-enzyme intermediate. Two members of this new class of inhibitor have been synthesized and kinetic studies performed with their target glycosidases. Thus 2,4,6-trinitrophenyl 2-deoxy-2,2-difluoro-α-glucoside is shown to inactivate yeast α-glucosidase with a second order rate constant of ki/Ki = 0.25 min−1 mM−1. The equivalent difluoromaltoside inactivates human pancreatic α-amylase with ki/Ki = 0.0073 min−1 mM−1. Competitive inhibitors protect the enzyme against inactivation in each case, showing reaction to occur at the active site. A burst of release of one equivalent of trinitrophenolate observed upon inactivation of human pancreatic α-amylase proves the required 1:1 stoichiometry. These are the first mechanism-based inhibitors of this class to be described, and the first mechanism-based inhibitors of any sort for the medically important α-amylase. In addition to having potential as therapeutics, compounds of this class should prove useful in subsequent structural and mechanistic studies of these enzymes.

Specific inhibitors of glycosidases have proved valuable in a number of applications ranging from mechanistic studies (Legler, 1990;Sinnott, 1990) through their use to study protein glycosylation (Elbein et al., 1984), to possible therapeutic uses such as the control of blood glucose levels via control of the degradation of dietary disaccharides and starch (Truscheit et al., 1981) or control of viral infectivity through interference with normal glycosylation of viral coat proteins (Elbein, 1984;Prasad et al., 1987). A number of naturally occurring reversible glycosidase inhibitors are known such as nojirimycin, castanospermine, swainsonine, and acarbose (Legler, 1990), and these have been subjected to intensive study including the synthesis and testing of a number of analogues. Another class of inhibitors that has been less well studied is that of the covalent, irreversible type, typically affinity labels. These are generally synthetic analogues of sugars containing reactive groups such as epoxides, isothiocyanates and ␣-halocarbonyls as reviewed recently (Legler, 1990;Withers and Aebersold, 1995). Less common are the more selective mechanism-based inhibitors whose efficacy depends upon binding and subsequent enzymatic action to generate a reactive species. These include the conduritol epoxides (Legler, 1968(Legler, , 1970, the quinone methidegenerating glycosides (Halazy et al., 1990;Briggs et al., 1992), and the glycosylmethyl triazenes (Marshall et al., 1980;Sinnott and Smith, 1976). Interestingly, two naturally occurring inhibitors of this class have now been described: the hydroxymethylconduritol epoxide, cyclophellitol (Atsumi et al., 1990;Withers and Umezawa, 1991), isolated from Phellinus sp.; and the putative quinone methide-generating glycoside salicortin, isolated from Salix (Clausen et al., 1990).
An additional, relatively recently described class of mechanism-based inhibitor that has proved successful is that of the 2-deoxy-2-fluoro (Withers and Aebersold, 1995;Withers et al., 1987Withers et al., , 1988Withers et al., , 1990. These function as excellent inactivators of retaining glycosidases (glycosidases that hydrolyze the glycosidic linkage with net retention of anomeric configuration) by formation of a stable glycosyl-enzyme intermediate which turns over to product only very slowly. As shown in Scheme 1 for an ␣-glucosidase, the normal mechanism of action of this class of enzyme involves the formation and hydrolysis of a glycosyl-enzyme intermediate with general acid/base catalytic assistance via transition states with substantial oxocarbenium ion character (Sinnott, 1987(Sinnott, , 1990. A combination of inductive destabilization of these positively charged transition states by the electronegative fluorine at C-2 and loss of crucial hydrogen bonding interactions with the 2-hydroxyl serves to substantially destabilize these two transition states, dramatically slowing both steps. Incorporation of a relatively reactive leaving group such as fluoride or 2,4-dinitrophenolate as aglycone accelerates the first (glycosylation) step sufficiently that the glycosyl-enzyme intermediate is formed, but then hydrolyzes only very slowly, thereby resulting in inactivation.
This strategy has proved very successful with a wide range of retaining ␤-glycosidases, and has allowed the characterization of this intermediate and the identification of the catalytic nucleophiles in a number of enzymes (Withers and Aebersold, 1995). These compounds have also proved effective in vivo, selectively inactivating the expected glycosidases in all organs tested in rats, including the brain (McCarter et al., 1994). However, as noted early on (Withers et al., 1988), and as has been confirmed in subsequent studies (McCarter et al., 1993), this approach has not been successful with ␣-glycosidases, despite the fact that ample evidence exists that equivalent mechanisms are followed in the two cases. 1 This evidence includes the 13 C NMR detection of a glycosyl-enzyme intermediate on an ␣-amylase (Tao et al., 1989), and the denaturation trapping of such an intermediate on a glycosyl transferase, a mechanistically analogous enzyme (Mooser et al., 1991;Mooser, 1992). Instead, the 2-deoxy-2-fluoro-␣-glycosides function as substrates, albeit poor, for the enzymes studied. Thus the lack of inactivation must be a consequence of the fluorine substituent not sufficiently slowing the deglycosylation step relative to glycosylation, and a possible stereoelectronic rationale for this * This work was supported by the Medical Research Council of Canada and the Natural Sciences and Engineering Research Council of Canada. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed.
observation has been proposed (Withers, 1995;Kempton and Withers, 1992). The generation of an effective inhibitor of this class for an ␣-glycosidase therefore requires further slowing of the deglycosylation step. This could be achieved via the incorporation of a second fluorine at C-2 to further inductively destabilize the transition state, in conjunction with the incorporation of a more reactive leaving group to ensure that glycosylation is not rate-limiting. Since the second fluorine is only slightly larger than the hydrogen it replaces, it would likely not result in any significant steric repulsive interactions upon binding. In this paper we describe such an approach which has led to the development of novel mechanism-based inactivators of both yeast ␣-glucosidase and human pancreatic ␣-amylase. This is the first mechanism-based inhibitor described for human pancreatic ␣-amylase, which should prove valuable as a probe of the structure and mechanism of this medically important enzyme. Compounds of this general class also have considerable potential as therapeutics agents, particularly in the control of post-prandial blood glucose levels by inhibition of digestive glycosidases.

MATERIALS AND METHODS
Synthesis-Syntheses of the two inhibitors will be described in detail elsewhere. Characterization data for the final compounds are provided below.
Inactivation Studies-These were performed by incubating each enzyme in its buffer in the presence of inactivator. Aliquots (5-10 l) were removed at time intervals and assayed for activity by addition to a large volume (0.25-1 ml) of substrate. This halts inactivation by dilution of the inactivator and by providing high concentrations of a competitive inhibitor, the substrate. First order rate constants for inactivation at each inhibitor concentration were determined by direct fit of the activity versus time data to a first order curve using GraFit (Leatherbarrow, 1990). Concentrations of inactivator were studied for ␣-amylase with TNPDFM (1.9, 2.5, 3.7, 5.6, 7.5 mM) and ␣-glucosidase with TNPDFG (1.1, 1.6, 2.2, 2.7, 3.2 mM). Protection of yeast ␣-glucosidase against inactivation by TNPDFG (1.6 mM) was monitored by measuring rates of inactivation in the absence and presence of 1-deoxynojirimycin (45 M). Protection of ␣-amylase against inactivation by TNPDFM (5.6 mM) was investigated with acarbose (64 M). A test that inactivation was truly a consequence of reaction with the difluoroglycoside and not with a low level contaminant was carried out. This involved reacting a large amount of enzyme (50 M) with the inactivator (0.9 mM) until completion of reaction, then removing the inactivated enzyme by centrifugal dialysis and re-testing the resultant inactivation mixture as an inactivator of a fresh batch of enzyme. Essentially identical inactivation rates were seen, showing that the inactivation could not have been due to a contaminant as this would have had to be present as more than 5% of the sample (this is the ratio of enzyme/inhibitor in the first reaction). Such levels of contamination would have been detected by the analytical techniques used.
Studies of Reactivation of Inactivated Enzymes-The inactivated enzyme was placed in an Amicon Centricon 30,000-Da molecular mass cutoff centrifugal membrane to remove excess inactivator, diluted into the appropriate buffer, and the concentration of enzyme determined spectrophotometrically. Aliquots of the reactivation mixture (incubated in the absence and presence of 50 mM maltose) were removed over a period of time and assayed.

RESULTS AND DISCUSSION
Synthesis of the required glycone portion posed no great synthetic problem since a method for the synthesis of a 2-deoxy-2,2-difluoroglucose derivative based upon the addition of fluorine to 2-fluoro-D-glucal had been published (McCarter et al., 1993). However, the incorporation of an aglycone of greater leaving group ability than fluoride or dinitrophenolate necessitated more careful synthetic considerations. The most attractive candidate as a leaving group was chloride. However, repeated attempts to synthesize this derivative via displacement chemistry were unsuccessful, with no reaction occurring. This is likely the consequence of the very effect sought, the resistance of 2,2-difluoroglycosides toward nucleophilic displace-ments. An alternative strategy for the installation of a good leaving group was therefore followed, which did not require the displacement reaction. This involved reaction of the protected hemiacetal of the 2-deoxy-2,2-difluoro sugar with fluoro-2,4,6trinitrobenzene, to yield the trinitrophenyl glycoside. Using this approach TNPDFG (1) and TNPDFM (2) were synthesized as shown in Structure 1.
Incubation of both yeast ␣-glucosidase with TNPDFG and human pancreatic ␣-amylase (␣-amylase) with TNPDFM resulted in time-dependent inactivation, as shown for ␣-amylase in Fig. 1a. Inactivation followed pseudo-first order kinetic behavior for at least two half-lives. However some deviation was observed at longer incubation times, which was shown to be due to the spontaneous decomposition of the inhibitor in the incubation mixtures (see below).
Inactivation data were analyzed according to the kinetic scheme shown below (Reaction 1).
Species E, TNP, and DFM represent, respectively, enzyme, 2,4,6-trinitrophenol and 2-deoxy-2,2-difluoro-D-maltose or 2-deoxy-2,2-difluoro-D-glucose. A reciprocal replot of the pseudofirst order rate constants (k obs ) at each inhibitor concentration, taken from slopes of the lines in Fig. 1a versus inhibitor concentration, is shown in Fig. 1b. The slope of this plot yielded a value of k i /K i ϭ 0.0073 min Ϫ1 mM Ϫ1 for reaction of human pancreatic ␣-amylase. Individual values of k i and K i could not be reliably determined because the solubility of the inhibitor (maximum concentration achieved ϭ 11 mM) precluded measurements at concentrations anywhere near the estimated K i value (based upon Fig. 1b) of 90 mM.
The active site-directed nature of the inhibition was tested by measuring the rate of inactivation at a fixed concentration of TNPDFM (5.6 mM) both in the absence and presence of the known competitive inhibitor acarbose. As is shown in Fig. 1c, the rate of inactivation was decreased from 0.044 min Ϫ1 to 0.008 min Ϫ1 by the addition of acarbose, indicating that inactivation is a consequence of modification of the active site. In a second control (not shown) TNP (picric acid) was shown not to result in inactivation of the enzyme, thereby removing any concerns that the true inactivator was the released product.
In order to demonstrate the stoichiometry of inactivation, the reaction of ␣-amylase, with the inactivator was studied by monitoring the release of TNP. ␣-Amylase (10 l, 52 g) in buffer was added to a solution of TNPDFM (10 l, 0.98 mM) in the same buffer, placed quickly into a microcell in a spectrophotometer, and the absorbance at 400 nm monitored. The biphasic time course shown in Fig. 2 was obtained. The first, fast phase corresponds to the release of TNP upon inactivation of the enzyme, and the second, slower phase to the breakdown of the substrate. TNP was released at an equivalent, slow rate from a second mixture to which was added buffer but no enzyme. Back extrapolation of the essentially linear steady state phase and of the initial burst phase to time zero revealed that 26 M TNP was released in the initial burst phase. This corresponds well to the concentration of ␣-amylase in the cell (23 M), indicating a stoichiometry of one TNP released per enzyme molecule. A repeat of this experiment with twice the amount of enzyme resulted, as required, in a burst of twice the magnitude 2,2-Difluoroglycosides as ␣-Glycosidase Inhibitors (data not shown). These data therefore show that inactivation of ␣-amylase is a consequence of the formation of a stable difluoroglycosyl-enzyme intermediate at the active site of the enzyme.
Similar data were also acquired with yeast ␣-glucosidase inactivated by TNPDFG, and kinetic parameters for this process along with other data are shown in Table I. In this case again, a competitive inhibitor, 1-deoxynojirimycin (45 M, K i ϭ 12.6 M), was shown to protect the enzyme against inactivation, reducing the rate of inactivation in the presence of 1.6 mM TNPDFG from 0.13 min Ϫ1 to 0.073 min Ϫ1 .
It is of interest to compare the rate reductions consequent upon introduction of the fluorine substituents at C-2 in the two cases. As can be seen in Table I, introduction of the first fluorine at C-2 reduces the glycosylation rate (from k cat /K m values) some 2500-fold for ␣-amylase but only 80-fold for the ␣-glucosidase. Comparison of parameters for the trinitrophenyl difluoroglycosides with those for the 2-fluoroglycosyl fluorides reveals a 330-fold reduction for ␣-amylase and again an 80-fold reduction for the ␣-glucosidase. This reveals a much greater sensitivity of the ␣-amylase-catalyzed reaction to the introduction of fluorine substituents than that for ␣-glucosidase, indicating either a greater degree of oxocarbenium ion character at the transition state for ␣-amylase, or more important interactions with the 2-hydroxyl. These massive reductions in glycosylation rate constants consequent upon the introduction of two fluorines (almost 10 6 -fold for ␣-amylase) were also reflected in reduced deglycosylation rate constants. No reactivation of either inactivated enzyme was seen when a sample of the inactivated enzyme was dialyzed to remove excess inactivator, then incubated for up to 30 days and aliquots removed for assay.
In summary, a new class of mechanism-based inactivator of human pancreatic ␣-amylase and of yeast ␣-glucosidase has been synthesized and shown to function via the stoichiometric trapping of a covalent glycosyl-enzyme intermediate. In addition to providing powerful evidence of the commonality of mechanisms of ␤and ␣-glycosidases, such inhibitors should prove useful in identifying the active site nucleophiles of these and other glycosidases, and compounds of this class may well prove valuable as therapeutic agents.  2,2-Difluoroglycosides as ␣-Glycosidase Inhibitors 26781