Báo cáo khoa học: Large conformational changes in the Escherichia coli tryptophan synthase b2 subunit upon pyridoxal 5¢-phosphate binding pot

tryptophan synthase b2 subunit upon pyridoxal5¢-phosphate binding Kazuya Nishio1, Kyoko Ogasahara1, Yukio Morimoto2,3, Tomitake Tsukihara1,4, Soo Jae Lee5and Katsuhide Yutani3 1 Institut

tryptophan synthase b2 subunit upon pyridoxal

5¢-phosphate binding

Kazuya Nishio1, Kyoko Ogasahara1, Yukio Morimoto2,3, Tomitake Tsukihara1,4, Soo Jae Lee5and Katsuhide Yutani3

1 Institute for Protein Research, Osaka University, Japan

2 Research Reactor Institute, Kyoto University, Japan

3 RIKEN SPring-8 Center, Harima Institute, Japan

4 Department of Life Science, University of Hyogo, Japan

5 College of Pharmacy, Chungbuk National University, Korea

Keywords

apo- and holo-forms; conformational change;

PLP-binding; tryptophan synthase b2

subunit; X-ray crystal structure

Correspondence

Katsuhide Yutani, RIKEN SPring-8 Center,

Harima Institute, 1-1-1 Kouto, Sayo, Hyogo

679-5148, Japan

Fax: 81-791-58-2917

Tel: 81-791-58-2937

E-mail: yutani@spring8.or.jp

Soo Jae Lee, College of Pharmacy,

Chungbuk National University, Sungbong-ro

410, Cheongju, Chungbuk, Korea

Fax: 82-43-268-2732

Tel: 82-43-261-2816

E-mail: sjlee@chungbuk.ac.kr

Database

Structural data are available from the Protein

Data Bank under the accession codes for

the holo- (2DH5) and apo- (2DH6) forms

(Received 16 October 2009, revised 24

February 2010, accepted 1 March 2010)

doi:10.1111/j.1742-4658.2010.07631.x

To understand the basis for the lower activity of the tryptophan synthase b2 subunit in comparison to the a2b2complex, we determined the crystal struc-tures of apo-b2 and holo-b2 from Escherichia coli at 3.0 and 2.9 A˚ resolu-tions, respectively To our knowledge, this is the first report of both b2 subunit structures with and without pyridoxal-5¢-phosphate The apo-type molecule retained a dimeric form in solution, as in the case of the holo-b2 subunit The subunit structures of both the apo-b2 and the holo-b2 forms consisted of two domains, namely the N domain and the C domain Although there were significant structural differences between the apo- and holo-structures, they could be easily superimposed with a 22 rigid body rota-tion of the C domain The pyridoxal-5¢-phosphate-bound holo-form had multiple interactions between the two domains and a long loop (residues 260–310), which were missing in the apo-form Comparison of the structures

of holo-Ecb2 and Stb2 in the a2b2 complex from Salmonella typhimurium (Sta2b2) identified the cause of the lower enzymatic activity of holo-Ecb2in comparison with Sta2b2 The substrate (indole) gate residues, Tyr279 and Phe280, block entry of the substrate into the b2subunit, although the indole can directly access the active site as a result of a wider cleft between the N and C domains in the holo-Ecb2subunit In addition, the structure around bAsp305 of the holo-Ecb2 subunit was similar to the open state of Sta2b2 with low activity, resulting in lower activity of holo-Ecb2

Structured digital abstract

l MINT-7712009 : Ecb2 (uniprotkb: P0A879 ) and Ecb2 (uniprotkb: P0A879 ) bind ( MI:0407 ) by x-ray crystallography ( MI:0114 )

l MINT-7712032 : Ecb2 (uniprotkb: P0A879 ) and Ecb2 (uniprotkb: P0A879 ) bind ( MI:0407 ) by biophysical ( MI:0013 )

Abbreviations

DSC, differential scanning calorimetry; Eca, tryptophan synthase a subunit from E coli; Ecb, tryptophan synthase monomer b subunit from

E coli; Ecb2, tryptophan synthase b2subunit from E coli; PLP, pyridoxal 5¢-phosphate; Sta, tryptophan synthase a subunit from

S typhimurium; Sta2b2, tryptophan synthase a2b2complex from S typhimurium; Stb, tryptophan synthase monomer b subunit from

S typhimurium; Stb 2 , tryptophan synthase b 2 subunit from S typhimurium; T d , denaturation temperature; bA, bB, two b subunits in the same Ecb2dimer.

Tryptophan synthase (EC 4.1.2.20) catalyzes the final

two steps in the biosynthesis of l-tryptophan The

bacterial enzyme, a multifunctional a2b2 complex

(Mr= 143 300), is composed of nonidentical a (Mr=

28 700) and b (Mr= 43 000) subunits The a2b2

complex can be isolated as a monomeric a subunit and

dimeric b2 subunits in solution The a and b2 subunits

catalyze different reactions, namely the a and b

reac-tions (Eqns 1, 2 respectively) The physiologically

important reaction is the ab reaction (Eqn 3), which is

catalyzed by the a2b2complex

a reaction:

Indole 3-glycerol phosphate$ indole

þD-glyceraldehyde 3-phosphate ð1Þ

b reaction:

Indole +L-serine!L-tryptophanþ H2O ð2Þ

ab reaction:

Indole 3-glycerol phosphate +L-serine!L-tryptophan

þD-glyceraldehyde 3-phosphateþ H2O ð3Þ

Combining the a and b2 subunits to form the a2b2

complex stimulates the enzymatic activity of each

subunit by one to two orders of magnitude [1,2]

This mutual activation of the two subunits is

thought to derive from conformational changes in

the subunits upon formation of the complex [3,4]

Therefore, tryptophan synthase is an excellent model

for using to study the relationship between functional

activation and conformational changes in proteins

The quaternary structure of the a2b2 complex from

Salmonella typhimurium (Sta2b2) is an extended linear

abba subunit arrangement The active sites of the a

and b subunits of Sta2b2 are connected by a 25–30 A˚

hydrophobic tunnel through which the indole is

trans-ferred from the a subunit to the b subunit [5,6]

Crys-tal structures of the Sta2b2 complex with allosteric

cations and⁄ or ligands, and of the Pfa2b2 complex

from the hyperthermophile Pyrococcus furiosus, have

been described [7–16] These structures provide

valu-able information to help us understand the allosteric

mechanism of the tryptophan synthase Recently,

structures have been solved of the tryptophan synthase

a subunit from Escherichia coli (Eca) [17] and of the

tryptophan synthase a [18], and b2 [19] subunits from

P furiosus To obtain a clear understanding of the

reason for a low enzymatic activity of the b2 subunit

in the absence of the a subunit, it is necessary to solve the structures of the b2 subunit from E coli (Ecb2) with and without its cofactor, pyridoxal-5¢-phosphate (PLP)

PLP-dependent enzymes catalyze multiple reactions during the metabolism of amino acids These enzymes have been classified into a, b and c families based on the chemical characteristics of the enzymatic reactions [20] The tryptophan synthase b2 subunit belongs to the b family, members of which catalyze b-replace-ment or b-elimination reactions This family has been distinctly classified into five-fold-types based on sequence and structural features [21] Fold-type II enzymes in the b family include the tryptophan syn-thase b2 subunit, O-acetylserine sulfhydrylase [22] and serine dehydratase [23] Several crystal structures of the apo- and holo-forms of PLP-binding enzymes exhibit only minor re-arrangements in the positions of residues lining the active site between the apo- and holo-enzymes [24,25] Other PLP-binding enzymes, however, display significant conformational changes between the apo- and holo-forms [23,26,27] Both the apo- and holo-types of serine dehydratase, isolated from the rat liver, form a homodimer In the apo-serine dehydratase dimeric form, a small domain inserts into the catalytic cleft of the partner subunit

so that the active site is closed when inactive Trypto-phan synthase, however, is a unique PLP-binding protein because the b reaction mediated by the b subunit is regulated via an allosteric mechanism trig-gered by association with the cognate a subunit PLP binds cooperatively to the apo-b2 subunit and nonco-operatively to the a2apo-b2 complex in E coli [28,29] Therefore, it is important to determine the b2 subunit structure of the apo-type without PLP, as well as that

of the holo-type, to elucidate the mechanism of PLP binding

We obtained crystals of the apo-form in the absence of PLP, as well as the holo-type bearing PLP, for Ecb2 The structures of the apo-b2 and holo-b2 subunits were solved at 3.0 and 2.9 A˚ resolutions, respectively We evaluated the thermal stabilities of the holo-Ecb2 and apo-Ecb2 forms using differential scanning calorimetry (DSC) In this communication

we will discuss the role of PLP in the stabilization of Ecb2, the mechanism of PLP binding to apo-Ecb2and the structural basis for lower enzymatic activity of Ecb2 in the absence of the a subunit, from a compari-son of the apo-Ecb2, holo-Ecb2 and Sta2b2 complex structures

Results and Discussion

Contribution of PLP to the stabilization of the

b2subunit

Using the holo-Ecb2, the apo-Ecb2 and the

reconsti-tuted holo-Ecb2 proteins, we confirmed the

contribu-tion of PLP to the stabilizacontribu-tion of the b2 subunits

Holo-Ecb2 bound to PLP demonstrated an absorption

spectrum bearing a peak at 413 nm, characteristic for

a PLP internal aldimine bound to a Lys in the b2

sub-unit [30], and a CD spectrum with a peak around

420 nm (data not shown) Dissociation of PLP from

the b2 subunit by dialysis against 0.1 m Mes buffer

(pH 6.5) containing 0.1 m Li2SO4 could be confirmed

by the disappearance of these peaks from repeat

analy-ses Ultracentrifugation analysis indicated that both

apo-Ecb2 and holo-Ecb2 remain in dimeric forms in

solution Reconstitution of holo-Ecb2 from apo-Ecb2

by dialysis against 50 mm potassium phosphate buffer

(pH 7.0) containing 0.2 mm PLP for 1 day at 4C was

also confirmed by the characteristic changes seen on

absorption and CD spectra, and DSC curves

Figure 1 displays the pH dependence of holo-Ecb2

and apo-Ecb2stabilities; the denaturation temperature,

Td, of the holo-protein increases by nearly 25C at

approximately pH 9 as a result of PLP binding The

difference in the Tdvalues between the holo- and

apo-proteins decreased with decreasing pH (Fig 1),

sug-gesting that the binding constant of PLP decreases at

pH 6.5 The denaturation temperature of the

reconsti-tuted holo-Ecb2 was similar to that of the native

holo-Ecb2 bound to PLP (Fig 1), indicating that the

dissociation of PLP from holo-Ecb2is reversible These

results confirmed that PLP plays an important role in stabilizing the b2 subunit of the tryptophan synthase For the b2 subunit from S typhimurium (Stb2), PLP dissociation also decreased the thermal stability [31] PLP binding to proteins has been reported to play an important role in protein stabilization [32,33] From the structural differences between the apo- and holo-Ecb2 forms, the stabilization of Ecb2 by PLP binding appears to be caused by increases in the number of hydrogen bonds, salt bridges and hydrophobic interac-tions, as described below

Overall structures of the tryptophan synthase

b2subunit from E coli The crystal structures of both apo- and holo-Ecb2form dimers with a crystallographic two-fold axis (Fig 2) The apo- and holo-Ecb2subunit structures are each com-posed of two domains, namely an N domain (residues 1– 206) and a C domain (residues 207–397); one stretch of the N-terminal sequence (residues 53–84), however, crosses over into the C domain A wide cleft in the inter-face between the two domains of holo-Ecb2corresponds

to the indole tunnel The holo-Ecb C domain contains a long stretched loop (residues 260–310) with a B-factor of 82.0 A˚2for the main chains, which includes a short helix and two strands By contrast, no electron density was observed for the long loop (residues 259–308) in apo-Ecb (Fig 3) The long loop forms seven salt bridges with residues in the N (Arg275-Glu11, Lys283-Glu11, Glu296-Lys167 and Asp305-Lys167) and C (His260-Glu266, His260-Asp329 and Lys272-Asp323) domains

Fig 1 pH dependence of the thermal stability (T d ) of holo-Ecb 2

and apo-Ecb2 The Tdvalue represents the peak temperature on

the DSC curve measured at a scan rate of 1 CÆmin)1 Closed

circles, open circles, and open triangles indicate the T d values for the

native holo-, reconstituted holo- and apo-b2subunits, respectively.

Buffers used were 50 m M potassium phosphate supplemented

with 1 m M EDTA at a pH of <8.5 and 50 m M Gly supplemented

with 1 m M EDTA at pH 9.

Fig 2 Schematic view of the holo-Ecb2dimeric structure The two holo-Ecb enzymes form a dimer relative to a crystallographic two-fold axis The symmetry-related molecule is colorless Magenta, blue and green regions represent the N domain, the C domain and the long loop (residues 260–310), respectively The PLP molecule is inserted in the CPK representation Figures displaying protein struc-tures were prepared using MOLSCRIPT software [58].

in addition to many hydrogen bonds: no such salt

bridges or hydrogen bonds were observed in the

apo-form These results suggest that the long loop interacts

with the N and C domains only upon PLP binding,

which contributes to the stabilization of the holo-form

Alignment of the secondary structures with the amino

acid sequences of the apo- and holo-Ecb subunits and

the b subunit of Sta2b2 (Stb) is shown in Fig 4 The

secondary structures were assigned with numbering as

referenced from Stb [5]

The overall topologies of the apo- and holo-Ecb

structures were equivalent to that of the Sta2b2b

sub-unit [5] Equivalent Ca atoms of the apo- and

holo-Ecb structures could be superimposed with an rmsd of

3.3 A˚, a remarkably large value in comparison to the

rmsd value of 0.5 A˚ for the superposition of holo-Ecb

and Stb (PDB code 1BKS) To investigate whether

this significant difference was caused by rigid

move-ment of the domains or induced-fitting, we

systemati-cally analyzed the domain movement caused by PLP

binding using DynDom software [34] This analysis

indicated that holo-Ecb can be divided into an

N-ter-minal domain (residues 1–52 and 85–206) and a

C-ter-minal domain (residues 53–84 and 207–397)

corresponding to the N and C domains of Stb,

respec-tively A rotation of 21.8 combined with a transition

of just 0.2 A˚ from the centroid of apo-Ecb permitted

good superposition of the C domain of the apo-form

with that of the holo-form This result indicated that

the conformational re-arrangement upon PLP binding

is predominantly a rigid body rotation of the C

domain with local conformational fittings of several

residues that accommodate PLP The structure of

apo-Ecb is an open form in which the N and C

domains are separated in comparison to holo-Ecb,

which is a closed form (Fig 3) PLP binding induces

the conformational change through a rigid body

rota-tion, inducing a transition from the open to the closed form at the deep cleft between the two domains of Ecb

Difference in structures between apo-Ec b2and holo-Ec b2

The differences in structures surrounding the active sites in apo-Ecb2 and holo-Ecb (shown in Fig 5A,B respectively) illustrate the conformational changes directly related to PLP binding The PLP coenzyme, located in a deep cleft between the two domains, forms

a Schiff base with Lys87 in holo-Ecb and is shielded from the solvent by the residues in the long loop (resi-dues 260–310) (Fig 5B) The resi(resi-dues interacting with PLP via hydrogen bonding are His86 and Thr190 in the N domain and Gly232, Gly234, Ser235, Asn236 and Ser377 in the C domain To explore the conforma-tional changes caused by PLP binding, we determined the rmsd values (A˚) of the Ca atoms for each N or C domain between apo-Ecb and holo-Ecb from the superposition of the respective domains Five signifi-cant conformational changes (peaks 1–5) were observed to result from PLP binding Three peaks – peak 2 (near Gly84), peak 3 (near Gly234) and peak 5 (near Arg379) – are close to the active site, while peaks

1 (near Tyr52) and 4 (near Thr319) are far from the active site Nine hydrogen bonds were observed between these peak regions and PLP in the holo-form that are absent in the apo-form Three sets of hinge residues are located at the sites connecting the N and

C domains: hinge 1 (residues 48–56), hinge 2 (residues 82–87) and hinge 3 (residues 206–207) (green boxes in Fig 4) Hinges 1 and 2 correspond to peaks 1 and 2, respectively, indicating that PLP binding also induces changes in the regions connecting the N and C domains

Fig 3 Schematic views of (A) the apo-Ecb structure and (B) the holo-Ecb structure Magenta, blue and green regions represent the N domain, the C domain and the long loop, respectively The PLP molecule is depicted by a CPK model The arrow in (A) denotes the motion of the C domain follow-ing PLP bindfollow-ing Figures displayfollow-ing protein structures were prepared using MOLSCRIPT

software [58].

Changes in the intersubunit interface of the b2

dimer as a result of PLP binding

Dimerization of the apo-Ecb and holo-Ecb subunits

occurs along a twofold crystallographic axis (Fig 2)

Apo-Ecb2 was confirmed by analytical

ultracentrifuga-tion, at pH 7.0 (as described earlier), to be a dimer in

solution The total areas buried in the subunit interface

for apo-Ecb2 and holo-Ecb2 were estimated at 1548

and 1711 A˚2, respectively The buried area of the b2

subunit within Sta2b2 (PDB code 1BKS) is 1619 A˚2

The hydrophobicity of the contact area at the

apo-Ecb2 and holo-Ecb subunit interfaces was estimated at

165 and 180 kJÆmol)1, respectively, using the human

lysozyme parameters obtained by Funahashi et al [35] This result indicates that the hydrophobic interaction

at the subunit interface of the dimer increases with PLP binding Upon binding of PLP to the active site

of one b subunit (bA), the side chain of bA-Glu350 moves in the direction of the subunit interface from the active site (Fig 6) By contrast, the side chain of bA-Lys382 moves from the subunit interface to the active site to form a salt bridge with bA-Glu350 The side chain of bA-Arg379 moves to the subunit interface

as a result of repulsion by the bA-Glu350 side chain The side chain of bA-Arg379, in concert with bB-Arg379 from another subunit (bB), creates an arginine–arginine short-range interaction, which in

Fig 4 Sequence alignments of apo-Ecb and holo-Ecb forms with Stb The first line represents the alias of the secondary segments as named by Hyde et al [5] The red bars, blue arrows and lines, which denote the a helices, b strands and the others, respectively, comprise the secondary structural elements of Stb (PDB code 1BKS), holo-Ecb and apo-Ecb, based on the definitions established by PROCHECK soft-ware [53] The red letters in the sequences indicate identical residues between Stb and Ecb The cyan squares in the third line represent hydrogen bonds between Stb and Sta, while the green squares in the fourth line indicate the hinge region in Ecb The hinge regions, includ-ing hinclud-inges 1 (residues 48–56), 2 (82–87) and 3 (206–207), were defined usinclud-ing D YN D OM software The asterisks in the sixth and seventh lines represent those residues binding PLP via hydrogen bonds (blue), hydrophobic interactions (black) or covalent bonds (red) in Ecb.

protein–protein interactions is often an important

fac-tor in stabilization and recognition [36] This

interac-tion induces large conformainterac-tional changes in the

empty bB subunit, facilitating the subsequent binding

of PLP The dissociation constant of PLP for the

iso-lated b2 subunit is 8.7· 10)6m for the first (bA) site

and 2.3· 10)7m for the second (bB) site [28] The

more efficient binding of the second PLP indicates that

this interaction occurs under different structural

cir-cumstances from the first The side chain of the shifted

bA-Arg379 contacts the bB-Asp381 side chain in

another subunit (data not shown) Each His82 turns to

the external surface of the molecule, forming a salt

bridge with Asp79 in the peak 2 region of the paired b

subunits The extent of the hydrophobic interaction,

and the number of hydrogen bonds and salt bridges

between the two subunits, are greater in the holo-Ecb2

when compared with the apo-Ecb2 These results

indi-cate that the two b subunits of Ecb2, bA and bB, are

more tightly associated in holo-Ecb2 than in apo-Ecb2;

the stronger subunit–subunit association for holo-Ecb2

results from local conformational changes caused by PLP binding

While the subunit–subunit interface did not change drastically between the crystal structures of dimeric apo-Ecb2 and holo-Ecb2, the crystal structures of sev-eral PLP-dependent b-elimination enzymes exhibit dif-ferent intersubunit interfaces for the apo-dimeric and holo-dimeric structures [23,27] In the apo-dimeric structure of serine dehydratase, the small domain rotates to open the active site cleft, allowing the small domain of the partner subunit to enter the opened cleft While bound PLP and potassium readily dissoci-ate from serine dehydratase by simple dialysis in the presence of cysteine [37], the binding of PLP to the tryptophan synthase b2 subunit is very tight (Kd of approximately 10)7m)

Mechanism of PLP binding Five regions (peaks 1–5) exhibited significant conforma-tional changes following PLP binding The N terminus

K87

S377

T190

S235

N236

H86

G232

SO42-K87

S377

T190

S235

G234 N236

H86

G232

SO42-Helix-9

Helix-8 Helix-7

Helix-12

Helix-9

Helix-8 Helix-7

Helix-12

PLP

long-loop

K87

S377

T190

S235

G234

N236

H86

POP4 OP2

OP3

OP1

G232

NZ C4A K87

POP4 OP2

OP3

OP1

NZ C4A

SO42-PLP

long-loop

K87

S377

T190 S235 G234

N236

H86 P

N1

OP4 OP2

OP3 OP1 G232

NZ C4A K87

POP4 OP2

OP3 OP1

NZ C4A

Helix-8 Helix-7

Helix-12

Helix-8 Helix-7

Helix-12

N1

G234

A

B

Fig 5 Schematic stereo view of the con-formations surrounding the active sites of (A) apo-Ecb and (B) holo-Ecb, shown from the same orientation Magenta, blue and green regions represent the N domain, the

C domain and the long loop, respectively The residues forming hydrogen bonds with PLP (yellow) are indicated by the ball-and-stick model, labeled in holo-Ecb (B) Figures displaying protein structures were prepared using MOLSCRIPT software [58].

of Helix 9 (peak 3 region) is structurally similar to a

region found in several other PLP-binding enzymes,

called the ‘anchoring alpha-helix’ [38] Therefore, the

first binding site of PLP must be the peak 3 region

(resi-dues 230–236) This region binds to the phosphate

group of PLP via multiple hydrogen bonds as well as by

the electrostatic effects of the Helix 9 dipole (Figs 4 and

5B) These interactions may rapidly induce the rigid

body rotation of the C domain, resulting in PLP binding

to holo-Ecb2via 10 hydrogen bonds, of which nine are

with the PLP phosphate group (Fig 5B) PLP binding

is also additionally stabilized by seven hydrogen bonds

and one hydrophobic interaction from the peak 3

region Schiff-base formation with Lys87 is not required

for the binding of PLP, as the K87T mutant of the

Sta2b2 b subunit [8] and the K41A mutant of S

ty-phimurium O-acetylserine sulfhydrylase both retained

the ability to bind PLP in the active site [39]

The peak 5 region near Arg379 also interacts with

the phosphate group of PLP Hydrogen bonds between

OG of Ser377 and N1 of PLP (Figs 4 and 5B) may

cause the conformational changes in this region upon

PLP binding The conformational changes in the peak

4 region near Thr319 are also affected by alterations

transmitted from the peak 3 region The C terminus of

the long-loop and the N terminus of Helix 10 are

pulled into the interior of the molecule by hydrogen

bonds between the N of Gly310 and the O of Gly233

and between the NE2 of His313 and the O of Gly234

The His86 side chain, which is located in the peak 2 region next to the Lys87 of the PLP-binding residue, shifted in location near the PLP upon PLP binding The H86L mutant of the Sta2b2 b subunit exhibited a PLP-binding ability that was reduced by approxi-mately 20-fold [40], indicating the importance of His86

in PLP binding The conformational changes in His86 upon PLP binding may correlate directly with interac-tions between the N and C domains, leading to the closure of the holo-b2 subunit PLP also forms a hydrogen bond with Thr190 in the N domain, located

in a loop between Sheet 6 and Helix 7 (Figs 4 and 5) The peak 4 region near Thr319 does not interact with PLP, but the conformation appears to be affected by the conformational changes in the peak 3 region The formation of these hydrogen bonds and the molecular rotation facilitate Schiff-base formation by bringing Lys87 close to PLP After the Schiff-base for-mation, residues in the long loop of the C domain form salt bridges and hydrogen bonds with residues in the N domain PLP binding to one b subunit (bA) induces a simultaneous conformational change in the other b subunit (bB) This conformational change in the bA subunit results in more efficient binding of PLP

to the bB subunit [28], generating an Ecb2 with both active sites bound to PLP Kinetic and calorimetric studies of PLP binding to the apo-b2subunit of trypto-phan synthase from E coli have suggested four bind-ing processes [29,41], which are consistent with these structural studies

Conformational changes in the b2subunit of tryptophan synthase caused by formation of the

a2b2complex

We compared the structures of holo-Ecb2 with the b2 subunits of Sta2b2 by superimposition of equivalent

Ca atoms (Fig 7) The rmsd values for a comparison between holo-Ecb2 and Sta2b2 without ligands bound

to either subunit active site (PDB code 1BKS), between holo-Ecb2 and Sta2b2 with a subunit ligands only (1A50), and between holo-Ecb2 and Sta2b2 with ligands bound to both the a and the b subunits (2TRS) were 0.5, 0.7 and 1.2 A˚, respectively The structure of holo-Ecb2 superimposed well on that of the b2 subunit from an unliganded Sta2b2 complex (1BKS) The structures of both b2 subunits were simi-lar despite the absence of residues 291–292 from the Ecb2 structure and differences between the long loop (residues 260–310) and the COMM (residues 102–189) domain The average B-factor value of the main-chain atoms for the long loop (82.00 A˚2) of Ecb2 was higher than that of the entire molecule (59.06 A˚2) By

con-Fig 6 Stereo view of the conformational changes in the

intersub-unit interface of the b2dimer following PLP binding The C domains

of apo-Ecb and holo-Ecb were superimposed The blue and

light-blue domains represent apo-Ecb and its pair subunit, respectively,

while the red and pink domains represent holo-Ecb and its pair

sub-unit, respectively Stick models are used to display the side chains

of the residues with conformational changes between the

apo-(cyan) and holo- (pink) b forms Black arrows denote the movement

of the residues following PLP binding Figures displaying protein

structures were prepared using PYMOL software [59].

trast, the B-factor of the long loop (26.50 A˚2) in the

Sta2b2lacking ligands was similar to that for the entire

molecule (28.30 A˚2) This result suggests that the

flexi-ble long loop becomes more rigid in the a2b2 complex

when compared with the b2 dimer as a result of the

formation of several hydrogen bonds between the a

and b subunits at their interface in the a2b2 complex

The long loop and the COMM domain constitute the

interface of the N and C domains that forms the wall

of the hydrophobic tunnel

Indole, the product of the a subunit, is transferred

via a 25–30 A˚ hydrophobic tunnel from the active site

of the a subunit to that of the b subunit [5,6] The

width of the holo-Ecb2tunnel was larger by 2–3 A˚ and

3–4 A˚ than those of Sta2b2 in the absence of ligands

(1BKS) and a-, b-liganded Sta2b2(2TRS), respectively

The tunnel gating residues bTyr279 and bPhe280,

which block the untimely passage of indole, were in a

closed conformation in the holo-Ecb2 and the

unli-ganded Sta2b2 (1BKS) As a hydrogen bond between

aAsp56 and bTyr279 seems to be essential for opening

the indole gate [16], it is likely that the tunnel gating

residues are also in a closed conformation in the

b2subunit In the holo-Ecb2, indole, the substrate, can

directly approach the active site because the width of

the cleft between the N and C domains in holo-Ecb2is

wide despite the indole gate being closed

An allosteric signal is transmitted between the a

and b subunits of tryptophan synthase via the salt

bridge-type interactions of bLys167 with bAsp305 and

aAsp56 The bLys167–aAsp56 salt bridge is more

important for the transmission of the allosteric signal

than the bLys167–bAsp305 salt bridge [42] The b

activities of bK167T, aD56A and bD305N mutants

are very low [43] These allosteric signals regulate the

transformation of the bienzyme complex from an open conformational state of low activity to a closed conformation state of high activity, which is triggered

by ligand bound to the a and⁄ or b active site Struc-tures of the unliganded wild-type (1BKS: a, open state; b, open state), IPP (indole propanol phosphate) bound wild-type (1A50: a, closed state; b, open state) and l-Ser-binding bK87T mutant (2TRS: a, closed state; b, closed state) proteins were compared with those of holo-Ecb2, especially around bAsp305 (Fig 8) The side chains of bLys167 and bAsp305, which are double conformers in the unliganded Sta2b2, exhibit different conformations in the struc-tures bearing different ligands (Fig 8A) The structure around bAsp305 of holo-Ecb2 is similar to the low-activity open state of Sta2b2 Holo-Ecb2 cannot trans-form into the higher-activity ‘closed-state’ from the

‘open-state’, resulting in a low-activity enzyme because

of the lack of interaction between bLys167 and aAsp56 Furthermore, the gate of the indole tunnel is blocked in the absence of any interaction with aAsp56 Therefore, aAsp56 appears to be one of the key residues for the activation of enzymatic activity upon complex formation

Conclusions

The crystal structures of apo- and holo-Ecb2 dimers were determined using X-ray crystallographic analysis

at 3.0 and 2.9A˚, respectively This is the first report of

b subunit structures with and without the PLP cofac-tor Holo-Ecb2 consists of two domains, the N and C domains, with a long loop in the C domain The long loop is not observed as a result of high flexibility in apo-Ecb2

Fig 7 Stereo view of the superimposed backbone structures of holo-Ecb and the ligand-unbound (1BKS) and ligand-bound (2TRS) Stb su-bunits Red, cyan, blue and yellow regions represent holo-Ecb, 1BKS, 2TRS and PLP, respectively This figure demonstrates that a cleft near the PLP-binding site between the N and C domains of holo-Ecb is wider than those seen in the other Stb subunits The COMM domain con-sists of the region from Gly93 to Gly189, which has few interactions with the rest of the protein, and plays an important role in allosteric communication between the a and b sites [11] Figures displaying protein structures were prepared using MOLSCRIPT software [58].

Comparison of the apo- and holo-Ecb2 structures

revealed a large conformational change in the C domain

between these two structures This conformational

change consisted of a rotation of the C domain by

approximately 22, which was triggered by the local

conformational changes induced by PLP binding to

the active site PLP binding to the apo-Ecb induced

the following major conformational changes (a) PLP

bound to the N terminus of Helix 9 induces a rigid

body rotation of the C domain This binding and

rota-tion results in Schiff-base formarota-tion involving Lys87

This interaction triggers a slow local conformational

re-arrangement in other regions, including residues of

the long loop of the C domain These changes were

primarily observed around the active site and hinge

residues connecting the N and C domains (b) PLP

binding to one b-subunit (bA) induces a

conforma-tional change in the other b subunit (bB) within a

dimer (c) PLP binding strengthens the b–b subunit

association as a result of increases in the hydrophobic

interactions, hydrogen bonds and salt bridges, resulting

in stabilization of the holo-form in comparison to the

apo-form, which was reflected in a higher denaturation

temperature of holo-Ecb2, as determined by the DSC

measurements, that was 25.3C greater than that of

apo-Ecb2at pH 9.0

Comparison of the holo-Ecb2 alone and Stb2 in

Sta2b2 structures revealed why the holo-Ecb2 enzyme

has lower b enzymatic activity in the absence of the a

subunit than Eca2b2 (a) The indole gating residues,

Tyr279 and Phe280, block entry of the substrate in the

b2 subunit alone, a situation similar to that seen for

the unliganded Sta2b2 complexed with Na+ (b) The

structure around bAsp305 of holo-Ecb2alone is similar

to the low-activity open state of Sta2b2 Holo-Ecb2

cannot transform into the higher-activity ‘closed-state’

from the ‘open-state’ because of the absence of an

interaction with aAsp56 (c) By contrast, the width of

the Ecb2 tunnel was larger, by about 2–3 A˚, than that

of unliganded Stb2 Furthermore, the width of the cleft

between the N and C domains in Ecb2 was wider,

resulting in better accessibility of the substrate (indole)

to the active site

Experimental procedures

Purification and preparation of holo-, apo- and

reconstituted holo-Ecb2

Holo-Ecb2 [44] was purified as described previously [45]

The purified protein was visible as a single band on

SDS⁄ PAGE Apo-Ecb2 was prepared by the dialysis of

holo-Ecb2 against 0.1 m Mes buffer (pH 6.5) containing

A

B

C

Fig 8 Stereo views of the structures surrounding the allosteric residue, bAsp305, in holo-Ecb2and the Sta2b2complex The sky-blue dotted lines in the figures represent hydrogen bonds (A) Red, holo-Ecb 2 ; gray, Sta 2 b 2 (1BKS, both the a and b subunits are in a low-activity open conformation state) (B) Red, holo-Ecb2; gray, Sta2b2(1A50, the a subunit is in a high-activity closed conformation state, while the b subunit is in the open state) (C) Red, holo-Ecb 2 ; gray, Sta2b 2 (2TRS, both a and b subunits are in closed states) Figures displaying protein structures were prepared using PYMOL

software [59].

0.1 m Li2SO4, for 3 days at 4C The reconstitution of

holo-Ecb2 was performed by dialyzing the apo-Ecb2

solu-tion against 50 mm potassium phosphate buffer (pH 7.0)

containing 0.2 mm PLP, for 1 day at 4C

Physicochemical properties of holo-Ecb2and

apo-Ecb2, monitored by CD, DSC and analytical

ultracentrifugation

The CD spectra were recorded using a Jasco J-720

spectro-polarimeter (JASCO Co., Hachioji, Tokyo, JAPAN) The

far-UV and near-UV⁄ visible CD spectra were scanned 16

and 36 times, respectively, at a scan rate of 20 nmÆmin)1,

using a time constant of 0.25 s The light path length of the

cell was 1 and 10 mm in the far-UV and near-UV⁄ visible

regions, respectively To calculate the mean residue

elliptic-ity [h] the mean residue weight was assumed to be 108.3

DSC was performed using a differential scanning

micro-calorimeter, VP-DSC (Microcal, Inc., Northampton, MA,

USA), at a scan rate of 1CÆmin)1 After dialysis against

the desired buffer, the sample was filtered through a

mem-brane with a pore size of 0.22 lm and degassed under a

vacuum The protein concentrations during measurements

were 0.2–1.4 mgÆmL)1

Sedimentation analysis was performed using a Beckmann

Optima mode XL-A centrifuge (Beckman Instruments, Inc.,

Fullerton, CA, USA) Sedimentation equilibrium

experi-ments used an An-60Ti rotor at a speed of 7658.3–8049.6 g

at 20C Before measurements, the apo- and holo-Ecb2

solutions were dialyzed overnight at 4C against 50 mm

potassium phosphate buffer, pH 7.9, containing 1 mm

EDTA, and the holo-Ecb2solution was dialyzed overnight

at 4C against 50 mm potassium phosphate buffer, pH 7.9,

containing both 1 mm EDTA and 0.2 mm PLP

Experi-ments at three different protein concentrations (ranging

from 0.92 to 0.31 mgÆmL)1) were run using Beckman

four-sector cells The partial specific volume was calculated,

from the amino-acid compositions, to be 0.754 cm3

Æg)1[46]

Analysis of the sedimentation equilibrium data was

per-formed using xlavel software (Beckman version 2.0)

Crystallization and data collection

For crystallization, purified holo-Ecb2 protein was

concen-trated to 20 mgÆmL)1in 20 mm potassium phosphate buffer

(pH 7.8) containing 0.1 mm dithiothreitol, 0.2 mm PLP and

a protease inhibitor mixture Holo-Ecb2 was crystallized

using the microbatch method with paraffin oil at 288K

Crystals were obtained with 0.5 lL of protein solution

combined with 0.5 lL of crystallization reagent [0.8 m

(NH4)2SO4, 0.1 m Tris buffer, pH 8.0] Light-yellow

hexag-onal rod-shaped crystals grew to maximal dimensions of

0.1· 0.1 · 0.8 mm over 1 month

Apo-Ecb2was crystallized using the hanging-drop

vapor-diffusion method at 288K Crystals were obtained by

equili-bration of a mixture containing 1 lL of protein and 1 lL

of crystallization reagent (0.25 m Li2SO4, 0.1 m Mes buffer,

pH 6.5) against 100 lL of the reservoir solution at 288K Needle-like clear crystals grew to maximal dimensions of 0.2· 0.02 · 0.02 mm in approximately 1 month All crys-tallizations were performed in the dark

Apo-protein crystals were soaked in cryoprotectant (0.1 m Mes buffer, pH 6.5, containing 0.4 m Li2SO4and 30% glyc-erol), while holo-protein crystals were soaked in 0.1 m Tris buffer (pH 8.0) containing 0.8 m (NH4)2SO4and 35% glyc-erol for several minutes before flash-cooling in liquid nitro-gen Data sets were collected at the beamline BL44XU at SPring-8 in Japan using the DIP6040 imaging plate detector (Bruker AXS, Inc., Karlsruhe, Germany) Data were pro-cessed using Mosflm [47] and ccp4 suite software [48] The apo-protein crystal belonged to space group P4322, with cell dimensions of a = b = 110.89, c = 102.21 A˚, one molecule

in an asymmetric unit, a solvent content of 66.1% and a specific volume (VM) of 3.7 A˚3Da)1[49] The crystal of the holo-form belonged to space group P6522, with the cell dimensions of a = b = 172.52, c = 82.59 A˚, one molecule

in an asymmetric unit, a solvent content of 69.9% and a specific volume (VM) of 4.1 A˚3Da)1 The statistics for data collection are summarized inTable1

Structure determination and refinement The apo- and holo-structures were determined by the molecular replacement method using cns software [50] The

b subunit of the Sta2b2 complex (PDB code 1BKS) was used as the initial search model The holo-Ecb structure was determined using a full-length search model The apo-Ecb structure, however, could be determined using an omitted search model without a long loop (residues 260– 310) All refinements were performed using cns software The sigma-A-weighted composite omit map was calculated

by CNS to reduce model bias The structure was visualized and modified using XtalView [51] and O [52] software Both the apo- and holo-Ecb refined models consisted of one molecule in the asymmetric unit The apo-Ecb model included residues 17–258, 309–397, three sulfate ions and

23 water molecules The holo-Ecb model included residues 3–290, 293–397, one PLP molecule, two sulfate ions, one glycerol molecule and nine water molecules The refinement statistics are summarized in Table 1 We have deposited the final coordinates of the tryptophan synthase b2 subunit from E coli into the Protein Data Bank

Analyses of amino acid sequences and protein models

The secondary structures of all coordinates were defined using procheck software [53] lsqkab software [54] in the CCP4 suite was used for superposition of all coordinates The potential hydrogen bonds and hydrophobic interactions