Báo cáo khoa học: Monomeric solution structure of the helicase-binding domain of Escherichia coli DnaG primase pdf

DnaG is composed of three main domains comprising an N-terminal zinc-binding domain for interaction with single-stranded DNA, a central domain responsible for primer synthesis, and a C-t

domain of Escherichia coli DnaG primase

Xun-Cheng Su, Patrick M Schaeffer, Karin V Loscha, Pamela H P Gan, Nicholas E Dixon

and Gottfried Otting

Research School of Chemistry, Australian National University, Canberra, Australia

All organisms replicate DNA by copying one strand

(the leading strand) in a continuous manner, whereas

the other DNA strand (the lagging strand) is replicated

in a discontinuous manner by the synthesis of short

Okazaki fragments that are later joined into a

continu-ous strand [1] During DNA replication, a helicase

sep-arates the double-stranded DNA into single strands,

and replication of the leading strand and synthesis of

the Okazaki fragments is initiated by RNA primers

made by the specialized RNA polymerase, primase

The first primase to be identified and characterized

was that from Escherichia coli

In E coli, the replicative helicase and primase are

encoded by the dnaB and dnaG genes, respectively

The DnaB helicase forms a hexameric ring structure

with up to three molecules of the DnaG primase attached [2–4] DnaG is composed of three main domains comprising an N-terminal zinc-binding domain for interaction with single-stranded DNA, a central domain responsible for primer synthesis, and a C-terminal domain (residues 434–581; DnaG-C) that binds to the DnaB helicase The binding interaction with DnaB locates DnaG in the correct position to lay down primers on newly formed single-stranded DNA

as the DnaB helicase progresses along the DNA Pri-mases are essential for DNA synthesis and are there-fore targets for the development of new antibiotics [5]

No 3D structure has been determined for full-length DnaG, but crystal structures have been obtained for the N-terminal domain from Bacillus stearothermophilus

Keywords

DnaB; DnaG; domain swap; NMR structure;

primase

Correspondence

G Otting, Research School of Chemistry,

Australian National University, Canberra,

ACT 0200, Australia

Fax: +61 2 61250750

Tel: +61 2 61256507

E-mail: Gottfried.Otting@anu.edu.au

Database

The NMR chemical shifts and coordinates of

the structure have been submiited to the

BioMagResBank (accession code 6284) and

Protein Data Bank (accession code 2HAJ)

(Received 28 July 2006, revised 7 September

2006, accepted 11 September 2006)

doi:10.1111/j.1742-4658.2006.05495.x

DnaG is the primase that lays down RNA primers on single-stranded DNA during bacterial DNA replication The solution structure of the DnaB-helicase-binding C-terminal domain of Escherichia coli DnaG was determined by NMR spectroscopy at near-neutral pH The structure is a rare fold that, besides occurring in DnaG C-terminal domains, has been described only for the N-terminal domain of DnaB The C-terminal helix hairpin present in the DnaG C-terminal domain, however, is either less sta-ble or absent in DnaB, as evidenced by high mobility of the C-terminal 35 residues in a construct comprising residues 1–171 The present structure identifies the previous crystal structure of the E coli DnaG C-terminal domain as a domain-swapped dimer It is also significantly different from the NMR structure reported for the corresponding domain of DnaG from the thermophile Bacillus stearothermophilus NMR experiments showed that the DnaG C-terminal domain does not bind to residues 1–171

of the E coli DnaB helicase with significant affinity

Abbreviations

DnaB(1–171), residues 1–171 of E coli DnaB helicase; DnaB-N, the N-terminal domain (residues 24–136) of E coli DnaB helicase; DnaG-C, the C-terminal domain of DnaG primase (residues 434–581 of the E coli protein); DTPA-BMA, diethylenetriamine pentaacetic

acid-bismethylamide; P16, the C-terminal domain of Bacillus stearothermophilus DnaG (residues 452–597).

[6], the central RNA polymerase domain from E coli

[7,8], the two-domain fragment comprising both the

N-terminal and RNA polymerase domains from

Aquifex aeolicus[9], and the C-terminal helicase-binding

domain from E coli [4] In addition, the structure of

the C-terminal domain from B stearothermophilus

(P16) has been determined by NMR spectroscopy [10]

Despite their conserved function, the crystal

struc-ture of E coli DnaG-C [4] and the subsequent solution

structure of B stearothermophilus P16 [10] show

sub-stantial differences, including a different number of

helices with different helix boundaries and a different

spatial arrangement of the C-terminal helices These

differences are important, because the C-terminal helix

hairpin is critical for the binding of DnaG to DnaB

[10,11] In P16, the C-terminal helices are only loosely

held in place by the rest of the structure [10] In both

structures, the N-terminal helices are packed in a fold

similar to that of the N-terminal domain of DnaB

(residues 24–136; DnaB-N) [12,13], and the DnaG-C

crystal structure shows the C-terminal helices from

different monomers entwined via intermolecular

con-tacts in a way reminiscent of the fold of DnaB-N The

dimer structure was distorted by crystal contacts,

resulting in noticeably different backbone

conforma-tions and different orientaconforma-tions of the C-terminal

heli-ces in each of the two monomers [4]

However, both gel filtration and analytic

ultracen-trifugation experiments at neutral pH showed that

DnaG-C was monomeric [4], and it was difficult to

ascribe any functional significance to the dimer In

addition, NMR spectroscopic analysis showed little

evidence for dimer formation in solution Some NOEs

were observed that were consistent with the dimer

interface observed in the crystal structure, and these

were interpreted as evidence for a monomer–dimer

equilibrium [4], but they could also arise from

intramo-lecular contacts in solution that are not present in the

monomers in the crystal structure In order to resolve

these difficulties and the discrepancies between the

structure of P16 (which is monomeric in solution) and

the different conformers in the crystal structure of

DnaG-C, we here report the solution structure of

E coli DnaG-C determined under conditions where

the protein is strictly monomeric

This new structure differs from the conformers

observed in the single crystal, reveals a fold even

closer to that of DnaB-N than the crystal

conform-ers, and shows no evidence for the presence of two

independent subdomains as in P16 The

conforma-tional rigidity of the monomeric DnaG-C structure

was confirmed by 15N-relaxation, coupling constant

and solvent accessibility measurements The structure

identifies the crystal structure of DnaG-C as a domain-swapped dimer that probably has no func-tional significance

The close fold conservation between DnaG-C and DnaB-N prompted us also to investigate a longer N-terminal construct of DnaB, DnaB(1–171), for the presence of a C-terminal helix hairpin as present in DnaG-C DnaB(1–171) comprises the complete N-ter-minal domain and hinge regions of DnaB identified by proteolysis [14], and includes peptide segments that have previously been shown by mutation analyses to modulate the interaction between DnaG and DnaB [3,11,15,16] Consequently, we also probed the interac-tion between DnaG-C and DnaB(1–171)

Results Aggregation state of DnaG-C DnaG-C is prone to self-aggregation at high protein concentration and in the absence of salt [17] Ultra-centrifugation experiments at 0.06 and 0.29 mm pro-tein concentration in the presence of 100 mm NaCl yielded Mr values of 16 500 and 14 100, respectively, indicating that the single species present was the monomer (calculated Mr¼ 16 701; supplementary Fig S1) To verify the monomeric state of the protein under the conditions used for NMR structure deter-mination (0.4 mm DnaG-C, pH 6.1, 100 mm NaCl,

25C), the rotational correlation time of DnaG-C was determined from the ratio of transverse and lon-gitudinal 15N relaxation rates The rotational correla-tion time sm was found to be 11 ± 1 ns, based on average values of R1¼ 0.99 ± 0.13 s)1 and R2¼ 20.41 ± 1.68 s)1 for the structurally well-defined part

of the protein (Fig 1) Increased R1 and decreased R2

relaxation rates indicated increased mobility and structural disorder for about 12 and three residues at the N-terminus and C-terminus of the construct, respectively, in agreement with the narrow 1H-NMR line widths reported earlier for these residues [17] Negative [1H]15N NOEs were observed for residues 437–441 at the N-terminus, demonstrating mobility

on the subnanosecond timescale, whereas the NOE was greater than 0.7 for residues 453–578, indicating structural rigidity for this part of the protein (data not shown)

The rotational correlation time of rigid protein structures can be predicted from the atomic coordi-nates using hydronmr [18] The rotational correlation times predicted for the individual monomers and the dimer in the crystal structure of DnaG-C [4] were about 17 and 36 ns, respectively, and thus much longer

than the value of 11 ns derived from the 15

N-relaxa-tion times in soluN-relaxa-tion However, rotaN-relaxa-tional correlaN-relaxa-tion

times of, respectively, 11 and 12 ns were predicted for

the corresponding domain from B stearothermophilus

[10] and for the monomeric DnaG-C structure

repor-ted here These data and the uniformity of the

relaxa-tion rates along the amino acid sequence (Fig 1)

supported the notion of DnaG-C being a monomeric,

structurally compact domain with no evidence for

seg-mentation into subdomains as observed in the crystal

structure [4] and reported for P16 [10]

Structure determination

The solution structure of E coli DnaG-C was

deter-mined using NOEs and backbone dihedral angle

restraints derived from chemical shifts All NOEs were

interpreted as intramolecular NOEs The resulting

monomeric structure fulfilled all assigned NOEs

with-out significant residual violations (Table 1) The fold

exposes all charged amino acid side chains to the

sol-vent and buries all hydrophobic side chains that are

highly conserved among different bacterial species

(Fig 2) The side chain solvent accessibility averaged

over the different NMR conformers is 16% or less for

any of the conserved hydrophobic side chains, except

for the side chain of Leu484, which is almost 30%

sol-vent exposed The conservation of Leu484 may be

explained by its contacts with Leu519, which is a

strictly conserved residue (Fig 2) Insertions and dele-tions in the sequence alignment of Fig 2 are all con-fined to loop regions, indicating that the secondary structure of DnaG-C is conserved among DnaG mole-cules from many different bacterial species

1.5

1.0

0.5

0.0

R1

s-1

440 460 480 500 520 540 560 580

Residue number

0

30

20

10

R2

s-1

Fig 1 15 N-relaxation rates measured for

15 N ⁄ 13 C-labeled DnaG-C The data were

measured at a 1 H-NMR frequency of

800 MHz, using a 0.4 m M solution of DnaG-C

in NMR buffer at 25 C Upper panel,

R1relaxation rates Lower panel, R2

relaxation rates Error bars indicate the error

reported by the fitting routine in SPARKY [40].

Table 1 Structural statistics for the NMR conformers of E coli DnaG primase C-terminal domain (DnaG-C).

Number of assigned NOE cross-peaks a 2400 Number of nonredundant NOE upper-distance

limits

2151 Number of dihedral-angle restraints 154 Intraprotein AMBER energy (kcalÆmol)1) ) 4575 ± 1176 Maximum NOE-restraint violations (A ˚ ) 0.17 ± 0.06 Maximum dihedral-angle restraint violations () 3.1 ± 3.1 rmsd for N, C a and C¢ (A˚) b,c 0.8 ± 0.2 rmsd for all heavy atoms (A ˚ ) b,d 1.2 ± 0.2 Ramachandran plot appearancee

Additionally allowed regions (%) 11.8 Generously allowed regions (%) 1.4

a Stereospecific resonance assignments were obtained for 26 pairs

of C b H2groups, two pairs of C c H2and C d H2groups, and six pairs

of CcH 3 and CdH 3 groups.bFor residues 449–576.c0.5 ± 0.1 A ˚ for residues 449–525 d 0.9 ± 0.1 A ˚ for residues 449–525 e From PRO-CHECK NMR [37] f All residues in disallowed regions were located in loop regions or at the C-terminus of the structure No residue was consistently found in disallowed regions.

The fold of DnaG-C comprises six helices arranged

as in the N-terminal domain of DnaB (Fig 3A,B)

Pairwise comparison using the CE server [19] gave an

rmsd between the two proteins of 3.3 A˚ for 101

aligned residues No other protein in the Protein Data

Bank has a similar fold (other than P16 from

B stearothermophilus; see below)

Comparison with the crystal structure of DnaG-C The crystal structure of dimeric DnaG-C [4] contains two DnaG-C molecules with different orientations and boundaries of helix 6 (Fig 3D,E), showing that this helix can be separated from the core of the structure The solution structure of DnaG-C identifies the crystal

Fig 2 Sequence alignment of DnaG-C with homologs from different bacterial species The sequence numbering of E coli DnaG-C is shown

at the top, together with the helix boundaries of DnaG-C determined in this work Conserved hydrophobic residues are shaded dark gray The amino acid sequence of DnaG-C from B stearothermophilus is shown at the bottom together with the helix boundaries reported by Syson et al [10] The following sequences from DnaG-C proteins are shown (abbreviation, species, GenBank number): E coli, Escherichia coli, 130908; S enterica, Salmonella enterica subsp enterica serovar Paratyphi A, str ATCC 9150, 56129407; Y pestis, Yersinia pestis CO92, 15978733; P luminescens, Photorhabdus luminescens subsp laumondii TTO1, 36787269; E carotovora, Erwinia carotovora subsp atroseptica SCRI1043, 49610155; B aphidicola, Buchnera aphidicola str Sg (Schizaphis graminum), 21622949; C blochmannia, Candidatus blochmannia pennsylvanicus str BPEN, 71795953; V parahe, Vibrio parahaemolyticus RIMD 2210633, 28805388; H somnus, Haemophilus somnus 2336, 46156266; P multocida, Pasteurella multocida subsp multocida str Pm70, 12721596; I loihiensis, Idiomarina loihiensis L2TR, 56180311; P profundum, Photobacterium profundum SS9, 46912067; X fastidiosa, Xylella fastidiosa Dixon, 71164362; L pneumophila, Legionella pneumophila, 1575484; P syringae, Pseudomonas syringae pv tomato str DC3000, 28851001; B stearo, Bacillus stearothermo-philus, 78101045 The sequences were identified and aligned in a BLAST search [41], except for the sequence of B stearothermophilus, which was aligned on the basis of its secondary structure elements.

structure of DnaG-C as a domain-swapped dimer,

where helix 6 from one protein molecule binds to the

core of the other in a manner similar to that in which

helix 6 binds to the core of the structure in the

mono-meric solution structure (Figs 3A and 4)

The two conformers in the crystal structure vary not

only with regard to helix 6 (Fig 3D,E) but also in the

part comprising helices 1–5, with a backbone rmsd of

2.0 A˚ for residues 449–525 The differences are mostly

due to a displacement of helix 5 and variability in the

loop region between helices 2 and 3 The backbone

rmsd for the same residues with respect to the solution

structure is 1.8 ± 0.1 and 2.4 ± 0.1 A˚ for crystal

con-formers I and II, respectively The largest differences

are in the loop region between helices 2 and 3,

suggest-ing that this region is flexible

Whereas 15N-HSQC spectra of DnaG-C at pH 4.6,

6.1 and 8.1 displayed virtually the same chemical

shifts, some of the cross-peaks in the spectrum

recor-ded at pH 4.6 (the pH used for crystallization) were

exceedingly weak, especially in the loop regions

between helices 2 and 7 (supplementary Figs S2 and S3) This indicates the presence of chemical exchange phenomena at low pH in the millisecond time regime Increased mobility of the loop regions at pH 4.6 and 8.1 was also suggested by the observation of enhanced

15N-relaxation rates (supplementary Fig S4) There-fore, the domain swap observed in the crystal structure may have been due to the use of a pH value below the isoelectric point of the protein (5.0) As comparable NMR line widths and 15N-relaxation rates were observed for the regular secondary structure elements

at all three pH values, the domain-swapped dimer is not the major conformational species even at low pH

Comparison with P16 from B stearothermophilus Except for the C-terminal helices, the solution struc-ture of P16, the DnaG-C domain from B stearo-thermophilus [10], shows the same overall fold as the present solution structure of E coli DnaG-C (Fig 3A,C) However, the similarity is less striking

h1

h2 h3

h4

h5

h6

h7

h1

h2 h3

h4

h5

h6

h6

h7

crystal conformer I crystal conformer II

DnaG-C DnaG-C

h1

h2 h3

h4

h5

h6

h7

h1

h2 h3

h4

h5

h2

h3

h4

h5

h6 h7

h8

solution structure

Fig 3 Ribbon representations of DnaG-C and related proteins (A) E coli DnaG-C The short 310helix between helices 2 and 3 was found in fewer than half of the NMR conformers and was therefore not labeled It was also found in conformer II but not conformer I of the crystal structure [4] (B) N-terminal domain of E coli DnaB (residues 24–136) [12] (C) B stearothermophilus DnaG-C (fragment P16) [10] (D) Con-former I of the crystal structure dimer of E coli DnaG-C [4] (E) ConCon-former II of the crystal structure dimer of E coli DnaG-C [4].

than anticipated based on the functional similarity of

DnaG-C domains, with a backbone rmsd of 3.2 A˚ for

88 aligned residues from the globular part of P16

(cor-responding to residues 449–543 of E coli DnaG-C),

which excludes helices 7 and 8 of P16 (Fig 1) Helices

6 and 7 of P16 do not form a single continuous helix

as in E coli DnaG-C, but are connected by a flexible linker, entailing a very different orientation of the C-terminal helix hairpin with respect to the core of the structure [10]

A

H541 H541

K447 K447

I530 I530

C

C

B

C

F535 L464 L454

E532 E532

L454 L464 F535

Fig 4 Stereo views of the solution and crystal structures of DnaG-C (A) Superposi-tion of the backbone atoms of residues 447–581 of the 20 NMR conformers of DnaG-C representing the NMR structure (Table 1) Numbers identify sequence posi-tions as in Fig 2 The 15 flexible N-terminal residues were not plotted (B) Stereo view

of the DnaG-C conformer closest to the mean structure of the 20 conformers shown

in (A), using a heavy atom representation The polypeptide backbone is drawn as a rib-bon and the flexible N-terminal 15 residues are omitted for clarity The following colors were used for the side chains: blue, Arg, Lys, His; red, Glu, Asp; yellow, Ala, Cys, Ile, Leu, Met, Phe, Pro, Trp, Val; gray, Asn, Gln, Ser, Thr, Tyr Darker-shaded bold lines indi-cate the side chains of Lys447, Lys448, Ile530 and His541 (C) Domain-swapped dimer in the crystal structure of DnaG-C [4] Only residues 447–528 of conformer I and residues 527–580 of conformer II of the crystal structure are shown, with white and magenta ribbons tracing the backbones of the respective conformers Darker-shaded bold lines indicate the side chains of Lys447, Lys448, Ile530 and His541 The side chain of Ile530 is buried in the dimer interface by packing against Ile530 from the other monomer (not shown) The side chains of Glu532, Phe535, Leu454 and Leu464 are labeled NOEs between these residues are explained by the monomeric solution structure, but are also predicted by intermolecular interactions in the dimer of the crystal structure [4].

The structural differences between P16 and E coli

DnaG-C may be explained by the low sequence

homology between the two proteins Although P16

fea-tures 14 of the 16 hydrophobic side chains found with

high conservation among DnaG-Cs from different

bac-terial species, the structure-based sequence alignment

of Fig 2 resulted in only 13% sequence identity

between P16 and E coli DnaG-C The low sequence

homology also explains why our structure-based

sequence alignment is very different from the sequence

alignment reported earlier [10]

The flexibility of the linker peptide connecting

heli-ces 6 and 7 in P16 (Fig 3C) and the different breaking

points in helix 6 of E coli DnaG-C observed in the

crystal structure (Fig 3D,E) raise questions about the

flexibility of helix 6 of E coli DnaG-C in solution

Structure verification of helix 6 of DnaG-C

An extensive set of Ha(i)-HN(i+3) NOEs, 3JHNHa

coupling constants smaller than 6 Hz, and chemical

shifts (15N,13Ca,13Cb,1Haand13C¢) indicative of

heli-cal secondary structure along the length of helix 6, all

suggest that a straight helix as depicted in Fig 3A is a

faithful representation of this helix in DnaG-C under

the conditions of the NMR experiments

Measure-ments of the 3JHNHa coupling constants at 20 lm

rather than 0.4 mm protein concentration (data not

shown) did not yield significantly increased coupling

constants, showing that the structure of helix 6 is not

stabilized by concentration-dependent self-association

Although the NMR structure of DnaG-C should be

a reliable representation of the average structure in

solution, this does not exclude the possibility of small

populations of conformers with spontaneously formed

transient breaks in helix 6 as a possible prelude to the

formation of a domain-swapped dimer We carefully

analyzed the NOESY spectra of DnaG-C with regard

to this question As NOEs strongly emphasize the

presence of short internuclear distances, NOE spectra

can convey the signature of minor conformational

spe-cies if short internuclear distances occur in a minor,

but not in the major, conformation However, the 3D

15N-NOESY-HSQC spectrum of DnaG-C recorded at

0.4 mm protein concentration on a 800 MHz NMR

spectrometer showed no significantly different NOE

patterns compared to the corresponding spectrum

recorded previously on a 600 MHz NMR spectrometer

with a 0.6 mm sample in the same NMR buffer [4] In

particular, strong sequential HN–HN NOEs and weak

sequential Ha–HN NOEs characteristic of helical

sec-ondary structure were found all along helix 6

Further-more, no evidence for a minor population of the

domain-swapped dimer could be found, as all NOEs previously thought to be indicative of the domain-swapped dimer [4] were in agreement with the present monomeric structure and independent of protein con-centration between 0.2 and 0.4 mm

The flexibility of helix 6 was further investigated by measurements of the solvent accessibility of amide pro-tons as evidenced by enhanced 1H-NMR line widths observed in the presence of a soluble paramagnetic relaxation agent Breaks in this helix would be expec-ted to interrupt the hydrogen bonding pattern and expose some of the amide protons to the solvent We used Gd[diethylenetriamine pentaacetic acid-bismethyl-amide (DTPA-BMA)] as an uncharged relaxation enhancement agent that does not change the chemical shifts of the protein signals [20] In addition, we used a low protein concentration (40 lm) to minimize the chance of any self-association Comparison of the peak heights measured in 15N-HSQC spectra recorded with and without Gd(DTPA-BMA) revealed pronounced solvent exposure only for loop regions between helices and for the flexible N-terminal residues (Fig 5) In contrast, the amide protons of helix 6 were among the protected protons In view of the uncertainty ranges associated with the data points, the slightly enhanced relaxation rates observed for the amide protons of resi-dues 541, 543, and 548 barely indicates significant temporary solvent exposure in a conformational equilibrium

Structure investigation of DnaB(1–171) The striking structural homology between DnaG-C and DnaB-N (Fig 3A,B) invites the question of whe-ther a longer construct of DnaB-N could display a C-terminal helix hairpin like DnaG-C, considering that

it is a feature of all DnaG-C conformers reported to date Secondary structure prediction of DnaB suggests

a helix for residues 153–169 and an extension of helix

6 by 11 amino acids to residue 145 As our original DnaB-N construct was truncated at Glu161, this could have caused the random coil behavior reported from residue 137 onwards [21]

A TOCSY spectrum recorded of DnaB(1–171), how-ever, displayed the same cross-peaks as the TOCSY spectrum reported previously of DnaB(1–161) [21] with additional cross-peaks for the 10 additional C-terminal residues (data not shown) Owing to the increased Mr

effected by dimerization of the DnaB-N domain [12], the TOCSY spectrum recorded with a long mixing time (80 ms) strongly emphasizes the signals from the mobile residues with narrow line widths In the

TOC-SY spectrum of DnaB(1–171), narrow line widths and

random coil chemical shifts were observed for the

entire polypeptide segment from residues 137 to 171

Therefore, the C-terminal helix hairpin observed in

DnaG-C is not a structural feature of DnaB(1–171)

Interaction of DnaG-C with DnaB(1–171)

Binding of DnaG-C to DnaB(1–171) was probed

by comparing the 15N-HSQC spectra of 0.13 mm 15N⁄

13C-labeled DnaG-C in the absence and presence of an

equal amount of unlabeled DnaB(1–171) No chemical

shift changes or changes in peak intensities were

detec-ted This indicates that any binding between these two

domains would be characterized by a dissociation

constant of at least 0.5 mm A dissociation constant of

4.9 lm has been reported for the complex between

DnaG-C and full-length DnaB from BIAcore studies

[4]

In agreement with the NMR results, no inhibitory

interaction between DnaB(1–171) and full-length

DnaG could be observed in a BIAcore assay, where a

5 lm solution of DnaB(1–171) was mixed with 285 nm

DnaG prior to its injection over a surface displaying

single-stranded DNA-bound DnaB hexamer, under

conditions used in our earlier studies [4] (data not

shown) Furthermore, there was no sign of toxicity of

DnaB(1–171) when overexpressed in E coli, as might

have been expected if tight binding of DnaB(1–171) to

DnaG were to compete with its interaction with the DnaB hexamer

Discussion The present structure determination of DnaG-C revealed a fold very similar to that of the N-terminal domain of the E coli DnaB helicase (DnaB-N) [12,13] The similarity includes helix 6, which is differently ori-ented in the conformers of the domain-swapped dimer (Fig 3) The structural similarity between DnaG-C and DnaB-N is intriguing, as no other protein is known with this particular fold, and DnaG binds to DnaB In view of the critical importance of the C-ter-minal helix hairpin of DnaG-C for the interaction with DnaB [4,10], it is tempting to speculate that the domain-swapped dimer observed in the crystal struc-ture of E coli DnaG-C might serve as a model for the interaction with DnaB-N

Many attempts have been made to pinpoint the interaction between DnaG and DnaB to protein sub-domains Whereas the interaction seems to be entirely confined to the C-terminal domain of DnaG [4,10], the situation is much less clear for DnaB For example, mutations in the N-terminal domain of E coli DnaB have been shown to interfere with the DnaB–DnaG interaction [22], but corresponding mutations in B ste-arothermophilus had much smaller if any effects [3,16]

440 460 480 500 520 540 560 580 0.0

0.5

1.0

h1 h2 h3 h4 h5 h6 h7

Residue number

Relative

intensity

Fig 5 Intensity ratio of backbone amide cross-peaks in 15 N-HSQC spectra of 0.04 m M15N ⁄ 13 C-labeled E coli DnaG primase (DnaG-C) in the presence and absence of 6.0 m M Gd(DTPA-BMA).

Apparently inconsistent results could arise from the

fact that E coli DnaB-N is only marginally stable

against unfolding [23,24] and easily destabilized by

mutations In B stearothermophilus, DnaG was found

to protect the linker residues between the N-terminal

and C-terminal domains of DnaB from digestion with

trypsin and pepsin [25] Mutations of linker residues

(I135N, I141T, L156P) also affected the interaction of

Salmonella typhimurium DnaB and DnaG [15] In

E coli, the interaction depends in addition on residues

of the C-terminal domain between Tyr210 and Val255 of

DnaB [26] Mutation analysis of linker residues and of

residues in the C-terminal domain of B

stearothermo-philus DnaB confirmed the importance of residues in

these parts of the protein [3,16] Unlike in the

wild-type protein, the individual N-terminal and C-terminal

domains of B stearothermophilus DnaB do not form a

complex with DnaG that is sufficiently stable for

isola-tion by gel filtraisola-tion [25] The emerging picture is one

of an extensive interaction interface between DnaB

and DnaG-C involving the N-terminal and C-terminal

domains of DnaB as well as the connecting linker

[3,16]

Interactions characterized by exceedingly weak

bind-ing affinities can be probed sensitively by NMR

spectro-scopy However, attempts to observe an interaction

between E coli DnaG-C and a shorter DnaB-N

frag-ment containing the N-terminal 161 residues by NMR

spectroscopy were unsuccessful [4] Our new fragment

DnaB(1–171), which includes many of the linker

resi-dues, equally showed no binding with DnaG-C or

DnaG, illustrating the critical importance of the

C-ter-minal domain of DnaB for this interaction Possibly,

the linker between the N-terminal and C-terminal

domains of DnaB also assumes a different secondary

structure in the full-length protein, considering that we

found the C-terminal 35 residues of DnaB(1–171) to be

disordered, although secondary structure predictions

show high helix propensity for more than half of them

The present structure of monomeric E coli DnaG-C

identifies the earlier crystal structure of the same

pro-tein as a domain-swapped dimer, in which helix 5 of

one monomer binds to the core of helices formed by

helices 1–4 of the other, in a very similar manner as in

the monomeric solution structure The present data

suggest that the domain-swapped dimer occurs only at

a pH value below the isoelectric point of the protein

and plays no role under physiologic conditions As the

present solution structure of DnaG-C accommodates

all the NOEs discussed previously [4] in a monomeric

structure, there remains no evidence for intermolecular

interactions across a dimer interface, and no

conform-ational exchange phenomena need to be invoked to

explain differences between the NMR data and the crystal structure [4]

The sensitivity of the DnaG-C structure with respect

to pH is reflected in much decreased peak intensities for loop residues observed in 15N-HSQC spectra at

pH 4.6 versus those recorded at pH 6.1 or 8.1, and in increased 15N-relaxation rates for amides in loop regions These exchange phenomena indicate the pres-ence of alternative conformations, especially at low

pH Considering that carboxylate side chains remain mostly deprotonated at pH 4.6, the low-pH form of the DnaG-C structure may be triggered by protonation

of histidine side chains Of the two histidine residues

in DnaG-C, His541 is located in helix 6 In the solu-tion structure, His541 is close to Lys447 and Lys448, which are located near the N-terminus of the domain, whereas these residues are much farther from His541

in the domain-swapped dimer (Fig 4A,B) Electro-static repulsion could thus drive the separation of helix

6 from the core of the structure Weak and missing

15N-HSQC cross-peaks observed for His541 and nearby residues, including residues 445–450, suggest that histidine protonation contributes to the exchange phenomena at pH 4.6 (supplementary Fig S3)

Comparison of the solvent-accessible surface of hydrophobic amino acid side chains in the monomer and the dimer shows only few significant differences, with the most notable difference involving the side chain of Ile530, which is highly solvent exposed in the monomer (Fig 4B) but buried in the dimer interface Neither His541 nor Ile530 are conserved in the amino acid sequence (Fig 2), suggesting that the phenomenon

of domain-swapping at low pH may be limited to DnaG-C from E coli Considering, in addition, the apparent absence of any interaction between DnaG-C and DnaB(1–171), the domain-swapped dimer of DnaG-C is unlikely to be a model of the DnaG–DnaB interaction

The equivalent DnaG-C domain from B stearother-mophilus (P16) [10] is a monomer in solution, but helix

6 in this structure is broken into two (Fig 3C) A flex-ible helix linkage is supported by the presence of Pro556 in P16, which may act as a helix breaker The corresponding residue in E coli DnaG-C is Met542, i.e a residue with high helix propensity None of the other DnaG-C domains shown in the sequence align-ment of Fig 2 features a proline residue at this posi-tion, suggesting that a break in helix 6 is not a general feature of DnaGs from different organisms Therefore, although the present solution structure of E coli DnaG-C is representative of DnaG-C domains from a large number of bacteria, significant structural variabil-ity seems to have evolved in less closely related species,

where the sequence divergence is sufficiently large to

render amino acid sequence alignments unreliable [10]

This observation highlights the fact that structures

determined for thermophilic or Gram-positive bacteria

are not necessarily faithful representations of their

homologs in E coli, the bacterium for which most

bio-chemical knowledge has been accumulated

Experimental procedures

Sample preparation

Unlabeled and uniformly13C⁄15N-labeled DnaG-C

contain-ing residues 434–581 was overproduced and purified as

pre-viously described [17] All samples for NMR measurements

were prepared in a buffer containing 90% H2O⁄ 10% D2O,

10 mm phosphate (pH 6.1), 100 mm NaCl and 1.0 mm

dithiothreitol The protein concentration was 0.4 mm except

where indicated otherwise

The DnaB(1–171) deletion mutant was amplified by PCR

from plasmid pPS562 containing the dnaB gene [27] An

NdeI site was present at the ATG start codon, and a TAA

stop codon followed by an EcoRI site was inserted

immedi-ately after codon 171 The amplified fragment was digested

and inserted between corresponding restriction sites in the

phage T7 promoter-based vector pETMCSI [28] and

trans-formed into E coli strain BL21(DE3)recA [23] for protein

expression Nucleotide sequences were confirmed using an

ABI 3730 sequencer (Biomolecular Resource Facility,

Aus-tralian National University, Canberra, Australia), following

the recommendations of the manufacturer (Applied

Biosys-tems, Foster City, CA, USA) DnaB(1–171) was produced

and the cells were lysed using a procedure established for

other DnaB-N domains [21] After cell lysis, the protein

was purified as described [12], except that the Sephadex

G50 column (Amersham Biosciences, Uppsala, Sweden)

was equilibrated with 50 mm Tris⁄ HCl (pH 7.6), 5 mm

MgCl2 and 100 mm NaCl Peak fractions containing

DnaB(1–171) were pooled (20 mL), diluted with an equal

volume of MonoQ buffer (50 mm Tris⁄ HCl at pH 7.6 and

5 mm MgCl2), and loaded directly onto a MonoQ (HR 5⁄ 5)

column (Amersham Biosciences) equilibrated in MonoQ

buffer A linear gradient of NaCl in MonoQ buffer was

applied (3.75 mmÆmin)1, at a flow rate of 0.5 mLÆmin)1)

DnaB(1–171) eluted as a sharp peak between 52 and

58 min The protein fractions were pooled and dialyzed in

NMR buffer ESI MS confirmed the identity of the protein

and the absence of an N-terminal methionine (observed

molecular mass, 18 919; calculated molecular mass 18 920)

Analytic ultracentrifugation

The molecular weights of DnaG-C samples were

deter-mined by equilibrium sedimentation using a Beckman

analytical ultracentrifuge XLI with An-60 rotor (Beckman Coulter, Fullerton, CA, USA) The samples were prepared

by dialysis against a buffer similar to that used for NMR studies, containing 10 mm sodium phosphate (pH 6.1),

100 mm NaCl, 1 mm dithiothreitol, and 0.1% (w⁄ v) sodium azide at two different concentrations (1.02 and 4.86 mgÆmL)1) The sedimentation equilibrium profile was recorded in triplicate at two different wavelengths (280 and

300 nm) after 18 h at 20 000 r.p.m and 25C Plots of

ln A versus r2 were linear (supplementary Fig S1), indica-ting the absence of an equilibrium mixture of species at both concentrations The average Mr was calculated by linear regression using ultrascan data analysis software Version 5 (Beckman Coulter), and an (assumed) partial specific volume of 0.72 mLÆg)1

NMR measurements

NMR measurements of unlabeled DnaB(1–171) were car-ried out in a buffer containing 10 mm Tris⁄ HCl (pH 6.5),

50 mm NaCl, 5 mm MgCl2 and 1 mm dithiothreitol Free DnaB(1–171) was measured at a concentration of 0.22 mm The interaction with DnaG-C was probed using the same buffer with each protein at 0.13 mm

All NMR spectra were recorded at 25C using a Bruker (Karlsruhe, Germany) AV 800 NMR spectrometer equipped with a TCI cryoprobe The previously reported backbone resonance assignments of DnaG-C [4] were veri-fied and supplemented with side chain resonance assign-ments using 3D CC(CO)NH, HNHA (H)CCH-TOCSY, NOESY-15N-HSQC (60 ms mixing time), 13 C-HSQC-NOESY (40 ms mixing time), and 2D C-HSQC-NOESY (40 ms mixing time), DQF-COSY, and TOCSY spectra

3JHNHacoupling constants were measured at protein con-centrations of 20 and 400 lm, in a CT-HMQC-HN experi-ment [29] The solvent exposure of protein backbone amides was probed by the decrease in peak intensities observed in15N-HSQC spectra caused by 6 mm Gd(DTPA-BMA) [20] The experiment was carried out at protein con-centrations of 20 and 40 lm

15

N-relaxation parameters (R2, R1, and [1H]15N-NOE) were measured [30], using relaxation delays of 3, 30, 80,

150, 250, 400, 600, 850 and 1200 ms in the R1experiment, and relaxation delays of 8.8, 17.6, 26.4, 35.2, 44.0, 52.8, 61.6, 70.4, 79.2 and 88.0 ms in the R2experiment The rota-tional correlation time sm was estimated from the R2⁄ R1

ratio [31]

A TOCSY spectrum of DnaB(1–171) was recorded under the same conditions, using a mixing time of 80 ms

Restraints used for the structure calculation

In total, 2400 NOE cross-peaks were assigned and integra-ted, resulting in 2151 meaningful distance restraints Most