Mechanism of inhibition of the class C β-lactamase of Enterobacter cloacae P99 by cyclic acyl phosph(on)ates: Rescue by return

Article abstract of DOI:10.1021/ja011094v

As previously described (Pratt, R. F.; Hammar, N. J. J. Am. Chem. Soc. 1998, 120, 3004.), 1-hydroxy-4,5-benzo-2,6-dioxaphosphorinone(3)-1-oxide (salicyloyl cyclic phosphate) inactivates the class C β-lactamase of Enterobacter cloacae P99 in a covalent fashion. The inactivated enzyme slowly reverts to the active form. This paper shows that reactivation involves a recyclization reaction that regenerates salicyloyl cyclic phosphate rather than hydrolysis of the covalent intermediate. The inactivation, therefore, is a slowly reversible covalent modification of the active site. The thermodynamic dissociation constant of the inhibitor from the inactivated enzyme is 0.16 μM. Treatment of the inactivated enzyme with alkali does not produce salicylic acid but does, after subsequent acid hydrolysis, yield one molar equivalent of lysinoalanine. This result proves that salicyloyl cyclic phosphate inactivates the enzyme by (slowly reversible) phosphorylation of the active site serine residue. This result contrasts sharply with the behavior of acyclic acyl phosphates which transiently inactivate the P99 β-lactamase by acylation (Li, N.; Pratt, R. F. J. Am. Chem. Soc. 1998, 120, 4264.). This chemoselectivity difference is explored by means of molecular modeling. Rather counterintuitively, in view of the relative susceptibility of phosphates and phosphonates to nucleophilic attack at phosphorus, 1-hydroxy-4,5-benzo-2-oxaphosphorinanone(3)-1-oxide, the phosphonate analogue of salicyloyl cyclic phosphate, did not appear to inactivate the P99 β-lactamase in a time-dependent fashion. It was found, however, to act as a fast reversible inhibitor (K_i = 10 μM). A closer examination of the kinetics of inhibition revealed that both on and off rates (9.8 × 10³ s^-1 M^-1 and 0.098 s^-1, respectively) were much slower than expected for noncovalent binding. This result strongly indicates that the inhibition reaction of the phosphonate also involves phosphylation of the active site. Hence, unlike the situation with bacterial DD-peptidases covalently inactivated by β-lactams, the P99 β-lactamase inactivated by the above cyclic acyl phosph(on)ates can be rescued by return. Elimination of the recyclization reaction would lead to more effective inhibitors.

Full text of DOI:10.1021/ja011094v

10436
J. Am. Chem. Soc. 2001, 123, 10436-10443
Mechanism of Inhibition of the Class C â-Lactamase of Enterobacter
cloacae P99 by Cyclic Acyl Phosph(on)ates: Rescue by Return
Kamaljit Kaur, Martin J. K. Lan, and R. F. Pratt*
Contribution from the Department of Chemistry, Wesleyan UniVersity, Middletown, Connecticut 06459
ReceiVed April 30, 2001
Abstract: As previously described (Pratt, R. F.; Hammar, N. J. J. Am. Chem. Soc. 1998, 120, 3004.), 1-hydroxy-
4,5-benzo-2,6-dioxaphosphorinone(3)-1-oxide (salicyloyl cyclic phosphate) inactivates the class C â-lactamase
of Enterobacter cloacae P99 in a covalent fashion. The inactivated enzyme slowly reverts to the active form.
This paper shows that reactivation involves a recyclization reaction that regenerates salicyloyl cyclic phosphate
rather than hydrolysis of the covalent intermediate. The inactivation, therefore, is a slowly reversible covalent
modification of the active site. The thermodynamic dissociation constant of the inhibitor from the inactivated
enzyme is 0.16 µM. Treatment of the inactivated enzyme with alkali does not produce salicylic acid but does,
after subsequent acid hydrolysis, yield one molar equivalent of lysinoalanine. This result proves that salicyloyl
cyclic phosphate inactivates the enzyme by (slowly reversible) phosphorylation of the active site serine residue.
This result contrasts sharply with the behavior of acyclic acyl phosphates which transiently inactivate the P99
â-lactamase by acylation (Li, N.; Pratt, R. F. J. Am. Chem. Soc. 1998, 120, 4264.). This chemoselectivity
difference is explored by means of molecular modeling. Rather counterintuitively, in view of the relative
susceptibility of phosphates and phosphonates to nucleophilic attack at phosphorus, 1-hydroxy-4,5-benzo-2-
oxaphosphorinanone(3)-1-oxide, the phosphonate analogue of salicyloyl cyclic phosphate, did not appear to
inactivate the P99 â-lactamase in a time-dependent fashion. It was found, however, to act as a fast reversible
inhibitor (Ki ) 10 µM). A closer examination of the kinetics of inhibition revealed that both on and off rates
(9.8 × 103 s-1 M-1 and 0.098 s-1, respectively) were much slower than expected for noncovalent binding.
This result strongly indicates that the inhibition reaction of the phosphonate also involves phosphylation of
the active site. Hence, unlike the situation with bacterial DD-peptidases covalently inactivated by â-lactams,
the P99 â-lactamase inactivated by the above cyclic acyl phosph(on)ates can be rescued by return. Elimination
of the recyclization reaction would lead to more effective inhibitors.
Introduction
include a variety of boronates5 and phosphonates. The latter
molecules have been investigated for some time in this
laboratory.6 They are particularly effective against class C
â-lactamases which are an acute problem at present.7 They act,
in general, by covalent phosphonylation of the nucleophilic
serine of the â-lactamase active site.8,9 The acyl phosph(on)-
ates (1) represent a new group of phosph(on)ate inhibitors,
having activity against both class A and class C â-lacta-
mases.10,11 Depending on the enzyme and the structure of the
The resistance of bacteria to â-lactam antibiotics largely arises
from the production of â-lactamases.1 These defensive enzymes
catalyze the hydrolysis of â-lactams and thus destroy the
antibiotic activity. The existence of â-lactamases was recognized
even before the introduction of these antibiotics into medicine,2
and the search for inhibitors of these enzymes began im-
mediately thereafter.3 Now, some sixty years later, the general
problem of â-lactamases has still not been solved. Although
many inhibitors have been discovered, few have proven suf-
ficiently effective to be introduced into clinical practice as
synergists of existing â-lactams. Many of these inhibitors,
including those presently in clinical practice, are mechanism-
based inhibitors and â-lactams themselves;4 â-lactamases are,
of course, well-known to be able to develop resistance to
â-lactams.
compound, either acylation or phosphylation, or both, of the
enzyme active site has been observed. The acyl derivatives can
slowly hydrolyze to free enzyme and carboxylate, and thus, in
this mode of reaction, compounds 1 are poor but inhibitory
substrates. Phosphylation, although less efficient than acylation
Another group of inhibitors that have been investigated in
recent years and which are not â-lactams are transition state
analogues. These acyclic molecules are actually analogues of
transition states for hydrolysis of acyclic ester substrates. They
(5) Weston, G. C.; Blazquez, J.; Baquero, F.; Shoichet, B. K. J. Med.
Chem. 1998, 41, 4577.
* Corresponding author. Phone: (860) 685-2629. Fax: (860) 685-2211.
E-mail: rpratt@wesleyan.edu.
(1) Medeiros, A. A. Clin. Infect. Dis. 1997, 24 (Suppl. 1), 519.
(2) Abraham, E. P.; Chain, E. Nature 1940, 146, 837.
(3) Kirby, W. M. M. Science 1944, 99, 452.
(6) Pratt, R. F. Science 1989, 246, 917.
(7) Rice, L. B.; Bonomo, R. A. Drug Resist. Updates 2000, 3, 178.
(8) Rahil, J.; Pratt, R. F. Biochemistry 1992, 31, 5869.
(9) Lobkovsky, E.; Billings, E. M.; Moews, P. C.; Rahil, J.; Pratt, R. F.;
Knox, J. R. Biochemistry 1994, 33, 6762.
(4) Pratt, R. F. In The Chemistry of â-Lactams; Page, M. I., Ed.; Chapman
& Hall: London, 1992; pp 229-271.
(10) Li, N.; Pratt, R. F. J. Am. Chem. Soc. 1998, 120, 4264.
(11) Kaur, K.; Pratt, R. F. Biochemistry 2001, 40, 4610.
10.1021/ja011094v CCC: $20.00 © 2001 American Chemical Society
Published on Web 10/05/2001

Mechanism of Inhibition of Class C â-Lactamase
J. Am. Chem. Soc., Vol. 123, No. 43, 2001 10437
with the compounds surveyed to date,10,11 produces a consider-
ably more refractory derivative, leading effectively to irreversible
inactivation of the enzyme. As would be expected on the basis
of chemical reactivity,12,13 phosphylation is more facile when
1 is a phosphonate than when it is a phosphate.
We have also described the reactivity of cyclic acyl phosphate
2 with these enzymes.5 This molecule reacts with representative
class A and C â-lactamases as well as with the DD-peptidase
of Streptomyces R61 to yield inert complexes. We had
anticipated that the breakdown of these complexes to free
enzyme would be slower than that of acyl enzymes derived from
1. This expectation, which was in fact realized, was based on
the (ex post facto) rationale for the effectiveness of penicillin
itself where acylation of transpeptidase active sites leads to an
inert acyl enzyme, partly at least because the thiazolidine leaving
group remains attached to the acyl moiety and obstructs
deacylation. The latter could also be true for an acyl enzyme
(or a phosphoryl enzyme) derived from 2.
Figure 1. Time-dependent changes in fluorescence intensity on mixing
acyl phosphate 2 (0.26 µM) with 0.1 µM (0), 0.2 µM (4), 0.3 µM
(O), 0.4 µM (3), and 0.5 µM (]) concentrations of the P99 â-lactamase.
The solid lines were calculated on the basis of Scheme 2 as described
in the text. The dotted line corresponds to the progress curve for reaction
of 0.26 µM 2 with 0.1 µM enzyme as calculated on the basis of Scheme
1.
Scheme 1
Our original report concerning the discovery of 2 did not
reach a conclusion as to whether the covalent complex of 2
with any particular enzyme was an acyl enzyme, 3, or a
phosphoryl enzyme, 4. The present paper addresses this point.
Scheme 1. The cyclic phosphate, I in this scheme, reacted in a
second-order process ([I] e 15 µM) to yield an intermediate
that slowly broke down, regenerating free enzyme and, implied
but not proven, the hydrolysis product of I. Evidence for the
covalent nature of E-I was provided by the decrease in
fluorescence of 2 on its reaction with the enzyme to form E-I,14
a change also observed on the spontaneous hydrolysis of 2.
Subsequently, we realized that complete turnover did not
accompany the addition of enzyme to 2. This is seen in the
results shown in Figure 1. It appears that the reaction of 2 with
the enzyme stops short of completion, a result seen very clearly
at the lower enzyme concentrations, for example, with 0.1 µM
enzyme in Figure 1. This is not solely a function of stoichi-
ometry and burst kinetics, because, although a decrease in
fluorescence occurs, representing formation of E-I, no subse-
quent decrease in fluorescence is seen which would represent
the expected breakdown of E-I and complete turnover of 2.
Scheme 1 would, in fact, predict the dotted line for reaction of
0.26 µM 2 with 0.1 µM enzyme. This clearly does not fit the
data (the upper trace in Figure 1).
As one approach to this problem, we prepared and examined
the enzyme inhibitory abilities of 5, the phosphonate analogue
of 2. The rationale here was that if 2 phosphylated the enzyme,
5 should do so more rapidly, whereas, if acylation were
involved, the reactivity of 5 should be very similar to or perhaps
less than that of 2. The pattern of reactivity of acyclic analogues
Fluorescence spectra of reaction mixtures at the end of the
reaction period shown in Figure 1 suggested that they contained
unreacted 2 (λex ) 310 nm, λf ) 370 nm). The reaction with
enzyme had therefore stopped prior to complete turnover of 2.
Further evidence that 2 was still present after the fluorescence
change had ceased was achieved by adding a KHCO3/K2CO3
buffer to the reaction mixture (pH 7.5) to achieve a 0.1 M total
carbonate concentration and a pH of 9.5. Figure 2A shows the
initial decrease in fluorescence intensity on reaction of 0.4 µM
enzyme with 0.26 µM 2. On then changing the pH to 9.5, a
further exponential decrease in fluorescence was observed
(Figure 2B) from which a rate constant of 8.4 × 10-4 s-1 was
derived. When 2 alone was treated with the same buffer, a very
similar fluorescence decrease was observed for which a rate
constant of 7.9 × 10-4 s-1 was obtained. These observations
support the proposition that after the reaction of 0.4 µM
â-lactamase with 0.26 µM 2 had apparently reached completion
at pH 7.5, a considerable amount of the original 2 remained in
solution and was not turned over by the enzyme in the time
frame expected from the kr value (Scheme 1).
6 and 7 was also examined for comparison. We conclude that,
in contrast to the acyclic analogues 6 and 7, cyclic compounds
2 and 5 primarily react with the class C â-lactamase of
Enterobacter cloacae P99 by phosphylation of the nucleophilic
active site serine residue. Further, dephosphylation to regenerate
the free enzyme occurs by recyclization to the cyclic phosph-
(on)ate.
Results and Discussion
Reaction Between 2 and the P99 â-Lactamase. Cyclic acyl
phosphate 2 has been shown to inactivate several â-lactam-
recognizing enzymes and, in particular, the class C â-lactamase
of Enterobacter cloacae P99.14 The results from the initial
experiments with this enzyme were interpreted in terms of
(12) Hudson, R. F.; Keay, L. J. Chem. Soc. 1960, 1859.
(13) Hays, H. R. J. Org. Chem. 1971, 36, 98.
(14) Pratt, R. F.; Hammar, N. J. J. Am. Chem. Soc. 1998, 120, 3004.

10438 J. Am. Chem. Soc., Vol. 123, No. 43, 2001
Kaur et al.
Scheme 2
Scheme 3
pseudo-first-order rate constant of 0.23 s-1. On addition of 8 to
the reaction mixture, the fluorescence change shown in Figure
2C was observed. The fluorescence intensity at 400 nm
increased exponentially with a rate constant of 9.4 × 10-4 s-1
.
(The reason for the subsequent slow linear increase in fluores-
cence, which was included as a linear term in the curve fitting,
is not known at present but may represent some denaturation
of the inactivated enzyme.) The emission spectrum at the end
of the exponential phase was again identical to that of 2. These
observations can be interpreted in terms of Scheme 3, where I′
represents 8. According to this scheme, I′ reacts rapidly and
irreversibly with free E, and E-I dissociates in the rate-
determining step to restore equilibrium; eventually, all E-I has
dissociated and all E is inactivated by I′. The observed rate
constant should be kr, because ki′[I′] . ki. The value of kr
obtained by this method is gratifyingly close to that obtained
from previous direct return of activity experiments, viz. 8.2 ×
10-4 s-1 14
. This result therefore provides convincing evidence
for Scheme 2.
The data of Figure 1 were then fitted to Scheme 2 by means
of the Dynafit program.15 The best fit obtained is indicated by
the solid lines in the figure. The values of ki and kr thus obtained
were 7350 s-1 M-1 and 1.14 × 10-3 s-1, respectively, in general
agreement with previous estimates,14 although slightly different
because the assumed reaction scheme is different. The thermo-
dynamic dissociation constant of E-I (Ki ) kr/ki) is, thus, 0.16
µM.
Chemical Nature of the Covalent Complex between 2 and
the P99 â-Lactamase. The covalent complex E-I may
represent either acyl enzyme 11 or a phosphoryl enzyme
(Scheme 4). Two types of phosphoryl enzyme, 9 and 10, may
be envisaged. Certainly, the carboxylate (from attack at phos-
phorus, upper path) would be expected to be a better leaving
group than the aryloxide (from attack at phosphorus, lower path),
and thus 9 the more likely phosphoryl enzyme, but the direction
of ring opening could be dictated by the active site.
Distinction between these possibilities was achieved by
treatment of E-I with sodium hydroxide, as described in the
Experimental Section. Acyl phosphate 10 would be expected
to yield salicylate directly. The HPLC analysis of the reaction
products gave no evidence of any salicylate on treatment of
E-I with hydroxide. Both 9 and 11 should give salicyl
phosphate 12, but by different routes. Alkaline hydrolysis of
11 should regenerate the active site serine residue after acyl
oxygen fission. In contrast, in alkaline solution, the facile route
to salicyl phosphate from 9 would be by way of elimination,
yielding a dehydroalanine residue on the protein as the other
product.8 As previous work in this laboratory has demonstrated,
generation of dehydroalanine at the P99 â-lactamase active site
by this type of elimination of phosph(on)ate leads to generation
of lysinoalanine.8 This product presumably arises by facile
addition of the adjacent Lys-67 amine group to dehydroalanine
derived from Ser-64 by elimination. Amino acid analysis of the
hydroxide-treated E-I revealed 1 mol of lysinoalanine per mole
Figure 2. Fluorescence intensity changes: A, on reaction of 2 (0.26
µM) with the P99 â-lactamase (0.4 µM); B, on adjustment of the pH
to 9.5 after the reaction in A; C, on addition of the phosphonate 8 (100
µM) to the reaction mixture after the reaction in A.
The implication of these results is that the reaction of E with
I proceeds, not to complete hydrolysis of I, but to a position of
equilibrium (Scheme 2). In this reaction scheme, kr corresponds
not to turnover but to reformation of I from E-I.
To provide further evidence for this revised view of the
interaction of 2 with the enzyme, the following experiment was
conducted. Enzyme (0.4 µM) and 2 (0.26 µM) were mixed in
a cuvette and the reaction monitored by fluorescence until no
further change occurred (Figure 2A). To this reaction mixture,
sodium m-nitrophenyl N-(phenylacetyl)aminomethyl phospho-
nate 8 (100 µM) was added. This is an irreversible, covalent
inhibitor of the P99 â-lactamase which phosphonylates the active
site serine residue.8,9 At a concentration of 100 µM, this
phosphonate would inactivate the P99 â-lactamase with a
(15) Kuzmic, P. Anal. Biochem. 1996, 237, 260.

Mechanism of Inhibition of Class C â-Lactamase
J. Am. Chem. Soc., Vol. 123, No. 43, 2001 10439
Scheme 4
of protein (1.20 and 1.14 mol on the basis of 44 alanine and 29
leucine residues, respectively). The enzyme alone, treated with
hydroxide in the same way, yielded no lysinoalanine. This result
strongly suggests that E-I is phosphoryl enzyme 9 and that
the active site serine hydroxyl is phosphorylated. Although it
is possible that 10 or 11 could rearrange into 9 on hydroxide
treatment, the simpler interpretation seems more likely.
Kinetics and Mechanism of Reaction of the Phosphonate
5 with the P99 â-Lactamase. The phosphonate analogue of 2
was prepared, as described in the Introduction, to probe further
into the inhibition mechanism and, perhaps, to obtain a more
effective inhibitor. The synthesis of 5 was straightforward, as
described in the Experimental Section (Scheme 8). The hy-
drolysis of 5 could be followed spectrophotometrically at 286
nm (ꢀ ) 1410 cm-1 M-1; the hydrolysis product absorbed
maximally at 271 nm, ꢀ ) 600 cm-1 M-1). The pseudo-first-
order rate constant of its disappearance in 50 mM sodium
hydroxide was 0.058 s-1. Under the same conditions, the rate
constant for hydrolysis of 2 (monitored at the same wavelength)
was 0.21 s-1. This difference, favoring 2, suggests nucleophilic
attack by hydroxide ion at the carbonyl group and C-O cleavage
(the phosphate of 2 would be a better leaving group -lower
conjugate acid pKa- than the phosphonate of 5; attack at
phosphorus should favor phosphonate 5 over phosphate 213,14,16).
This conclusion was supported by the finding that the products
obtained by addition of 2 to a mixture of 0.1 M sodium
hydroxide and methanol (1:4) had 1H NMR spectra containing
a singlet peak at 3.52 ppm, which disappeared on brief heating
in aqueous hydroxide ion. A similar peak arose and fell during
the reaction of 5. This strongly suggests a methyl carboxylate
product (13, Scheme 5) in each case, arising from nucleophilic
attack at the carbonyl group, rather than methyl phosph(on)ate
14 which should be more stable and give rise to a methyl
doublet, coupled to phosphorus, in the NMR spectrum. Thus,
alkaline solvolysis of both 2 and 5 appears to involve C-O
cleavage. This result is in accord with precedent for acyclic acyl
aryl phosphates.17
Scheme 5
phosphate buffer, pH 7.5), is considerably greater than that for
2 (2.11 × 10-5 s-1). This is closer to the result that would be
expected for phosphyl attack and P-O bond cleavage.13,14,16
Solvolysis of 2 and 5 in a mixture of 20 mM phosphate at pH
7.5 and methanol (1/1) yielded different results. The product
mixture from 2 contained both hydrolysis and carbonyl metha-
nolysis products, the latter (13, X ) O) distinguished by a singlet
at 3.91 ppm which disappeared on heating the solution in the
NMR tube. In contrast, the solvolysis mixture from 5 initially
contained two products in roughly equal amounts. One, distin-
guished by a singlet at 3.93 ppm and a doublet at 3.47 ppm (J
) 21.5 Hz), is likely to be the carbonyl methanolysis product
(13, X ) CH2), while the second, characterized by doublets at
3.47 ppm (J ) 10.4 Hz) and 3.36 ppm (J ) 21.2 Hz), probably
corresponds to the phosphonyl transfer product, 14, X ) CH2.
Both of these compounds slowly hydrolyzed to a common
product, characterized by a doublet at 3.17 ppm (J ) 20.8 Hz),
characteristic of the complete hydrolysis product, 15. Although
it is not known at present whether the phosphyl transfer product
from 5 arises from attack by water alone or from buffer
catalyzed hydrolysis, there is clear evidence that nucleophilic
attack at the phosphonyl group of 5 may be as facile as at the
carbonyl center under certain circumstances. It is likely, for
example, that neutral nucleophiles (e.g., water) may show a
greater preference than anionic nucleophiles (e.g., hydroxide
ion) for attack at the anionic phosphonyl center.
In very sharp distinction to the results with 2,14 the phos-
phonate analogue 5, when incubated with the P99 â-lactamase,
did not cause inactivation of the enzyme. At first, this seemed
very surprising. The structural similarity between 2 and 5 made
it seem unlikely that 5 would not react with the enzyme at all.
In contrast to the above result, the rate constant for hydrolysis
of 5 at around neutral pH, viz. 1.59 × 10-4 s-1 (20 mM
(16) Rahil, J.; Pratt, R. F. Biochem. J. 1993, 296, 389.
(17) DiSabato, G.; Jencks, W. P. J. Am. Chem. Soc. 1961, 83, 4400.

10440 J. Am. Chem. Soc., Vol. 123, No. 43, 2001
Kaur et al.
Scheme 6
then be combined with the Ki value above to give the rate
constant ki for association of the enzyme and 5 (Scheme 2); a
value of 9.8 × 103 s-1 M-1 for ki was thus obtained. The
magnitude of these rate constants is certainly more suggestive
of a covalent interaction than a noncovalent interaction between
enzyme and inhibitor.
Acyclic Analogues. Compounds 6 and 7 represent, to some
degree, acyclic analogues of 2 and 5, respectively. They appear
to interact differently, however, in a quantitative sense at least,
to their cyclic counterparts. Compound 6 is a substrate of the
P99 â-lactamase by an acyl-transfer reaction where a transiently
stable benzoyl enzyme intermediate is generated.10,11 Phospho-
nate 7, synthesized as described in the Experimental Section,
turned out, like other acyclic acyl phosphonates,10,11 to also be
a substrate, although poorer than 6 (kcat/Km values of 6 and 7
are 6070 s-1 M-1 and 226 s-1 M-1), but forming the same
benzoyl enzyme intermediate (kcat ) 0.011 s-1). Like other
acyclic phosphonates, 7 also slowly and irreversibly inactivated
the enzyme, presumably by phosphonylation11 (ki ) 22 s-1
M-1). Acylation by 6 is thought to be facile because of favorable
interaction of the negatively charged phosphate leaving group
with the electropositive active site.9,11 Phosphylation of the
active site by 2 and 5 is clearly much more facile than by 6 and
7. Possible reasons for this are discussed below. It is interesting
to note, however, that the second-order rate constant for enzyme
acylation by 6 is very similar to that of phosphylation by 2.
Figure 3. Double reciprocal plot for inhibition of the P99 â-lactamase-
catalyzed turnover of nitrocefin by acyl phosphonate 5 at concentrations
of 0 µM (]), 3 µM (4), 8 µM (O), 15 µM (3), and 25 µM (0). The
lines represent the fit to the data of a competitive inhibition model as
described in the text.
Figure 4. Return of â-lactamase activity on mixing a mixture of
enzyme (1.5 µM) and 5 (20 µM) with the substrate cephalothin (300
µM). The absorbance of cephalothin at 284 nm as a function of time
is shown. The change in slope with time reflects the increase in enzyme
activity. The line was calculated on the basis of Scheme 6 as described
in the text.
Molecular Modeling
Computational models for the tetrahedral intermediate of
acylation 16 and the pentacoordinated intermediate of phos-
phorylation 17 of the P99 â-lactamase by 2 were constructed
as described in the Experimental Section. These were studied
If the enzyme were to attack at phosphorus, as the emerging
results with 2 suggested would be likely, the rationale originally
employed in the design of 5 suggested that it should be a more
effective inhibitor than 2. The enzyme did not appear to catalyze
hydrolysis of 5 either.
This dilemma was resolved when it was found that 5, although
not a slowly reversible inhibitor as was 2, behaved, under
manual-mixing steady-state conditions, as a fast reversible
competitive inhibitor (Figure 3), with a Ki of (10.1 ( 0.6) µM.
In view of all precedents, this would represent surprisingly tight
noncovalent binding for a molecule of structure 5. The unlikeli-
hood of this explanation was confirmed by experimental
measurements of initial rates of cephalothin turnover by the P99
â-lactamase in the presence of 2 (0-100 µM; see Experimental
Section for details) where no indication of inhibition was
observed. It seemed unlikely that 5 but not 2 would bind
noncovalently at micromolar concentrations. Consequently, the
possibility of fast, reversible covalent binding by 5 was
considered. Evidence for this was obtained from a stopped-flow
experiment when a time-dependent return of activity of the
â-lactamase against cephalothin was observed on mixing a
combination of the enzyme and 5 with cephalothin (Figure 4).
Interpretation of these data in terms of Scheme 6 yielded the
value of the rate constant, kr, for dissociation of the complex of
5 with the â-lactamase. The value obtained, 0.098 s-1, could
as approximations to the transition states of the acylation and
phosphorylation reactions, respectively. Molecular dynamics and
several starting conformations were employed to find the most
likely structures, as also described in the Experimental Section.
Calculation of the Eint parameter18 for a variety of energy-
minimized structures led to the structures of Figure 5 as the
most favorable intermediates for the two reactions. Both appear
to involve strong interactions with the active site functional
groups of the enzyme. For example, the acylation tetrahedral
intermediate (Figure 5A) displays nominal hydrogen bonding
(heavy atom distances < 3 Å) between Tyr-150 Oê and Ser-64
Oγ, Lys-67 Nꢀ and Ser-64 Oγ, and Lys-67 Nꢀ and a phosphoryl
O-. The Asn-152 side chain also seems well-placed to hydrogen
bond to a phosphoryl oxygen, although the N-O distance is
3.17 Å in the structure shown. The tetrahedral oxyanion is firmly
held in the oxyanion hole (backbone N-H moieties of Ser-64
(18) Curley, K.; Pratt, R. F. J. Am. Chem. Soc. 1997, 119, 1529.

Mechanism of Inhibition of Class C â-Lactamase
J. Am. Chem. Soc., Vol. 123, No. 43, 2001 10441
Figure 5. Energy-minimized model structures of an acylation tetrahedral intermediate (A) and a pentacoordinated phosphorylation intermediate
(B) which may occur on interaction of the acyl phosphate 2 with the P99 â-lactamase.
Scheme 7
and Ser-318). The intermediate thus appears to be well-stabilized
by the enzyme, and activation of the leaving group would be
provided by its interaction with Lys-67.
Similarly, the pentacoordinated intermediate of phosphor-
ylation (Figure 5B) appears to interact closely with the enzyme
active site. There appear to be hydrogen bonds between Lys-
67 Nꢀ and Ser-64 Oγ and from Tyr-150 Oê to the phosphoranyl
POH; the phosphoranyl PO- is accommodated in the oxyanion
hole. No functional group appears to interact with the leaving
Summarizing Discussion
group, but a carboxylate, directed as it is, out of the active site
The evidence described above strongly suggests that the P99
into solution, should not require general acid catalysis to depart.
â-lactamase does not catalyze the hydrolysis of cyclic phosph-
(on)ates 2 and 5 but is reversibly phosphylated, and thus
inhibited, by them. Quantitatively, the results are summarized
by Scheme 7. These results show that 5 does phosphylate the
It is not possible to choose between 16 and 17 on the basis
of the structures in Figure 5 or the semiempirical calculations
leading to them. Experiment, as discussed, strongly supports
17. It is possible, however, that this result is obtained by default
enzyme faster than 2, but not by a significant amount. Given
in the sense that it is only the phosphorylation product that is
that the reaction does involve enzyme phosphylation, this result
may reflect a weaker productive noncovalent binding of 5 to
inert enough to be detected. It is still possible, in view of the
results from the solvolysis experiments where acylation appeared
the enzyme prior to reaction. The greater reactivity of the
to be the dominant reaction of 2, and of the apparently well-
phosphonate is seen, however, in kr which, if enzyme phos-
stabilized structure in Figure 5A, that acylation of the P99
phylation is involved, also describes a phosphyl transfer reaction
â-lactamase by 2 is very facile. If, for example, ki for acylation
were 10-fold larger than for phosphorylation, say 7 × 104 s-1
M-1 (cf. Scheme 6) and kr were greater than 7 s-1, which would
preclude its detection in manual mixing experiments, then Ki
for 2 as a fast, covalent, acylation reagent would be >100 µM.
The latter number is in accord with the upper limit established
by our stopped-flow experiment (see above). A rapid intramo-
lecular deacylation reaction would certainly be in accord with
the structure of Figure 5A. The phosphate leaving group of the
acylation reaction is directed down into the protein and is likely
to be held there in position for facile deacylation by the Lys-67
ammonium ion. Even if this scenario were true, however, the
enzyme would be predominantly phosphorylated rather than
acylated when both reactions had reached equilibrium.
(18). The rather counterintuitive result of the relative sizes of
these rate constants, therefore, is that the phosphonate is a poorer
inhibitor of the enzyme, both kinetically (it forms a more labile
complex) and thermodynamically (Ki).
The reversible phosphylation of the P99 â-lactamase by 2
and 5 is reminiscent of the reversible formation of the cyclic

10442 J. Am. Chem. Soc., Vol. 123, No. 43, 2001
Kaur et al.
phosphate intermediate in ribonuclease catalysis.19 Analogies
among enzyme inhibitors are the reversible inhibition of
chymotrypsin20 and papain21 by cyclic sulfonates and of
chymotrypsin and thrombin by various lactones.22,23 The â-lac-
tams themselves supply a counterexample, however. They
covalently inhibit bacterial DD-peptidase by acylation of the
active site serine to give the inert complex, 19. This slowly
deacylates by hydrolysis or fragmentation,24 but there is no
evidence of recyclization. This may be due, in part, to the high
thermodynamic barrier to ring closure of bicyclic â-lactams,25
but it may also, in some cases at least, reflect other factors. For
Scheme 8
(Porton Down, Wiltshire, U.K.) and used as received. Chemical reagents
for synthesis were generally purchased from Aldrich Chemical Co.
Lysinoalanine was purchased from Sigma Chemical Co. Salicyloyl
cyclic phosphate 2 and its acyclic analogue 6 were synthesized
previously in this laboratory.10,14
1-Hydroxy-4,5-benzo-2-oxaphosphorinanone(3)-1-oxide Dicyclo-
hexylammonium Salt (5). This compound was prepared by a route
beginning with ethyl o-toluate (Scheme 8). Bromination of this material
with N-bromosuccinimide gave substituted benzylbromide 20. This was
reacted with triethyl phosphite to give phosphonate diester 21 which
was hydrolyzed to diacid 22. The synthesis of 22 was largely as
described by Gasuti et al.29 The acid was cyclized by azeotropic
dehydration to give cyclic acyl phosphonate 5, which was isolated as
the dicyclohexylammonium salt.
example, on acylation of the Streptomyces R61 DD-peptidase
by cephalosporins, rotation around the C6-C7 bond of the
â-lactam occurs (19), presumably to relieve steric stress between
the â-substituents on C6 and C7.26 Recyclization would not
directly be possible from this acyl enzyme conformation. On
the other hand, an acyl enzyme generated from the reaction of
penicillin binding protein 2x of Streptococcus pneumoniae with
cefuroxime appears to be well-placed for recyclization27 (al-
though it should be noted that this structure, like the R61
examples, does not represent the initial acyl enzyme but that of
a subsequent complex where the 3′-leaving group has been
eliminated and a ∆1-dihydrothiazine ring formed). The phosphyl
enzyme intermediates generated from 2 and 5 (18) apparently
retain access to a conformation where recyclization is possible.
A more effective inhibitor than 2 might include functionality
to promote conformational relaxation after phosphylation that
would better hinder or prevent recyclization and regeneration
of the enzyme.
Ethyl o-toluate (20 g, 122 mmol) and N-bromosuccinimide (21.7 g,
122 mmol) were mixed with 81 mL of carbon tetrachloride. Benzoyl
peroxide (20 mg) was added as a catalyst. The flask was irradiated
with visible light from a 150 W tungsten lamp, while the mixture was
heated under reflux for 3 h. At that time, the reaction mixture was
cooled to room temperature. The precipitated solid was removed by
vacuum filtration, and the remaining solvent was removed by rotary
evaporation to give the brominated product 20 as a deep yellow liquid
1
(33.1 g). H (CDCl3) δ 7.90 (d, J ) 7.5 Hz, 1H), 7.49 (t, J ) 7.5 Hz,
1H), 7.48 (t, J ) 7.5 Hz, 1H), 7.40 (d, J ) 7.5 Hz, 1H), 4.99 (s, 2H),
4.42 (q, J ) 7.1 Hz, 2H), 1.44 (t, J ) 7.1 Hz, 3H). This material was
used without further purification.
Acyl phosph(on)ates 2 and 5 thus inhibit the P99 â-lactamase
most likely by phosphylation of the active site. The enzyme is
unable to catalyze the hydrolysis of these species to restore the
enzyme, but the latter can be achieved by a recyclization reaction
which restores the cyclic acyl phosph(on)ates: rescue by return.
These compounds therefore function as reversible inhibitors.
Experiments to expand the kinetic and thermodynamic repertoire
of these molecules in the direction of more effective inhibitors
are in progress. There is evidence that even 2 does enhance
â-lactam activity in in Vitro microbiological assays.28
Ester 20 was heated under reflux with an equimolar amount of
triethyl phosphite at 160 °C for 3 h. The reaction mixture was cooled
overnight and then vacuum distilled to give phosphonate ester 21 (bp
170-174 °C, 0.1 Torr) as a clear yellow liquid in 82% yield (two steps).
1H NMR (CDCl3) δ 7.91 (d, J ) 7.7 Hz, 1H), 7.44 (d, J ) 7.7 Hz,
1H), 7.41 (t, J ) 7.7 Hz, 1H), 7.30 (t, J ) 7.7 Hz, 1H), 4.39 (q, J )
7.6 Hz, 2H), 4.00 (quint, J ) 7.6 Hz, 4H), 3.82 (d, J ) 23 Hz, 2H),
1.41 (t, J ) 7.6 Hz, 3H), 1.21 (t, J ) 7.6 Hz, 6H).
The phosphonate 21 was added to 500 mL of 6 M HCl and the
mixture heated under reflux for 3 days. The HCl/H2O was removed by
rotary evaporation, and the white solid thus obtained was recrystallized
from water in 62% yield. The product diacid 22 exhibited a broad
Experimental Section
1
melting point of 194-202 °C (lit.29 mp 177 °C). H (D2O) δ 7.83 (d,
Materials. The â-lactamase of Enterobacter cloacae P99 was
purchased from the Centre for Applied Microbiology and Research
J ) 8.1 Hz, 1H), 7.56 (t, J ) 8.1 Hz, 1H), 7.41 (m, 2H), 3.57 (d, J )
21.9 Hz, 2H). νmax (KBr) 1659s (CdO), 1271s (PdO).
Solid acid 22 (200 mg) was heated under reflux in 50 mL xylene in
a Dean and Stark apparatus. The solid dissolved in boiling xylene, and
drops of water were noted in the collection arm of the apparatus after
5 h, signifying cyclization. At this stage, an equimolar amount of
dicyclohexylamine was added to the reaction mixture. A colorless salt
precipitated as the solution was cooled. The solid was recrystallized
from cyclohexane/benzene, yielding the amine salt of pure cyclized
(19) Eftink, M. R.; Biltonen, R. L. In Hydrolytic Enzymes; Neuberger,
A., Brocklehurst, K., Eds.; Elsevier: Amsterdam, 1987; Chapter 7.
(20) Heidema, J. H.; Kaiser, E. T. J. Am. Chem. Soc. 1970, 92, 6050.
(21) Campbell, P.; Kaiser, E. T. Biochem. Biophys. Res. Commun. 1972,
48, 866.
(22) Hedstrom, L.; Moorman, A. R.; Dobbs, J.; Abeles, R. H. Biochem-
istry 1984, 23, 1753.
(23) Weir, M. P.; Bethell, S. S.; Cleasby, A.; Campbell, C. J.; Dennis,
R. J.; Dix, C. J.; Finch, H.; Harren, J.; Mooney, C. J.; Patel, S.; Chi-Man,
T.; Ward, M.; Wonacott, A. J.; Wharton, C. W. Biochemistry 1998, 37,
6645.
(24) Fre`re, J.-M.; Joris, B. Crit. ReV. Microbiol. 1985, 11, 299.
(25) Page, M. I. AdV. Phys. Org. Chem. 1987, 23, 165.
(26) Kuzin, A. P.; Liu, H.; Kelly, J. A.; Knox, J. R. Biochemistry 1995,
34, 9532.
1
product 5. Melting point 163-165 °C. H NMR (CDCl3) δ 9.06 (br,
2H), 8.21 (d, J ) 8 Hz, 1H), 7.51 (t, J ) 8 Hz, 1H), 7.38 (t, J ) 8 Hz,
1H), 7.28 (d, J ) 8 Hz, 1H), 3.29 (d, J ) 18 Hz, 2H), 2.82 (br, 2H),
1.1-2.0 (m, 20H). 31P NMR (2H2O) δ 10.52 (t). νmax (KBr) 1713s
(CdO), 1251s (PdO). Anal. Calcd for C20H29NO4P: C, 63.31; H, 7.97;
N, 3.69; P, 8.16. Found: C, 63.40; H, 8.08; N, 3.64; P, 8.80. ESMS
(H2O) m/z 560.9 (M + 2Cl2H23NH+) 560.8.
(27) Gordon, E.; Mouz, N.; Due´e, E.; Dideberg, O. J. Mol. Biol. 2000,
299, 477.
(28) Besterman, J. M.; Rahil, J.; Pratt, R. F. PCT Int. Appl.; Chem. Abstr.
1999, 131, 82947.
(29) Garuti, L.; Farranti, A.; Roberti, M.; Katz, E.; Budriesi, R.; Chiarini,
A. Pharmazie 1992, 47, 295.

Mechanism of Inhibition of Class C â-Lactamase
J. Am. Chem. Soc., Vol. 123, No. 43, 2001 10443
Sodium Benzoyl Benzylphosphonate (7). This compound was
prepared by a general method described previously,2 beginning with
the silver salt of benzylphosphonic acid and benzoyl chloride in a 1:2
molar ratio. The diester thus obtained was hydrolyzed to give the desired
sodium salt of the monoester, which was recrystallized from water. 1H
NMR (D2O) δ 7.98 (d, J ) 7.7 Hz, 2H), 7.69 (t, J ) 7.4 Hz, 1H), 7.52
Scheme 9
(t, J ) 7.7 Hz, 1H), 7.23-7.33 (m, 5H), 3.39 (d, J ) 21 Hz, 2H). 31
P
To obtain the inactivation rate constant (ki) and turnover number,
cat, the absorbance of 7 (50 and 100 µM, in separate experiments)
(2H2O) δ 18.73. νmax (KBr) 1701s (CdO), 1286s (PdO).
k
with time in the presence of the enzyme (1.07 µM) was monitored at
240 nm. The data were analyzed by means of the Dynafit program
using Scheme 9 as described previously.11 In the fitting of these progress
curves, Km was fixed at the value obtained as described above.
Characterization of the â-Lactamase-Inhibitor (2) Complex.
HPLC Analysis. Product analysis of the enzyme-inhibitor (2) complex
after base hydrolysis was obtained from a HPLC method using a Rainin
Rabbit-HPX solvent delivering system, a Machery-Nagel Nucleosil
C18 column, and a Gilson HM/HPLC UV-vis detector. The mobile
phase was 0.05 M acetonitrile, adjusted to pH 3.5 with phosphoric acid.
P99 â-lactamase (20 µM) was partly inactivated by incubation with
2 (10 µM) for 30 min at room temperature. Excess â-lactamase was
used to ensure that essentially all of 2 was converted to the enzyme-
inhibitor complex. The sample was then treated with NaOH (final
concentration 0.2 M) and incubated for another 90 min. It was then
filtered through an ultracentrifuge tube (Ultrafree-MC filters 5000
NMWL from Millipore). The filtrate (15 µL aliquots) was then injected
into the HPLC column at a flow rate of 1.5 mL/min with the detector
set at 232 nm. Authentic salicyclic acid was used as a standard. Thus,
salicyclic acid (10 µM) was treated the same as 2 before passing through
the HPLC column; that is, it was treated with NaOH and filtered through
an ultracentrifuge tube. The retention time of salicylic acid was 7.9
min. Also, 15 µL of 2 (10 µM) alone was injected into the column as
a control.
Lysinoalanine Analysis. The P99 â-lactamase (10 µM) was
inactivated by incubation with 2 (20 µM) for 10 min at room
temperature. Excess of 2 was used to ensure that essentially all of the
enzyme was converted to the enzyme-inhibitor complex. The enzyme,
diluted with MOPS buffer to 10 µM final concentration, was used as
a control. The solutions were then made 0.2 M in NaOH and held at
room temperature for 1.5 h. Each of these was then dialyzed separately
against 0.01 M HCl for 24 h (two changes) to remove excess of 2.
Aliquots (100 µL) were then freeze-dried in acid washed tubes and
submitted for amino acid analysis to the Yale Protein and Nucleic Acid
Chemistry Facility. Standard solutions of lysinoalanine were prepared
for quantitation. Lysinoalanine was identified by its retention time (41.7
min) on the amino acid analysis column.
Analytical and Kinetic Methods. The concentration of enzyme in
stock solutions was determined spectrophotometrically.11 All enzyme
kinetics experiments were performed at 25 °C in 20 mM 3-(N-
morpholino)propanesulfonic acid (MOPS) buffer, pH 7.5.
Kinetics of Reaction of the P99 â-Lactamase with 2. Fluorescence
measurements were obtained from a FluoroMax-2 (Instruments S.A.,
Inc.) spectrofluorimeter. Samples of salicyloyl cyclic phosphate 2 were
excited at 310 nm, and the emission was monitored between 340 and
450 nm, where the maximum was at 370 nm. After addition of the
P99 â-lactamase, emission intensity at 400 nm was followed as function
of time. The data from these experiments were fitted to Scheme 2 using
the program Dynafit.15
Initial velocities of cephalothin (200 µM) hydrolysis by the P99
â-lactamase (40 nM) in the presence of several concentrations (0-
100 µM) of 2 were measured with a Durrum D-110 stopped-flow
spectrophotometer at 278 nm over 2 s. This experiment was designed
to detect any fast noncovalent binding of the enzyme to 2 at these
concentrations without the complication of the covalent reaction.
Kinetics of Reaction of the P99 â-Lactamase with 5. The inhibition
constant (Ki) for 5 as an inhibitor of the P99 â-lactamase was obtained
from spectrophotometric measurements of initial velocities of nitrocefin
(at several concentrations: 40, 50, 60, 80, 160, and 280 µM) hydrolysis
in the presence of several concentrations of 5 (0-25 µM). Nitrocefin
hydrolysis was monitored by the absorbance of product formation at
482 nm. Initial velocities of nitrocefin hydrolysis were fitted to a
competitive inhibition model by the method of Cleland.30
The rate of breakdown of the complex of the enzyme with 5 to restore
free enzyme was obtained from the following experiment. Return of
activity of the â-lactamase against cephalothin was monitored spec-
trophotometrically at 284 nm. Thus, a solution of 5 (20 µM and 40
µM, in separate experiments) and â-lactamase (1.5 µM) was mixed
with cephalothin (300 µM), and the decrease in absorbance due to
cephalothin hydrolysis was measured for 5 s in the stopped-flow
spectrophotometer. A rate constant, kr, for the return of activity against
cephalothin, as the complex dissociated, was obtained by fitting the
experimental data to Scheme 6 using the program Dynafit as described
previously.11,15 In this Scheme, S represents cephalothin, EI the
enzyme/5 complex, and P and P′ represent the respective products of
cephalothin hydrolysis and dissociation of I from EI. A control
experiment where cephalothin (300 µM) hydrolysis in the presence of
the â-lactamase (1.5 µM) alone was performed gave kcat and Km values
for cephalothin of 780 s-1 and 20 µM, respectively.
Molecular Modeling. The computations were set up essentially as
previously described11,18 and run on an SGI Octane 2 computer with
Insight II 2000 (MSI, San Diego, CA). The starting point for the
simulations was the crystal structure of the P99 â-lactamase with a
phosphonate inhibitor covalently attached to the active site serine residue
(PDB file 1bls9). A variety of initial conformations were explored by
means of molecular dynamics, and the significantly populated confor-
mations arising were selected and subjected to energy minimization.
The conformations shown in Figure 5 were the lowest energy carbonyl
and phosphoryl adducts found, where the energies compared were Eint
values18 which take into account interactions between ligand and protein,
residue-residue interactions, and internal residue energies.
Kinetics of Reaction of 7 with the P99 â-Lactamase. Kinetic
parameters for interaction of the â-lactamase with 7 were determined
as described previously for similar acyclic acyl phosphonates.11 Thus,
Km was obtained as the inhibition constant (Ki) from an experiment
where 7 was used to inhibit turnover of a good substrate (cephalothin).
Initial velocities of cephalothin (200 µM, Km ) 20 µM, monitored
spectrophotometrically at 278 nm) hydrolysis in the presence of several
concentrations of 7 (0-200 µM) were fitted to a competitive inhibition
model by the method of Cleland.30
Acknowledgment. This research was supported by the
National Institutes of Health.
(30) Cleland, W. W. Methods Enzymol. 1979, 63, 103.
JA011094V