An Iron-Sulfur Cluster in the C-terminal Domain of the p58 Subunit of Human DNA Primase*

DNA primase synthesizes short RNA primers that are required to initiate DNA synthesis on the parental template strands during DNA replication. Eukaryotic primase contains two subunits, p48 and p58, and is normally tightly associated with DNA polymerase α. Despite the fundamental importance of primase in DNA replication, structural data on eukaryotic DNA primase are lacking. The p48/p58 dimer was subjected to limited proteolysis, which produced two stable structural domains: one containing the bulk of p48 and the other corresponding to the C-terminal fragment of p58. These domains were identified by mass spectrometry and N-terminal sequencing. The C-terminal p58 domain (p58C) was expressed, purified, and characterized. CD and NMR spectroscopy experiments demonstrated that p58C forms a well folded structure. The protein has a distinctive brownish color, and evidence from inductively coupled plasma mass spectrometry, UV-visible spectrophotometry, and EPR spectroscopy revealed characteristics consistent with the presence of a [4Fe-4S] high potential iron protein cluster. Four putative cysteine ligands were identified using a multiple sequence alignment, and substitution of just one was sufficient to cause loss of the iron-sulfur cluster and a reduction in primase enzymatic activity relative to the wild-type protein. The discovery of an iron-sulfur cluster in DNA primase that contributes to enzymatic activity provides the first suggestion that the DNA replication machinery may have redox-sensitive activities. Our results offer new horizons in which to investigate the function of high potential [4Fe-4S] clusters in DNA-processing machinery.

DNA polymerase ␣-primase (pol-prim) 2 associates with eukaryotic replication forks in S-phase during the initiation of DNA replication (1,2). pol-prim synthesizes a chimeric RNA-DNA primer of ϳ30 nucleotides that is then extended by more processive DNA polymerases that synthesize the leading and lagging strands. pol-prim is composed of four subunits (p180, p68, p58, and p48). The p180 subunit has the DNA polymerase catalytic activity and binds to both the p68 and p58 subunits. The p68 subunit has a regulatory function that is not completely understood. It is required for initiation of yeast chromosomal replication (3,4) and cell-free SV40 DNA replication (5). In addition, phosphorylation of p68 alters the activity of polprim in SV40 replication (6 -9).
The two smallest subunits, p48 and p58, together function as the DNA primase by creating an RNA primer of 7-10 nucleotides (10,11). The p48 subunit contains the catalytic site (12). The p58 subunit stabilizes p48 and participates in initiation, elongation, and "counting" the ribonucleotides polymerized (13). Interestingly, p58 is also involved in transferring the RNA strand directly into the active site of the associated p180 subunit, which extends the growing nucleotide with dNTPs to complete the formation of the RNA-DNA primer (1,14,15). Knowledge of the molecular basis for regulation of the length of RNA portion of the primer and internal transfer to the p180 subunit is very limited.
Despite the fundamental importance of primase in DNA replication, the only structural information available for a heterodimeric primase is for an archaeal (Sulfolobus solfataricus) primase that does not form a pol-prim complex (16,17). Multiple sequence alignments reveal homology between the p48 subunits and the N-terminal half of the p58 subunit. In the crystal structure of the S. solfataricus primase core, the p48 subunit assembles with the N-terminal half of p58 (16). However, in human p58, both the N-and C-terminal regions have contacts with p48 (18). Interestingly, the C-terminal half of p58 also contains a region with homology to a DNA polymerase ␤ domain; this region was determined to be important for primer synthesis (13), but how it functions is not known.
DNA primase serves as a key target for regulation of DNA replication initiation, telomere maintenance, and response to DNA damage or fork stalling, in part through its physical interactions with other proteins involved in DNA replication and in checkpoint signaling (19). Primase interacts physically with the viral helicase SV40 large T antigen, eukaryotic replication protein A (20,21), and GINS, a recently identified component that plays a central role in establishment and progression of eukaryotic and archaeal replication forks (22)(23)(24). Primase activity is essential for optimal checkpoint signaling at stalled replication forks (25)(26)(27)(28) and possibly in rescuing stalled replication fork progression (29), but its interaction partners are not known.
To better understand the role of human primase in these pathways, it will be vital to elucidate its structure and interactions with partner proteins. This strategy has been useful in determining the roles of the SV40 large T antigen-replication protein A interaction in the context of SV40 DNA replication (30,31). To facilitate similar experiments with human DNA primase, we sought to characterize the domain architecture of DNA primase. The p58 and p48 subunits can be expressed and purified independently of the other two subunits and retain primase activity in vitro at levels similar to those observed for the intact heterotetramer (12,18). Working from bacterially expressed primase protein, a structured domain in the C terminus of the p58 subunit (p58C) was identified. Biophysical analysis of this construct showed that the domain is folded and has the characteristics of a [4Fe-4S] high potential iron protein (HiPIP). The conservation of four cysteines across several species suggests a critical role for the cluster, and this was confirmed by in vitro experiments that demonstrate that the [4Fe-4S] cluster is required for primase activity.

EXPERIMENTAL PROCEDURES
Primase Construct Design-The recombinant human p48/ p58 primase expression plasmid used in this study has been previously described (18). Full-length p48 cDNA was amplified using the dimer plasmid as the template. Using BamHI and EcoRI restriction enzymes, it was then inserted into the inhouse pBG100 vector (L. Mizoue, Center for Structural Biology, Vanderbilt University), which contains an N-terminal His 6 tag. A p58-(266 -509) construct (p58C) was subcloned from the dimer plasmid into pET15b (Novagen) using NdeI and XhoI. This construct also contains an N-terminal His 6 tag. Mutations were generated in the p58C expression plasmid by site-directed mutagenesis (QuikChange, Stratagene), and verified by DNA sequencing. An SphI-BglII fragment of p58C containing the mutation was then used to replace the corresponding wild-type fragment in the dimer p48/p58 expression plasmid.
Protein Expression and Purification-Each construct was expressed in BL21(DE3) cells. Cells were grown at 37°C in LB to an A 600 of ϳ0.6. The temperature was then lowered to 22°C, and the cells were allowed to equilibrate for 30 min. Expression was induced using 1 mM isopropyl thio-␤-D-galactopyranoside. Cells were harvested by centrifugation 4 h postinduction. Pelleted cells were resuspended in lysis buffer containing 50 mM Tris-HCl (pH 8), 300 mM NaCl, 20 mM imidazole, 3 mM 2-mercaptoethanol (BME), 1% Nonidet P-40, 0.5 mg/ml lysozyme, ϳ10 mg of DNase I, and one Complete Mini EDTA-Free protease inhibitor mixture tablet (Roche Applied Science). Cells were lysed by sonication at 4°C. Insoluble material was removed by centrifugation.
The primase polypeptides were purified using nickel-nitrilotriacetic acid affinity chromatography. The bound proteins were eluted using a linear imidazole gradient ranging from 20 mM to 250 mM. Fractions containing the primase polypeptides were pooled and dialyzed overnight at 4°C into buffer containing 30 mM MES (pH 6.5), 50 mM NaCl, and 3 mM BME. The sample was then further purified using a Mono S column (Amersham Biosciences) equilibrated in the same buffer and eluted with a linear gradient to 1 M NaCl.
Limited Proteolysis-p48/p58 and p48 primase preparations were exchanged into buffer containing 20 mM Tris-HCl (pH 7.5), 50 mM NaCl, and 3 mM BME. Digestions using proteinase K were performed at room temperature using a 1:1000 ratio of protease:protein. At several time points over the 24-h course of the digestion, an aliquot was removed, mixed with SDS sample loading buffer, and boiled for 10 min. The aliquots were then stored on ice until the end of the experiment, when they were analyzed by SDS-PAGE. Bands of interest were excised and submitted for in-gel trypsin digestion and mass spectrometry analysis, or they were transferred to polyvinylidene fluoride membrane and submitted for Edman degradation.
CD-p58C was buffer exchanged into 20 mM sodium phosphate (pH 6.5), 50 mM NaCl, and 3 mM BME. The protein concentration was ϳ10 M. The far-UV CD spectrum was scanned at room temperature from 190 to 260 nm using a Jasco J-180 spectrophotometer (Easton, MD). Secondary structure content was estimated using the K2d web server (32).
NMR-To obtain 15 N-enriched p58C, the protein was expressed in M9 minimal media with 15 NH 4 Cl as the nitrogen source. The protein was buffer-exchanged into 20 mM sodium phosphate (pH 6.5), 50 mM NaCl, 3 mM BME, and 5% D 2 O. 15 N-1 H heteronuclear single quantum coherence (HSQC) spectra were acquired at 25°C on a Bruker Avance 600-MHz spectrometer equipped with a cryoprobe. A total of 64 scans was recorded in the direct ( 1 H) dimension for each of the 128 points sampled in the indirect ( 15 N) dimension. The data were processed using NMRPipe (33) and analyzed and displayed using NMRView (34).
UV-visible Spectrophotometry-Spectra were recorded using a Varian Cary 100 Bio spectrophotometer. Samples were scanned from 550 to 250 nm at room temperature in buffer containing 20 mM Tris (pH 7.5), 50 mM NaCl, and 3 mM BME.
EPR-Samples were prepared at 0.5 mM p58C in 20 mM MES (pH 6.5) and 50 mM NaCl, then freshly oxidized with 2.5 mM potassium ferricyanide. Spectra were recorded at X-band (ϳ9.05 GHz) on a Varian E-122 spectrometer. The data were acquired from frozen glasses at ϳ15 K using an Air Products Helitran cryostat with liquid helium. The magnetic fields were calibrated with a Varian NMR Gauss meter, and the microwave frequency was measured with an EIP frequency meter.

RESULTS
Primase Domain Architecture-Proteins involved in DNA replication are often modular, containing several independent domains tethered together (36). In many cases, it is possible to map the location of these structural domains by subjecting the intact protein to very limited proteolytic digestion, then using mass spectrometry and N-terminal sequencing to identify the fragments. To test whether DNA primase contained any structured domains, the p48/p58 dimer was subjected to limited proteolysis with proteinase K. Digestion products were monitored as a function of reaction time by using denaturing gel electrophoresis. As shown in Fig. 1A, two stable fragments were produced corresponding to molecular masses of ϳ42 kDa (band a) and ϳ28 kDa (band b). Bands were excised from the gel and characterized by matrix-assisted laser desorption ionization time-of-flight mass spectrometry, liquid chromatographymass spectrometry, and N-terminal sequencing to determine the identity of the two fragments. The 42-kDa band was found to be the result of a C-terminal truncation of roughly 60 residues from p48. Proteinase K digestion of isolated p48 produced the same fragment (Fig. 1B), which confirmed that the 42-kDa fragment is a stable domain of p48. The 28-kDa fragment corresponds to the C-terminal half of p58, residues Gly-266 to Ser-509.
Production of Primase Domain Expression Constructs-To evaluate the potential primase domains further, they were subcloned into bacterial expression vectors. Two p48 constructs (Met-1-Ser-361 and Met-1-Glu-366) were subcloned into bacterial expression vectors but did not produce soluble protein, suggesting the C terminus is an integral part of the protein.
Apparently, p48 contains a solvent-exposed loop in the vicinity of the proteinase K cleavage site. Although a stable fragment was not detected in the experiments for the N-terminal half of p58, sequence analysis suggested it should form a structured domain: sequence alignment reveals homology between the N-terminal half of p58 and the region of the large subunit from S. solfataricus that has previously been crystallized (16). In an effort to produce and characterize a p58N construct, a wide range of sub-cloning experiments was performed, but none yielded soluble protein. A similar observation was made for the full-length p58 subunit. In contrast to p58N, p58C has no significant homology to proteins whose structure is known. However, an extensive amount of secondary structure was predicted for this stable fragment. In fact, the p58C domain expressed well and was very soluble, even to very high concentrations (ϳ30 mg/ml).
A combination of CD and NMR spectroscopy was used to characterize the structural integrity of p58C. Minima observed at 208 and 222 nm in the far-UV CD spectrum ( Fig. 2A) indicate a significant amount of ␣-helical content in p58C. Analysis of the CD spectrum using the K2d program provided an estimate of 31% helix, 14% ␤-sheet, and 55% coil. The 600-MHz 1 H-15 N HSQC NMR spectrum of p58C is shown in Fig. 2B. The spectrum contains relatively narrow line widths and very good dispersion, which are indicative of a well folded structural domain.
An Iron-Sulfur Cluster in p58C-Interestingly, when either the primase dimer or p58C was purified, the protein solution had a golden-brown color. The color intensified as the protein was concentrated, becoming very dark at high concentrations.  A UV-visible spectrum of p58C contained a broad peak at 400 nm, similar to spectra from proteins containing iron-sulfur clusters (Fig. 3A) (e.g. Ref. 37). The primase dimer also had this property, whereas isolated p48 did not (data not shown).
To further investigate the presence of an iron-sulfur cluster in p58C, the protein was analyzed by inductively coupled plasma mass spectrometry, which enables the content of metal ions to be quantified. This analysis provided an estimate of 3.09 mol of iron per mole of protein, which is consistent with the presence of 4 iron atoms in the cluster. A stoichiometry considerably lower than 4 is typical for proteins containing [4Fe-4S] clusters (38,39) and is attributable to the loss of one iron in the cluster during protein purification and analysis.
EPR spectroscopy is now a well accepted means to confirm the presence of iron-sulfur clusters in proteins and assign them to specific cluster types and oxidation states. The 9.05-GHz X-band EPR spectra of p58C acquired directly on the purified protein lacked any appreciable signal. Reduction of p58C with sodium dithionite produced only a weak, broad signal. However, oxidation of the protein resulted in a strong signal characteristic of a low spin S ϭ 1 ⁄ 2 iron-sulfur cluster (Fig. 3B). The signal is highly rhombic with g 1 ϭ 2.087, g 2 ϭ 2.040, and g 3 ϭ 2.013. Simulation of the EPR spectrum with SIMPOW6 software (40) revealed two slightly different forms were present in solution, one with g 1 ϭ 2.0870, g 2 ϭ 2.0405, and g 3 ϭ 2.0126 and a second weaker form with g 1 ϭ 2.0872, g 2 ϭ 2.0311, and g 3 ϭ 2.0094 and broader line widths. Similar results reported for oxidized HiPIPs have been interpreted as resulting from the presence of multiple isomeric states (e.g. Ref. 41). The average EPR g-factor of 2.046 observed for p58C is much more typical of a [4Fe-4S] 3ϩ cluster than a [3Fe-4S] ϩ cluster, because the latter normally exhibit an average g-factor of 2.015 and relatively smaller g anisotropy (39,42,43).
The EPR spectrum has many properties that suggest the [4Fe-4S] cluster is HiPIP-like. Like typical HiPIP clusters, the iron-sulfur cluster in p58C is only visible by EPR when oxidized, has an average g-value greater than two, is observed best below 30 K, and the signal does not readily saturate. Although the observed g-values deviate somewhat from the prototypical HiPIP cluster, such differences may be attributable to changes in the environment around the [4Fe-4S] cluster or distortion of the cluster by the protein environment (44,45). In summary, the biophysical data on p58C are consistent with assignment to the class of high potential iron proteins, thus adding to the growing list of DNA-processing proteins with this unique cofactor.
Identification of Iron-Sulfur Ligands in p58C-Iron-sulfur clusters are typically bound to proteins via four cysteine residues, and p58C contains six cysteine residues. Following the strategy used for other DNA-processing proteins that contain an iron-sulfur cluster, potential cysteine ligands were identified from a multiple sequence alignment of p58 from five different species using ClustalX (35). The alignment in Fig. 4A reveals that four of the cysteine residues (Cys-287, Cys-367, Cys-384, and Cys-424) are conserved. These are residues most likely responsible for cluster binding.
To test this hypothesis, each of the four cysteine residues in p58C was individually mutated to serine. Although the mutant constructs expressed well, they were poorly soluble and degraded rapidly (data not shown). Consequently, an alternate strategy was used involving mutation in the context of the p48/ p58 dimer. Data are shown here for the C367S mutant primase dimer, which expressed at levels comparable to the wild-type primase, remained soluble, and co-purified with the p48 subunit (Fig. 4B). Analysis of this mutant primase dimer by UVvisible spectroscopy provided a spectrum in which the broad peak at 400 nm was clearly absent (Fig. 4C), confirming this mutant primase does not contain an iron-sulfur cluster.
A Role for the Iron-Sulfur Cluster in Primase Function-To initially assess the functional relevance of the iron-sulfur cluster in p58, the primase activity of wild-type and p48/C367S-p58 primase dimers on a natural ssDNA template was assayed as a function of protein concentration. Radiolabeled CTP was incorporated into RNA primers of 8 -10 nucleotides by the wild-type primase (Fig. 5A, lanes 1-3), and small amounts of larger products were detectable, as observed previously (14,18). No RNA primers were observed in the absence of enzyme (Fig.  5A, lane 7). Products of the mutant primase that lacks the ironsulfur cluster were barely detectable above background (Fig.  5A, lanes 4 -6), and the level of reaction product was not proportional to the amount of mutant primase in the reaction. Quantification of the products as a function of primase concentration revealed that the specific activity of the mutant primase was reduced at least 5-fold (Fig. 5B). These data suggest that the iron-sulfur cluster in p58 is crucial for the primase activity of p48/p58 in this assay.

DISCUSSION
Our studies show that DNA primase contains a structured domain in the C terminus of the p58 subunit and that this domain contains an iron-sulfur cluster. This is the first report of an iron-sulfur cluster in a DNA replication protein. Analysis of p58C by ICP-MS, UV-visible, EPR, and phylogenetic amino acid sequence comparisons are consistent with the presence of a HiPIP [4Fe-4S] cluster coordinated by four conserved cysteine residues.
Iron-sulfur clusters have been reported in DNA repair proteins, including DNA glycosylases MutY (46), endonuclease III (47), and family 4 uracil-DNA glycosylases (38). Recently, several members of a family of DNA helicases involved in DNA repair, including the nucleotide excision repair helicase xeroderma pigmentosum complementation group D (XPD) and the Fanconi J cross-link repair helicase, have also been shown to contain an iron-sulfur cluster that is essential for helicase activity but not for ssDNA-dependent ATPase activity (37).
Remarkably, each of these proteins contains a [4Fe-4S] cluster coordinated by four conserved cysteines like DNA primase.
Iron sulfur clusters in proteins are traditionally associated with electron transport and redox chemistry (48). The presence of clusters in proteins involved in several aspects of DNA metabolism points to a different and possibly common function. However, a specific role for iron-sulfur clusters in DNA repair proteins has not yet been determined. Evidence has accumulated showing that the clusters are required for enzymatic activity, e.g. for MutY (49) and XPD (37). In addition, x-ray crystal structures of MutY (50) and endonuclease III (51) have been determined. These structures show that the iron-sulfur clusters are too far from the active site to participate directly in catalysis. Thus, the influence of the cluster on enzymatic activity appears to arise from an allosteric effect.
In support of this proposal, studies conducted on MutY revealed that the cluster was critical for orienting key residues that contact the distorted DNA (49,50,52). Also, modeling studies on the uracil-DNA glycosylases point to a role in substrate recognition rather than catalysis (38). The lack of structural data on XPD makes determining the role of the iron-sulfur cluster considerably more difficult. Hence, the bulk of the data available to date point to the iron-sulfur clusters having an influence on structural features of these proteins as opposed to participating directly in aspects of protein chemistry (37). Because iron-sulfur clusters are invariably integrated into protein structure, it makes sense that they modulate the structure and stability. However, it should be noted that it is not readily possible to manipulate the redox state of HiPIP proteins in vivo or in cells. Moreover, it is difficult to design experiments that directly address functional questions that do not result in complete loss of the cluster. Thus, although the importance of ironsulfur clusters in the structure DNA processing proteins is evident, additional redox-mediated functional roles cannot yet be ruled out.
If the role of the iron-sulfur cluster is purely structural, why was such a complex cofactor chosen to serve for this purpose? Iron sulfur clusters are inserted into proteins via a multistep process involving several proteins (53), which seems like an excessive utilization of cellular resources for a purpose that could be attained at a lower energetic cost. Recently, studies on MutY have led to another theory of iron-sulfur cluster function in DNA glycosylases. Initially, the cluster was not thought to be involved in redox chemistry, as it is not redox active in vitro (54). However, a recent report that binding of DNA shifts the redox potential of the cluster (55). This observation led to a complex model of glycosylase function in which electron transfer between iron-sulfur clusters on separate glycosylases occurs through the DNA; the cluster in the remote glycosylase is reduced, and the enzyme dissociates from the DNA. Aberrant DNA would not have this property, and the glycosylase would remain oxidized and bound to DNA, promoting DNA repair. Although this model has a number of appealing features, it remains controversial.
Although we have shown that the iron-sulfur cluster in DNA primase is required for enzymatic activity, the specific function of the cluster remains unclear. The decreased solubility and stability of the cysteine-to-serine mutants in p58C suggest that the cluster does provide some level of structural stability to the protein. However, full-length p58 is still sufficiently structured in the absence of the iron-sulfur cluster to bind p48. Interestingly, a budding yeast mutant pri2-2 that encodes Tyr instead of one of the conserved Cys ligands (C434Y, Fig. 4A) in p58 displays a temperature-sensitive, slow growth phenotype (56,57). This substitution would be expected to result in loss of the iron-sulfur cluster from p58. The observation that the mutation is not lethal suggests that, in the context of the pol-prim complex in vivo, the cluster may serve primarily a regulatory as opposed to a purely structural function in p58.
The possibility that the DNA primase iron-sulfur cluster becomes redox active when bound to DNA cannot be ruled out. The idea that this might account for some of the unique properties of p58, such as its ability to regulate the length of unitlength primers, is intriguing. The above noted 5Ј-3Ј repair helicases as well as both subunits of primase (14) bind ssDNA. Moreover, a 5Ј-overhanging, 3Ј-recessed primer-template junction is known to be especially important for processing DNA by these proteins. The helicases have ssDNA-dependent ATPase motor domains that translocate the protein 5Ј to 3Ј along ssDNA until it encounters the 3Ј-end of the complementary strand in duplex DNA. It is known that, without the ironsulfur cluster, the helicase cannot unwind the DNA, perhaps because its interaction with the junction is weak or defective. Similarly, pol-prim bound to ssDNA template polymerizes NTPs into an oligoribonucleotide-template that remains bound to p58 (14). These authors proposed that the completed RNA primer-template likely remained bound to p58C, based on the sequence homology to the 8-kDa domain of DNA pol-␤ that is known to enhance pol-␤ processivity.
Negative regulation of primer polymerization beyond unit length requires a stable primer-template (14,58), implying that the unit-length primer-template forms a stable complex with p58C prior to internal transfer of the primer to the polymerase active site in intact pol-prim. Thus, there is a pause in the reaction until the transfer occurs, relieving the negative regulation. In the absence of the polymerase subunit, primase activity resumes only after the primase dissociates from the primer- template and rebinds to ssDNA. Taken together, these studies strongly suggest that p58C interacts with primer-template during and after primer synthesis. If the iron-sulfur cluster in p58C is important for primer-template binding, the C367S primase might dissociate too frequently to enable efficient polymerization to create the full-length primer. This interpretation could also partially explain the temperature-sensitive phenotype of the pri2-2 p58 yeast mutant, which would be expected to dissociate from the DNA more frequently at higher temperature and lose efficiency in generating the primer. In terms of a redox function for the iron-sulfur cluster in p58C, one could speculate that control of the redox state may enable primase to be retained on the growing primer-template either until the unitlength primer is completed and transferred to p180 for elongation, or primase dissociates from the primer-template.
The discovery of an iron-sulfur cluster in DNA primase offers new horizons in which to investigate the function of [4Fe-4S] clusters in DNA replication and repair machinery. Our findings imply that, in addition to modulating the structural stability, the cluster in eukaryotic DNA primases may well function in some form of regulatory role, perhaps in controlling the length of the primer strand. As is evidenced by the study of MutY and endonuclease III, high resolution structural analysis would aide significantly in investigating the role of the ironsulfur cluster in DNA primase function. To this end, further integrated structural and functional analyses are currently in progress in our laboratories.
Addendum-During the review of the manuscript, Klinge and colleagues (59) reported the presence of an iron-sulfur cluster in DNA primase from S. solfataricus and S. cerevisiae. Our findings are in agreement with their data and extend them by identifying the cluster in human DNA primase and by showing that the cluster is contained in a distinct and well folded structural domain within the p58 subunit.