J. Mol. Biol. (1992) 223, 317-335

Structure of a. Ternary Complex of an Allosteric Lactate Dehydrogenase f!om Bacillus stearothermophilus at 2.5 A Resolution Dale B. WigleylY3-j-, Steven J. Gamblin’, Johan P. Turkenburg3, Eleanor J. Dodson3 Klaus Piontek4$, Hilary Muirhead’ and J. John Holbrook2 ‘Department of Biochemistry, University Leicester LEl YRH, U.K.

of Leicester

2Molecular Recognition Centre and Department of Biochemistry University of Bristol School of Medical Sciences Bristol BS8 ITD, U.K. ‘Department

of Chemistry, University York YOl 5DD. U.K.

of York

4Department of Biological Sciences, Purdue University West Lafayette, IN 47907, F.S.A. (Received 30 January

1991; accepted 30 September 1991)

We report the refined structure of a ternary complex of an allosterically activated lactate dehydrogenase, including the important active site loop. Eightfold non-crystallographic symmetry averaging was utilized to improve the density maps. Interactions between t)he protein and bound coenzyme and oxamate are described in relation to other studies using site-specific mutagenesis. Fructose 1,6-bisphosphate (FruP,) is bound to the enzyme across one of the P-fold axes of the tetramer, with the two phosphate moieties interacting with two anion binding sites, one on each of two subunits. across this interface. However, because FruP* binds at this special site, yet does not possess an internal 2-fold symmetry axis. t#he ligand is statistically disordered and binds to each site in two different orientations. Binding of FruP, to the tetramer is signalled to the active site principally through two interact’ions with His188 and Arg173. His188 is connected to His195 (which binds the carbonyl group of the substrate) and Arg173 is connected to Arg171 (the residue that binds the carboxylate group of the substrate).

Kpyu,ord.s: ternary

complex;

allosteric;

lactate dehydrogenase: active site loop

1. Introduction dehvdrogenase

(LDHase$)

stearothermophjtua is one of several

7 Buthor to whom correspondence should be addressed at present’ address: Department of Chemistry. University of York, York YOl 5DD, U.K. fPresent address: Laboratorie fiir Biochemie. ETH-Zentrum, CH-8093 Ziirich, Switzerland. 9:Abbreviations used: LDHase, lactate dehydrogenase; FruP,. fructose 1.6-bisphosphate; E. enzyme. OOZZ-1836/92/010317-19

$03,00/O

stearothermoph.il,u,,s;

stability (Wigley et al., 1987a) and allosteric activation by fructose 1,6-bisphosphate (FruP,: Clarke, et al., 19876; Schroeder et al., 1988) have all been investigated. From these studies, rules governing substrate specificity have been developed (Wigley et al., 19876) that were used to alter the specificity of the enzyme to accept malate rather than lactate as substrat,e (Clarke et al., 19876: Wilks et al.. 1988). The enzyme has been engineered further to accept substrates larger than malate (H. Wilks & J.J.H., unpublished results), such that the enzyme is now a more general keto-acid dehydrogenase. Such an enzyme is of use in the stereospecific reduction of carbonyl groups during the synthesis of organic compounds, an important step in t,he production of many pharmaceuticals.

from Bacillus enzymes that have been the subject of intensive study by sitedirected mutagenesis. The functions of residues involved in catalysis (Clarke et al., 1986a, 1988), substrate binding (Hart et aZ.. 1987a,b), thermal Lactate

Bacillus

317

(c, 1992 Academic

Press Limited

3 I8

1). 8. CVigley et, al.

---

.

Table 1 ( ‘rystal ,form,s

Space-group

(‘4

Orthorhombic

dimensions

per unit

(I = 86 A

p8,2,2

h = 105 c=136p\

Monoclinic

4 h c B

P2,

Monoclinic p2,

= = = =

A

2.27

4 ( I35,M)0

849 A 118.2 4 135.6 A 96.07’

2.52

n=112A h=85A c = 136 /I = 91’

Da) 8

(270,000

Da)

(27”.&~

T)a)

240 .k

Although these studies were based on existing crystal structures of pig and dogfish LDHases (Grau et nl., 1981; White et al., 1976; Rossmann et al.. 197I), some assumptions concerning the structure of the B. stearothermophilus enzyme had to be made. Tn addition, the study of existing mutant enzymes and the design of further mutants with new specificities would be considerably aided by the availability of a high resolution structure of the wild-type enzyme. The ideal subject for such structure studies would be the “activated” enzyme complex comprising FruP,, NADH and oxamate (a close structural and electronic analogue of pyruvate, which is a tight binding inhibitor of LDHase). For these reasons, the structure of such a complex of wild-type B. stearothermophilus LDHase was solved as a prelude to further study of some of the mutant enzymes. Many baterial LDHase, unlike those from eukaryotic sources, are allosterically activated by FruP, (Garvie, 1980). Consequently, several groups have attempted to solve the crystal structure of a LDHase from a bacterial source. Crystals of the apo-enzyme (E/Co’+/FruP,) from Lactobacillus casei (Buehner et al., 1982) and of both the apoenzyme and a binary complex (E/NADH/FruP,) from B. stearothermophilus (Schar et al.. 1982) have been reported, all crystals having been obtained using ammonium sulphate as precipitant. The structures of all of these different complexes have now been solved (Buehner et al., 1982; Piontek et al., 1990). Here, we describe the solution of the crystal structure of a ternary (E/NADH/oxamate/FruP,) complex of B. stearothermophilus LDHase obtained from polyethylene glycol. 2. Experimental

Subunits asymm.

(*&ton)

Procedures

(a) Crystallization Crystallization conditions have been described elsewhere (Wigley et al., 1988). At least 5 different crystal forms were grown under similar conditions from polyethylene glycol 6000, of which 3 were subjected to X-ray analysis(Table 1). Although more suitable from a crystallographic viewpoint, the orthorhombic crystals did not diffract as strongly as other crystal forms. Of the 2

monoclinic forms, the type V crystals were only rarely obtained, so the type TV form was chosen for analysis. It is interesting to note the similarity in cell dimensions between the 3 crystal forms. As noted by Wigley rt al. (1988), the hOZ zones of the monoclinic crystal forms exhibit unexpected pseudo mm symmetry. Several other features of these photographs are worthy of comment: (1) The striking similarity between equivalent zones of different crystal forms and, to a lesser extent. between different zones of the same crystal form. (2) The tendency for OOE reflexions of even index to be stronger than those where 1 is not equal to 2n (suggesting an approximate 2, screw along c). (3) The non-primitive appearance of the reciprocal lattice as viewed down the a axis (k + 1 = 2n, strong) in the type TV crystals particularly for reflexions of low I index. This is indicative of a displacement of a half in h and in c, between 2 of the molecules in the cell. The separation on h is apparently closer to a half than is the displacement along c, as the relationship k + I = 2n is maintained at higher values of k than of 1. These features of the low resolution diffraction patterns proved to be important hints that aided interpretat#ion of the results of the rotation and translation functions. (b)

Data

collection

Two data sets were collect,ed, hoth at the Synchrotron Radiation Source at Daresbury, U.K. The 1st was collected from a single cLrysta1 using the 0% A (1 A = @I nm) wavelength radiation available on the Wiggler station. Data were collected, using B-film packs. to 47 w resolution by a 90” rotation about c* of the

Table 2 Data analysis No. of photographs No. of crystals Resolution (A) R merge cxl) All data Data > I,,,,,, No. of measurements No. of independent Percentage possible

where

I, is the

mean

(4.7 A)

i.5 FP.5 24,036 12,874 91

reflexions data

of Zhi observations

of reflexion

h

Ternary

Complex

of LDHase

Table 3 Data

analysis

(2.5

A)

No. of photographx No. of cT,Yst~als Kesolut~ion (A) L-rge CC!” 1 So. of mvasurement,s So. of indrpmdent

rrflexions

so. of Resolutic~n

(A) 30~0& I :iw 15~oc 1om 104K) 7.50 7~5Ok500 5xK-3~50 3~*5&3W 3Nb2~75 I.75 -r’~50

F > k

independent

reflexions

(9;,)

339 961 1852 7451 19.717 15.836 10.348 5942

(‘ompletrness (“0)

99

92

98 99 99 99 96 93 92

96 95 93 91 82 66 26

crystal. The films were deliberately underexposed to ensure that most of the strongest reflexions did not exceed the dynamic range of the film and thus were recorded accurat,rly, improving signal-to-noise in the rotation and translation functions (both of which are based on Pat&r-son functions, and are strongly dependent upon the reflexions of high intensity). The 2nd data set (to 2.5 A) was collected using 1.488 A wavelength and comprised data from 3 crystals. Tn bot’h cases, the films were scanned and digitized using an Optronics PlOOO rotating drum film scanner. with a 50 pm raster step. The low resolution data were processed using the CCP4 suite of programs (CCP4. Daresbury, C.K.), while the 2.5 L% data were processed using the Purdue programs (Rossman, 1979; Rossmann et al., 1979). Statistics for the 2 data sets are shown in Tables :! and 3.

3. Molecular

Packing

In order to facilitate the reader’s understanding of this rather complex structure determination, the molecular packing is presented in Figure 1. The rotation and translation function studies that support t.his model are presented in subsequent sections.

4. Molecular

Orientations

Determination of the molecular orientation and positions was achieved using the 4.7 A data, which contained most of the very strong reflexions. Because the LDHase tetramer shows 222 molecular symmetry (Adams et al., 1969, 1970), inspection of the K = 180” section of a self-rotation function map should reveal the positions of these molecular S-fold axes. With two tetramers in the asymmetric unit. six peaks corresponding to the three 2-fold axes of each tetramer were expected. These peaks should, of course, fall into two mutually orthogonal sets. In addition, it might be expected that certain other peaks in the self-rotation function would relate one tetramer t,o the other of the same crystallographic

from

B. stearothermophilus

:319

asymmrt,ric unit. Such peaks nerd not result from rotations of 1X0”. but in view of the pseudoart horhombic symmetry of the crystals wta might exprc$ them t,o lie close to the K = 1 HO“ src*tion. wit,h 4 close to 0” or 90” and $ close to 90 Orthogonal axes (X. Y, Z) were defined with the crystal axis c along X, a* along Y. h along Z. X11 rotation function results are expressed relative t,o t,his axial system. A self-rotation function was calculated (using the program PC)LAK.RFN from thr (VT’4 suit,r. Daresbury, I’.K.) using the data with amplitudes greater than the mean. between 15 and 6 A spacing. A radius of integrat’ion of 25 A was chosen. The rotation function was sampled at 5” intervals on 4 a,nd $, while K wa,s hrld at 180”. LVith t.wo tetramers in the asymmetric unit in spacegroup P2,, the t’heoretical height of‘ arl)- peak corresponding to an intramolecular 2-fold axis should be 2iioo of that of the origin. Hence a rotation function was calculated and the map produced was rontoured at every 57; int,erval from l.?“,, (Fig. 2). Several features were immediat,rly apparent. t)he most, obvious being the very large peaks at, $ = !a()“, 4 = 5” and, by symmetry. 9.5” (peaks I and 2. height = 71 ?lo of origin). There were also two large peaks at $ = 40”. 4 = 90” (peak 3. 500,, of origin) and at II/ = SO”, 4 = 95” (peak 4. 4P”,, of origin). Thr heights of all four of these peaks were around twice that expected for a single molecular dyad axis. The pseudo-orthorhombic symmet)ry of the crystals is also evident in this self-rotat,ion function. such that, there is an approximate mirror plane on the ti = 180” section at (b * 90”. This symmetry is not expected for monoclinic space-groups. This observat,ion provides further strong e;idence that at least t’he orientations of the two molecules of the asymmetric unit are related by approsimat,r orthorhombic symmetry. This self-rotation function could be interprrted such that the S-fold axes of the two t’et ramers are in vrry similar positions, either placing the two molecules in very similar orientations. or placing t)he molecular 2-fold axes coincident but> different in each case. In either sit,uation, self-rotat’ion functions should reveal peaks that place one tetramer onto the ot’her of the asymmetric unit. In the first instance. these rotations will be coincident with the molecular 2-fold axes (or their symmctr\; equivalents). so cannot~ be distinguished tram them. Tn the la,tter case, t)here must be rot,ationb placing one tthtramer in a given orientation onto another in a different orient’ation. Such rotations are not con&rained to lie on the K = 180” section. Hence a search on IC from 0 to 180” was conducted to search for any such peaks. The only signifira,nt peaks found in the rest of the self-rotation function were symmetry equivalents of peaks 1 to 4 on t)he IC = 180” section. These result)s suggested the two tthtramers of the asymmetric unit to be in very similar orientations, an interpretation t,hat, seemed t,o be supported by cross-rot,ation function st’udies (Wigley et al., 1988) using the unrefined pig H, NAD-S-lactate co-ordinates from t’hr Iirookhaven

II ,I 1 --*-----------I---------TI (0.365)

l

T4 (0.185)

-----

I I I

I

12. (0.815)

Y

: j

T3 (0.635:

Z 4

(b)

Figure 1. The molecular packing in the type IV carystals. (a) Schematic diagram indicating the positions of’ the 4 LDHase tetramers in the unit cell (labelled TI to T4; the asymmetric unit comprises Tl and TZ). Figures in parentheses refer to the relative heights of the molecules along the J axis. (b) Stereo diagram illustrating the molecular orientations. The view is approximately the same as in (a).

Protein Data Bank (Grau et al., 1981). However, if this were indeed the case, then a native Patterson function should reveal a large peak corresponding to the translation vector between the two noncrystallographically related tetramers. No significant peak was found, even at very low resolution (25 to 15 A data).

The solution of the structure of the binary (enzyme/NADH/FruP,) complex of B. Ytearothermophilus LDHase (Piontek et al., 1990) gave us access to a partly-refined (R-factor = 28%) set of coordinates. A model was constructed that was based on the binary structure but in which the active site loop (not, visible in the binary structure) was

Ternary

Complex

of LDHase from

321

B. stearothermophilus

2700 Figure

2. A 15-6

A self-rotation

function.

Stereographic

modelled in the “down” position observed in other ternary complexes of LDHase (Grau et al., 1981; White et al., 1976; Rossman et al., 1971). Structure factors were calculated for this B. stearomolecule placed in a large thermolphilus “binary” (160 A x 160 A x 160 A) P222 cell. These data were then used in a cross-rotation function using the MERLOT suite of programs (Fitzgerald, 1988). The fast rotation function (Crowther, 1972) was employed using all of the data between 8 and 4.7 A spacing with Fj>3o. The terms were modified for removal of the origin, and a radius of integration of 25 A was chosen. The whole of the asymmetric unit of rotation space was explored at 2.5” on CI and 5” on j? and y. The known model was oriented with its molecular P, & and R axes (Rossmann et al., 1973) aligned initially along X, Y, and 2, while the ternary crystal was oriented to place c, a*, b along X, Y and Z. The cross-rotation results are presented in Table 4. The rotation function produced two very significant peaks. The ROTSYM option within MERLOT was used to determine the relationship between these two peaks (and their symmetry equivalents). This analysis revealed two sets of peaks; those related to others by exact 2-fold axes, and those related by approximate S-fold axes (Table 4). This latter set again confirmed the pseudo-orthorhombic nature of the crystals.

projection

of rotations

with

K = 180’

These results prompted a re-examination of the self-rotation function calculated with data between 8 and 4.7 A resolution (Fig. 3). Originally these selfrotation results were discarded because the IC = 180” section seemed to show essentially the same features as the function using 15 to 6 A data, but the peaks were less sharp, making interpretation rather difficult. Interpreted in the light of the new crossrotation results, based on a set of rejined co-ordinates, it was possible to identify a set of peaks consistent with these results (Table 5). From the set of peaks that were related by exact 2-fold axes, it was possible to extract two orthogonal sets, each corresponding to an individual tetramer. In addition, those rotations that were not exact 2-fold axes corresponded to rotations that placed one orthogonal set onto another. In this way, all of the peaks observed in the self-rotation function could be accounted for. It appears that the lower resolution terms were dominated by the intermolecular dyad axes, swamping those from the intramolecular 2-fold axes. The dominance of such “Klug peaks” has been noted in other rotation function studies (see e.g. Johnson et al., 1975). A more accurate determination of the orientation of the two independent tetramers was obtained by Lattman’s (1972) rotation function as implemented in MERLOT. A fine search was conducted around

(‘alculations RCW made using data Radius of integration = 25 A.

with

F > 30 in the wsolution

Subunits of tetramrr

I-4 I (1’1)

Subunits of t>etramer

I- 4 3 (1’3)

Subunits of tetramer

14 2 (1’2)

Suhunits of tetramrr

1~4 4 (T4)

~‘itnge 8 to 4.i 4

Table 5 Mf-f-rotation Rotation ..-__

4

A. Relationship 1.4-11~

K

*

between

the cross-rotation

results

and

srlf-rotation

peaks

lA-l(‘ IA-ID

16X 274 555

70 53 44

180 1 x0 1 I x0

Subunit 1 onto the other 3 subunits of Tl (i.e. 2-folds of T1)

24-2 H 2A-2( ’ PA-“I)

19X 93 302

i5 47 41

I80 1x0 I 1x0 I

Subunit 1 onto the other 3 subunits of T% (i.e. I-folds of T2)

IA--PA IA-213 I A-2(.” I Ap”D

355 9% lH5 180

IX 87 61 90

151 IX4 i 73 80 I

4

*

K

Peak 13. Self-rotation

peaks

Subunit subunits

Height

1 ofT1 onto 1 to 4 of T2

Local

symmet,ry

8 to d.7 .d)

I 2 3

.55 254 168

44 53 70

1x0 180 1x0

20 1.5 22

- P axis Q axis fi axis i

4

47 47 75 40 Xl 19 89

I80 1x0 I80 180 178 I50 174

3X 27 22

$aiI axis -I’

9 10 11 12

302 93 198 92 93 1 75 3 95 2

30

132

90

X0

65 20 6.5 I 20 47

14 13

855

60 50

IO8 49

21 19 /

set of peaks

is listed.

.i

6

7 x

Only

the unique

‘1’ I

T2 L_I

17

Peaks corresponding t,o rotations that place at least 1 trtramer onto anot,hrr

Ternary

qf LDHase

Complrx

from

B. stearothermophilus

323

2700 Figure

3. A R-4.7

A self-rotation

function.

Stereographir

t,he peak positions on a &5” grid. It is important to note that the relative heights of the two peaks were more similar at this stage (Table 6), showing the small discrepancy in the peak heights obt’ained from t,he fast rotation function ho be a consequence of the rat,her than of real differences coarse sampling. between the two t,etramers. Cross-rotation function studies were performed using refined dogfish M, NADH/oxamate LDHase co-ordinates (Brookhaven Protein Data Bank) and with P. Q and R-axis dimers of R. stearothermophilus binary complex LDHase. All result,s confirmed those described above with the P222 (i.e. whole tetramer) structure factor set.

5. Molecular Of the frequent,ly

of rot,ations

with

K = 180’

is particularly true where there are two independent molecules in t’he asymmetric unit. For t#he P2, case, this results in a five-dimensional search problem (the I- co-ordinate of 1 molecule fixes the origin along the unique axis). Searches based on R-factor minimization procedures (see e.g. Derewenda et al., 1981) in t’his particular case were shown to be unsuccessful even with the benefit’ of hindsight. The translation function of Crowther & Blow (1967), on the other hand, uses a modified Patterson function t’o look for cross-vectors between different molecules in the unit cell. This approach is eminently more suit,able for the case where there are two crystallographically independent’ molecules. Hence, this translation function was employed (available as TRNSUM within MERLOT). A variety of searches were carried out to look for vectors between both crystallographically related, and independent molecules. Searches between crystallographically

Positions

rotation and translation functions, the latter that is the most difficult.

projection

it is This

Table 6 Cross-rotation d( 1 2

167.5 19.5 (‘alculations

used

data

(Jine search)

B

Y

Absolute

value

6%5

3&w 455

0277

x 10’5 x 1Ol5

106.5 with

peaks

F> 3~ in the

resolution

03279 range

8 to 53

.%

Relative

height, 99.3 100-O

324

I). H. Wiglcy

Table 7 7’ransEation Search

1

Z

0.73 040

0.50 0.50

021

9.43

IO0

0.81

3.49

37t

‘l-4 + T2

063 0.93

050 050

079 0 I0

x-99 3.62

I(Hl Sot

T2-+Tl

w.55

0.40

071

028

0.01

014

0.18 0.57

1)lO 0.71

0.50

‘1’3 + 1’ I

T3 + T2

r.tn.s.

Height

loo

Il.59 4.7%

-.

6. Refinement

41t

12.68

025

-

procedure resulted in very small overall shifts. brrl reduced the H-factor to 3!,‘iC,. showitlg t hc H. stearothermophilus binary structure tc) 1~ a got~ti model for the ternary complex. its might IW expected. The R-factor between the high rcsoiut lot! (2.5 A) data and the model was calculat~ctl to t)c 42 0’0 t,o the same resolution.

function

X

et al.

loo 37t

4.73

r.m.s. root-mean-square. t The height of the highest noise peak for each search is given. (Calculations used data with F>3a in resolution range 8 to

5.3 A.

related molecules could be restricted to the section where Y = 0.5, whereas those between independent tetramers were made over the entire unit cell. All searches employed a grid sampling of l/lOOth of a cell edge. The results were very convincing (Table 7), and could be interpreted to define the molecular packing (Fig. 1). The molecules were correctly oriented, placed at their known positions in the cell, and structure factors calculated to 4.7 A. At this stage, the R-factor between the binary LDHase model structure factors and the ternary LDHase data was 44 O/o. The molecular positions and orientations were refined by a least-squares procedure (RMINIM in MERLOT) in 0.25” intervals on c(, B and y, and by 0.1 A on X, Y and 2. This

Electron density (SF0 - r,; 2.5 A resolution) maps calculat,ed at this stage were generally of good quality. Side-chain densities were ck:arly distinguishable from that of the main-chain (t,hough atoms were not always in density at this stage). and densit,y attributable to the NADH, oxamate and FruP, was observed even though these molecules had been omitted from the calculations. Consequently, the maps did not appear to bc biased by the starting model. With over 20,000 non-hydrogen atoms in the asymmetric unit, the refinement problem was not trivial and required several approaches. Manual model building (which was potentially verv t#ime consuming for such a large asymmetric unit) was kept to a minimum and until the final stage of the refinement was only carried out upon one of the non-crystallographically related subunits, with the other subunits being produced by rotat’ion and translation of the rebuilt subunit to the other positions. The course of the refinement is summarized in Table 8, and deviations of the final model from ideal geometry are shown in Table 9.

Table 8 Course of refinement Non-crystallographic symmetry constraints

Clycles 130

Resolution range (A)

B-factor applied (A2)

loG2~5 1OG2.5

150 indiv

&fold &fold

20 Model 90

rebuilt

Model

rebuilt

with

with

respect

respect

60

to &fold &fold

averaged

to 8-fold 8-fold

averaged

Xon-protein atoms included

2F,, - F, map 100-4~5

indiv.

26.9

2F, - Fc map 1 OS2.5

indiv.

Simulated 1

annealing

refinement

lo+25

8-fold

indiv.

8 NADH 8 oxamate

(B = 25.4) Model rebuilt Least-squares 10

with respect minimization

Model -5

with

rebuilt

to X-fold

2F,, - F, map

averaged

8-fold (restraints) respect

to unaveraged None

lOG25

2li0

indiv (B = 29.1)

X NADH 8 oxamate 568 H,O

IN2

indiv.

8 NADH 8 oxamate 512 H,O 8 FruP,t

14.7

- FC map 100-25

(i5 = 264)

t Eight FruP, The calculations LDHase subunit.

24.4

8 NADH X oxamate

(F? = 203)

molecules were included but with used all data within the specified

an occupancy of @5 (see the text). range and included 316 out of 317 amino

acid

residues

in each

B. stearothermophi1u.s

Ternary

Complex

of LDHase

from, R. stearothermophilus

325

Table 9 Deviations

from

ideal geometry

Parameter Distances (A) Hands (I-2 neighbour) Angles (l--3 neighbour) 1 nterplanar ( 14 neighbour) Planar groups (A) Chiral centres (A3) Torsion angles (“) Planar (0,180) Staggered (+ 60, 180) (kthonormal ( k 90) Sowbonded contacts (A) Single torsion Multiple torsion Possible H-bond

TlIe first stage involved a conventional positional refinement using the X-PLOR package (Briinger et al., 1989) mounted on the CRAY X-MP/48 at the Rutherford Appleton Laboratory (Chilton, U.K.). The starting model comprised two tetramers of the binary B. stearothermophilus model (protein atoms only). Strict, 222 molecular symmetry was imposed on both tetramers at’ this stage of the refinement. and an initial overall B-value of 15.0 A2 was applied. Cycles of X-ray restrained energy minimization refinement were repeated until convergence had been achieved. After a few cycles of individual B-factor refinement, the R-factor had fallen to 28.8%. Electron density maps (2F, - F,) calculated at this stage showed a great improvement when compared to the maps based on the original model. Most of the side-chains previously out of density were now surrounded by it. However, it was decided that the 8-fold redundancy within the asymmetric unit should be utilized to improve the maps further. Hence the density covering each of the eight subunits was skewed (Bricogne, 1976; Buehner et al., 1974) such that all of the molecules were in the same orientation. The density was then averaged over all eight maps. The averaged map was of superb quality (Fig. 4), Both main-chain and side-chain densities were unambiguous and atoms correctly positioned within them for over 95% of the molecule. Bulges corresponding to main-chain carbonyl groups were obvious in almost every case. The active site was beautifully resolved, including residues 98 to 110 (the active site loop (numbering according to Eventoff et al. (1977))). Density corresponding to EADH and oxamate was strong (even though these molecules had still been omitted from the calculations) and allowed unambiguous interpretation of their conformations. The density corresponding to the FruP, was surprisingly good, though showed evidence of some disorder in this map because the two FruP, molecules bound per LDHase tetramer cannot> obey the 222 symmetry of the whole molecule. The implications of this are discussed below. There were t,wo areas of the map in which the

of the jinal

Number

“0.392 27.704 7000 3488 3112 2656 3416 296 7206 8567 1524

model

Target

value

0.020 0040 0.060 0020 0,120 590 1500 2000

Final

value

0011 004.5 0054 0.010 0.115 1.84 23.45 33.04 0233 0296 0.264

model did not follow the density. In one of these regions (residues 75 to 82) the density could easily be interpreted by a shift in the sequence by one residue and the model was corrected manually using FRODO (Jones, 1978, 1982) on an Evans and Sutherland PS300 computer graphics system. This region could not be fully interpreted in the apo or binary structures, such that residues 81 and 83 had to be omitted from these models (Piontek et al., 1990). The antigenic loop region (residues 210 to 225) proved to be more problematic to rebuild. This loop connects a /?-strand @J) to a helix (alG). The B. stearothermophilus binary model structure had the alG helix beginning at’ residue 223. However, the electron density maps suggested t’hat the helix began several residues earlier than in our initial model. The model was therefore adjusted, such that’ t’he helix now began at residue 220. The density for the remaining six residues of the loop (214 to 219) was weaker but could be followed and the model built into the density. It should be noted, however, that this region is involved in crystal contacts in some (but not all) of the eight subunits of the asymmetric unit. After this first round of model building, further cycles of refinement were run (still imposing strict &fold non-crystallographic redundancy) until convergence was reached. A second round of model building was applied to the protein atoms, and this time the coenzyme and oxamate were built into the electron density maps. After a few cycles, the refinement of the bound ligands had converged. However, the R-factor at this stage was still 24.4%, so refinement was continued using the simulated annealing mode of X-PLOR with non-crystallographic symmetry constraints maintained. This involved a heating stage in the dynamics procedure during which the structure was heated at 3000 K for 1000 steps of 0.001 picosecond, followed by cooling at 300 K for 500 steps of @OOl picosecond. After a further 50 cycles of positional refinement, another round of model building was undertaken using g-fold averaged 2F, - F, and F,, - F, electron density maps. Solvent molecules were also fitted at

(b)

Figure 4. Stereo electron density (contoured at lo) corrrsponriing to (a) a /&strand in the c~ornzy~~~~-~)i~lditry tlorn;tin. and (II) t.he region around residues Xl and X3. illustrating the ciuality of the averaged 2b’” - FC rlec~tron density rrrirl).

this stage. Further refinement employed a conventional least-squares minimization procedure (Hendrickson & Konnert., 1981) hut maintjainrd strict non-crystallographic symmetry restraints on main-chain atoms. hut, less strict for side-chains. The &factor at this stage was 1820,,. The final stage of the refinement allowed a release of t’he noncrystallographic symmetry restrain&. and required a further round of model building. The hound FruP,

molecules were also included ate this &age. Krc~ausc of the problem of statistical disorder. two FruP2 molecules were included at each site but’ the occupanties were set’ t,o 0.5. The R-factor of the tinal model using all data between 10 and 2.5 ‘4 rcsolution was 14.7’?,,. Details of the deviations from ideal geometry of the tinal model are presented in Table 9. Ramachandran and atomic B-factor plots are presented in Figure 5.

-100 1

0

-150 -I

60-

-60

-

I ” 0

”

I ” 50

”

I ” 100

”

I ”

”

I ”

150

200 Residue

”

I ” 250

“I”” 300

no.

(b)

Figure 5. (a) Ramachandran plot (Ramachandran et aE., 1963) of the refined B. stearothemophilus LDHase coordinates, prior to the release of non-crystallographic symmetry restraints. The allowed region for poly-r-alanine is marked, and circles denote glycine residues. (b) Plot of mean B-factors for main-chain atoms (above) and side-chain atoms (below) for each residue of the final model prior to release of the non-crystallographic symmetry restraints.

U. B. U’igley et, al

32X 7. Discussion (a) General

This is the first, LDHase crystal structure to he obtained using polyethylene glycol as precipitant. All of t,he other LDHase structures reported have been grown from solutions of high ionic strength. usually ammonium sulphate. One consequence of this is that many of these LDHase structures contain bound anions. The significance of this to the observed structure is unclear, but these bound ions (usually sulphate) often occupy important sites in the protein such as the active site, in apo and binary forms of the enzyme? and the FruP2 site in R. stearothermo$iZus apo-LDHase crystals grown from ammonium sulphate in the absence of FruP,. It is known that sulphate ions can increase the thermal stability of B. stearothermophilus LDHase dimers (D.B.W.. unpublished results) in a manner similar to that observed for FruP, (Wigley et nl.. 1987a), and that t,he enzyme is at least partially activated by sulphate ions (I. Badcoe & J.J.H.. unpublished results). Hence the structural consequencesof bound sulphate ions may cloud the issue. Such ambiguities do not arise for crystals grown from polyethylene glycols (particularly those of high molecular weight). The large size of the hydrated polyethylene glycol molecules will result in their exclusion from the crystal lattice. Hence protein molecules within the crystal will be bathed in a solution that does not contain either polyethylene glycol or protein (which would, of course, also be excluded). This leaves a solution of dilute buffer and ligands that is rather more similar to the conditions used to study the enzyme in solution than is the 1 t,o 2 M-ammonium sulphate surrounding the LDHase molecules in other crystal forms. The second anion-binding site identified by Piontek et al. (1990), which is occupied by a structure of the sulphate ion in the R. stearothermophilus enzyme crystallized from ammonium sulphate, is occupied by a water molecule in four of the subunits of the structure reported here, and by the side-chain of Lys75 in the ot,her four subunits. Differences between the binary and ternary structures that are a consequence of bound sulphate ions are obscured because of the different interpretation of the electron density in this region (residues 75 t’o 82, see above). The Ramachandran plot (Fig. 5) indicates very few deviations of non-glycine residues from the allowed regions. In particular, none of the residues that were identified as having suspect main-chain conformations in the binary structure (Piontek it al., 1990) falls into the disallowed region for the ternary complex. (b) Molecular

packing

and pseudo-symmetry

Any proposed packing model must agree with the strong pseudo-symmetry features of the precession photographs described earlier. The approximate 2, screw along c is a consequenceof the relationship of

molecules 1 to 4, and of 2 to 3. At low rrsolutlotl. thlb orient,ations of these pairs of molec~ulrs is vet’> similar but more importantly. thcl 1ranslat,io1i between them is exactly a half on %. Thr approsi mate screw axis along c breaks down at highc~r resolution only when the difference in orientatiotl between the two molecules of a given pair (‘at1 1)~ distinguished. The approximate centriny of’ t hr lattice, when viewed in projection down rr. givrs risck to t’he appearance of the non-primitive lat,tire &&t observed in the OkZzone. As predict,ed. t hp displacr~ment of those molecules most closely relatrtl in S (e.g. Tl and T2) is closer to a half in thtt l- tiirec~tion than along Z ( Y = @5 k 0.1. Z = W5 + 02). In view of the similarity bet,ween the precession photographs of the orthorhombic c.r\rst,alx and both of the monoclinic crystal forms. it seemslikely t,hat t,he molecular packing is similar in all three carystals. The observed pseudo-orthorhombic nature of the packing in t,he type IV crystals would support this. and it would be easy to obtain true orthorhombic symmetry by small changes in t)hck tnolcc~ular packing. Similarly it would be easy to oht8ain thr type V monoclinic packing from the t#ype IV 1~~ very small changes that would alter the definition of t’he unique axis, since there is an approximatth scr(A’u axis along both a and b. Because of the pseudo-orthorhombic: packing. t,he molecular contacts between LDHase tetramrrs are rather similar. In most cases, cont’acts involvcl thtb C-terminal helix (aH) and the antigenic loop. Because of tht eight-fold non-crystallographic redundancy. however, no region of polypeptidr is involved in contacts in all eight subunits. This minimizes distortions in the final st’ructure that arise from crystal contacts. A least-squares analysis of the a-carbon backbones of all eight subunits it1 the final structure indicated that there were indeed differences between them. Not surprisingly. thestl differences were usually a consequence of csrystal contacts. However, it is interesting to note that the deviations from the mean positions were as much as 3 A for main-chain atoms in the act’ive site loop of one of the molecules (Fig. 6) and in thr antigenica loop of another subunit (the other 7 subunits adopted conformations that were very similar t,o each other in these regions), after t,he non-crystallographic symmetry restraints had been released, (c) Active

site loop

The active sit,e was well resolved, including the active site loop (residues 98 to 110). The conformation of this loop is important for the design of dehydrogenase enzymes with new specificities (Scawen et al., 1989). Sequences for this loop vary somewhat between species (Fig. 7) and consequently the conformation of this loop may also be different in these different enzymes. This is true for the B. stearothermophilus enzyme, when compared to the structure of the dogfish M, NADH/oxamate ternary complex. In the bacterial enzyme the loop adopts a more open conformation. with the tip of

Ternary

Complex

.of LDHase

from

B. stearothermophilus

Figure 6. Diagram in stereoillustrating the large changein the structure of The with

conformation of the loop residue numbers), while

prior to release the conformation

of non-crystallographic of the loop involved

the loop being some 2 w further away from Tyr237 (which represents the lip of the active site pocket) result’ing in the widening of a hole at the top of the active site. This hole is made even more accessible because of the smaller bulk of the side-chain at the tip of the loop that would obscure the hole in the eukaryotic enzymes. This residue (Glu105 in most eukaryotic enzymes) is replaced by a proline residue in the bacterial enzymes, which tightens the turn at the tip of the loop. This turn is further tightened by two additional hydrogen bonds within the loop structure (compared to the dogfish enzyme) between AsnlOl and the main-chain carbonyl group of residue 102, and between Asp111 and the mainchain amide group of residue 108. These bonds serve to “pin-back” the upturned loop and restrain it in the more open conformation. It is known that the B. stearothermophilus enzyme shows a broader specificity for substrates than do the eukaryotic enzymes, and will accept substrates with larger sidechains (H. Wilks & J.J.H., unpublished results). The specificity is reduced further when Pro105 is replaced by the even lessbulky serine residue, which

Dogfish M PigM PigH Chicken H Mouse X Ratx

L. casei B. megaterium B.stearothemwphilus

Figure 7. Sequences of the active site loop in a variety of LDHases, with the standard LDHase numbering according to Eventoff et al. (1977). The following sequences were obtained: dogfish M (Taylor, 1977), pig H and M (Kiltz et al., 1977), chicken H (Torff et al.; 1977), mouse and rat X (Li et al, 1983; Tanaka & Fujimoto, 1986), L. casei (Hensel et aZ., 1983), B. megaterium and B. ~~tearothermophilus (Wirz et al., 1983). Regions of absolute homology are boxed.

symmetry in crystal

329

the active site loop in I of the subunits. restraints is shown in bold lines (together contacts is overlaid in fainter lines.

suggeststhat this hole is important for the design of a keto-acid dehydrogenase with a very broad specificity (Clarke et al., 1987b).

(d) NADH-binding

site

The density corresponding to bound NADH is very well defined (Fig. 8). Not surprisingly, the overall conformation of the coenzyme is very similar to that described for the B. stearothermophilus binary (enzyme/NADH/FruP,) complex (Piontek et al., 1990). Hydrogen-bonding requirements of the coenzyme are satisfied by contacts with the protein as well as with a number of water molecules (Fig. 9). The major difference between the binary and ternary structures concerns the proposed hydrogen bond between the nitrogen atom of the carboxamide side-chain of the coenzyme and the hydroxyl of Ser163. This was measured to be 2.9 A in the binary complex, vet is 3.2 a in the ternary complex. This distance “is 3.6 A in both the dogfish M, (Brookhaven Protein Data Bank) and pig M, (C.R. unpublished Dunn, results) enzyme/NADH/ oxamate structures. In all of the ternary complexes, this distance is too great to constitute a hydrogen bond of any significance. This is supported by the observation that a site-directed mutant of LDHase in which Ser163 is changed to alanine (Wigley et al., 1987c) shows wild-type properties. In particular, the affinity of the apo-enzyme for NADH is unaltered by the mutation. This suggests that a hydrogen bond between Ser163 and the coenzyme is not important for binding of the coenzyme t.o LDHase. However, the carboxamide side-chain of NADH is in contact with the enzyme in the ternary complex. Rather than being hydrogen bonded through the nitrogen atom to Serl63, there appears to be a hydrogen bond between the nitrogen atom and the main-chain carbonyl group of residue 138. This distance is 3-O A in the B. stearothermophilus ternary complex compared to 3.3 A in the binary complex. This same hydrogen bond is 2.8 A in the dogfish ternary complex. The utilization of hydrogen bonds via main-chain rather than side-chain atoms makes good evolutionary sense. as the

Figure 8. Stereoscopic

view

of the coenzyma electron

enzyme is less susceptible to spontaneous mutations. This is a strategy that has also been adopted in dihydrofolate reductase (Filman et al., 1982). The affinity of LDHase for the 3-acetyl pyddine analogue of NADH (where the possibility of this hydrogen bond between the enzyme and coenzyme has been removed) is less than that for NADH (Kaplan et al., 1956). This suggests that there is indeed a hydrogen bond between the carboxamide side-chain of NADH and the enzyme, but we believe this to be with the main-chain carbonyl group of residue 138 rather than with the side-chain of Serl63. Ser163 is conserved in all known LDHase sequences, but is not present in the closely homologous malate dehydrogenase enzymes. However, the function of this residue in LDHase remains

density in the final 2F0 - 6’, maps (contoured

unclear, though, together wibh the main-chain carbonyl of residue 164, it, is involved in binding a water molecule in the active site that appears to he close enough to hydrogen bond to the O-atom of the carboxamide side-chain of the coenzymr. It is not known whether this water molecule is also present in the Ala163 mutant enzyme. or what the role of this water molecule might be. (e)

Oxamate

binding

site

The oxamate binding site in this structure is similar to that described for the dogfish ternary complexes (White et al.. 1976). However. one residue that. seems to be very important in R. stearothermophilus LDHase was not noted in the dogfish structures. Thr246 forms a good hydrogen

i Figure 9. Interactions

at IO)

between the bound NADH and the enzyme that constitute

r the coenzyme-binding

site, in stereo.

Ternary

Complex

of LDHase

from

B. stearothermophilus

:

Figure

10. Residues

that

contribute

bond with oxamate in this structure (Fig. 10). This interact*ion explains the large reduction in the affinity of the enzyme for pyruvate in a mutant enzyme in which Thr246 has been replaced by a glycine residue (Wilks et al., 1988). This amino acid substitution proved to be very important for the design of a maiate dehydrogenase from the LDHase enzyme (Wilks et al., 1988). The general role of this residue in determining substrate specificity in LDHase has been discussed by Bur et al. (1989). (fl FruP,

binding

site

It has been demonstrated that only two FruP, molecules bind per B. stearothermophilus LDHase tetramer (Clarke et al., 1986b). Consequently, it was

Figure calculated For clarity,

11. Eightfold using the

averaged the partly refined final co-ordinates

to the

oxamate

binding

331

!

site

in stereo

suggested that these FruP, molecules might, bind to the anion-binding site identified in eukaryotic LDHases (Adams et al., 1973), bridging two such sites across the P-axis (Clarke et al., 19866). Support for this hypothesis came from studies using sitedirected mutagenesis to probe the function of some of the residues in the anion-binding site (e.g. Arg173 (Clarke et al., 1987a); His188 (Schroeder et al., 1988)). Crystal structures of the binary complex of B. stearothermophilus LDHase (Piontek et al., 1990) and now the ternary complex: confirm that FruP, does indeed bind between two anion-binding sites. In the ternary complex, density corresponding to the FruP, molecule is clearly visible in both the symmetry-averaged and unaveraged maps (Fig. 11). The molecule curls around two histidine

2F, - F, electron density in the region of the bound FruP,, co-ordinates (R = 182%) and the FruP, atoms were omitted of the FruP, have been superimposed.

in stereo. The map was from the calculations.

Figure

12. Interactions

between

the FruPz

and the 2 His188 residues from adjacent. subunits,

residues (His188 residues from symmetry-related subunits) such that the two imidazolium rings are parallel and separated by just 3.5 A (Fig. 12). This energetically unfavourable configuration is stabilized by the two negatively charged phosphate groups of the FruP, such that each phosphate group forms an ionic interaction with both histidine residues. Presumably it is this unusual interaction that gives rise to the strong pH dependence of activation by FruP, (Clarke et al., 1985). Binding of the phosphate groups is further strengthened by interactions with two symmetry-related Arg173 residues, but in this case each phosphate group

in stereo

interacts only with the arginine from one subunit. As predicted by Clarke et al. (19866) the increased affinity of the enzyme for pyruvate (and oxamate), as a result of binding of FruP,, is probably the result of linkages between Arg173 and His188 to Arg171 and His195, respectively, at the active site (Fig. 13), which are two of the main contributors to substrate binding. Because the FruP, molecule does not have an exact 2-fold axis of symmetry, yet binds across a molecular 2-fold axis in the tetramer, the electron density at the FruP, site is expected to show signs of disorder corresponding to binding of the FruP, in

‘BP

Figure 13. Stereo diagram illustrating the linkage between residues in the FruP, binding site and the oxamate site. For simplicity, only 1 of the 2 possible orientations of the FruP, molecule is shown.

binding

Ternary

Complex

of LDHase

from

B. stearothermophilus

Figure 14. Interactions between the bound FruP, and the enzyme. including the main-chain carbonyl group of residue 185. in st&eo. either

of two possible

orientations.

In fact, because

the FruP, molecule almost has 2-fold symmetry that is broken only by the hydroxyl group at C-2 of the furanose ring, it is only the density for this hydroxyl group that shows evidence of disorder. Interestingly, the only apparent interaction between the furanose ring and the enzyme is a hydrogen bond between the main-chain carbonyl of Gln185 and the 2-hydroxyl of the sugar ring. This interaction is independent of the orientation of the bound FruP,, but is with a different subunit across the P axis in each case (Fig. 14). There is no evidence, in this structure, of any interaction between the side-chain of Gln185 and the bound FruP,. The electron density for this residue is unambiguous and shows a single conformation, rather t’han the two conformations observed in the binary complex (Piontek et al., 1990). This residue is arginine in L. casei LDHase, an enzyme that binds four, rather than two, FruP, molecules per tetramer (M. Buehner, Brookhaven Protein Data Bank). The 3- and 4-hydroxyl groups do not interact with the enzyme, and do not, appear to have discretely ordered solvent molecules within hydrogen bonding distance. The affinity of eukaryotic LDHases for pyruvate is not regulated by FruP,, and the pyruvate afinity of these enzymes resembles most closely that of the “activated” bacterial enzymes. It appears that the bacterial enzymes are best regarded as inactivated in the absence of FruP,, rather than activated by its presence. Although this appears to be a subtle distinction it is important when one considers the fact that the anion binding site in eukaryotic enzymes is always occupied. If attempts are made

the hydrogen

333

bond between

FruP, and

to remove this bound anion (e.g. by dialysis) then the pig heart enzyme becomes irreversibly inactivated (J.J.H., unpublished results). Thus, it would seemthat in prokaryotes a system has evolved that enables LDHase to be fully active only when FruP, levels are also high (e.g. rapid rate of glycolysis), but that under less favourable activity is reduced allowing

conditions, LDHase pyruvate to pass into the citric acid cycle rather than being converted to

lactate and then excreted. Whether this system evolved in prokaryotes and has been lost in eukaryotes, or vice versa, is unclear. The structural basis for the differences between these two classesof enzyme must be t’hat the affinity of the site for anions has been reduced in the bacterial enzymes. One obvious way to achieve this is to make the site less positivelv charged. Those residues that link the FruPz site w:t,h the active site obviously have to be conserved. However, from the position of the bound FruP,, it seemsthat, there are two side-chains that are positively charged in the eukaryotic enzymes but’ uncharged in the bacterial enzymes. These are Arg267 and His269 (Ala and Leu in B. stearothermophilus LDHase). These residues were shown to be in contact with bound citrate in the dogfish apo-enzyme (Abad-Zapatero et al., 1987), but at that time it was not known how FruP, was bound at this site in the bacterial enzymes. The structure reported here confirms that these changes are likely to account for the reduced affinity

of the site for anions

such as citrate.

We thank M. G. Rossmann, S. Krishnaswamy and C. R. Dunn for helpful discussions, and Smith-Kline, Reecham for travel funds between Purdue and Leicester.

Jn addition. I).H.LV. was supported JI. (:. Rossmann. while on a visit, This work was funded by SERC. Advanc~I Fellow and .I .P.T. holds

tly an XIH grant to to Purdue I-niversity. 1j.H.D is an SEKC’ an EN’ studentship.

References Abad-Zapatero, Rossmann, dogfish M,

C’., (Xffith. J. P., Sussman. ,J. I,. 8: M. C. (1987). Refined crystal structure ot ape-lacatate dehydrogenase. ,I. Mol. Biol.

198. 44%467. Adams. M. .J.. Haas, I). ,I., .Jeffery. B. A., .McPherson. ‘4.. Mermall, H. L.. Rossmann, M. G.. Schevitz. R. \Y. & LVonarott. 9. *I. (1969). Low resolution study of crystalline L-lacstate dehydrogenasr. .J. ~Mol. Hiol. 41. 159--I 88. Adams. M. .J.. Ford. C:. C., Koekoek, R.. Lentz. 1’. .I McPherson. A.. Rossmann. M. G.. Smiley. I. E.. Schevitz. R. 1Y. X: Wonacott, A. ,J. (1970). Structure of lactate dehydrogenase at 2.8 A resolut’ion. Snturr’ (London), 277. 1098-l 103. Adams. M. ,J.. Lilijas. A. & Rosstnann, M. (:. (1973). Functional anion-binding sites in dogfish M, lac’t,attx drhydrogenase. J. ,WliloZ. Hiol. 76, 51!&631. Kricopnr. C. (1976). Methods and programs for direct space exploit,ation of geometric redundant-ies. Z4f&l (‘rystallogr.

wt.

A,

32,

832-837.

Krlnger. A. T.. Karplus, M. & Petsko. (:. A. (1989). (‘rystallographic refinement by simulated annealing: applicatjion to (*rambin. Acta ~,“rystaZlogr. srct. A, 45. .5&6 1 Burhner. M.. Ford. (:. (1.. Moras, I).. Olsen. K. I$-. h Rossmann, M. (:. (1974). Structure determination of c*rystalline lobster o-glyceraldehyde-Z-phosphate dehydrogenase. J. Mol. Hiol. 82. :i63-;iXfi. Kuehner. M.. Hecht. H.-.J., Hensel. R. & Mayr, Il. (1982). f’rystallization and preliminary carystallographic analysis at low resolution of the allostrric I,-lacatate drhydrogenasr from Lactobacillu.9 cns~i. J. ,lfol. Kiof. 162, 819-838. Eur. I).. (“larke. A. R,.. Friesen. .I. I).. Cold. Al., Hart. K. I\:.. Holbrook. tJ. *J.. ,Jones. ,J. K., Luyten. $1. .A. & Wilks. H. M. (l!J89). On the effect) on spetrifiritj. of Thr246-+(Gly tnutation in L-lactate dehydrogenase ot Bacillus (‘ommun.

(‘larke.

Biophys.

Krs.

Acta,

829,

387-396.

A. R., Wigley. D. R.. Chia. W. X., Barstow. 1). A.. Atkinson. T. & Holbrook. ,J. *J. (1986a). Bite-directed mutagenesis reveals the role of a mobile arginine residue in lacatatr dehydrogenase catalysis. :Va,tvrr (London),

(Yarke,

Hiochrm.

161. 5c3&63.

A. R,.. Atkinson. T.. (‘ampbell. -1. IV. B Holbrook. .I. .I. (1985). The assemb1.v mechanism of the lactate dehydrogenase trtramrr from Hacillus Sstearothermophilus: the equilibrium relationships between quatrrnary structure and the binding of fructose 1.6. bisphosphatr. XAI>H and oxamatr. Riochim. Biophys.

(‘larkr,

stearothrrmophilus.

324.

699-702.

A. R., Evington. J. R. Iv., Dunn, c‘. R., Atkinson. T. &, Holbrook. ,I. ,J. (I 9866). The molecular pathway by which frucatose 1,6-bisphosphat,e induces t,he assembly of a bacterial lactate dehydrogenase. Hiochim. Riophys. Acta. 870, 11% 126. (Yarke. A. R., Wigley. D. B.. Barstow. D. A.. (‘hia. \V. N.. Atkinson, ‘I’. & Holbrook. J. ,J. (1987~). X single amino acid substitution deregulates a bacterial lactate dehydrogrnase and stabilises its tetrameric structure. RioAim. Biophys. A eta, 913, 7dWO.

(‘larke,

A. I~clrox> acid drhydrogenasr with new sub&at+% speciticil y. Hiochrn/, Hiophys. Krs. ~‘orwrr un. 148. I5 “:\. (‘larke. A. K.. R’ilks. H. >I.. Barstow. I). :I.. :\tkinsott, ‘I’.. (‘hia. \V. N. & Holbroc)k. .I .J. (198X). ;irr itl\esti~ gatiott of the contrit~utiotr made by thta ~~art~oxylstr group of an active site hist)idinr aspartat~ (~~uplr t,o binding and c,at,alysis in lactat,? tlrh~tlroMI,nasf~. Kiochrn,icstry, 27. I617 ISZ?. C’rowther. R. :\. (195%). In Thr ‘WotPculor Krplrrrr’nlr’nt M&hod (Rossmann. Yl. (;.. etl.). pp. I73 IX. (Gordon and Breach. TVrw York. (‘rowther. I