Molecular basis for the subunit assembly of the primase
from an archaeon Pyrococcus horikoshii
Nobutoshi Ito
1,2
, Ikuo Matsui
3
and Eriko Matsui
3
1 Cellular Physiology Laboratory, Discovery Research Institute, RIKEN, Wako, Saitama, Japan
2 Division of Biofunctional Science, Tohoku University Biomedical Engineering Research Organization, Sendai, Japan
3 Biological Information Research Center, National Institute of Advanced Industrial Science and Technology, Ibaraki, Japan
In chromosomal replication, DNA polymerases require
the 3¢-OH group of a primer terminus to initiate syn-
thesis of a nascent DNA chain, as they cannot catalyze
phosphodiester bond formation between two deoxyri-
bonucleotides (dNTPs) [1]. In bacteria, viruses, arch-
aea, and eukaryotes, DNA polymerases add dNTPs to
the termini of short oligoribonucleotide primers,
thereby producing RNA-DNA hybrid molecules [1–3].
Primases, single-stranded DNA-dependent RNA
polymerases, are classically divided into bacterial, viral,
and archaeal ⁄eukaryotic groups [4]. In addition, it was
recently reported that a unique bifunctional DNA pri-
mase-polymerase is encoded by an archaeal plasmid
[5]. X-ray crystal structures of the catalytic domains of
both primase groups have been reported. These include
bacterial primases (Escherichia coli [6,7], T7 bacterio-
phage [8] and Aquifex aeolicus [9]), archaeal plasmid
primases (Sulfolobus islandicus [10]) and archaeal ⁄euk-
aryotic primases (Pyrococcus furiosus and Pyrococcus
horikoshii [11,12]).
Archaeal ⁄eukaryotic primases form a heterodimer
that consist of a small catalytic subunit (PriS) and a
Keywords
Archaea; primase; Pyrococcus horikoshii;
replication; X-ray crystallography
Correspondence
I. Matsui, Biological Information Research
Center, National Institute of Advanced
Industrial Science and Technology, 1–1
Higashi, Tsukuba, Ibaraki 305–8566, Japan
Fax: +81 298 616151
Tel: +81 298 616142
E-mail: ik-matsui@aist.go.jp
Database
The atomic coordinates and structure fac-
tors have been deposited in the Protein
Data Bank (accession code 2DLA), Research
Collaboratory for Structural Bioinformatics,
Rutgers, the State University of New Jer-
sey, New Brunswick, NJ (http://www.rcsb.
org/).
(Received 29 November 2006, accepted 11
January 2007)
doi:10.1111/j.1742-4658.2007.05690.x
Archaeal ⁄eukaryotic primases form a heterodimer consisting of a small cat-
alytic subunit (PriS) and a large subunit (PriL). The heterodimer complex
synthesizes primer oligoribonucleotides that are required for chromosomal
replication. Here, we describe crystallographic and biochemical studies of
the N-terminal domain (NTD) of PriL (PriL
NTD
; residues 1–222) that bind
to PriS from a hyperthermophilic archaeon, Pyrococcus horikoshii, at 2.9 A
˚
resolution. The PriL
NTD
structure consists of two subdomains, the helix-
bundle and twisted-strand domains. The latter is structurally flexible, and
is expected to contain a PriS interaction site. Pull-down and surface plas-
mon resonance analyses of structure-based deletion and alanine scanning
mutants showed that the conserved hydrophobic Tyr155-Tyr156-Ile157
region near the flexible region is the PriS-binding site, as the Y155A ⁄
Y156A ⁄I157A mutation markedly reduces PriS binding, by 1000-fold.
These findings and a structural comparison with a previously reported
PriL
NTD
–PriS complex suggest that the presented alternative conformations
of the twisted-strand domain facilitate the heterodimer assembly.
Abbreviations
CTD, C-terminal domain; dNTP, deoxyribonucleotide; GST, glutathione S-transferase; NTD, N-terminal domain; PriL, DNA primase large
regulatory subunit; PriS, DNA primase small catalytic subunit; RU, resonance unit; SeMet, seleno-L-methionine; SD, subdomain.
1340 FEBS Journal 274 (2007) 1340–1351 ª2007 The Authors Journal compilation ª2007 FEBS
large subunit (PriL). PriS has strictly conserved cata-
lytic aspartate residues, which are essential for the
nucleotidyl transfer reaction [13–15]. Archaeal PriL is
composed of two structural domains, the N-terminus
(PriL
NTD
; amino acid residues 1–222) and the C-termi-
nus (PriL
CTD
; residues 223–396). The former domain
interacts tightly with PriS, and the latter is responsible
for binding ssDNA [15–17]. For efficient primer syn-
thesis, PriS–PriL complex formation is required, sug-
gesting that PriL
NTD
links the DNA binding function
of PriL
CTD
and the catalytic activity of PriS [17,18].
Biochemical data has indicated the functional import-
ance of PriL, but a detailed understanding of its
molecular structure has not been achieved.
To determine the molecular basis for PriS binding
by PriL
NTD
, we report crystallographic and biochemi-
cal studies of PriL
NTD
from a hyperthermophilic arch-
aeon, P. horikoshii (Pho). Recently, the crystal
structure of the PriL
NTD
–PriS complex from the arch-
aeon Sulfolobus solfataricus (Sso) has been reported
[19]. These studies suggest that the hydrophobic sur-
face of the twisted-strand domain is the PriS binding
site, which undergoes a conformational change during
heterodimer formation. These results indicate the
molecular mechanism for the assembly of the PriS–
PriL complex from their components.
Results
PriL
NTD
is a PriS-binding module
We initially investigated the physical interaction of
PriS with full-length PriL, PriL
NTD
, and PriL
CTD
fused
to glutathione S-transferase (GST) using a pull-down
assay with glutathione Sepharose resin. PriL
NTD
and
full-length PriL bound to PriS, indicating that
PriL
NTD
, but not PriL
CTD
, contributes to the interac-
tion with PriS (Fig. 1, lanes 6, 8 and 10). This prefer-
ential binding of PriS to PriL
NTD
is in accord with a
previous report showing that PriL
NTD
has a stronger
binding constant to PriS than PriL
CTD
, as measured
by surface plasmon analysis [17]. We conclude that the
interaction of PriS with PriL occurs mainly through a
binding site in PriL
NTD
.
Overall structure of PhoPriL
NTD
The X-ray crystal structure of PriL
NTD
was solved by
the method of multiwavelength anomalous diffraction
[20] (Table 1 and Fig. 2A). The structure of PriL
NTD
comprises two subdomains, SD1 and SD2 (Fig. 2A,B).
SD1 (residues 1–100 and 166–222) is entirely helical,
consisting of a core three-helix bundle structure
(helices H5, H6, and H11) surrounded by proximal
short helices H1-H4 (Fig. 2A,B). Helices H5 and H6
form an antiparallel helix-bundle structure that is
intersected by the long H11 helix. Helices H1, H2, and
H3 wrap around helix H5 from three directions. Helix
H12 protrudes from SD1, and it is likely that PriL
CTD
can be positioned around the end of the helix. SD2
(residues 101–165) is inserted between helices H6 and
H11 of SD1 and is composed of four helices (H7, H8,
H9, and H10) and four b-strands (S1, S2, S3, and S4)
(Fig. 2A,B). Overall, SD2 forms a twisted, four-stran-
ded antiparallel b-sheet that is encompassed by the
four short helices.
Conformational change and molecular disorder
of PriL
NTD
To compare the three noncrystallographic symmetry
PriL
NTD
structures, rmsd values were calculated for
each. The rmsd values for the overall structures were
1.9 A
˚(molB to molA for 214 C
a
atoms), 1.3 A
˚(molC
to molA for 200 C
a
atoms), and 0.9 A
˚(molC to molB
for 197 C
a
atoms), respectively. In contrast, the rmsd
values obtained by comparing isolated SD1 structures
were 0.9 A
˚(molB to molA for 149 C
a
atoms), 0.7 A
˚
(molC to molA for 152 C
a
atoms), and 0.7 A
˚(molC to
molB for 149 C
a
atoms), suggesting that each SD1 is
more ordered than the overall structure. The merging
of each SD1 helix showed that the relative position of
each SD2 is distinct (Fig. 3A). In molA and molB, the
loop S3-S4 was rotated toward the loop S1-H8 by 15
degrees and, for example, both Caatoms of Glu117
and Glu159 are shifted by 8A
˚(Fig. 3A,B). In molC,
Fig. 1. PriL
NTD
physically interacts with PriS. Pull-down analysis of
interactions of PriS with full-length PriL and PriL
NTD
and PriL
CTD
fragments are shown. GST-tagged proteins were bound to glutathi-
one epharose and incubated without (–) or with (+) PriS. Lanes 1
and 12, molecular weight markers; lanes 2 and 11, PriS input
(2.5 lg); lanes 3 and 4, GST without (–) or with (+) PriS used as
negative controls; lanes 5–10, GST fused to full-length PriL and
with PriL
NTD
and PriL
CTD
proteins as indicated above each lane. Pro-
teins bound to the beads were resolved by SDS ⁄PAGE and stained
with Coomassie blue.
N. Ito et al. PriL
NTD
structure and PriS-binding site
FEBS Journal 274 (2007) 1340–1351 ª2007 The Authors Journal compilation ª2007 FEBS 1341
the helix H8 region (residues 117–133) following the
loop S1-H8 seems to be in an intermediate position rel-
ative to molA and molB; however, this region is miss-
ing due to its higher temperature factors relative to
neighboring residues, including strands S3 and S4. This
observation led to the idea that the helix H8 region of
PriL
NTD
is flexible in the absence of PriS and stabil-
ized in its presence suggestive of a disorder-to-order
mechanism.
The alternative conformation of SD2 in molA and
molB seems to be coupled with the conversion of
intramolecular hydrogen bonds. In molA, four hydro-
gen bonds can be observed {Glu150-Tyr155, Arg171-
Asp168, Arg161-Asp111, and Asp142-Asp111n [where
n is the nitrogen atom of the protein main-chain
(mc)]} (Fig. 3C). Three interactions can be identified in
molB [Glu150-Tyr155, Arg166-Lys100o, and Arg166-
Lys101o (where o is the oxygen atom of mc)]
(Fig. 3D). The Glu150–Tyr155 interaction is conserved
and seems to maintain the overall structure, whereas
other residues can exchange hydrogen bond partners
in concert with structural changes (Fig. 3C,D).
Table 1. Crystallographic data and refinement statistics.
X-ray source
SeMet Native
Peak
SPring-8 ⁄BL41XU
Remote2 SPring-8 ⁄BL44B2
Edge Remote1
Diffraction data and phasing statistics
Wavelength (A
˚) 0.9793 0.9795 0.9840 0.9717 1.000
Resolution (A
˚) 50–3.1 50–2.9
Unique reflections 35154 35169 35378 35934 23431
Redundancy 14.3 14.3 14.3 14.3 7.2
Completeness (%)
a
99.0 (99.2) 99.0 (99.2) 99.0 (99.2) 99.0 (99.2) 99.2 (100)
I⁄r(I)
a
14.0 (6.8) 13.6 (6.1) 12.0 (4.1) 12.7 (5.0) 17.0 (7.3)
R
merge
(%)
a,b
9.0 (36.1) 7.5 (39.9) 7.9 (57.6) 8.3 (48.4) 4.4 (21.8)
PhP
cCen ⁄Acen
0.23 ⁄0.14 – ⁄– 0.75 ⁄0.46 0.95 ⁄0.57
FOM
dCen ⁄Acen
0.44 ⁄0.55
Refinement statistics
Resolution (A
˚) 20.0–2.9
R
work
⁄R
free
(%)
e
24.8 ⁄30.9
Protein atoms 5271
Water atoms 18
Overall B-factor (A
˚
2
) 55.24
rmsd
Bond length (A
˚) 0.008
Bond angles () 1.363
a
Numbers in parentheses correspond to the values in the highest resolution shell.
b
R
merge
¼S
h
S
i
|I
hi
-<I
h
>| ⁄S
h
S
i
|I
hi
|, where hare unique
reflection indices, and iindicates symmetry equivalent indices.
c
PhP, phasing power is f
rms
⁄E
rms
, where f
rms
¼[(S
fh2
)⁄n]
1⁄2
and E
rms
¼
[S(F
PH
-| F
P
+F
H
|)
2
⁄n]
1⁄2
.
d
FOM, figure of merit.
e
R
work
and R
free
¼S|F
o
)F
c
|⁄S F
o
for all reflections or calculated with 7% of data excluded
from refinement.
AB
Fig. 2. Structure of PhoPriL
NTD
. (A) Ribbon
representation of PriL
NTD
. Subdomains 1
(SD1) and 2 (SD2) are labeled and colored in
blue and yellow, respectively. Primary sec-
ondary structures are labeled. The figure
was generated with PYMOL (http://www.
pymol.org). (B) Topology of the secondary
structure of the PriL
NTD
molecule with the
corresponding residue numbers. Tubes and
arrows represent helices and strands,
respectively.
PriL
NTD
structure and PriS-binding site N. Ito et al.
1342 FEBS Journal 274 (2007) 1340–1351 ª2007 The Authors Journal compilation ª2007 FEBS
PriS binding is mediated by SD2 in PriL
NTD
We first explored the PriS-binding region by dividing
the PriL
NTD
structure into three regions: the N- and
C-terminal regions of SD1 (SD1N, residues 1–100;
SD1C, residues 166–222) and SD2 (residues 101–165).
N-Terminal GST fusions of these peptides were subjec-
ted to pull-down analysis. As shown in Fig. 4A, SD1N
and SD1C could not bind to PriS, although further
experimental studies are required to determine whether
both GST-fusion proteins are folded correctly. On the
contrary, a specific interaction between PriS and GST-
SD2 was clearly suggested (Fig. 4A, lane 6).
We tested whether the flexible helix H8 is directly
involved in PriS binding. We prepared GST fusion
proteins with residues 116–133 (GST-PriL
116)133
), and
the rest of the PriL
NTD
region, GST-PriL
NTDD116)133
,
to mimic structures containing only helix H8 (Fig. 4B,
lane 3) or structures lacking only helix H8 (Fig. 4C,
lane 2), respectively. Pull-down analysis showed that
GST-PriL
NTDD116)133
, but not GST-PriL
116)133
, also
associates with PriS, suggesting that PriL
NTDD116)133
was sufficient for the interaction with PriS (Fig. 4B,
lane 4, and Fig. 4C, lane 3). This result indicates that
PriS binds to SD2, but the helix H8 region is not
essential. All data indicates that the binding site is else-
where on the surface of SD2 except helix H8.
Cooperative function of the Tyr155-Tyr156-Ile157
residues to PriS interaction
There is prior evidence that hydrophobic interactions
are important for interaction of PriL with PriS [17].
To further define the PriS-binding surface, we focused
on hydrophobic residues on the surface of SD2. We
examined whether replacement with alanine (Ala) of
the hydrophobic residues located on SD2 reduced the
ability of PriL to associate with PriS (Fig. 5A).
Phe118, Tyr130, Phe143, Tyr155, Tyr156, Ile157,
Tyr158, Val162, Tyr163, and Leu164 were selected as
mutation sites. Interestingly, among 12 mutants ana-
lyzed (Fig. 5B), only the Y155A ⁄Y156A ⁄I157A triple
mutant showed a severe PriS binding defect (Fig. 5B,
lane 7). It is noteworthy that the other single or mul-
tiple Ala mutants, except for this mutant, as well as
the wild-type (WT) protein, all seemed to bind tightly
Fig. 3. Molecular elasticity and flexibility.
(A) Superimposed acarbon representation
of SD1 clarifies the conformational change
in SD2. Noncrystallographic symmetry mole-
cules of PriL
NTD
are colored in blue (molA),
green (molB), and pink (molC). Glu117 and
Glu159 show the most pronounced struc-
tural deviation, as indicated by the arrow.
(B) Close-up view of SD2. Pink dots repre-
sent the molecular disorder of molC. (C) and
(D) Intramolecular hydrogen bonds (orange
dots) observed in molA and molB.
N. Ito et al. PriL
NTD
structure and PriS-binding site
FEBS Journal 274 (2007) 1340–1351 ª2007 The Authors Journal compilation ª2007 FEBS 1343
to PriS. The result suggests that cooperative function
of the Tyr155-Tyr156-Ile157 residues to PriS facilitates
subunit interaction.
The CD profiles of the WT and the Y155A ⁄Y156A ⁄
I157A mutant of PriL
NTD
were measured to confirm
that no significant changes of the molecular structure
had occurred by site-directed mutagenesis. As shown
in Fig. 6, the CD spectrum of the mutant was the
same as that of the WT.
The Tyr155-Tyr156-Ile157 motif of sheet S3 is a
key determinant for the dissociation constant of
the PriL
NTD
–PriS complex
To investigate the precise role of the putative binding
surface, the interaction between PriS and GST-
PriL
NTD
was analyzed by the surface plasmon reson-
ance method. PriS was bound to the CM5 chip, and
the interaction was measured with 500 nmWT
or mutant (Y155A ⁄Y156A ⁄I157A, V162A ⁄Y163A ⁄
L164A, Y130A ⁄F143A, and F118A ⁄Y158A) PriL
NTD
proteins. Association and dissociation rates were
measured for 3 min and for a prolonged period,
30 min, respectively, because WT PriL
NTD
dissociated
slowly from PriS (Fig. 7A). Sensorgrams of the
V162A ⁄Y163A ⁄L164A, Y130A ⁄F143A, and F118A ⁄
Y158A mutants showed little difference compared with
that of the WT protein (Fig. 7A). The WT dissociation
rate was very low and remained constant for 30 min.
However, the sensorgram of the Y155A ⁄Y156A ⁄I157A
mutant was clearly distinguishable from that of the
WT protein, and it showed an enhanced rate of
dissociation of the Y155A ⁄Y156A ⁄I157A mutant
from PriS. The dissociation constant (K
D
) of the
Y155A ⁄Y156A ⁄I157A mutant was 2.4 ·10
)7
, which
was 1.3 ·10
3
-fold higher than that of the WT protein,
indicating a marked decrease of the affinity of the
Y155A ⁄Y156A ⁄I157A mutant for PriS (Fig. 7B).
We further investigated the function of the sequence
Tyr155-Tyr156-Ile157 by constructing single and dou-
ble alanine mutants. The sensorgram of the double Ala
mutant Y156A ⁄I157A was different from that of the
WT protein, although the mutant seemed to dissociate
more slowly than Y155A ⁄Y156A ⁄I157A. In contrast,
the sensorgrams of Y155A ⁄Y156A and of the single
Ala mutants (Y155A, Y156A, and I157A) were similar
to that of the WT protein. The kinetic parameters
are summarized in Table 2. The K
D
value of
Y156A ⁄I157A was 14-fold higher than that of the WT
protein. The kinetic parameters of Y155A, Y156A,
A
BC
Fig. 4. Structure-based molecular dissection of PriL
NTD
. Truncated PriL
NTD
proteins were tested for interaction with PriS. (A) Binding assay
with GST-fused SD1N, SD2, and SD1C fragments (residues 1–100, 101–165, and 166–222, respectively) without (–) or with (+) PriS (lanes
3–8). The truncated proteins were sensitive to proteolysis, resulting in putative PriL
NTD
breakdown products, as indicated by asterisks.
SD1N, SD2, and SD1C and the residues 116–133 of the PriL
NTD
fragment are colored in navy, green, cyan and orange, respectively. (B) Pull-
down analysis of PriS with GST-PriL
116)133
(lanes 3 and 4). (C) Binding assay of PriS with an internally deleted PriL
NTD
protein, GST-
PriL
NTDD116)133
(identical to PriL
1)115 ⁄134)222
) (lanes 2 and 3). Gels were visualized by silver (A and C) or Coomassie (B) staining. Lanes 1
and 11 in (A), 1 in (B), and (C), molecular weight markers; lane 2, 10 in (A), 2 in (B), and 5 in (C), PriS input (2.5% of total); lane 9 in (A), 2 in
(B), and 4 in (C), GST without PriL
NTD
used as a negative control.
PriL
NTD
structure and PriS-binding site N. Ito et al.
1344 FEBS Journal 274 (2007) 1340–1351 ª2007 The Authors Journal compilation ª2007 FEBS