J. MOE.BioE. (1991) 217, 701-719

Refined Crystal Structure of fl-La$amase from Staphylococcus aureus PC1 at 2-OA Resolution Osnat Herzberg Center for Advanced Research in Biotechnology Maryland Biotechnology Institute, University of Maryland 9600 Cudelsky Drive, Rochwille, MD 20850, U.S.A. (Received 24 July

1990; accepted 10 October 1990)

The crystal structure of a class A B-lactamase from Staphylococcus aureus PC1 has been refined at 2.0 A resolution. The resulting crystallographic R-factor (R = &llF,I - IFG,ll/ChlF,I, where IF,1 and lF,I are the observed and calculated structure factor amplitudes, respectively), is 0163 for the 17,547 reflections with I2 2a(I) within the 8.0 A to 2.0 A resolution range. The molecule consists of two closely associated domains. One domain is formed by a fivestranded antiparallel P-sheet with three helices packing against a face of the sheet. The second domain is formed mostly by helices that pack against the second face of the sheet. The active site is located in the interface between the two domains, and many of the residues that form it are conserved in all known sequences of class A /I-lactamases. Similar to the serine proteases, an oxyanion hole is implicated in catalysis. It is formed by two main-chain nitrogen atoms, that of the catalytic seryl residue, Ser70, and that of Gln237 on an edge P-strand of the major /?-sheet. Ser70 is interacting with another conserved seryl residue, Serl30, located between the two ammonium groups of the functionally important lysine residues, Lys73 and Lys234. Such intricate interactions point to a possible catalytic role for this second seryl residue. Another key catalytic residue is Glul66. There are several unusual structural features associated with the active site. (1) A cis peptide bond has been identified between the catalytic Glu166 and Ile167. (2) Ala69 and Leu220 have strained 4, + dihedral angles making close contacts that restrict the conformation of the active site P-strand involved in the formation of the oxyanion hole. (3) A buried aspartate residue, the conserved Asp233, is located next to the active site Lys234. It is interacting with another buried aspartyl residue, Asp246. An internal solvent molecule is also involved, but the rest of its interactions with the protein indicate it is not a cation. (4) Another conserved aspartyl residue that is desolvated is Aspl31, adjacent to Ser130. Its charge is stabilized by interactions with four main-chain nitrogen atoms. (5) An internal cavity underneath the active site depression is filled with six solvent molecules. This, and an adjacent cavity occupied by three solvent molecules partially separate the R-loop associated with the active site from the rest of the protein. A total of 207 solvent molecules have been identified in the crystal. Of these, several are located in the active site. In particular, one water molecule occupies the oxyanion hole in a manner found in some of the serine protease structures. A second water molecule is located between Glu166, Ser70 and Asnl70, indicating that it may play a role in the deacylation of a hydrolyzed substrate.

1. Introduction /?-Lactamases (EC 3.5.2.6) are a family of bacterial enzymes that inactivate /%lactam antibiotics by hydrolyzing the /?-la&am bond typical of this group of compounds (Abraham & Chain, 1940). The abundance of the enzymes among pathogenic bacteria has considerably reduced the usefulness of /I-lactam antibiotic therapy. Intensive research addressing the clinical aspects of the problem has been carried out. In addition, these enzymes have also been studied as models for understanding protein structure and function, and as useful genetic 0022-2836/91/040701-19

$03.00/O

markers in molecular biology. Thus, substantial information is available about their variability within the bacterial kingdom and their physical and chemical properties (Hamilton-Miller & Smith, 1979; Coulson, 1985; Frere & Joris, 1985; Knowles, 1985). Of the various classifications of the enzymes, that based on sequence homology proves most comprehensive. This approach was first used by Ambler (1979) to define the class A /I-lactamases. Three of the four classes defined to date involve a serine residue in the active site (classes A, C and D). Although this is reminiscent of the serine proteases, 701 0 1991 Academic Press Limited

702

0. Herzberg

b-lactamases lack the catalytic histidyl residue common to all known serine proteases. Instead, a conserved lysine is located three residues upstream from the active site serine and has been shown by site-directed mutagenesis studies to be essential for the activity of the enzyme (Madgwick & Waley, 1987). A similarity in the catalytic mechanism of the three classes of P-lactamase is implied by this invariance. Moreover, the sequences of the /?-lactam-sensitive cell wall peptidases known to date also share the active site serine and the lysine residues. Although the sequence homology between the cell wall peptidases and the P-lactamases is rather poor (Joris et al., 1988), this identity of catalytic residues supports the proposal put forward by Tipper & Strominger (1965) based on functional similarity, that the P-lactamases have evolved from the cell wall peptidases. Crystallography on a D-ala-D-alanine transpeptidase/carboxypeptidase (Kelly et al., 1986), on four class A p-lactamases (Samraoni et al., 1986; Herzberg & Moult, 1987; Dideberg et al., 1987; Mowes et al., 1990) and on one of the class C fl-lactamases (Oefner et ah, 1990) has established that, the functional similarity corresponds to similarity in the overall fold of these molecules. Traditionally, the class A p-lactamases have been considered more effective against penicillins than against cephalosporins, whereas the class C p-lactamases have been considered better cephalosporinases. Careful examination shows that such a generalization is inaccurate (Matagne et al., 1990). Even within one class, the varitibility in substrate A striking example has profile is substantial. emerged recently, with the characterization of plasmid-mediated TEM and SHV class A &4actamases that hydrolyze the so-called third generation cephalosporins (Sougakoff et al., 1988; Barth&my et al., 19$8). The development of new effective antibiotics resistant to p-lactamases hinges upon understanding the subtle structural differences and their effect on substrate profile. Thus, the structure determination of several fl-lactamases is desirable. The number of newly sequenced fi-lactamases is increasing rapidly. The amino acid sequences of 15 class A @-lactamases have been aligned, showing they are evolutionary related (A. Coulson, personal communication), The identity between pairs of sequences varies between 30% and 70 y0 ~ The enzymes from Gram-positive bacteria are more closely related to each other than to those from Gram-negative bacteria. So far, the crystal structure determination of three class A enzymes have been reported: the detailed structure of j?-lactamase from Staphyloaoccus aureus PC1 has been described at 2.5 A resolution (1 .& = 0.1 run; Herzberg & Moult, 1987), and that from Bacillus bicheniformis 749/C at 2.0 a resolution (Mowes et aE., 1990). Partial interpretation bas been reported of the structure of P-laotamase from Stre@omyces albus G at 3.0 A resolution (Dideberg et al., 1987). The overall fold and the disposition of the active site residues in these structures are extremely similar. The aotive site

architecture implies a catalytic mechanism that is related to that of the serine proteases. Following the analogy to the serine proteases we have proposed that an oxyanion hole is involved in the stabilization of the negatively charged enzyme-substrate tetrahedral transition state (Kraut, 1977). Since the /&lactamases lack the conserved histidyl residue of the serine proteases, the mechanism of proton transfer from the active site serine to the leaving group nitrogen must differ. We have suggested (Herzberg & Moult,’ 1987) that it may involve the conserved residue Lys73, in a manner that is not yet structure fully understood. The high resolution described here implicates Serl30 as well. We have also proposed that the deacylation step differs from that of the serine protease family, involving a solvent molecule with an enhanced nucleophilicity due to interaction with the conserved 61~166. The active site serine residue is located at the amino terminus of a helix in a manner similar to that of subtilisin, suggesting a role for the helix dipole (1101, 1986). I describe here the results of the crystallographic refinement of j?-lactamase from S. aureus PC1 at 2 A resolution. Structural details are summarized? and features of interest are highlighted. Although some reference is made here to the oatalytic mechanism, the detailed account of mechanism-related computational studies will be published elsewhere. The refined co-ordinates have been deposited in the Brookhaven Protein Data Bank (Bernstein et al., 1977), accession number 3BLM. 2. Experimental

Procedures

(a) Crystallization, data collection and structure determination Crystals of S. aureus /?-lactamase were obtained by the hanging drop method, from solutions containing 75 to 78% saturated ammonium sulfate, 0.3 M-MCI, 0.1 MNH,HCO, buffered at pH 8.0, and @3 to O%% (w/v) polyethylene glycol 1000. They belong to space group 1222, with cell dimensions a=53*98, b=940.&, G= 139.1 & and 1 protein molecule in the asymmetric unit. The crystal structure was determined initially at 25 a resolution. The multiple isomorphous replacement (m.i.r.) method was employed at 3 A resolution, using 3 heavy-atom derivatives and improving the phases by the solvent-flattening technique as described (Her&berg 8r. Moult, 1987). The binding of the heavy-atom derivatives is summarized in Table 1. The gold and the platinum derivatives bind in the interface between molecules, and the mercurial derivative binds to an asparagine side-chain in a large cavity in the interior of the prot,ein. Following the building of the initial model the data were extended to 2.5 L%and a preliminary restrained-parameter reciprocal space least-squares refinement (Hendrickson & Konnert, 1980) was carried out without any manual re-fitting, resulting in a crystallographic R-factor of 0.284 (I?= ~hllF,l-IIp,II/~hl~~t, where IF,/ and IFJ are the observed and calculated structure factor amplitudes, respectively). For the high resolution refinement X-ray data were collected on a Nicolet area detector mounted on a Huber 4-circle goniostat. CuK, X-rays were generated with a fine focus sealed tube mounted on a Philips generator. Two

703

P-Lactamase at 2.0 d ReSoWon

Table 1 Binding

of heavy-atom derivatives to /Mactamase

Number of sites

Heavy atom

WUJ’WW,

Relative occupancyt

Co-ordinate1 2, y, 2 (4

1

67.4

-1.0;

-21.5;

-153

2

364

-6.3;

-12.0;

-2.5

@O; @O; -7.2

97

KAu(CN),

Interactions

4

140

-254;

-456;

-541

191

-253;

-38.4;

-650

-8.6;

-67

88 94

-3.9;

0.0; 00; -6.9

In the cave Asn76/polar Vall48/hydrophobic Prol29/hydrophobic Site 1: intermolecular Va1103 mol l/hydrophobic Phe171 mol Z/hydrophobic Lys270 mol 2/polar Ile239 mol 2/hydrophobic Thr240 mol2/hydrophobic Site 2: intermolecular TyrlO5 mol l/hydrophobic Tyr105 mol 2/hydrophobic Site 1: intermolecular TyrlO5 mol l/polar symmetry-related Au Site 2: intermolecular Ile167 mol l/hydrophobic Ile239 mol l/hydrophobic Ala104 mol 2/hydrophobic Site 3: intermolecular Tyrl”rl/hydrophobic Site 4: intermolecular like Pt site 2

t The occupancy is expressed in number of electrons. These values were obtained by the heavy-atom refinement program, and are not accurate. $ The positions correspond to the co-ordinate set deposited in the Brookhaven Data Bank. They were obtained by the heavy-atom refinement program. The heavy-atom derivative data were not refined independently by PROLSQ. $ The distances between the Pt sites and the neighboring protein residues are rather long, varying between 3.6 and 48 A.

crystals were used, such that there was large redundancy in the data. A unique set of structure factors at 2.0 A resolution was obtained using the XENGEN data processing package (Howard et al., 1987). Details of the crystal parameters and data collection are summarized in Table 2 and some statistics of the data processing in Table 3. (b) Crystallographic

rejinement

The restrained-parameter least-squares refinement program of Hendrickson & Konnert (1980), to which a

Table 2 /I-Lactamase data collection Space group Cell dimensions No. of crystals Generator Goniostat Collimator Incident beam Diffracted beam Detector Detector swing angle Oscillation steps Data acquisition Data processing software Data processing computer

I222 a=539A;b=940A;c=1391A 2 Philips RG3100 at 40 kV 44 mA Huber 4-circle 65 mm Monochromated fine focused CuK, 10 om crystal-counter air path Nicolet area detector 225” 025” contiguous steps Cadmus 9000 XENGEN package (Howard et al., 1987) MicroVAX II

fast Fourier transform calculation was added (Finzel, 1987), was used to refine the b-lactamase structure. The structure factors were scaled to absolute values with the computer program ORESTES written by W. E. Thiessen and H. A. Levy. The overall indicated temperature factor was 23.2 A2. The starting co-ordinates were those derived from the initial refinement at 2.5 A resolution. Data for which 12 a(l) were used for the refinement. At each stage, cycles of refinement were carried out until the applied shifts did not appear to improve the model. Atomic co-ordinates and temperature factors were allowed to vary. The overall scale was fixed until an advanced stage of the refinement, to avoid unrealistic shifts of temperature factors. Thus, all through the refinement the overall temperature factor of the protein remained close to that calculated based on Wilson statistics, and when the overall scale was refined towards the end of the procedure, it only changed by 7%. When solvent molecules were incorporated in the refinement their

occupancies

and

temperature

factors

were

varied

alternately. For manual fitting, the model was displayed together with 3 types of electron density maps: (1) a composite omit map (Bhat, 1988); (2) a map computed with the coefficients ZIP01-IE‘,I and calculated phases; (3) a map with coefficients jF,l-lF,l and calculated phases. The E C S PS390 interactive graphics system was used, with the program FRODO (Jones, 1982). After 35 cycles of refinement, when the crystallographic R-factor was 0.247, the model was considered sufficiently reliable to start including solvent molecules. These were assigned by a peak search of the difference Fourier map, and were

704

0. Her&erg Table 3 /3-Lactamase data processing statistics No. of reflections: Possible Missing

Shell lower limit (A)

No. of observations

Fraction with I> 2a(Z)

(I/G)>

&p?

363 2.88 2.52 2.29 2.13 2.00

4232 4078 4025 4034 3992 3978

10 0 2 2 5 441

55,030 27,114 17,369 15,998 14,246 8201

111.8 352 135 7.9 4.7 23

0.97 092 082 0.72 0.59 0.44

0047 PO86 0145 0193 0263 0.321

Total

24,339

460

137,958

305

0.75

0073

Absolute scaling Overall scale: Overall temperature

0.033 23.2 dz

factor:

f or sYmmetry-related

t Rsym= ~h~il(Z)-Z(h)ill~~~i~(h)i

added to substantial

the

co-ordinate

set

only

if

there

were

not

negative peaks in the vicinity, indicating possible errors in the model. The initial cutoff level for

inclusion of solvent was 4 standard deviations of the rootmean-square (r.m.s.t) density in the difference Fourier

map. Towards the end of the refinement the cutoff was lowered to 3.5 standard deviations. Solvent molecules were refined as neutral oxygen atoms. None of them refined to unusually low temperature factors, which would

indicate

sites of ions heavier

than

water.

At cycle

84 the solvent assignment was checked by removing a quarter of the solvent molecules at a time, and refining the model for 5 cycles. New difference Fourier maps of types (2) and (3), described above, were calculated. Those solvent

molecules

for

which

peaks

did

not

reappear

in

both maps were eliminated. Only a few molecules were excluded in this manner. Next, the electrostatic energy of the structure was analyzed by the program ENEANA (M. Toner and J. Moult, unpublished results). Residues with unusually high electrostatic energy were examined for alternative conformation. Using this approach, the side-chain x2 dihedral angles of some asparagine residues were rotated by 180” to provide better electrostatic interactions (note that nigrogen, oxygen and carbon atoms cannot be discriminated in electron density maps calculated at 2 a resolution). The side-chains of a few surface lysine residues were also modified. These were frequently associated with weak electron density ‘and high crystallographic temperature

factors such that their conformation could not be determined reliably from the crystallographic information. The side-chains of the 2 histidyl residues in the structure were found to be properly oriented to start with, since both are involved in salt bridges. Finally, the structure was refined for 10 more cycles, after which the procedure was considered complete. Table 4 describes the progress of the refinement.

3. Results and Discussion (a) Crystallization Reproducing

resolution possible

the crystals

structure only

t Abbreviation

used to obtain

the low

observations.

and adding small amounts of organic solvent. The importance of the organic solvent additive has been emphasized in the structure determination (Herzberg & Moult, 1987). Here it is worthwhile to note that once crystals start growing (4 days to 2 weeks), the growth is very rapid, resulting in imperfect morphology. Frequently the growth habit is such that an empty cavity extends from one edge of the crystal towards its center. Clusters of small crystals tend to grow in this cavity. Separating those from the main crystal is tedious, and sometimes destructive. It now appears that bettershaped crystals can be obtained by initially ineubating the crystallization trays overnight at 35”C, or until crystals appear. However, this procedure often results in many nucleation sites, and hence in smaller final crystals, and requires further optimization. It has also been found that dissolving lyophilized protein in water just before setting the

Progress

Table 4 of the rejinement

No. of solvent molecules

Cycle

Resolution (4

l-26 27-35 36-42 43-49

6-2 6-2 6-2 10-Z

41 69

SO-56 57-61 62-65 66-71 72-78 79-84 85-95

10-Z 10-Z 8-2 8-2 8-2 8-2 8-2

130 154 154 175 186 214 207

Final

Rf

0291 0247 0204 0207

0.190 0184 0180 0.175 0.171 0.168 0169 0.1631:

Remarks Major adjustments Tighten geometry, GlulSB-Ilel67 peptide changed to eis Tighten geometry Tighten geometry Solvent check Tighten geometry

at 5 a (Moult et al., 1985) was

by modifying

the original

conditions

used: r.m.s., root-mean-square.

1 R-factor along the refinement is quoted for reflections for which Z 2 C(Z). In the final cycle the R-factor is also given for data for which Z 2 L%(Z).

/3-Lactamase

at 2.0 A Resolution

705

Figure 1. Stereoscopic view of the electron density map in the region of the cis peptide Calculated phases and the coefficients 2F,-Fc are used.

bond

betn

TeenGlu166 and

Ile167.

crystallization experiment yields substantial precipitation and only a few crystals. In contrast, if the protein is equilibrated in 60 y. saturated ammonium sulfate solution for a few days prior to crystallization, less protein precipitates in the drops and better crystals are obtained. Even better results are obtained if the protein is not lyophilized after purification but kept in 60% saturated ammonium sulfate solution. (b) Quality of the re$nement Most of the adjustments were made to the model in the beginning of the refinement. In particular, the original chain tracing of a stretch of eight amino acid residues (Asp100 to Prol08) was shifted by one amino acid residue. In addition, about 5% of the main-chain carbonyl oxygen atoms were modified to Table 5 Deviation from ideal geometry at the end of the re$nement

move them away from negative to positive density in the difference Fourier map. Support for the new interpretation was provided by the drop in temperature factor values of the corrected and neighboring amino acid residues during following refinement cycles. The initial crystallographic R-factor for the 2 A data was 0.41. Ninety-five least-squares cycles of refinement have been carried out. The final structure has an R-factor of 0.163 for the 17,547 reflections between %O and 2.0 A resolution, for which 12 20(I). The R-factor is 9169 for the 19,149 reflections, for which I> o(I). In the final cycle the indicated r.m.s. shift in atomic co-ordinates was 0008 A. The stereochemical parameters (Table 5) are well within the range known from crystal structures of small peptides. The mean error in co-ordinates of P-lactamase was estimated from the slope of a least-squares line fitted to the points describing the dependence between ln(o,) and (sin f3/n)2 (Read, 1986). A value of 93 A was obtained. The variation of the R-factor with the resolution of the data is shown in Table 6. The increase in the

r.m.s. deviation from ideal values? Distance restraints (8) Bond distance Angle distance Planar l-4 distance Plane restraint (A) Chiral-center restraint (A3) Non-bonded contact restraints Single torsion contact Multiple torsion contact Possible hydrogen bond Tram peptide torsion angle restraint w (deg.)

Crystallographic 0023(0023) @045(@033) @049(@040) @026(@030) 0213(0200) (A) 0230(0400) 0195(0400) @227(@400) 44(&O)

t The values in parentheses are the input estimated standard deviations that determine the relative weights of the corresponding restraints (Hendrickson I% Konnert, 1980).

Table 6 R factor as a function

of resolution

R-factor? I2 20(Z)

I2 o(Z)

39 33 28 2.4 2.2 20

0.220 0125 0132 0.158 0166 0184 0.209

0.222 @125 0.133 0161 0173 0196 0.228

Overall

0163

0169

Shell lower limit (A)

5.0

t R = ChllF,I-I~cll/~hl~oI.

0. Herzberg

706 -____

Side-chain

-1

50 40 30 “+J

20 IO

j

0

7

IO

Q

20 30 40

50I’ 601

1--._ 20

40

60

80

Maln-chain ._._ _--. .~. _ . ~. ?-~. -iOO 120 140 160 180 200 220 240

Figure 2. Variation in crystallographic temperature factors along the polypeptide chain of p-lactamase. The lower lines correspond to an average over the main-chain atoms of each amino acid residue; the upper lines correspond to an average over the side-chain atoms. Key

The peptide bond between 61-1~166and Ilel67 was initially refined in the trans conformation, and with left-handed a-helix main-chain dihedral angles of the isoleucyl residue. However, the side-chain sf Ilel67 could not be fitted into the electron density by manual adjustments of the amino acid residues around. Several interpretations were attempted but residual positive and negative electron density always appeared in the difference Fourier maps. After 42 cycles of refinement the peptide bond between Glu166 and Ilel67 was modeled as a cis-peptide. The side-chain of Ile167 could now be readily fitted into the electron density, and in subsequent refinement cycles its crystallographic temperature factor values dropped by 10 A2. The t’emperature factors of other residues in the vicinity also dropped significantly, including those of the active site Glu166. The electron density of this region is well defined, as can be seen in Figure 1.

residues in the active site are indicated.

(d) Overall struekre value of R at lower resolution

is probably due to the incomplete description of the solvent and the omission of hydrogen atoms from the refined model (Phillips, 1980). (c) &uality of the structure The final model of /3-lactamase includes all 257 amino acid residues (2029 atoms) and 207 water molecules. Figure 1 shows a representative section of the electron density map with its associated model. The crystallographic temperature factors and occupancy values for the solvent molecules do not indicate discrete positions of ions heavier than water. Note, though, that water molecules and ammonium ions cannot be discriminated in the electron density maps. The final overall temperature factor including the solvent is 24-l pi2, similar to the value obtained by the absolute scaling. This rather high value may be associated with the high solvent content of the crystals (about 65%). Figure 2 shows the variation in isotropic temperature factors along the polypeptide chain. Residues that form the active site depression have some of the lowest temperature factors. There are a total of 43 lysine residues, many of which are on the surface of the molecule and are associated with electron density of poor quality. Most of these refined with relatively high temperature factors and should not be considered reliable. In contrast, lysine residues that are involved in specific interactions, for example those in the active site, are well defined.

J-Laetamase consists of two closely associated domains and has overall dimensions of 41 A x 46 a x 58 a (Fig. 3). The fold of one domain (the bottom domain in Fig. 3) is that of an open face p-sandwich (Richardson, 1981), formed by a five-stranded p-sheet and t,hree helices that pack against one of its faces. The second domain (the top domain in Fig. 3) packs against what should be the open face of the sandwich. It consists of six a-helices and four short segments of S1,,-helices (segments shorter than 4 residues are not, considered). The chain crosses twice from one domain to the other: one link is perpendicular to the P-strand direction and across the second face of the sheet (residues 61 to 69). The second link passes over an edge of the sheet (residues 214 to 220). The active site is located in the interface between the domains, with the active site serine (Ser70) at the amino terminus of a helix that follows the first inter-domain cross-over. The novelty of this folding topology has been discussed (Herzberg & IMoult, 1387). Although the exact assignment of the start and end of secondary structure elements have been modified following the high resolution refinement, the overall fold has not changed. The structural assignments determined by the DSSP computer program (Kabsch & Sander, 1983) are given in Table 7. On the basis of this analysis, 35% of the structure is a-helical, 7 y. of the structure has 310 conformation, and 12% of the residues participate in the formation of three antiparallel b-sheets. The major sheet (A in Table 7) is formed by four ladders (A to D in Table 7). Figure 4

Figure 3. The structure of /?-lactamase from S. aureus PC1 at 2.0 A resolution. (a,) The overall fold, highlighting secondary structure motives: helices are shown as cylinders, P-strands as ribbons. The active site Ser70 is located at the N terminus of the golden helix at the interface between the open face p-sandwich domain and the a-domain. The model was generated on a Silicon Graphics Iris/4D workstation with the computer program RASTER3D written by David Eaton. (b) Stereoscopic representation of the whole molecule. The numbering is according to Ambler (1979). Bonds between main-chain N, C”, C atoms are drawn in thick lines; side-chain and main-chain carbonyl bonds are drawn in thin lines. The view is similar to that shown in (a).

/3-Lactamase at FO A Resolution

(a)

(b)

Fig. 3.

707

BB

EE

+--+-------4. cccccccc 1) D I1 D D D D AAAAAAAA

RNDVAFVYPK EFFEEFPF il _liJ_(

ss 9s ++-+-+-i--i-

PKSKKDTSTP TT TTEE 3333
3

S

T

253

>>

H

183

AAAAA DDDDD AAAAA

s s +-+------++

263 CQSEPIVLVI XEJ$:EJi: T s 3
9>9

SSSS + + - - + -- + + -. .AAA DD AAA

273

HHH

PTNKDNKSDK EEE SRTT 9331

sssss +++-+--+++

ii5555

143

SS

88688888 --+-t-F+++++

PNDK.LJHETA THHNHHHH >93>>>xxxx

ssss8ssss8 +-t-t+++++++

xxxxxxxxxxxx>x >

EE

133

ss

>

61

>5

1

153

>>3

81

SSSSY i--i+++

x3x3< < >>44 >5555< ssssssss8 -++-I-++-+++

GIKKVKQRLK HHHHHHHHH >33< 9999xXxXxxxx55

193

123

+++++++i-++

9>XXXXX>>444
>>>XX>3444


173

103

The structural analysis was done with the Kabsch $ Sander (1983) program. The numbering of amino acid residues follows Ambler (1979). The sequence is given in l-letter code. The row labeled Sumnnxy is interpreted: I;, $-helix (a-helix); E:, extended strand in P-ladder; G, a-helix (3,,-helix); T, hydrogen-bonded turn; S, bend. Tn cases of overlap, priority is given bo the structure first in the above list. The 3-turn, g-turn and S-turn rows show the hydrogen bonding pattern for the turns and helices; the first and last residues of a turn whose backbone CO makes a hydrogen bond with the beckbouc NH are indicated by > and < , respectively. The residues bracketed by such a hydrogen bond arc denoted 3, 4 or 5, unless they arc also the ezld points of another hydrogen bond. X indicates that both CO and NH are hydrogen-bonded. Bend row: S is a region of 5 residues centered at residue i with high ourvat,ure. Chirality row: the sign of the dihedral angle defined bar CT- 1 to C7,,. Bridge I and Bridge 2 70~s: the names of the @adders in which each residue psrticipetes, A, H. etc. for antiparallel ladders. Sheet row: the “.>nln nCtL% A-&t,,.+ :n .r,lr:,& CL,, -.&A..- nnmi:r:--+^m a..- ii^L.*.^L Y. U....A^.. ,lllU’)\ C^.. “-^i.^ i^‘^:li

Ambler Sequence Summary 3-Turn 4-Turn B-Turn Bend Chirality Bridge I Bridge 2 Sheet

Bend Chirality Bridge 1 Bridge 2 Sheet

5TUlYl

Ambler Sequence Summary 3mTurn 4-Turn

Ambler Sequence Summary 3-Turn 4-Turn B-Turn Bend Chirality Bridge 1 Bridge 2 Sheet

Ambler sequence Summary 3-Turn 4-Turn 5-Turn Bend Chirality Bridge 1 Bridge 2 Sheet

Secondary

709

P-Lactarnase at 24 A Resolution

K267

K267 >\

(b)

Figure 4. Stereoscopic representation of the main-chain tracing of the major P-sheet. The polypeptide chain segment that crosses the domain perpendicular to the direction of the strands and leads to the active site Ser70 is also shown. Bonds between main-chain atoms are filled and those between selected side-chain atoms are open. The hydrogen bond interactions between the side-chains of Glu37 and As&l and the main-chain atoms of the neighboring strand are shown in broken lines. (a) Face-on view; (b) edge-on view. The helical domain is located to the right of the P-sheet in this orientation, whereas the 3 helices involved in forming the open face P-sandwich domain are to the left of the sheet.

710

0. Herzberg

Figure 5. Cp,$ plot of j?-lactamase. Glycine residues (0); all others (+). The continuous lines enclose regions of equi-energy for an alanine dipeptide (N-formyl-alanylamide) according to Peters & Peters (1981). The contouring is at 40, 60 and 8.0 kcal/mol (1 cal = 4184 J).

shows two views of this five-stranded /?-sheet, face on and edge on. The first two strands are contained within a continuous polypeptide chain segment close to the N terminus, and the last three are

formed by a continuous polypeptide chain segment close to the C terminus. An interesting aspect of the architecture of the sheet is that the edge strand of the N-terminal segment (residues 56, 57, 59, 60: Amber (1979) numbering scheme) is four residues shorter than the second strand (residues 43 to 50). All potential /?-sheet hydrogen bonds from the

longer strand to its adjacent P-strand at the C-terminal segment are satisfied. However, because of the shortness of the edge strand of the N-terminal segment, Ile44 is not involved in B-sheet interactions. Instead, Asn61 turns sharply at the start of the first inter-domain cross-over, and adopts leftbanded helix 4, II/ values. Thus, the side-chain nitrogen atom of this asparagine is able to hydrogen bond to the main-chain oxygen of He44 of the second strand in a manner that mimics an extension of the strand. To complete the hydrogen bonding pattern, the main-chain nitrogen of Ile44 is hydrogen-bonded to the carboxylate moiety of Glu37, a conserved residue in all class A fi-lactamases. As shown in Table 7; in addition to sheet A, there are two short ladders, E and I?, two and t.hree residues long, respectively. They correspond to two small sheets, B and C. Sheet B (not shown in Fig. 3(a) because of its shortness), involving residues 66-67 and 180-181, is located in the interfa,ee between the two domains. It links the first interdomain cross-over segment to the bottom of an R-loop (residues 163 to 178), a structural motive (Leszczynski & Rose, 1986) that flanks one side of the active site. The close packing of sheets A and B creates a crowded environment for residues 66 to 69, and this helps anchor the active site Ser70. The second small sheet C stabilizes a large loop structure (residues 83 to 119) that flanks the top of the structure as viewed in Figure 3. The 4-@ dihedral angle distribution of the S. aureus P-lactamase is shown in Figure 5 within the context of regions of equi-energy for an alanine dipeptide according to Peters & Peters (1981). All but two residues that have conformations outside the energetically favorable regions are glycines. The P-carbon-containing residues that lie in the high energy region of the 4, $ surface are Ala69 (42”?

Figure 6. Stereoscopic representation of the location of conserved amino acid residues from 15 sequences of p-lactamases in the structure of the 8. aureus enzyme. Virtual bonds between a-carbon atoms are filled and the sidechain bonds of the conserved residues are open.

/I-Lactamase at 2.0 d Resolution

711

Table 8 Proposed role for the conserved residues in class A fi-lactamases Residue Glu37 Gly45

Phe66 SW70 Lys73 Leu81 Ser130 Asp131 Ala134 Am136

Role

Location First a-helix in p-sheet domain First /?-strand of the main /?-sheet

On the cross-over /?-sheet domain u-domain Active site Active site On a buried helix cl-domain Active site Close to the active

from the to the

in the site

N-Helix in the helical domain Close to the active site

Gly144

On a turn at the C terminus of a helix in the cc-domain

Glyl.56

C Terminus of a helix in the u-domain Asx turn leading to the active site Q-loop On the R-loop

Asp157 Arg164 Glu166 Leu169 Asp179 Thr180 Leu207 Asp233

Active site On the n-loop, adjacent to the active site End of the n-loop Close to the Q-loop On an a-helix in the cc-domain Edge b-strand, by the active site

Lys234

Active site

Gly236

Active site

Strand stabilization.

Described in Results and Discussion, section (d)

Core packing requirements. A /3-carbon-containing residue would clash with the CO of Am61 (which adopts a left-handed helical conformation such that the polypeptide chain turns to form the first inter-domain cross-over), with the side-chain of Pro183, and with the side-chain of the conserved Phe66 Core packing requirements; same environment as Gly45

Involved in catalysis Involved in catalysis Core packing requirements. Should be replaceable provided that the mutation is coupled to other compensating mutations The structure suggests involvement in catalysis (Results and Discussion, section (f)) This buried charge interacts with 4 main-chain nitrogen atoms, 2 of the N terminus of an a-helix and 2 of a 31,,-helix, assuring the precise positioning of Ser130 Core packing requirements. Contacts the CO of Leu122, and the side-chains of Ala125 and LeulO9. Should be replaceable by a glycine residue The side-chain interacts with the catalytic Glu166 which is involved in a non-proline cis peptide bond (Results and Discussion, section (f)) Folding requirements. Its 4,1(1 dihedral angles are loo”, 161”, a region not usually occupied by /l-carbon-containing residues. This conformation results in 2 main-chain nitrogen atoms forming hydrogen bonds to a carbonyl oxygen atom at the C terminus of the helix (Schellman, 1980) 4, $ = 91”, 18”. On the edge of the left-handed a-helix region. Should be replaceable Exposed to solvent, and does not interact with other charges. Should be replaceable at least by Asn, Ser or Thr Makes a salt bridge with the buried Asp179, thus stabilizing the Q-loop (Results and Discussion, section (f)) The structure suggests involvement in catalysis Packing requirements. But should be replaceable As Arg164 No special requirement; should be replaceable Similar environment to Leu81 A buried carboxylate, interacting with a buried solvent molecule, and a second buried carboxylate (Results and Discussion, section (g)). This unusual arrangement in the vicinity of the active site suggests this residue is essential for the integrity of the structure A model of substrate binding suggest it interacts with the essential negatively charged moiety of all b-lactams Packing requirements. Close to the catalytic Ser70. A P-carbon-containing residue would clash with the seryl residue

- 140”) and Leu220 (- lOSo, - 125”). The temperature factors of their main-chain atoms are lower than 20 A’, indicating that the structure is reliable in that region. Each of these residues is located on a cross-over between the two domains in proximity to the active site. The association of such steric strain with binding and catalysis has been discussed elsewhere (0. Herzberg & J. Moult, unpublished results). We have attributed it to -the greater precision necessary for ligand binding and catalysis, compared with the requirements of satisfactory folding? (e) Conserved residues As more sequences of class A /I-lactamase become available, the number of absolutely conserved amino acid residues reduces. Some of those

remaining appear to be conserved because of functional requirements, and others because they are located in positions crucial to the folding of the enzyme. Out of 15 sequences (A. Coulson, personal communication), 21 residues are invariant. Figure 6 shows their spatial location and Table 8 lists their structural/functional role as proposed by inspecting the crystal structure. Obviously, amino acid residues that are directly involved in catalysis or that are crucial to maintaining the shape of the active site depression are conserved. However, at least six of the other conserved residues could tolerate replacement. Three would require other compensating mutations to obtain satisfactory cored packing (Leu81, Ala134, Leu207), and three may be readily replaceable (Gly156, Asp157, ThrlSO). In some cases, the packing requirements seem more restrictive. Such an example is the

0. Herzberg

712

Figure 7. Stereoscopic view of the residues that form the active site depression of /I-lactamase. Bonds between mainchain atoms are filled and between side-chain atoms open. The positions of some water molecules that are closely associated with the active site are double-circled.

environment of the two neighboring residues, Gly45 and Phe66. There, any addition of side-chain atoms to residue 45 would also require altering the position of main-chain atoms. This may also affect enzyme activity, since Phe66 is located on the inter-domain cross-over chain leading to the active site. Although the invariant Asp233 is adjacent in sequence to the active site Lys234, its buried sidechain is not involved directly with the active site. Its carboxyl moiety is interacting with a buried solvent molecule, as well as with a second buried carboxylate, Asp246 (see section (g), below). The role of these two buried charges is not clear. The sequence alignment shows that in Gram-negative sidebacteria Asp246 is replaced by a hydrophobic chain (Ile or Leu), whereas all Gram-positive bacteria have an aspartyl residue at this position. At this stage, no structural rationale can be provided to explain this difference without the knowledge of a structure of a fl-lactamase from Gram-negative bacteria. Recently, the sequence of a carbenicillinase, PSE4, has been determined (Boissinot & Levesque, 1990), revealing that Lys234 can be replaced by an arginine still residue, maintaining function. Apparently, the change to another positively charged residue at position 234 is accommodated in some cases. However, subst,rate specificity may be affected through subtle modifications of the depression shape and the relative positions of key functional groups. (f) The active site The active site depression is located at the interface between the two domains, with the primary catalytic residue, Ser70, at the bottom of the depression. Residues that form the active site are located on the N terminus of the second a-helix (sequentially); on part of an edge /?-strand of the major P-sheet (3rd strand, sequentially), on the Q-loop, and on two turns: residues 104-105 of the

large loop spanning residues 83 to 119, and residues 130 to 132 between the two helices 119 t,o 129 and 132 to 142 (Fig. 7). Most of the key residues that are important for function and the implication of their disposition to the catalytic mechanism have been described (Herzberg & Moult: 1987). These include Ser70, Lys73, Glul66, Lys234 and the oxyanion hole formed by Ser70 and Gln237. The refinement at high resolution provides additional information that’ will be discussed here. The conserved residue Serl30 is interacting with the active site Ser70 (Oy-Oy, 3.3 A), with Ljrs73 (OY-Nr, 3-7 A), and with Lys234 (OY-NC, 3.0 A), suggesting that it plays a role in catalysis. In the original structure determination (Herzberg & Moult, 1987) we have shown that, although the overall fold of the polypeptide chain of P-lactamase does not resemble that of any serine protease of known structure, the spatial disposition of the respective active site seryl side-chains and the oxyanion hole nitrogen atoms is very similar. Moreover, this similarity extends to the location of the /?-lactam bond in modeled substrate-enzyme complex and that, of t,he peptide bond of an inhibitor bound to a serine protease. The alignment of the two structures also places the catalytic histidine of the serine protease between the ammonium group of Lys73 and the Oy atom of Serl30 in /?-lactamase. Recently, the structure of a class C P-lactamase has been determined where t,he analogous position of the seryl hydroxyl group is occupied by a tyrosine hydroxpl group (Tyr150). The authors suggest that the tyrosine hydroxyl is negatively charged, acting as a general base in the hydrolysis of the p-lactam (Oefner et al. 1990). For Ser130 to be negatively charged would require a dramatic drop in the pK, value, never demonstrated in any other system. The charge distribution of the enzyme-substrate complex, wit,h two lysine residues (Lys73 and Lys234), a glmamic acid (Glu166), and a carboxylate moiety of the p-lactam compound, a.11in close proximity, is rather complicated. In addition, there is an intrica.ts

P-Lactamase at 2-O d Resolution

713

N170

Figure 8. Stereoscopic representation of the environment of GM66 and Ile167. Bonds between main-chain atoms are filled and between side-chain atoms open. Important stabilizing interactions are shown as broken lines.

arrangement of dipoles, which also affect the overall electrostatics of the active site. Although the spatial position of Ser130 is suggestive of its crucial functional role, a careful electrostatic analysis is needed in order to decide whether it could carry a negative charge at any stage of the catalytic process. The peptide bond between Glu166 and Ile167 is in the cis conformation (Figs 7 and S), which is rare for non-proline residues in known crystal structures. Both residues appear important for function: the position of Glu166 suggests it is important for catalysis. We have suggested that Glu166 plays a role by enhancing the nucleophilicity of a water molecule positioned between the carboxylate group and the active site seryl residue, and that the function of the water molecule is to deacylate the substrate. Ile167 provides hydrophobic character to the active site gully, together with Ile239 and Ala104. We have proposed that the hydrophobic tip of the /I-lactam side-chain substituent would interact favorably with such hydrophobic environment. The cis bond is instrumental in helping to define the shape of the gully as well as the precise positioning of Glu166. The cis peptide bond is located on the Q-loop, a structural unit that is somewhat isolated from the rest of the structure by two internal spaces filled with solvent molecules. The stability of the R-loop appears to be marginal because of its isolation from the rest of the structure and the presence of the nonproline cis peptide bond. On the other hand, a salt bridge between Arg164 and Asp179 at the base of this loop should enhance its stability. Other contributions to the stability of the R-loop are: the interaction of the cis peptide with the side-chain of the conserved Asn136; the hydrogen bond between Glu166 and Asn169; the helical hydrogen bond between the CO of He167 and the N of Asnl70; the salt bridge between Glu166 and Lys73. The balance between all these interactions is apparently sufficiently favorable for the protein to maintain a unique conformation for the R-loop. Other experimental data that suggest the presence of a marginally stable piece of structure associated with the active site are: (1) the identification of non-active

stable folding intermediates close to the native structure under denaturing conditions (Creighton & Pain, 1980); (2) the inactivation of the enzyme by substrates with bulky side-shains (Citri et al., 1976). The transition between ordered and disordered conformation of the R-loop is consistent with the above phenomena. The side-chain of Tyrl05 is located along one wall of the active site depression (Fig. 7). It is a conserved residue in 12 of the 15 p-lactamase sequences. The remaining three sequences contain large deletions around the tyrosyl residue, implying a different conformation of this part of the structure. The tyrosyl side-chain may interact with the substrate in a manner that is not clear, but there is no evidence that this should be an essential interaction. Ala69, located before the active site Ser70, and Leu220 on the second cross-over from the helical domain to the /?-sheet-containing domain both adopt strained 4, $ dihedral angles (see section (d), above). The j-carbon atom of Ala69 makes van der Waals’ interactions with the main-chain atoms of Gln237, a residue participating in the oxyanion hole (Fig. 9). The b-carbon atom of Leu220 contacts the P-carbon of Ser235, also part of the active site B-strand. Such a crowded environment restricts the conformation of the P-strand, presumably assuring the precise positioning of the oxyanion hole. The conserved Asp131 lies between the active site Serl30 and Asn132 with its carboxylate moiety buried in the interior of the protein (Fig. 10). This charge is stabilized by interactions with four mainchain nitrogens: two on the following helix (Thr133 and Ala134), and two on the N terminus of a 3,,-helical segment (Ile108 and LeulO9). Such multiple interactions fix the position of Aspl31, and therefore help in maintaining the orientation of Ser130 relative to Ser70. The active site depression

is occupied

by several

solvent molecules. Clearly, most of these have to be displaced for a substrate to bind to the protein. The determination of the positions of solvent molecules in the native structures of proteins may help in

714

0. Herxberg -_

Figure 9. Stereoscopic representation of the environment of Ala69 and Leu220, the 2 residues with strained 4, $ dihedral angles. Bonds between main-chain atoms are filled, and those between side-chain atoms open. van der Waals’ contacts to the active site p-strand are shown in broken lines.

identifying those that play a role in catalysis, and provide insight into the mechanism. The water molecule bound between the two aspartyl residues in the aspartyl proteases is such an example (James & Sielecki, 1987). Of particular interest in fl-lactamase is a solvent molecule interacting with the sidechains of Glul66, Ser70 and Asnl70 (Fig. 7). That solvent molecule (Wats1 in the co-ordinates deposited in the Brookhaven Data Bank) has a temperature factor and occupancy consistent with it being a water molecule. It is located exactly in the position that has been predicted by us to provide the water that would deacylate the acyl-enzyme complex (Herzberg & Moult, 1987).

A second water molecule (Wat22 in the deposited co-ordinates) is located between the two main-chain nitrogen atoms that form the oxyanion hole (Ser70 and Gln237), providing an energetically favorable interaction that mimics the interaction of the carbonyl oxygen of a p-la&am substrate previously proposed by us. The distance from Wat22 to the Ser70 main-chain nitrogen is 3.2 d, and that to Gln237 nitrogen is 3.0 8. An equivalent water molecule has been identified in the oxyanion holes of some serine proteases: proteinase A (BSGA; Sieleeki et al., 1979), proteinase K (ZPRK; Betzel et al., 1988), and a-chymotrypsin (4CHA; Tsukada & Blow, 1985). However, this analogy is complicated

Klll

Kill El1

1108

Figure 10. Stereoscopic representation of the environment of the buried Asp131 Bonds between main-chain atoms are filled, and those between side-chain atoms open. The broken lines indicate the interactions between the carboxyl. oxygen atoms and main-chain nitrogen atoms.

/?-Lactamase at 2.0 ,d Resolution

715

Table 9 Charge-charge interactions in S. aureus PC1 fi-lactamase Inter-molecular

Intra-molecular

Salt bridge

d (4

Positivepositive

d (4

Carboxylcarboxylate

d (4

K39-E281 H43-D268 K73-El66 H96-D116 K120-El24 R151-ES2 R151-El54 R164-D179 K222-D233 R244-D276 K277-D223

29

R164-K178

34

D233-D246

2.4

Salt bridge K31-D116 K38-DlOl K93-E32 K215-El24

2.5 24 25 29 29 30 28

d (4 34 32 3.2 3.2

crystal contacts Positivepositive None

Carboxylcarboxylate None

32 2.9 33

by the possible binding of peptides resulting from self-digestion of the protease, as has been observed in y-chymotrypsin (Dixon & Matthews, 1989). Clearly, fi-lactamase does not have such proteolytic activity. Three solvent molecules are located between Ser70, Serl30, Lys234 and Ser235 (Wat64, Wat71 and Watlll in the deposited co-ordinates). They are associated with a flat three-lobe electron density feature. The possibility that an anion is occupying this position has been investigated, since it is flanked by two positively charged residues: the guanidinium group of Arg244 (3.4 A to Watl 11)) and the ammonium group of Lys73 (3.6 A to Wat64). A sulfate anion was excluded, since the shape of the density is not tetrahedral, and the solvent atoms refined with too high temperature factors to be consistent with a sulfur-containing anion. Since the crystallization media contains bicarbonate buffer, bicarbonate ion was incor-

D246

porated into the refinement. However, it also refined with a very high temperature factor. Thus, it was concluded that a model of three water molecules is most appropriate. (g) Charge-charge interactions Charge-charge interactions in B-lactamase are summarized in Table 9. There are 11 intramolecular salt bridges and one of each of a carboxylcarboxylate interaction and a guanidiniumammonium interaction. There are four intermolecular salt bridges, and no other charge-charge interactions shorter than 3.5 8. Of the intramolecular salt bridges, three occur between pairs of helices and two between adjacent P-strands. Two are between a residue on a helix and a residue on a b-strand, and two are i to i+4 intra-helical interactions. One salt bridge occurs from an a-helix to the R-loop (between the active site Lys73 and

D246

Figure 11. Stereoscopic view of the environment of the buried Asp233 and Asp246. Bonds between main-chain atoms are filled, and bonds between side-chain atoms open. The position of the solvent molecule Wat4 is double-circled. The carboxy-carboxylate interaction is shown with a broken line, as well as the salt bridge between Asp233 and Lys222. The interactions between Wat4 and the protein are also indicated by broken lines.

716

0. Herzberq

(a)

Figure 12. Stereoscopic view of histidyl&aspartate interactions in /3-la&amass. Bonds between main-chain atoms are filled, and between side-chain atoms open. The salt bridge interaction is indicated by a broken line, as is the main-chain interactions between adjacent P-strands. (a) His43 to Asp268; the interaction between Giu37 carboxylate and the mainchain nitrogen of the P-strand residue Ile44 is shown with a broken line. (b) His96 to Aspll6.

B-Lactamase

at 2.0 ,d Resolution

Glu166), and one is lying between the two ends of the Q-loop (Arg164 and Asp179). The intermolecular salt bridges are exposed to a large solvent channel. There are plenty of positively charged residues in close proximity to each other on the surface of the molecule, though most of them are more than 3.5 A apart (/3-lactamase from S. aurezcs PC1 has 43 lysine, 4 arginine and 2 histidine residues, in contrast to 18 aspartic acid and 14 glutamic acid residues). The crystals have been obtained at pH 8. Although not seen in the electron density map, hydroxyl or other counter ions in the solvent channels are expected to stabilize the excess positive charges otherwise crystals would not be formed. Considering the high pH, the carboxylcarboxylate interaction is unusual. Such an interaction is associated with an elevated pK, value of about 6 (Sawyer & James, 1982; Herzberg & James, 1985), but not as high as 8. Figure 11 shows the environment of Asp233 and Asp246. In addition to the carboxyl-carboxylate interaction the conserved Asp233 makes a salt bridge with Lys222. In all but one of the 15 sequences of class A /l-lactmases residue 222 is positively charged, suggesting that this is an essential interaction. In addition, a wellordered internal solvent molecule (WaM) interacts with a carboxyl oxygen atom of Asp246, with the carbonyl oxygen of Gly217, and with the mainchain nitrogen atom of Lys222. The temperature factor (8 8’) and occupancy (699) is consistent with it being either a water molecule or an ammonium ion. A cation would balance the overall net charge of this environment; however, the interaction with a main-chain nitrogen atom conflicts with such an interpretation, and the co-ordination of the solvent is not tetrahedral as expected for an ammonium ion. The deposited co-ordinates list the solvent molecule as water, in which case its polarization is expected to be substantial due to the interactions with the carboxylate. Both of the two histidyl residues in p-lactamase make close interactions (2.5 A) with aspartate sidechains (Fig. 12). Each interaction occurs across antiparallel b-strands, contributing to the sheet integrity. In one case the b-nitrogen atom of the imidazole ring is involved, and in the other it is the a-nitrogen atom. These interactions and the pH of the crystals suggest that the pK values of the histidyl residues are elevated by at least one pH unit. Most of the salt bridges are located on the surface of the protein; however, in two salt bridges (Arg164-Asp179 and Lys222-Asp233, each of which occurs in the vicinity of the active site) the aspartate residues are buried. Asp233 is discussed above. Arg164 and Asp179 are conserved residues in all class A /I-lactamases, suggesting that this salt bridge is essential for function. Support for this proposal is provided by the observation that the mutant P54 of S. aureus /I-lactamase in which Asp179 is replaced by an asparagine has very low

717

activity (Ambler, 1979). This salt bridge stabilizes the Q-loop, as discussed in section (f), above. The active site salt bridge (Lys73-Glu166) is located at the bottom of the depression, such that it is only marginally solvated. Clearly, substrate binding results in the complete desolvation of these charges. This potentially complicates site-directed mutagenesis experiments in which these residues are replaced one at a time, attempting to determine their involvement in catalysis. For example, Madgwick & Waley (1987) have shown that replacing Glul66 by a glutamine residue inactivates the enzyme. However, it is conceivable that the active site conformation is altered to permit solvation of the remaining single charge. Thus the catalytic properties may be affected by changes in structure, not just by the replaced functional group. (h) Solvent structure A total of 207 solvent molecules were located in the crystal structure of /I-lactamase. The electron density map associated with all of them is consistent with the scattering of water molecules. In some cases the possibility of a binding site for an ion was checked, but the type of the interactions involved or the refined temperature factors did not justify a change in interpretation. The water molecules have been arranged in descending order of reliability, where the reliability is defined by the quality factor: 100*occupancy2/B (James & Sielecki, 1983). The striking observation is that less than a third of them have a quality factor above 2.5 (equivalent to B = 40 and an occupancy of 1). This is associated with the high frequency of lysine residues on the surface of the protein whose temperature factors are high as well, and with the large solvent channels in the crystal indicating possible high mobility of water molecules within the channels. The /I-lactamase crystal structure has 16 internal solvent molecules (Table 10). In the original publiTable 10 Internal

Cavity 1

solvent

Solvent

molecules in /I-lactamase

100* Occupancy2/B

Wat14

48

WatlB

4.32

Wat20

3.9

S. aureus PC1

Interactions 06’ 0” 0 C” Na2 Oy Na2 0 0 Cd2 0 0 0 0 C6’ Cd2

Asp179 Tyr68 wat20 Pro162 Asn76 Ser172 As11135 wat20 Wat62 Led69 Argl64 Wat132 Wat14 WatlB Ile145 Led69

d (4 2.6 27 2.8 3.4 2.8 30 31 30 3.4 3.5 2.6 27 28 3.0 3.4 3.4

0. Herzberg

718 Table 10 (continued) 100* Cavity

Solvent Wat62

OccupancyZ/B 2.4

Wat124

1.7

wat132

1.6

Wat2

9.0

Watll

51

Wat13

4.8

Wat5

7.3

Wat37

2.9

Watl

121

Wat8

59

Wat7

64

wat4

8.1

Wat90

2.0

Interactions Oy 0”’

Ser72 GM66

0 0 Od’ Nd2

Ala69 W&l5 Am76 Asn76

0 0 0 Cd1 CD N 0 0 N 0 0 0 0 N 0 0 0 N”’ C” Oy 0 0 N 0 0 0 6” 0 Od’ 0 0 Oy Ndz N 0 Oy 061 Oy 0 C” 0 06’ N 06’ N cy 0

Wat132 Wat124 Wat20 Ile145 Pro162 Ala69 Asp179 Led69 Tyr68 Wat13 Lys178 Ser176 Ser173 Ser173 Wats:! Led69 Lys178 Arg164 Tyr172 Thr265 ,41a242 Wat37 Am275 hsn275 Wat5 Wat185 Pro274 Thr265 Asn245 Ala67 Wats Ser243 Asnv66 i

0

Met127

Cy

Met211

Ala67

Watl Ser126 Asn135 Ser77 Lys73 He138 Gly217 Asp246 Lys222 Asp233 He221 Lys222 Asp233

d (4 F5 2.6 3.3 34 28 33 2.8 28 39 31 33 2.9 2.9 30 3.5 3.5 2.9 2.9 33 3% 2.8 2.9 31 32 32 2% 2.9 26 27 33 %6 35 33 2.8 2.8 3.2 2.6 2.8 29 30 26 2.5 2.9 3.1 3.1 3.3 2.4 2.7 2.9 32 32 30 30 32 35

solvent molecules lie in a space that corresponds to the narrow exit to solvent on the side of the Q-loop. These cavities are associated with the loose packing between the Q-loop and the rest of the structure. the internal solvent molecules are Finally, involved in intricate hydrogen bonding interactions with protein residues and adjacent solvent molecules. Some close non-polar contacts are also observed. Most of these water molecules have temperature factors similar to the surrounding protein, and nearly full occupancies, with the exception of two molecules in the largest cavity. I thank John Moult, Tom Poulos, Gary Gilliland and Walter Stevens for many helpful discussions. I am grateful to Gary Galliland for guiding me through a new data collection process when I first came to CARB, and to Andrew Coulson for the sample of protein prepared in his laboratory. This work was supported by NIH grant ROl-A127175.

References Abraham, E. P. & Chain, E. B. (1940). Nature (London), 146, 837. Ambler, R. P. (1979). In Beta-Lactamases (HamiltonMiller, J. M. T. 8: Smith, J. T., eds), pp. 99-125, Academic Press, London. BarthBlBmy, M., PAduzzi, J., Ben Paghlane, H. & Labia, R. (1988). FE&S Letters, 231, 217-220. Bernstein, F. C., Koetzle, T. F., Williams, 6. J. B., Meyer, E. F., Jr, Brice, M. D., Rodgers, J. R., Kennard, 0.. Shimanouchi, T. & Tasumi, M. (1977). J. MOE. Bid. 112, 535-542. Betzel, C., Pal, G. P. &. Saenger, W. (1988). Acta Crystallogr. sect. A, 44; 163-172. Bhat, T. N. (1988). J. Appl. Crystallogr. 21, 279-281. Boissinot, M. 8: Levesque, R. C. (1990). J. Biol. Chem. 265, 1225-1230. Citri, pu’., Samuni, A. C Zyk, N. (1976). Yroc. Na.t. Acad. Sci., U.9.A. 73, 1048--1052. Coulson, A. (1985). Biotechnol. Genet. Eng. Rev. 3, 219-253.

Creighton: T. E. & Pain, R. H. (1980). J. ,woZ. Biol. 134, 431-436. Dideberg, O., Charlier, P., W&y, J. P., Dehottay, P.. Dusart, J., Erpicum, T., Fr&re, J.-M. & Ghuysen, J.-M. (1987). Biochem. J. 245, 911-913. Dixon, M. M. & Matthews, B. W. (1989). Biochemistry, 28, 1033-1038. Finzel, B. C. (1987). J. Appl. Crystallogr. 20, 53-55. Frere, J.-M. & Joris, B. (1985). CRC Grit. Rev. Microbial.

11, 299-396. Hamilton-Miller, J. M. T. & Smith, J. T. (1979). Editors of Beta-Lactamases, Academic Press, London. Hendrickson, W. A. & Konnert, J. H. (1980). In

cation of the structure determination we noted the existence of a large cavity, t,ermed the cave, because of its narrow exit to solvent. Refinement at high resolution resulted in a reduction of the cave volume. It is also segmented to two parts by protein atoms that shifted during refinement. Nine solvent molecules lie within this volume. Six of them fill the larger segment located underneath the active site depression, surrounded mainly by residues on the active site helix and on the !&loop. The other three

Biomolecular Structure, Bvolution (Srinivasan,

Function, R., ed.),

Conformation and vol. 1, pp. 43-75.

Pergamon Press, Oxford. Herzberg, 0. & James, M. r\‘. G. (1985). Nature (London?, 313, 653-659. Herzberg, 0. & Moult, J. (1987). Science, 236, 694-701. Hol, W. G. J. (1985). In Prog. Biophys. Mol. Biol. 45; 149-195.

Howard, A. J., Gilliland, 6. L., Finzel, B. C., Poulos, T., Ohlendorf, D. 0. & Salemme, F. R. j1987). J. Appl. Crystallogr.

20, 383-387.

P-Lactamase

at 8.0 A Resolution

James, M. N. G. & Sielecki, A. R. (1983). J. Mol. Biol. 163, 299-361. James, M. N. G. & Sielecki, A. R. (1987). In Biological Macromolecules and Assemblies; Active Sites oj Enzymes (Jurnak, F. A. & McPherson, A., eds), vol. 3, pp. 413-482, John Wiley & Sons, New York. Jones, T. A. (1982). In Computational Crystallography (Sayre, D., ed.), pp. 303-317, Oxford University Press, London and New York. Joris, B., Ghuysen, J.-M., Dive, G., Renard, A., Dideberg, O., Charlier, P., Frere, J.-M., Kelly, J. A., Boyington, J. C., Mowes, P. C. & Knox, J. R. (1988). Biochem. J. 250, 313-324. Kabsch, W. & Sander, C. (1983). Biopolymers, 22, 2577% 2637. Kelly, J. A., Dideberg, O., Charlier; P., W&y, J. P., Libert, M., Mowes, P. C., Knox, J. R., Duez, C., Fraipont, Cl., Joris, B., Dusart, J., Frere, J.-M. & Ghuysen, J.-M. (1986). Science, 231, 1429-1431. Knowles, J. R. (,1985). Accts. Chem. Res. 18, 97-102. Kraut, J. (1977). Annu. Rev. Biochem. 46, 331-358. Leszczynski, J. F. & Rose, G. D. (1986). science, 234, 849-855. Madgwick, P. J. & Waley, S. G. (1987). Biochem. J. 248, 657-662. Matagne, A., Misselyn-Baudin, A.-M., Joris, B., Erpieum, T., Granier, B. & Frere, J.-M. (1990). Biochem. J. 265; 131-146. Moult, J., Sawyer, L., Herzberg, O., Jones, C. L., Coulson, Edited

719

A. F. W., Green, D. W., Harding, M. M. & Ambler, R. P. (1985). Biochem. J. 225, 167-176. Mowes, P. C., Knox, J. R., Dideberg, 0.; Chalier, P. & F&e, J.-M. (1990). Proteins: Struct. Funct. Genet. 7, 156-171. Oefner, C., D’Arcy, A., Daly, J. J., Gubernato, K., Charnas, R. L., Heinze, I., Hubschwerlen, C. & Winkler, F. K. (1990). Nature (London), 343, 284-288. Peters, D. & Peters, J. (1981). J. Mol. Struct. 85, 1077123. Phillips; S. (1980). J. Mol. Biol. 142, 531-554. sect. A, 42, 140-149. Read, R. J. (1986). Acta Crystallogr., Richardson, J. (1981). Advan. Protein Chem. 34, 167-339. Samraoni, B., Sutton, B. J., Todd, R. J., Artymiuk, P. J., Waley, S. G. & Phillips, D. C. (1986). Nature (London), 320, 378-380. Sawyer, L. & James, M. N. G. (1982). Nature (London), 296, 79-80. Schellman, G. (1980). In Protein Folding (Jaenicke, R., ed.), pp. 53-61, Elsevier/North-Holland, New York. Sielecki, A. R., Hendrickson, W. A., Broughton, C. G., Delbaere, L. T. J., Brayer, G. D. & James, M. N. G. (1979). J. Mol. Biol. 134, 781-804. Sougakoff, W., Goussard, S., Gerbaud, G. & Courvalin, P. (1988). Rev. Infect. Dis. 10, 879-884. Tipper, D. J. & Strominger, J. L. (1965). Proc. Nat. Acad. Xci., U.S.A. 54, 1133-1141. Tsukada, H. & Blow, D. M. (1985). J. Mol. Biol. 184, 703-711.

by R. Huber