Background
The overuse of antibiotics in the clinic and for
agricultural uses has resulted in a tremendous selective
pressure for antibiotic resistant bacteria. These bacteria
become resistant by a number of mechanisms, such as
producing enzymes that hydrolyze or inactivate the
antibiotics, producing efflux pumps that transport the
antibiotic out of the cell, or modifying their cell wall
components so they no longer bind effectively to the
antibiotics [ 1 2 3 ] . The most common, least expensive,
and effective antibiotics are the β-lactam containing
antibiotics, such as the penicillins, cephalosporins, and
carbapenems [ 2 4 5 ] . These antibiotics are
mechanism-based inhibitors of transpeptidase, a bacterial
enzyme required for the production of a strong viable cell
wall [ 6 7 ] . In response to the widespread use of
β-lactam containing antibiotics, bacteria have acquired the
ability to produce β-lactamases, which are enzymes that
hydrolyze and inactivate β-lactam containing antibiotics.
There are over 300 distinct β-lactamases known, and these
enzymes have been grouped by a number of classification
schemes [ 8 9 10 11 12 13 14 15 ] . For example, Bush has
developed a scheme, based on the enzymes' molecular
properties, that has four distinct β-lactamase groups [ 10
15 ] . One of the more alarming groups are the Bush group 3
enzymes, which are Zn(II) dependent enzymes that hydrolyze
nearly all known β-lactam containing antibiotics and for
which there are no or very few known clinical inhibitors [
9 14 16 17 18 19 ] . The metallo-β-lactamases have been
further divided by Bush into subgroups based on amino acid
sequence identity: the Ba enzymes share a >23% sequence
identity, require 2 Zn(II) ions for full activity, prefer
penicillins and cephalosporins as substrates, and are
represented by metallo-β-lactamase CcrA from
Bacteroides fragilis, the Bb enzymes
share a 11% sequence identity with the Ba enzymes, require
only 1 Zn(II) ion for full activity, prefer carbapenems as
substrates, and are represented by the metallo-β-lactamase
imiS from
Aeromonas sobria, and the Bc enzymes
have only 9 conserved residues with the other
metallo-β-lactamases, require 2 Zn(II) ions for activity,
contain a different metal binding motif than the other
metallo-β-lactamases, prefer penicillins as substrates, and
are represented by the metallo-β-lactamase L1 from
Stenotrophomonas maltophilia [ 9 ] .
A similar grouping scheme (B1, B2, and B3) based on
structural properties of the metallo-β-lactamases has
recently been offered [ 41 ] . The diversity of the group 3
β-lactamases is best exemplified by the enzymes' vastly
differing efficacies towards non-clinical inhibitors; these
differences predict that one inhibitor may not inhibit all
metallo-β-lactamases [ 18 20 21 22 23 24 25 26 27 28 29 ] .
To combat this problem, we are characterizing a
metallo-β-lactamase from each of the subgroups in an effort
to identify a common structural or mechanistic aspect of
the enzymes that can be targeted for the generation of an
inhibitor. It is hoped that this inhibitor, when given in
combination with an existing antibiotic, will prove to be
an effective therapy against bacteria that produce a
metallo-β-lactamase. This work describes our efforts on
metallo-β-lactamase L1 from
S. maltophilia.
S. maltophilia is an important
pathogen in nosocomial infections of immunocompromised
patients suffering from cancer, cystic fibrosis, drug
addition, AIDS and in patients with organ transplants and
on dialysis [ 30 31 32 ] . This organism is inherently
resistant to most antibiotics due to its low outer membrane
permeability [ 33 ] and to β-lactam containing antibiotics
due to the production of a chromosomally expressed group 2e
β-lactamase (L2) and a group 3c β-lactamase (L1) [ 34 35 ]
. L1 has been cloned, over-expressed, and partially
characterized by kinetic and crystallographic studies [ 36
37 ] . The enzyme exists as a homotetramer of
ca. 118 kDa in solution and in the
crystalline state. The enzyme tightly binds two Zn(II) ions
per subunit and requires both Zn(II) ions for full
catalytic activity. The Zn
1 site has 3 histidine residues and 1
bridging hydroxide as ligands, and the Zn
2 site has 2 histidines, 1 aspartic
acid, 1 terminally-bound water, and the bridging hydroxide
as ligands. Spencer and coworkers used the crystal
structure and modeling studies to propose a substrate
binding model and identified several active site residues
that were involved in substrate binding (Figure 1) [ 37 ] .
However, this model has not been tested experimentally. In
order to prepare tight binding inhibitors of the
metallo-β-lactamases, knowledge about how substrate binds
to the enzymes is needed so that all substrate-enzyme
binding contacts can be maintained in any proposed
inhibitor. This work describes our efforts at understanding
how substrates bind to metallo-β-lactamase L1. Several
site-directed mutants of L1 were generated and
characterized, and the results from these studies reveal
that none of the active site residues predicted from
earlier computational studies [ 37 ] are essential for
tight substrate binding.
Results
Wild type L1
Wild-type L1 was over-expressed in
Escherichia coli and purified as
previously described [ 36 ] . This procedure produced an
average of 50-60 mg of >90% pure, active protein per 4
L of growth culture. Circular dichroism spectra were
collected on wild-type samples to ensure L1 expressed
using the pET26b expression system had the correct
secondary structure. The CD spectrum of wild type L1
showed an intense, broad feature at 190 nm and a smaller
feature at 215 nm (see Additional file 1: CD spectra).
These features are consistent with a sample with
significant α/β content. The Compton and Johnson
algorithm [ 38 ] was used to estimate secondary structure
in the samples; wild-type L1 was estimated to have 38.3%
α-helix, 26.7% β-structure (9.3% antiparallel β-sheet,
2.1% parallel β-sheet, and 15.3% β-turn), and 34.9% other
structure. These estimates are in excellent agreement
with the crystallographically determined secondary
structure of ~40% α-helix and 30% β-structure [ 37 ] .
Metal analyses on multiple preparations of wild-type L1
demonstrated that the enzyme binds 1.9 ± 0.2 Zn(II) ions
per monomer (Table 2), in agreement with previous results
[ 36 ] .
Steady state kinetic studies were performed on
multiple preparations of wild type L1, and the resulting
kinetic data are shown in Tables 3, 4, 5. When using
nitrocefin as substrate and 50 mM cacodylate, pH 7.0, as
buffer, wild-type L1 exhibited a
k
cat value of 38 ± 1 s -1and a K
m value of 12 ± 1 μM. The inclusion of
100 μM ZnCl
2 in the assay buffer resulted in
slightly lower values of K
m and higher values for
k
cat [ 36 ] . The inclusion of higher
concentrations of Zn(II) did not further affect the
steady-state kinetic constants. Apparently, the purified,
recombinant enzyme does not bind its full complement of
Zn(II); therefore, 100 μM Zn(II) was included in all
subsequent kinetic studies.
Wild-type L1 exhibited
k
cat values of 41 ± 1 s -1, 1.9 ± 0.1 s
-1, 42 ± 1 s -1, and 82 ± 5 s -1for the cephalosporins,
nitrocefin, cefoxitin, cefaclor, and cephalothin. For
these same substrates, the K
m values were 4 ± 1 μM, 1.1 ± 0.1 μM,
13 ± 1 μM, and 8.9 ± 1.5 μM, respectively. Two
penicillins were tested as substrates, and penicillin G
and ampicillin exhibited K
m values of 38 ± 12 μM and 55 ± 5 μM
and
k
cat values of 600 ± 100 s -1and 520 ±
10 s -1, respectively (Table 4). Three carbapenems were
also used as substrates for L1, and biapenem, imipenem,
and meropenem exhibited K
m values of 32 ± 1 μM, 57 ± 7 μM, and
15 ± 4 μM and
k
cat values of 134 ± 4 s -1, 370 ± 5 s
-1, and 157 ± 9 s -1, respectively (Table 5). L1's
preference for penicillins and carbapenems over
cephalosporins, as exemplified by the
k
cat values, is in agreement with
previous studies and supports L1's placement in the
β-lactamase 3c family [ 9 ] .
Rapid-scanning visible spectra of 25 μM wild-type L1
with 5 μM nitrocefin demonstrated a decrease in
absorbance at 390 nm, an increase at 485 nm, and a rapid
increase and slower decrease in absorbance at 665 nm.
These spectra are similar to those previously reported
for wild-type L1 and nitrocefin [ 39 ] , and the features
can be attributed to substrate decay, product formation,
and intermediate formation and decay, respectively. Under
these conditions, 2.2 μM intermediate was formed during
the first 10 milliseconds of the reaction (Figure 2), and
the rate of decay of this intermediate corresponds to the
steady-state
k
cat (Table 3). To probe further the
binding of nitrocefin to wild-type L1, stopped-flow
fluorescence studies were conducted as previously
described [ 40 ] (Figure 3). The reaction of wild-type L1
with nitrocefin under steady-state conditions at 10°C
resulted in a rapid decrease in fluorescence followed by
a rate-limiting return of fluorescence (Figure 3A).
Fitting of the data, as described by Spencer
et al. [ 40 ] , resulted in a K
S value for wild-type L1 of 38 ± 5 μM
(Figure 3B).
Ser224 mutants
(the BBL numbering scheme proposed in reference 41 was
used throughout this manuscript). All sequenced subclass
Ba and Bb metallo-β-lactamases (except VIM-1) have a
lysine residue at position 224 [ 41 ] , and all
computational models for substrate binding to the
metallo-β-lactamases assume that the invariant
carboxylate on substrates forms an electrostatic
interaction with this lysine. In L1, the residue at
position 224 is a serine [ 35 ] , and the
substrate-binding model for L1 predicts that this serine
residue interacts with the carboxylate on substrate via a
water molecule [ 37 ] . To test the proposed role of
Ser224 in L1, serine was changed to an alanine (S224A),
aspartic acid (S224D), and lysine (S224K), and these
mutants were characterized using metal analyses, CD
spectroscopy, steady-state kinetics, and pre-steady state
kinetic studies.
Small-scale growth cultures showed that all three
mutants were over-expressed at levels comparable to those
of wild-type L1. Large-scale over-expression and
purification of the mutants showed that all three mutants
were isolatable at levels comparable to those of
wild-type L1. Metal analyses of the S224A and S224D
mutants showed that both mutants bind nearly two Zn(II)
ions (Table 2), like wild-type L1 [ 36 ] ; however, the
S224K mutant binds only 1.0 Zn(II) per protein. CD
spectra of the mutants were similar to those of wild-type
L1 (see Figure in Additional materials). Steady-state
kinetic studies were conducted with all three mutants in
buffer containing 100 μM ZnCl
2 to ensure that both Zn(II) binding
sites were saturated in these studies. Addition of higher
concentrations of Zn(II) did not result in different
values for the steady-state kinetic constants in Tables
3, 4, 5.
When the cephalosporins were used as substrates, the
S224A and S224K mutants exhibited 2- to 4-fold changes in
K
m values (Table 3). In studies with
cefoxitin, cefaclor, and cephalothin as substrate, the
observed
k
cat values for the S224A and S224K
mutants were 2- to 7-fold lower; however, the
k
cat values when using nitrocefin as
substrate were slightly higher (< 2-fold). On the
other hand, the S224D mutant exhibited 3- to 50-fold
higher K
m values and 2- to 20-fold lower
k
cat values for the cephalosporins
tested. A similar trend was observed in kinetic studies
when using penicillins as substrates (Table 4).
Generally, the S224A and S224K mutants exhibited small
changes in K
m and
k
cat , while the S224D mutant yielded
20- to 40-fold increased values for K
m and >10-fold decreases in
k
cat when using the penicillins as
substrates. When the carbapenems were used as substrates
however, the changes in K
m values were relatively smaller than
with the other substrates, and 2- to 37-fold changes in
k
cat were observed (Table 5).
Rapid-scanning Vis studies of the S224X mutants were
conducted to probe whether the mutations caused changes
in the amount of intermediate that accumulates during
catalysis. When 50 μM S224A was reacted with 5 μM
nitrocefin, 1.7 μM intermediate formed during the first
10 milliseconds of the reaction (Figure 2), and rate of
decay of this intermediate was equal to the steady-state
k
cat (Table 3). In spite of utilizing a
number of reaction conditions, the S224K and S224D
mutants yielded rapid-scan spectra with no detectable
absorbances at 665 nm (Figure 2), indicating that the
intermediate is not stabilized as well in these mutants
as in wild-type L1. Stopped-flow fluorescence studies at
10°C with the S224A, S224D, and S224K mutants and
nitrocefin as the substrate resulted in K
S values of 39 ± 10, 213 ± 63, and 33
± 5 μM, respectively.
Asn233 mutants
Two-thirds of all sequenced metallo-β-lactamases have
an Asn at position 233 [ 41 ] , and this residue was
predicted [ 42 ] and shown [ 43 ] to be involved with
substrate binding and activation by interacting
electrostatically with the substrate β-lactam carbonyl.
However, in L1, Asn233 is 14 Å away from the modeled
position of the substrate β-lactam carbonyl [ 37 ] . To
test the role of Asn233 in substrate binding, the Asn was
changed to a leucine (N233L) and to an aspartic acid
(N233D), and these mutants were characterized by using
metal analyses, CD spectroscopy, steady-state kinetics,
and pre-steady state kinetic studies.
Small-scale growth cultures showed that both mutants
were over-expressed at levels comparable to that of
wild-type L1. Large-scale over-expression and
purification of the mutants showed that both mutants were
isolatable at levels comparable to that of wild-type L1.
Metal analyses of the N233L and N233D mutants showed that
both bind nearly two Zn(II) ions (Table 2), like
wild-type L1 [ 36 ] . CD spectra of the mutants were
similar to those of wild-type L1. Steady-state kinetic
studies were conducted with both mutants in buffer
containing 100 μM ZnCl
2 to ensure that both Zn(II) binding
sites were saturated in these studies. Addition of higher
concentrations of Zn(II) did not result in different
values for the steady-state kinetic constants in Tables
3, 4, 5.
With all substrates tested, the N233L and N233D
mutants exhibited K
m values that differed less than a
factor of 4 than that observed for wild-type L1 (Tables
3, 4, 5). The
k
cat values exhibited by these mutants
for all substrates also differed by less than a factor of
4, except when biapenem and meropenem were used as
substrates for the N233D mutant. With these two
substrates, there was a 19-fold and 45-fold decrease in
the
k
cat values when using biapenem and
meropenem, respectively (Table 5). The steady-state
kinetic data generally support the prediction that Asn233
does not play a large role in binding or catalysis.
However, rapid-scanning Vis studies of N233L and N233D
with nitrocefin demonstrate that no detectable amounts of
intermediate are formed during the reaction, even when
using a wide number of reaction conditions (Figure 2).
Stopped-flow fluorescence studies at 10°C with the N233L
and N233D mutants and nitrocefin as substrate resulted in
K
S values of 26 ± 9 and 25 ± 8 μM,
respectively.
Tyr228 mutants
The substrate-binding model showed that Tyr228 in L1
was position-conserved with Asn233 in the other
crystallographically characterized metallo-β-lactamases [
37 42 44 45 46 ] . Spencer and coworkers postulated that
Tyr228 is part of an oxyanion hole that interacts with
the β-lactam carbonyl on substrate and helps to stabilize
the putative tetrahedral intermediate formed during
substrate turnover [ 37 ] . To test this hypothesis,
Tyr228 was changed to an alanine and to a phenylalanine
to afford the Y228A and Y228F mutants, respectively.
Small-scale growth cultures showed that both mutants
were over-expressed at levels comparable to those of
wild-type L1. Large-scale over-expression and
purification of the Y228A and Y228F mutants showed that
both mutants were isolatable at levels comparable to
those of wild-type L1. Metal analyses of the mutants
showed that both bind nearly two Zn(II) ions (Table 2),
like wild-type L1 [ 36 ] , and CD spectra of the mutants
were similar to those of wild-type L1. Steady-state
kinetic studies were conducted with both mutants in
buffer containing 100 μM ZnCl
2 to ensure that both Zn(II) binding
sites were saturated in these studies. Addition of higher
concentrations of Zn(II) did not result in different
values for the steady-state kinetic constants in Tables
3, 4, 5.
When cephalosporins were used as substrates, the Y228A
and Y228F mutants exhibited K
m values that were 6- to 45-fold
higher than those observed for wild-type L1 (Table 3).
The largest change in K
m was observed when cefaclor was used
as substrate, and the smallest change was observed when
nitrocefin was used as substrate. The Tyr228 mutants
exhibited < 4-fold change in
k
cat values for the cephalosporins
tested (Table 3), suggesting that Tyr228 is not playing a
large role in catalysis. When penicillins were used as
substrates, the Tyr228 mutants exhibited 3- to 13-fold
increased K
m values and < 2-fold changes in
k
cat , as compared to the values
ascertained using wild-type L1 (Table 4). On the other
hand when carbapenems were used as substrates, the Tyr228
mutants exhibited < 6-fold increases in K
m values as compared to those values
for wild-type L1 (Table 5). Interestingly, there was a 2-
to 8-fold drop in
k
cat values for the Tyr228 mutants, as
compared to values observed for wild-type L1, when using
the carbapenems as substrates.
Rapid-scanning Vis spectra of the reaction of the
Y228A and Y228F mutants with nitrocefin demonstrated a
marked decrease in the amount of intermediate formed with
these mutants (Figure 2). In reactions with 50 μM mutant
and 5 μM nitrocefin, only 0.75 and < 0.30 μM
intermediate formed for the Y228F and Y228A mutants,
respectively. The concentration of mutants were varied
between 25 to 150 μM to ensure that all of the substrate
was bound; however, none of the reactions resulted in the
detection of intermediate at levels observed for
wild-type L1 (data not shown). Stopped-flow fluorescence
studies at 10°C of nitrocefin hydrolysis by Y191A and
Y191F resulted in K
S values of 85 ± 9 and 22 ± 6 μM,
respectively.
Ile164 and Phe158 mutants
All crystallographically characterized
metallo-β-lactamases have a flexible amino acid chain
that extends over the active site [ 37 42 44 45 46 47 48
49 ] . Previous NMR studies on CcrA have shown that this
loop "clamps down" on substrate or inhibitor upon
binding, and there is speculation that the distortion of
substrate upon clamping down of the loop may drive
catalysis [ 50 ] . The crystal structure of L1 showed
that there is a large loop that extends over the active
site, and modeling studies have predicted that two
residues, Ile164 and Phe158, make significant contacts
with large, hydrophobic substituents at the 2' or 6'
positions on penicillins, cephalosporins, or carbapenems
[ 37 ] . To test this prediction, Ile 164 and Phe158 were
changed from large, hydrophobic residues to alanines to
afford the I164A and F158A mutants.
Small-scale growth cultures demonstrated the I164A and
F158A mutants were over-expressed at levels comparable to
that of wild-type L1 (data not shown). Large-scale
over-expression and purification of the mutants resulted
in comparable quantities of isolatable enzymes, which had
identical CD spectra as wild-type L1 and bound slightly
less Zn(II) than wild-type L1 (Table 2). All steady-state
kinetic studies were conducted in buffers containing 100
μM ZnCl
2 to ensure that both metal binding
sites were saturated during the studies.
When using the cephalosporins, nitrocefin, cefoxitin,
and cephalothin, as substrates and the I164A mutant,
there were 2- to 10-fold (only for cefoxitin) increases
in K
m and 2- to 4-fold increases in
k
cat observed (Table 3). However when
cefaclor was used as substrate, the I164A mutant
exhibited a 3-fold decrease in K
m and a 1.5-fold decrease in
k
cat (Table 3). On the other hand, the
K
m and
k
cat values for the I164A mutant when
the penicillins or carbapenems were used as substrates
were very similar to those numbers exhibited by wild-type
L1 (Tables 4and 5).
When the cephalosporins were used as substrates for
the F158A mutant, the K
m values observed were 7- to 31-fold
higher than those determined for wild-type L1, and
surprisingly, the
k
cat values were 2- to 31-fold higher
than those exhibited by wild-type L1 (Table 3). As with
the I164A mutant, the changes in K
m and
k
cat for the penicillins and
carbapenems were relatively small, as compared with the
values obtained with the cephalosporins (Table 4).
Rapid-scanning Vis studies on nitrocefin hydrolysis by
I164A and F158A showed a marked decrease in intermediate
accumulation, with the I164A mutant generating < 0.30
μM intermediate and the F158A producing no detectable
intermediate (Figure 2). Stopped-flow fluorescence
studies at 10°C resulted in a K
S value of 31 ± 11 μM for the I164A
mutant. The reaction of F158A with nitrocefin was so
rapid, we could not determine a K
S value for this mutant.
Discussion
β-Lactam containing antibiotics constitute the largest
class of antibiotics, and these compounds are relatively
inexpensive to produce, cause minor side effects, and are
effective towards a number of bacterial strains.
Nonetheless, bacterial resistance to these antibiotics is
extensive, most commonly due to the bacterial production of
β-lactamases [ 10 51 ] . In fact, there have been over 300
distinct β-lactamases reported, and most of these enzymes
utilize an active site serine group to nucleophilically
attack the β-lactam carbonyl, resulting in a hydrolyzed
product that is covalently attached to the active site. To
combat these enzymes, β-lactamase inhibitors such as
clavulanic acid, sulbactam, and tazobactam have been given
in combination with a β-lactam containing antibiotic to
treat bacterial infections [ 52 ] . One class of
β-lactamases that are particularly unaffected by the known
β-lactamase inhibitors and have been shown to hydrolyze
almost all known β-lactam containing antibiotics including
late generation carbapenems at high rates are the
metallo-β-lactamases [ 14 15 16 17 18 19 ] . Although there
are no reports of metallo-β-lactamases isolated from major
pathogens [ 51 53 ] , these enzymes are produced by
pathogens such as
B. fragilis, S. maltophilia, and
P. aeruginosa. It is inevitable that
the continued and extensive use of β-lactam antibiotics
will result in a major pathogen that produces a
metallo-β-lactamase.
Efforts to solve the crystal structure of one of the
metallo-β-lactamases with a bound substrate molecule have
failed, most likely due to the high activity of the enzymes
towards all β-lactam containing antibiotics [ 37 54 ] .
Therefore, computational studies have been used extensively
to study substrate binding, the role of the Zn(II) ions in
catalysis, the protonation state of the active site, and
inhibitor binding [ 37 42 55 56 57 58 59 ] . All of the
substrate binding models have made assumptions before the
substrate was docked into the active site [ 37 42 ] , and
some of these assumptions have been shown to be invalid for
certain substrates [ 43 ] . With L1, two key assumptions
were made: (1) the bridging hydroxide functions as the
nucleophile during catalysis and (2) Zn
1 coordinates the β-lactam carbonyl [ 37
] . With these assumptions and after energy minimizations,
Ser224 was predicted to hydrogen bond to the substrate
carboxylate [ 37 ] , reminiscent of the role predicted for
Lys224 in CcrA [ 42 ] . Ullah
et al. predicted that Phe158 and Ile
164 form hydrophobic interactions with bulky substituents
on the substrate, suggesting that the loss of these
residues would only affect binding of substrates with large
aromatic substituents [ 37 ] . In the modeling studies on
CcrA [ 42 ] , Asn233 was predicted to interact with the
β-lactam carbonyl on substrate, and mutagenesis studies
have supported this prediction [ 43 ] . Although Asn233 is
sequence conserved in L1 [ 35 ] , it is located 14Å away
from the modeled position of the β-lactam carbonyl and was
predicted not to play a role in substrate binding to L1 [
37 ] . On the other hand, the substrate-binding model
predicted that Tyr228 was in position to offer a hydrogen
bond to the β-lactam carbonyl and participate in an
oxyanion hole that was proposed to form as the substrate
was hydrolyzed [ 37 ] . By using the crystal structure and
modeling studies on L1, Ullah
et al. proposed a reaction mechanism
for the enzyme [ 37 ] . To test this proposed mechanism and
the proposed roles of the amino acids discussed above,
site-directed mutagenesis studies were conducted on
metallo-β-lactamase L1 and reported herein.
The overlap extension method [ 60 ] was used to prepare
the site-directed mutants, and a variety of studies were
used to probe whether the single point mutations resulted
in large structural changes in the mutant enzymes. (1) The
over-expression levels of mutants were analyzed with
SDS-PAGE to ensure that the mutations did not result in
changes in the over-expression levels of the enzymes. With
a few L1 mutants and with other enzyme systems in the lab,
single point mutations often result in depressed levels of
over-expression [ 61 ] . In the case of the mutants
described here, all of the mutants over-expressed at levels
comparable to wild-type L1 (data not shown). (2) The total
amounts of the mutants isolatable after chromatography were
compared with wild-type L1 levels. We have found, in
particular with metal binding mutants of L1 (G. Periyannan,
R.B. Yates, and M.W. Crowder, unpublished results) and
glyoxalase II [ 61 ] , that single point mutations can
result in over-expressed mutants being processed into
inclusion bodies and unisolatable as soluble proteins. In
the case of the mutants described here, all of the mutants
were isolated at levels comparable to wild-type L1. (3) CD
spectra were collected for all mutants and compared to the
spectrum of wild-type L1. Although we did not expect a
large change in the secondary structure of L1 upon single
point mutations, CD spectroscopy is the most common
structural technique to characterize site-directed mutants.
All of the mutants described here exhibited CD spectra that
were very similar, or identical, to that of wild-type L1
(see Additional file 1: CD spectra). (4) Metal analyses on
the mutants were used to probe whether point mutations
caused a significant change to the metal binding site as to
preclude metal binding. The crystal structures of the
metallo-β-lactamases reveal a complex and far-reaching
hydrogen-bonding network around the metal binding sites,
and disruption of this network is predicted to affect metal
binding [ 37 42 44 45 48 49 62 63 ] . With all of the
mutants described here except the S224K mutant, each mutant
binds wild-type or near-wild-type levels of Zn(II) after
purification. The S224K mutant exhibited a 50% reduction in
metal binding (Table 2), and we postulate this is due to
electrostatic repulsions between the newly introduced Lys
with Zn
2 . In spite of the mutants binding
significant amounts of Zn(II), we included 100 μM ZnCl
2 in all of the kinetic buffers to
ensure saturation of the metal binding sites and to
facilitate direct comparison of the kinetic data. (5) All
mutants were stable to multiple freeze/thaw cycles and to
prolonged storage (> 3 weeks) at 4°C, retaining > 95%
of their activity. With these five lines of evidence, we
were confident that none of the point mutations resulted in
large structural changes in L1 and that any kinetic
differences could be attributed to the changed amino
acid.
As a first approximation of substrate binding, we
examined the steady-state kinetics of 4 cephalosporins, 2
penicillins, and 3 carbapenems (Tables 3, 4, 5) and
compared the K
m values of the mutants with those of
wild-type L1. The substrates tested were chosen because
they exhibited low K
m values in previous kinetic studies [
36 ] , and we believed that we could saturate the enzymes
with substrate even if there was large change in binding
with the point mutations. The Tyr228 mutants exhibited
increased K
m values for 8 of the 9 substrates
tested, with the smallest changes in K
m observed when the carbapenems were
used as the substrate. This result supports the proposed
role of Tyr228 in substrate binding. In contrast, the
results on the Ser224 mutants suggest that this residue is
not important in substrate binding, since the S224A and
S224K mutants did not exhibit any significant increases (by
a factor of ≥ 10) in K
m for any of the substrates tested. Only
when Ser224 was replaced with an Asp residue was there
significant increases in the observed K
m value for 6 of the 9 substrates
tested, and the largest changes were exhibited when the
penicillins were used as substrates. This result supports
the observation of differential binding modes of substrates
to the β-lactamases, depending on the structure of the
substrate [ 43 64 65 ] . The only remaining mutants that
exhibited significant changes in the K
m values were the I164A and F158A
mutants. The I164A mutant exhibited increased K
m values only when using cefoxitin as
the substrate, suggesting an interaction of the isoleucine
group with the methoxy group on cefoxitin. The F158A mutant
exhibited higher K
m values when using the cephalosporins
as substrates, suggesting an interaction of the
cephalosporins' substituents with the phenylalanine on the
loop that extends over the active site. None of the other
mutants exhibited vastly different values for K
m with any of the substrates tested.
An examination of the
k
cat values of the mutants revealed some
surprising results. The S224D mutants displayed decreased
k
cat values for 7 of the 9 substrates
tested. Since similar results were not observed with the
S224K and S224A mutants, we do not propose a catalytic role
for Ser224. Instead, we predict that the insertion of an
aspartic acid into the active site at position 224 results
in a change in the hydrogen bonding network in L1; this
hydrogen bonding network is extensive in all
metallo-β-lactamases that have been characterized
crystallographically [ 37 42 44 45 48 49 62 63 ] . The
N233D mutant also exhibited greatly reduced
k
cat values for biapenem and meropenem
but not for imipenem or any of the other substrates tested.
This mutation is also predicted to affect the hydrogen
bonding network around the active site, and apparently,
interactions of the enzyme with the 4' substituent of the
carbapenems has an effect on catalysis. More surprisingly
are the increases in
k
cat of the F158A mutants. We are
uncertain why the mutation of residues on the loop that
extends over the active site would affect
k
cat , since substrate and product
binding have been predicted to be very fast in the reaction
of nitrocefin with L1. However, we do note that the k
cat /K
m values of wild-type L1 and F158A
differ by a factor less than 2.
The inability to propose a consistent binding model also
supports the recent proposal that different substrates of
L1 are hydrolyzed by different mechanisms and further
suggests that using steady-state kinetic constants may not
be a valid way to probe substrate binding to L1. In
addition, the minimal kinetic mechanism of nitrocefin
hydrolysis by L1 has been reported, and this mechanism
predicts that K
m does not accurately reflect substrate
binding. By using this mechanism [ 39 ] , K
m is equal to {(k
-1 + k
2 )k
3 k
4 } / {k
1 (k
3 k
4 + k
2 k
4 + k
2 k
3 }. To probe more directly the
reaction, stopped-flow absorbance studies were conducted,
and the substrate decay rates (390 nm) were studied as a
function of nitrocefin concentration. While nitrocefin is a
nontypical substrate, as a result of the
dinitro-substituted styryl substituent [ 40 ] , it is the
substrate about which the most is known about its
hydrolysis mechanism. Therefore, kinetic studies with
nitrocefin as substrate allowed for us to evaluate the
effect of point mutations on the reaction mechanism of L1.
There was no clear dependence on substrate decay rates with
nitrocefin concentration (data not shown). We did note
though that the amount of intermediate formed during the
reactions varied considerably depending upon which mutant
of L1 was used in the study. All of the mutants exhibited
decreases in intermediate formation, and the S224D, S224K,
F158A, N233D, and N233L mutants yielded rapid-scanning data
consistent with no detectable intermediate. These same
mutants exhibited vastly differing K
m values. Clearly there is no
correlation of K
m with the presence of the reaction
intermediate. Apparently, the ability to observe the
intermediate is not governed entirely by the choice of
substrate [ 40 ] , and it also depends on precise
arrangement of active site residues. It is also possible
that the site-directed mutants could be utilizing a
different mechanism to hydrolyze nitrocefin [ 66 ] .
Recently, Spencer and co-workers reported that
stopped-flow fluorescence studies can be used to monitor
the reaction of L1 with nitrocefin and that an initial
binding step can be directly monitored [ 40 ] . By
increasing the concentration of nitrocefin, the rate of the
initial binding step increased to a maximum, and fitting of
these data yielded a binding constant (called K
S herein) for nitrocefin. Each of the L1
mutants were studied using the stopped-flow fluorescence
studies, and the resulting data were fitted as reported by
Spencer
et al. [ 40 ] (Table 6). All of the
mutants exhibited K
S values identical, within error, to
wild-type L1, except the S224D and the Y228A mutants. The
placement of a negative charge at position 224 drastically
affects nitrocefin binding and results in a 6-fold decrease
in binding affinity (Table 6). To a lesser degree, the
aromatic portion of Tyr228 must have an effect on the
binding site as the K
S value for nitrocefin binding to this
mutant is decreased by a factor of 2; however, the hydroxyl
group probably does not form a hydrogen bond to the
substrate as proposed. By using nitrocefin as substrate and
K
m values alone, a completely different
conclusion is reached regarding important substrate binding
residues. The results presented here suggests that none of
the residues in this study are essential for tight
nitrocefin binding, possibly because other parts of the
active site accommodate the loss of certain binding
contacts.
Spencer
et al. also reported stopped-flow
fluorescence studies when using cefaclor and meropenem as
substrates, and K
S values for these substrates were
reported to be 710 ± 180 and 272 ± 112 μM, respectively [
40 ] . However in our hands, the rates of substrate
hydrolysis were so fast when using wild-type L1 that we
could not use substrate concentrations high enough to
saturate the enzyme. Similarly, we could not determine K
S values for penicillin G or ampicillin
because the observed rates of hydrolysis at low substrate
concentrations were too fast to observe data, even at
10°C.
Conclusions
The results presented herein indicate that none of the
active site residues identified with computational studies
are essential for tight substrate binding. These data also
indicate that the use of K
m values to describe substrate binding
to L1 is unreliable and that there is no correlation
between intermediate accumulation and substrate binding
affinity. These results demonstrate that new computational
studies are now needed to probe substrate binding to L1,
and these studies are currently underway. The results
presented herein can be used to guide these new
computational studies, which will lead to the design of
potential inhibitors and hopefully a way to combat
penicillin resistance in bacteria.
Materials and Methods
Materials
E. coli strains DH5α and BL21(DE3)
pLysS were obtained from Gibco BRL and Novagen,
respectively. Plasmids pET26b and pUC19 were purchased
from Novagen. Primers for sequencing and mutagenesis
studies were purchased from Integrated DNA Technologies.
Deoxynucleotide triphosphates (dNTP's), MgSO
4 , thermopol buffer, Deep Vent DNA
polymerase, and restrictions enzymes were purchased from
Promega or New England Biolabs. Polymerase chain reaction
was conducted using a Thermolyne Amplitron II unit. DNA
was purified using the Qiagen QIAQuick gel extraction kit
or Plasmid Purification kit with QIAGEN-tip 100 (Midi)
columns. Wizard Plus Minipreps were acquired from
Promega. Luria-Bertani (LB) media was made following
published procedures [ 67 ] . Isopropyl-β-thiogalactoside
(IPTG), Biotech grade, was procured from Anatrace.
Phenylmethylsulfonylfluoride (PMSF) was purchased from
Sigma. Protein solutions were concentrated with an Amicon
ultrafiltration cell equipped with YM-10 DIAFLO membranes
from Amicon, Inc. Dialysis tubing was prepared using
Spectra/Por regenerated cellulose molecular porous
membranes with a molecular weight cut-off of 6-8,000
g/mol. Q-Sepharose Fast Flow was purchased from Amersham
Pharmacia Biotech. Nitrocefin was purchased from Becton
Dickinson, and solutions of nitrocefin were filtered
through a Fisherbrand 0.45 micron syringe filter.
Cefaclor, cefoxitin, and cephalothin were purchased from
Sigma; penicillin G and ampicillin were purchased from
Fisher. Imipenem, meropenem, and biapenem were generously
supplied by Merck, Zeneca Pharmaceuticals, and Lederle
(Japan), respectively. All buffers and media were
prepared using Barnstead NANOpure ultrapure water.
Generation of site-directed mutants of L1
The over-expression plasmid for L1, pUB5832, was
digested with
Nde I and
Hind III, and the resulting
ca. 900 bp piece was gel purified
and ligated using T4 ligase into pUC19, which was also
digested with
Nde I and
Hind III, to yield the cloning
plasmid pL1PUC19. Mutations were introduced into the L1
gene by using the overlap extension method of Ho
et al. [ 60 ] , as described
previously [ 68 ] . The oligonucleotides used for the
preparation of the mutants are shown in Table 1. The
ca. 900 bp PCR products were
digested with
Nde I and
Hind III and ligated into pUC19.
The DNA sequences were analyzed by the Biosynthesis and
Sequencing Facility in the Department of Biological
Chemistry at Johns Hopkins University. After confirmation
of the sequence, the mutated pL1PUC19 plasmid was
digested with
Nde I and
Hind III, and the 900 bp, mutated
L1 gene was gel purified and ligated into pET26b to
create the mutant overexpression plasmids. To test for
overexpression of the mutant enzymes,
E. coli BL21(DE3)pLysS cells were
transformed with the mutated over-expression plasmids,
and small scale growth cultures were used [ 68 ] .
Large-scale (4 L) preparations of the L1 mutants were
performed as described previously [ 36 ] . Protein purity
was ascertained by SDS-PAGE.
Metal content
The concentrations of L1 and the mutants were
determined by measuring the proteins' absorbance at 280
nm and using the published extinction coefficient of ε
280 nm = 54,804 M -1•cm -1 [ 36 ] or
by using the method of Pace [ 69 ] . Before metal
analyses, the protein samples were dialyzed versus 3 × 1
L of metal-free, 50 mM HEPES, pH 7.5 over 96 hours at
4°C. A Varian Inductively Coupled Plasma Spectrometer
with atomic emission spectroscopy detection (ICP-AES) was
used to determine metal content of multiple preparations
of wild type L1 and L1 mutants. Calibration curves were
based on three standards and had correlation coefficient
limits of at least 0.9950. The final dialysis buffer was
used as a blank, and the Zn(II) content in the final
dialysis buffers was shown to be < 0.5 μM (detection
limit of ICP) in separate ICP measurements. The emission
line of 213.856 nm is the most intense for zinc and was
used to determine the Zn content in the samples. The
errors in metal content data reflect the standard
deviation (σ
n-1 ) of multiple enzyme
preparations.
Steady-state kinetic studies
Steady-state kinetic assays were conducted at 25°C in
50 mM cacodylate buffer, pH 7.0, containing 100 μM ZnCl
2 on a HP 5480A diode array UV-Vis
spectrophotometer at 25°C. The changes in molar
absorptivities (Δε) used to quantitate products were (in
M -1cm -1): nitrocefin, Δε
485 = 17,420; cephalothin, Δε
265 = -8,790; cefoxitin, Δε
265 = -7,000; cefaclor, Δε
280 = -6,410; imipenem, Δε
300 = -9,000; meropenem, Δε
293 = -7,600; biapenem, Δε
293 = -8,630; ampicillin, Δε
235 = -809; and penicillin G, Δε
235 = -936. When possible, substrate
concentrations were varied between 0.1 to 10 times the K
m value. In kinetic studies using
substrates with low K
m values (cefoxitin, nitrocefin, and
cephalothin) or with small Δε values (penicillin and
ampicillin), we typically used substrate concentrations
varied between ~ K
m and 10 × K
m and used as much of the ΔA versus
time data (that was linear) as possible to determine the
velocity. Steady-state kinetics constants, K
m and
k
cat , were determined by fitting
initial velocity versus substrate concentration data
directly to the Michaelis equation using CurveFit [ 36 ]
. The reported errors reflect fitting uncertainties. All
steady-state kinetic studies were performed in triplicate
with recombinant L1 from at least three different enzyme
preparations.
Circular dichroism
Circular dichroism samples were prepared by dialyzing
the purified enzyme samples versus 3 × 2 L of 5 mM
phosphate buffer, pH 7.0 over six hours. The samples were
diluted with final dialysis buffer to ~75 μg/mL. A JASCO
J-810 CD spectropolarimeter operating at 25°C was used to
collect CD spectra.
Stopped - flow/Rapid-Scanning UV-Visible
Spectrophotometry
Rapid-scanning Vis spectra of nitrocefin hydrolysis by
L1 and the L1 mutants were collected on a Applied
Photophysics SX.18MV stopped-flow spectrophotometer
equipped with an Applied Photophysics PD.l photodiode
array detector and a 1 cm pathlength optical cell. A
typical experiment consisted of 25 μM enzyme and 5 μM
nitrocefin in 50 mM cacodylate buffer, pH 7.0 containing
100 μM ZnCl
2 , the reaction temperature was
thermostated at 25°C, and the spectra were collected
between 300 and 725 nm. Data from at least three
experiments were collected and averaged. Absorbance data
were converted to concentration data as described
previously by McMannus and Crowder [ 39 ] . Stopped-flow
fluorescence studies of nitrocefin hydrolysis by L1 were
performed on an Applied Photophysics SX.18MV
spectrophotometer, using an excitation wavelength of 295
nm and a WG320 nm cut-off filter on the photomultiplier.
These experiments were conducted at 10°C using the same
buffer in the rapid-scanning Vis studies. Fluorescence
data were fitted to
k
obs = {(
k
f [S]) / K
S + [S])} +
k
r as described previously [ 40 ] or to
k
obs =
k
f [S] +
k
r by using CurveFit v. 1.0.
Abbreviations
AES, atomic emission spectroscopy; bp, base pairs; CD,
circular dichroism; ε, extinction coefficient; ICP,
inductively coupled plasma;
k
cat , turnover number; kDa, kilodaltons;
K
m , Michaelis constant; K
S , substrate binding constant; LB,
Luria-Bertani media.
