Dps proteins are the structural relatives of bacterioferritins and ferritins ubiquitously present in the bacterial and archaeal kingdoms. The ball-shaped enzymes play important roles in the detoxification of ROS (reactive oxygen species), in iron scavenging to prevent Fenton reactions and in the mechanical protection of DNA. Detoxification of ROS and iron chaperoning represent the most archetypical functions of dodecameric Dps enzymes. Recent crystallographic studies of these dodecameric complexes have unravelled species-dependent mechanisms of iron uptake into the hollow spheres. Subsequent functions in iron oxidation at ferroxidase centres are highly conserved among bacteria. Final nucleation of iron as iron oxide nanoparticles has been demonstrated to originate at acidic residues located on the inner surface. Some Dps enzymes are also implicated in newly observed catalytic functions related to the formation of molecules playing roles in bacterium–host cell communication. Most recently, Dps complexes are attracting attention in semiconductor science as biomimetic tools for the technical production of the smallest metal-based quantum nanodots used in nanotechnological approaches, such as memory storage or solar cell development.
INTRODUCTION AND OVERVIEW: IRON FLOW IN GRAM-NEGATIVE BACTERIA
Iron in biology is essential as a cofactor for many enzymes, but simultaneously is a major threat to cell survival under oxidative conditions . In bacteria, the uptake of iron is therefore important, but strongly compromised through the low availability of iron under most physiological conditions [2–4]. Hence acquisition is a constant battlefield between micro-organisms and their specific environment [5–9]. After uptake of iron through FeSid (iron–siderophore) complexes, the free Fe2+ is released from these complexes into the cytoplasm (see Figure 1 for an overview) [10–15].
The iron-uptake pathway in Gram-negative bacteria is redundant
The iron distribution (concentrations and number of iron enzymes with iron cofactors) inside most bacterial cells is largely unknown since microbial metalloproteomes are still insufficiently characterized . To date, only a handful of bacterial proteomes have been characterized with respect to iron/metal distribution and only very basic parameters are accessible. As an example, in Pyrococcus furiosus ~97% of the identified metal ions are zinc or iron, and additional elements are present only in spurious amounts . In the archaeon Ferroplasma acidiphilum, iron is the terminal acceptor of the electron redox chain and 87 proteins contain iron; however, none of the representatives of the Dps superfamily has been identified among those . A limited analysis of the iron distribution in Bacillus anthracis cells revealed that most of the cellular iron was quantitatively stored in three unrelated enzymes: ferredoxin, Dps2 and two superoxide dismutases are the major representatives of iron-containing enzymes [19–21]. Using Mössbauer spectroscopy of entire Escherichia coli cells grown to exponential phase, most of the iron was traced in a hexagonally co-ordinated environment that excludes cytochromes, Bfrs (bacterioferritins) or iron–sulfur proteins. Furthermore, the total amount of iron in these cells was estimated to be 59–168 μg/g of cells (equivalent to a cytoplasmic concentration of 0.5–0.8 mM) [13,22]. Interestingly, also by Mössbauer spectroscopy combined with biochemical methods, an iron-enriched protein complex of ~155 kDa was observed. This complex, consisting of two protein species with an apparent molecular mass of 15 and 17 kDa respectively (judged from SDS/PAGE) each carrying 13 irons/subunit (~160 irons/dodecameric complex), was described to include most of the cytoplasmic E. coli iron. The molecular mass of this complex (from size-exclusion chromatography) and the individual components coincide well with E. coli Dps complexes, which, when isolated from whole cells, also show two different masses due to N-terminal processing .
Three subfamilies of proteins representing the ferritin fold are observed in bacteria: the ferritins, Bfrs and the ferritin-like Dps proteins [23–26]. These subgroups are clearly related by structure and sequence but show an evolutionarily divergent development in 12- and 24-mers. On the basis of their sequence, Bfrs are more closely related to the Dps proteins than ferritins (Figure 2A). Whereas bacterial ferritins and Bfrs are formed by 24 identical subunits expressing the archetypal four-helix bundles plus a short C-terminal helix, the Dps class of proteins is missing this terminal helix and assembles to a dodecameric state (Figure 2B). Consequently, the number of atoms stored in Dps cages is much smaller (~500 per dodecamer) when compared with ferritins or Bfrs (where 2000–4000 atoms are estimated) respectively [27–30]. Notably, the fold of this four-helix bundle has recently been used to trace relatives based on a pure structural classification scheme, and this analysis revealed a large number of metal-binding proteins from all kingdoms of life . In eukaryotes, ferritins are the only iron-storage proteins in the cytoplasm, whereas mitochondria import frataxin, a structurally unrelated protein forming a stoichiometrically variable protein complex cage (up to 48 subunits are reported) for iron storage by an as yet unknown mechanism [32–34].
Sequence relationship of Dps proteins to proteins of the ferritin superfamily deposited in the PDB and their structural architecture
During their normal life cycle, bacteria face a variety of potentially harmful circumstances such as nutrient limitation, heat-shock pH changes, osmotic pressure or oxidative stress (Figures 3A and 3B). A class of proteins that can functionally assist to cope with such threats was discovered and termed ‘Dps’ proteins according to their DNA protection during starvation properties . Once the structural analogy of these proteins to ferritins became obvious they were initially termed ferritin-like Dps proteins (Figure 2B). However, owing to the many differences in biological function, assembly, structure and active-site geometry, these proteins are nowadays termed Dps proteins. Proteins of this class maintain the mechanical stability of DNA during the stationary phase through the formation of protein–DNA complexes (Figures 3A and 3B), but also enhance cellular stability against radical damage through detoxification of ROS (reactive oxygen species) and rapid removal of free Fe2+ [25,27,35–37]. The protection of macromolecules in bacterial cells, in particular DNA, from irreversible damage caused by the formation of Fe2+-induced hydroxyl radicals OH• (also known as the Fenton reaction, see Figure 2A) together with a strongly expressed peroxidase activity are the best-studied functions of Dps proteins [38–40]. Moreover the assembly of Dps with DNA for mechanical protection of DNA has convincingly been shown for the protein from E. coli under in vivo and in vitro conditions . Furthermore, plasmid–DNA complexes in vitro have successfully being analysed by electron microscopy of highly ordered DNA/protein arrays (Figure 3B) [41–43]. More recently, an enzymatic function in the synthesis and degradation of amino-acyl glutamines was reported, but this function may occur only in a very small population of Dps proteins (Figure 2C) .
Dps proteins and their diverse functions in biology
A continuously growing number of Dps structures from all kingdoms of the bacterial life have been determined at high resolution (Figure 2A). Currently, ~30 Dps structures of different organisms have been characterized by X-ray crystallography since the first structure of Dps from E. coli was discovered in 1998 (Figures 2A and 2C) . Aside from variations in the monomeric structures at N- or C-termini, all structures contain the general ferritin-like fold of a four-helix bundle with only small structure variations [1,44–46]. Some of these structures were determined also in the presence of iron and related divalent metal ions such as zinc [1,40,47–49]. The first iron-enriched structure of a Dps enzyme was reported in 2000 with the Dps complex from Listeria innocua containing iron located at FOCs (ferroxidase centres) of the enzyme . Mechanistically very intriguing are iron translocation pathways into Dps complexes, redox processes at FOCs and the formation of initial iron-oxo nuclei. Previously published crystal structures provide insights into the variability of such mechanisms, and the geometry of FOCs and storage of iron oxide at nucleation centres [1,40,48,49].
Ferritin and Dps protein-based nanotools play a role as templates for the formation of defined metal oxide patterns for floating gateway nanodot structures [FNGM (floating nanodot gate memory)] with a strong potential for technological application, including memory storage on surfaces [50–52]. Currently, these structures are the smallest ‘biominerals’ fabricated by nature-derived cages with high reproducibility and strong plasticity to variations of the metal core that can assemble e.g. iron oxide, cobalt oxide, cadmium selenide and more related nuclei for specific technological approaches [50,52–54]. Related nanotechnological studies based on Dps enzymes reported the production of H2 using Dps from L. innocua after covalent attachment of a chromophore and Pt-cluster formation on their surface in the presence of triethanolamine and light .
In the present review, I present a comprehensive overview on principles of iron metabolism and the importance of Dps enzymes in Gram-negative bacteria leading to biomineralization. Recent crystal structures of these enzymes transiently trapping the uptake and storage routes are the experimental basis for the deduction of mechanisms.
IRON ACQUISITION BY GRAM-NEGATIVE BACTERIA
Iron uptake and metabolization in all bacterial species is a major prerequisite for their cell survival due to the importance of this metal in many essential (redox) enzymes. Consequently, the development of pathogenicity and virulence in bacterial strains such as Neisseria meningitidis, Staphylococcus aureus or B. anthracis is critically dependent on the presence of sufficient iron [9,20,56–58]. Regulation of iron uptake is orchestrated through signalling cascades by proteins encoded in operons such as the Fec (ferric citrate uptake) operon, which contains the uptake receptor FecA and a two-domain inner membrane sensory protein FecR that binds the cytoplasmically located iron-specific sigma factor FecI (Figure 1) [2,59,60]. Acquisition of iron is essential, but is compromised by the low bioavailability of Fe3+ under physiological conditions (dissociation constant 10−23) [9,11].
Sophisticated and redundant strategies for iron uptake have evolved in bacteria [9,61–64]. Under the pressure of environmental conditions, where access to iron is limiting (e.g. no unbound iron available), bacteria, and in particular many human pathogens, were reported to have additional sophisticated and redundant capture strategies. Human pathogens often target the extracellular human iron-transport proteins lactoferrin and transferrin, or snatch FeSid complexes from haemoglobin by the secretion of extraplasmic haemoglobin proteases [65,66]. The most successful and general strategy of bacteria is to acquire FeSid complexes. This pathway is initiated after synthesis and excretion of bacterial siderophore molecules and their further re-uptake as iron siderophores [9,67–70]. A unique diversity of siderophore complex chemistry developed in bacterial niches ensures iron acquisition from many sources [71–73]. Owing to a diversity of their uptake systems, bacteria can also cope with iron deficiency using siderophores synthesized by other microbes or they may even use siderophores, which were artificially added to the growth medium [40,74].
This impressive diversity of siderophore complex chemistry ensures iron acquisition from variable sources [71–73]. Uptake of these FeSids into the cytoplasm is a multistep process that is initiated by the binding of small organic iron complexes to receptor proteins (Figure 1). Large concentrations of siderophores in bacterial cultures can occur, with up to 200 mg/ml as determined, for example, for aerobactin (affinity constant to iron of 10−22) produced by E. coli [60,75]. These siderophores functionally resemble chelator molecules known from organic chemistry, such as citric acid with an affinity constant of 10−15 for the [FeCit2]5− complex [11,76]. In contrast, naturally optimized chemical structures allow for much higher binding affinities towards Fe3+ that are reported to achieve 10−52 for enterobactin, representing the strongest binding constants in biology . Thus siderophores can easily acquire iron even from solutions with insoluble iron-oxo minerals (solubility of Fe3+ from iron oxide in solution 10−23 M) and FeSid complexes are subsequently taken up into the bacterial periplasm by siderophore receptors of the outer membrane.
In E. coli, seven iron-uptake receptors in the outer membrane serve for FeSid binding, all of which have a specific affinity to chemically variant FeSid complexes (Figure 1) . These receptors are all 22-stranded outer membrane β-barrel proteins with a copy number of several hundred per cell . The C-terminal β-barrel domain encloses the N-terminal plug (or cork domain), which tightly locks the barrel with a diameter of 3.5 nm (Figure 1) [78,79]. These N-terminal domains need to functionally accomplish three major challenges: (i) binding of iron-loaded siderophores; (ii) the formation of a stable physical link to the TonB complex mediated by the N-terminal TonB box; and (iii) the reversible movement of the plug between the open and closed conformations [80,81]. The Ton protonmotive-force-driven complex of the inner membrane consisting of TonB, ExbB and ExbD is the energizing source in charge of the uptake of small-molecule complexes involved in iron and zinc binding and the binding of further small siderophore-like molecules [60,82–84] (Figure 1). Upon binding of the FeSid complex to the outwards-directed portion of the cork domain, FeSid attachment signals binding across the cork domain to render the N-terminus (TonB box) accessible to the periplasmic TonB protein. After the periplasmic TonB domain has bound to the receptor N-terminus, the driving force for the movement of the cork can be exerted to temporarily move the cork in complex with FeSid to finally release this substrate into the periplasm (Figure 1) [11,81]. FeSid complexes are further shuttled through the periplasmic space by FeSid-binding proteins that allow for the selective directional transfer of the FeSid components to inner membrane transporters (Figure 1) [85,86]. Further transport of FeSid complexes over the inner membrane is the second energy-dependent step and accomplished by ATP-dependent transporter complexes. These complexes interact with the related periplasmic-binding protein and orchestrate translocation of the FeSid complex to transfer the FeSid complex into the cytoplasm upon passage through several ATP-dependent domain movements [85,87]. Once the FeSid complex has reached the cytoplasmic destination, Fe3+ is reduced to Fe2+ and can thereby be released from most siderophores. Reduction of Fe3+ depends on the FeSid complex stability due to differences of up to 10−20 in Fe2+/Fe3+ binding constants (highest potential for enterobactin −750 mV). Therefore some FeSid complexes need further enzymatic processing to release iron. Siderophore and related iron uptake allows for an intracellular iron concentration of ~10−6 M (or 105–106 iron atoms per bacterial cell; see also the above values determined by Mössbauer spectroscopy ), sufficient for bacterial cell survival due to the incorporation of iron into cofactors such as iron–sulfur clusters, haem proteins, catalases or for storage by proteins such as ferritins, Bfrs or Dps proteins (Figure 1) .
LOW COPY IRON SCAVENGERS IN E. COLI: FERRITIN AND BFR
Three different systems for iron oxidation and storage exist in E. coli. The 24-meric Bfr and ferritin complexes are constitutively expressed at low levels. Both are considered to be the actual iron storage proteins with a high capacity of up to ~4500 atoms per complex accumulating iron as amorphous iron oxide [27,29,88]. Although the role of Bfr in bacterial iron metabolism is still unexplored, ferritin A from E. coli has been demonstrated to collect up to 50% of the cellular iron during a normal growth phase . Phenotypes of ferritin-knockout strains are weak and only slightly influence the overall metabolic iron levels [27,89]. This lack in a defined phenotype is most likely due to the lower importance of ferritin in Fenton and peroxidase reactions. Given the large storage capacity of these complexes, a small number of them is sufficient to collect and provide accelerated access to iron at concentrations needed for the maintenance of iron protein turnover, which is reflected by the low and stable expression of ferritin regardless of changes in the environment [9,11]. It is therefore reasonable to conclude that ferritin can provide rapid access to captured iron for the bacterium if large amounts of this element are requested, and that ferritin fulfils the classical role in iron storage and release . In contrast, the importance and function of Bfr in the iron metabolism of E. coli is largely undiscovered (e.g. no knockout phenotype) and remains enigmatic.
REGULATION OF DPS IN E. COLI: A HIGHLY INDUCIBLE ENZYME
Regulation of Dps proteins in bacterial cells is an important task due to the critical and diverse functions these proteins carry out and to control and eventually down-regulate the large copy number that can be produced during specific cell cycles. Under physiological conditions, the stationary phase can be considered as the growth phase with the highest natural expression profiles of Dps proteins, reaching up to 2% of the cellular proteome (12000 dodecamers per cell; see also below) [36,90,91].
Regulation of Dps expression in response to fluctuations of, for example, iron/FeSid levels is also observed, but generally not well understood, and, only for E. coli, sufficient experimental data have accumulated to draw an approximate picture. Dps expression in E. coli is regulated by the general starvation transcription factor RpoS (most importantly acting during the stationary phase) and OxyR, the positive regulator of H2O2-inducible genes (during the exponential phase of bacterial growth) [92,93]. Negative regulators of Dps are the Fis and H-NS proteins, which recognize different motifs on the dps promoter region, leading to differential down-regulation of Dps molecules [94,95]. The numbers of Dps monomers between exponential and stationary phase vary significantly, depending on the experimental technique used for determination and the bacterial species used, from 6000 to 180000 copies (the equivalent of 500 and 12000 dodecamers) [94,96,97].
The up-regulation of Dps proteins can also occur in response to external stimuli, e.g. an increase or decrease in Fe2+/Fe3+, chelator molecules or FeSid complexes induces protein expression at any growth phase [37,40]. The presence of increased FeSid concentrations, as, for example, reported for experiments using the Gram-negative bacterium Microbacterium arborescens induce Dps up-regulation after expression of uptake receptors in the outer membrane (analogous to the Fec system in E. coli ). This activation can, for example, lead to the replacement of Fur by the σ factor FecI at the DNA level as described previously [60,74]. Uptake of FeSids and the release of free iron into the cytoplasm require the increase of Dps levels in response to enhanced ROS production. ROS and H2O2 concentrations can be sensed via the OxyR regulator and returned as a positive signal upstream to enhance the expression of Dps [92,99].
The adjustment and down-regulation of Dps proteins in E. coli is controlled at the post-transcriptional level through proteolysis by two ATP-dependent proteases. The removal of Dps proteins before entry into the stationary phase is mostly regulated by the ClpXP system . During the stationary phase and upon exit, the removal of Dps is accomplished by the major proteolytic machineries ClpAP and ClpXP [91,100,101]. Interestingly, both chaperone systems need the N-terminus of Dps for substrate recognition, although at different processing states. Recognition and degradation of excessive Dps as proteolytic substrates of ClpAP requires previous processing through the removal of six N-terminal amino acid residues to generate a substrate protein of the ClpAP adaptor protein ClpS [101–104].
THE DIVERSE FUNCTIONS OF DPS PROTEINS: PEROXIDASE REACTION AND DNA PROTECTION
During cellular respiration, ROS are produced as side products. These include hyperoxide anions (O2−) and H2O2 with a limited destructive potential. However, both species can destroy iron–sulfur clusters of important enzymes, which in turn can cause the release of Fe2+. Notably, the most important in vivo mediator of iron reduction during iron-induced redox stress has been identified as free flavin molecules (FAD) . Intracellular free Fe2+ eliminated from enzymes or after uptake from the extraplasmic space can progressively react with H2O2 to produce ROS radicals (e.g. OH• via the Fenton reaction; Figure 3A). Most harmful are the hydroxyl radicals, which can react with unsaturated lipids, protein side chains or irreversibly generate DNA lesions (Figure 3A) . Bacteria are protected against ROS and NO• radicals [known as RNS (reactive nitrogen species)] and H2O2 by enzymes such as superoxide dismutases, reductases, catalases and Dps enzymes [59,106]. Here, Dps has vital functions in (i) the general detoxification of the bacterial cell through peroxidase activity and (ii) the rapid removal of iron from the Fenton reaction followed by storage as insoluble Fe2O3 (Figure 2A) [88,107,108]. These functions requiring rapid activity of Dps enzymes and cellular protection may temporarily also induce the expression of a large number of dodecameric enzymes (e.g. as induced by the addition of FeSids into the medium), even under conditions when Dps is not required for DNA condensation (Figures 3A and 3B) [39,109,110].
In the stationary phase, the role in DNA condensation by Dps proteins becomes effective (Figure 3B). DNA protection through direct physical interaction with Dps proteins and condensation was the initial function assigned to Dps proteins [24,111]. This protective DNA interaction was originally described as being mediated by magnesium ions that were believed to bridge the negatively charged Dps surface with the polyanionic DNA [41–43,112]. More recently, it turned out that protein–DNA interactions more critically depended on the presence of a positively charged and extended N-terminus, which was shown to be the prerequisite for DNA binding in E. coli (Figure 3B2). Residues involved were identified through mutational analysis of three lysines that are crucial for Dps–DNA complex formation (…AKLVKSKAT…) [113,114]. DNA interactions have also been demonstrated for Dps from B. subtilis, B. anthracis, S. aureus, Deinococcus radiourans, Mycobacterium smegmatis and Helicobacter pylori. The last species harbours a Dps enzyme that lacks the N-terminal elongation but instead shows a positively charged outer molecular surface [23,110,113,115–123]. In contrast with E. coli, the Dps enzyme from M. smegmatis carries a C-terminal extension that is of importance for DNA in vitro [124,125]. Interestingly, in analogy to the N-terminus of E. coli, this extension carries five positively charged residues (…KGAADKARRK…). Both modes of interaction allow protection of DNA and maintain stability against mechanical stress as well as enzymatic degradation (e.g. by nucleases) during the bacterial life cycle. Visualization of Dps–DNA interactions were conducted by Kolter and colleagues, who demonstrated these complexes by electron microscopy . In this condensed Dps–DNA state, the chromosome (or plasmid DNA) is rendered DNase resistant, a similar behaviour already shown for histone-protected DNA complexes from eukaryotes [24,122,126]. Owing to this analogous functional behaviour of Dps in DNA protection, these complexes were also termed ‘bacterial chromatin’ .
Current models of protein–DNA complexes are based on structure data acquired at limited resolution, but the precise mode of the putative cellular three-dimensional arrangement remains speculative . In eukaryotes, chromosomal DNA is organized by nucleosome particles and protein scaffolds for DNA winding [127–131]. This model could also apply to DNA–Dps interactions with DNA winding around the roundish complex. Assuming a number of 3×106 bp covering a typical bacterial chromosome and an approximate number of 200 bp per Dps molecule (assuming a single winding of DNA around Dps plus connecting base-pairs), ~1–1.5×104 Dps dodecamer molecules would be sufficient for the protection of the chromosome (see Figure 3B3 for a model). These numbers agree well with the experimental number of ~12000 dodecamers expressed in the stationary phase of E. coli (1–2% of whole-cell protein content) [90,91].
MORE FUNCTIONS OF DPS PROTEINS: CATALYSIS OF ENZYMATIC REACTIONS
An additional enzymatic function of a Dps proteins belonging to a small and evolutionary distant group of bacteria was discovered recently . A Dps protein derived from the Gram-negative bacterium M. arborescens, detected in the insect gut of the herbivore Spodoptora exigua was characterized biochemically. This protein has been shown to catalyse both the synthesis and hydrolysis of N-acyl amino acids. In particular, N-acyl glutamines as products of biocatalysts are reported to be elicitors and stimulate defence reactions (e.g. production of terpenes) in plants after they are recognized, e.g. during the feeding process by the insect . The finding of this reaction carried out by a bacterium co-existing in the microbial gut raises the question of the functional importance of this by-product in the ecological system formed through plant–insect–bacterium interactions .
DPS PROTEINS AS IRON CHAPERONES
Aside from the important Dps function in ROS suppression and the DNA condensation properties leading to the biocrystallization of protein–DNA complexes, iron biomineralization, including uptake and storage by Dps enzymes, is another field of intensive structural research. Although a function of Dps proteins in iron storage and release, e.g. under low-iron conditions, has never been demonstrated in vivo, these data from in vitro characterization are of general importance for the understanding of iron-scavenging kinetics (translocation and nucleation) and the activity of Dps enzymes in the detoxification of ROS and H2O2. Moreover, for technological applications of Dps (and also ferritin) biochemical, biophysical and structural studies regarding the specific influence of the biological scaffold on the formation of nanoparticles with defined size are an essential prerequisite (see also the final paragraph).
Most of our current knowledge on iron chaperoning in bacteria, including Dps, ferritin and Bfr, has accumulated on the basis of high-resolution crystal structure analysis combined with biochemical experiments. A database search of Dps sequences against sequences in the PDB yielded a plot of three ferritin-like clusters (Figure 2A). The largest number of structures containing the fold of the ferritin superfamily localizes to the Dps protein family, which shows a closer evolutionary sequence and structure-based connection to Bfrs than ferritins. Dps proteins assemble into dodecameric shells via hexamerization of dimers or tetramerization of trimers (Figures 2C and 2D) . To date, all Dps protein structures determined follow essentially this principle of the cubic 23-point symmetry comprising three two-fold and four three-fold axes. In the dodecamer, each monomer forms five interfaces to neighbouring subunits with different surface areas, the most extended interface being along the C2 dimer axis [1,40]. This interface between two monomers along the two-fold axis of the four-helix bundle is also architecturally maintained in ferritin and Bfr molecules and may resemble the initial assembly driving the entire assembly process [1,40]. Although the stability of Dps complexes is reported to be high, point mutations in the vicinity of the trimer interface can significantly destabilize the Dps complex from E. coli, leading to the disassembly of the complex into dimers .
The major driving force attracting iron atoms is the unequal distribution of negatively charged residues between the outer and inner surface, which generates an overall negative potential inside relative to outside the hollow sphere. For the Dps complex from M. arborescens, the difference in charges is ~48 for the oligomer (for E. coli Dps this difference is ~96) . Since Fe2+ after entry into the cavity is continuously oxidized at the FOCs, the charge gradient between outside and inside is largely maintained. Together with the translocation channels typically charged through a surplus of aspartates and glutamates, these two prerequisites force the rapid diffusion and nucleation of iron in these hollow shells. Inner volumes of Dps enzymes vary over the range 55000–65000 Å3 (1 Å=0.1 nm), providing space for approximately 500 iron oxide units. If oxidizing molecules such as H2O2 are not present in the cell, Fe2+ remains at the FOC, yielding a positive charge distribution of 12×2 charges in maximum. Thermodynamically, the overall process of Fe2O3 formation is driven by the strongly exergonic contribution of ~800 kJ/mol .
IRON-UPTAKE PATHWAYS: LIFESTYLE VERSUS EFFICIENCY?
Iron translocation in Dps enzymes has long been proposed to occur along the channels formed at the molecular three-fold axis also considered as ‘ferritin-like’ pores due to their similarity in local symmetry and the guiding residues. These channels of ~1 nm diameter typically express negatively charged residues at the outermost surface and additional residues supporting hydrogen bonds and are sequentially organized in the channel interior (Figures 4A and 4B). Inside the channel of C3 symmetry, typically three conserved aspartate or glutamate residues are involved in the translocation process (termed a type I channel [26,44,136]) (Figures 2D and 4A). These general principles are observed in all Dps structures except the structure of DpsA from Halobacterium salinarum (see Figures 4A and 4B for the Dps from M. arborescens) [44,137]. The importance of single aspartate residues in iron guiding along these channels has been demonstrated previously for ferritin and by mutational studies of L. innocua Dps that provided insights into the influence of individual residues on the kinetics of iron-oxo uptake and cluster formation [89,137–139]. Most of these residues are conserved between Dps proteins, e.g. the two pairs Asp131/Asp121 and Asp139/Asp130 are structurally identical residues of Dps from M. arborescens and Dps L. innocua respectively [40,137]. In particular, residues Asp139/Asp130, which are located at the inwards-directed pore entry site of the pore, appear to play an important role in the control of diffusion kinetics, even if mutated against related asparagines (see also Figures 4A and 4B). Hence this motif of negatively charged residues lining these channels is important and functionally conserved between Dps proteins and ferritins (see also the final part of the present review on technological applications).
Iron-uptake pathways of Dps proteins from M. arborescens and H. salinarum
These ferritin-like pores are the only openings allowing ion diffusion into the inner sphere of the protein complex. In a superposition of a dodecameric Dps complex and a trimeric ferritin subcomplex it becomes obvious how strongly these pore geometries invented for iron translocation in quaternary architecture are actually maintained (Figure 5A). However, it was only recently that iron atoms could be traced by X-ray crystallographic methods inside these type I channels of a Dps protein from M. arborescens . Two iron atoms were located in the channel interior separated by ~0.5 nm along the channel axis (Figures 4A and 4B). Unexpectedly, these iron atoms appeared not as free iron but as hexa-hydrated iron complexes that were oriented inside the channel via interactions between the water shell molecules and mostly negatively charged residues of the channel interior (Figures 4A and 4B). Distances between iron atoms and the channel walling are clearly too long to be strongly involved in salt bridge formation. The entire distance between pore entry of iron and the pore exit into the cavity is approximately 20 Å in length and (charged) residues guiding this way are located ~6 Å distant from each other (Figure 4B). All of the residues involved are located on the short loop connecting helix 4 and helix 5 and on helix 5 (see Figure 4C). Although only two iron-binding sites could be localized in the electron density, it is tempting to speculate that Glu132 at the pore entrance and the conserved Asp139 at the pore exit to the inner sphere may play additional roles in guidance of the hexa-aquo complexes (in the case of the entrance) or free iron after removal of the hexa-aquo shell (in the case of the exit) respectively (see the model in Figure 4B). The two hexa-aquo complexes are almost perfectly aligned relative to each other with three water molecules on an upper and three on the lower layer (Figures 4A–4C). The chemical environment of the two iron hexa-aquo complexes changes along this channel; nevertheless the orientations of the individual hexagonal complexes are not disturbed. Asp139 of the channel strongly constricts the diffusion pore and, consequently, the water atoms need to be stripped off at this point. Presumably the iron atom follows the path guided by further charges inside the cavity as ‘naked’ ion and finally reaches the FOCs for oxidation. Interestingly, a recently published structure of Dps from Streptococcus pyogenes shows a sodium ion in the diffusion pore, which was modelled near the iron position we observed in the vicinity of the Dps from the M. arborescens iron-uptake channel . The position of this ion together with surrounding water molecules is similar to that of the first iron atom observed in our Dps structure and may reflect traces of iron rather than sodium.
Comparison of iron-translocation systems: ferritin/Dps, the Dps from M. arborescens and DpsA from H. salinarum
Although the iron-uptake scenario reported for Dps from M. arborescens can be delineated to most of the currently known Dps type I enzyme structures, another uptake path was observed when we investigated the DpsA protein from H. salinarum (type II) . Here, the uptake along channels at the three-fold axis is clearly occupied by bulky side chains and a phosphate ion that is co-ordinated by side chains of the three-symmetry-related arginines. However, in iron-uptake studies using crystals with low iron content, we achieved the tracking of another diffusion pathway which followed a narrow channel that was formed at the non-symmetrical interface of three subunits (Figures 4D and 4E). A total of 12 of these narrow channels are observed in the dodecameric shell, which facilitates the transfer of only dehydrated iron atoms (type II channels). Each channel disembogues into one of the 12 individual FOCs. Three iron atoms can be identified in these channels, co-ordinated by residues from three adjacent subunits, not related by symmetry (Figures 4E and 4F). Fingerprints of this channel architecture are negatively charged residues at the outer surface, similar to the Dps from type I, whereas glutamine and histidine residues with a lower affinity towards Fe2+ are mostly responsible for the iron translocation inside the channel . So far, only one Dps structure of this type has been determined, although the structure of Thermosynechococcus elongatus may also follow this pathway. This protein is most closely related in sequence among all of the Dps proteins currently available in the PDB (Figure 2A). It also contains similar residues known to guide iron inside the asymmetric channel and the entry at the three-fold axis of the complex is largely buried, as is the case in DpsA [48,141].
Why has nature invented two different iron-uptake systems in Dps enzymes? From the classification of proteins and their specific bacterial lifestyle, it seems that in particular halobacterial Dps enzymes, such as DpsA, allow only for the uptake of largely dehydrated iron atoms. These enzymes occur under conditions of high salt: in the case of the archaeon H. salinarum, 4–5 M intracellular [1,142,143]. Consequently, the hydration of all intracellular ions is highly restricted and alternative uptake pathways excluding hexa-aquo complexes in solution must be favourable. Although these channels do not form strongly charged funnel-like entry channels as in most Dps proteins from eubacteria, their individual uptake potential is enhanced by two factors: the number of diffusion channels (12 compared with four), and the large charge difference between cavity and outer surface. Although these channels may provide a rather limited diffusion for iron due to their small average diameter, the increased number of 12 such channels compared with four in type I may account for a similar uptake rate. Moreover, the channel residues can interact much less strongly with iron (or related divalent ions) due to their chemical nature (histidine and asparagine residues). Another advantage of this particular type II architecture is the proximity between channel exit and the FOCs and the immediate delivery of iron to the FOCs (distance between translocation site T3 and FOC of 8 Å; Figures 4E and 4F). In case of the type I uptake channels, this exit of the channel is ~2 nm away from the FOCs, and additional transfer steps, presumably via transient hydrogen bonds to the matrix, are required (Figures 5B and 5C).
FOCS AND THEIR PLASTICITY
FOCs are the structure entities where Fe2+ is oxidized to Fe3+ by molecular oxygen or H2O2. These sites were originally discovered in ferritins and later in Bfrs, where they are located on single subunits. In the E. coli ferritin soaked with Fe2+, three iron atoms per monomer were observed in the crystal structure, two of which belonged to the di-iron site (A and B sites; ‘active centre’) with a distance of 3.8 Å in between, while the third site (named the C site) was located approximately 6 Å apart [144–146]. The iron-binding site A was arranged by side chains of two glutamates and one histidine (Glu27, Glu62 and His65) with Glu62 as the bridging residue between sites A and B (ligated by Glu62, Glu107 and Glu144). Under oxidative conditions, the two iron atoms are Fe3+ connected by molecular oxygen O2− according to Mössbauer spectroscopy .
In Dps enzymes the iron-binding sites at FOCs are localized at the interface of two-symmetry-related monomers at a distance of approximately 2 nm (Figure 5) and were firstly discovered in the Dps structure of L. innocua with a single iron atom ligated to a glutamate and aspartate residue from one subunit and a histidine residue from the adjacent subunit (His31, Asp47 and Asp58; Figure 2D and Figure 6A) . This catalytic site with two histidine and two aspartate residues marked the archetypical fingerprint of this enzyme class as the basis for binding and oxidation of Fe2+ to Fe3+ (Figures 6A–6C). From the local environment of this iron site and due to the local structural similarity with the di-iron sites observed in ferritins, a second hypothetical iron site was hypothesized and modelled accordingly (Figure 6A, site F2). This second site in the Dps ‘di-iron’ has first been confirmed experimentally in the DpsA structure after soaking these crystals with Fe2+ for 30 min under reducing conditions . This second site was observed at the place where the second site in the Listeria Dps was predicted and structurally at the same distance as the B site observed in ferritins. Furthermore, in this DpsA structure a third iron atom was visible located in the vicinity of the FOC, similar in distance and arrangement to the C site located in ferritins.
FOCs of Dps proteins and iron storage in DpsHS
Recently, the plasticity of the FOC centres in Dps enzymes was investigated further and it turned out that in the Dps from T. elongatus the aspartate residue is replaced by histidine . In a crystal structure of this enzyme, which can bind both iron and zinc, it appears that the aspartate/histidine mutation in the active site leads to a higher affinity towards zinc, and similar to the F2 site of DpsA a second zinc atom in the neighbourhood of F1 was observed. The co-ordination of Zn2+ instead of iron is presumably favoured by the presence of this third histidine . Although zinc may mimic the binding sites of iron at the FOC, it is questionable if this has a physiological function since zinc cannot be reduced or oxidized under physiological conditions. Consequently, zinc ions are therefore not electro-neutral and would allow for a limited number of atoms taken up until the surplus of negative charges at the inner walling is neutralized. Furthermore, both the iron reduction and peroxidase activity of the enzyme from T. elongatus was demonstrated and hence Zn2+-binding can be considered as being artificial due to in vivo reaction conditions.
In our own studies with DpsA, we could show that not only zinc, but also also manganese, nickel, cobalt and copper can bind the FOC centres (R. Albrecht and K. Zeth, unpublished work). Although copper as another element with redox potential (Cu+/Cu2+) in the physiological range has been proposed to be one additional element for possible storage and removal from the Fenton reaction, it is rather questionable whether Dps proteins play an important role in the detoxification of copper in bacterial cells [147,148].
NUCLEATION OF IRON OXIDE
After iron is oxidized at FOCs, the storage of the element as polymorphic Fe2O3 can progress. Although a large number of Dps structures have been published in the last few years, only the DpsA complex from H. salinarum showed formation of iron oxide nuclei outside the FOCs. Two different types of nuclei were discovered: type I is located at the two-fold axis of the protein, whereas type III nuclei are discovered at the three-fold axis. The distances between the FOCs and the type I and II nuclei are 13 and 12 Å respectively (Figure 6E). Three iron atoms at the type I centre may form an initial seed for further growth of a nucleus into the inner sphere of the complex. Symmetry-related glutamate residues (Glu72 and Glu75) of two adjacent subunits provide the structural basis for iron fixation. A similar architecture has been observed for the five-atom iron cluster at the type II centre. Three-symmetry-related Glu154 residues form the basis for the formation of an iron oxide nanocluster (Figures 6E and 6F). Interestingly, in a low-iron form of this protein, which was isolated from H. salinarum lysate, only a single iron atom was present at NII (nucleation centre II), which solely organized the formation of a large water cluster comprising 13 water molecules (results not shown). Some of these water molecules were later observed to occupy the same positions as the iron-oxo atoms found in the high-iron form .
BNP (BIO-NANO-PROCESS) APPROACHES USING FERRITINS AND DPS ENZYMES
Nanobiotechnology is a growing research area aiming, for example, to produce surface properties with laterally reproducible and re-occurring structures. Ferritins, the larger cage-shaped ‘cousins’ of Dps proteins, have been successfully introduced into semiconductor devices since they (i) form highly reproducible nanostructures at the close to atomic scale dimension; (ii) can self-assemble on surfaces as two-dimensional crystalline arrays to form the functional basis for nanostructures; and (iii) provide biological architectures for the functional deposition of inorganic material which can be prepared under a wide range of conditions (pH, salt, metal ions and temperature) . Furthermore, these processes can be tightly controlled and specific properties optimized through mutations to optimize surface properties. Using these nanoblocks in a controlled manner can be considered as the basis for the development of the method called BNP to produce regular top-down structures on inorganic surfaces as nanofunctional structures (see Figures 7A and 7C) . One of the major requirements to generate FNGM is the generation of highly ordered two-dimensional arrays of proteins (e.g. two-dimensional crystals; see Figure 7B) which can separate small inorganic functional units made up of lead sulfide, cadmium sulfide, nickel oxides (NiO) or cobalt oxides (CoO) etc. Although open protein templates such as chaperones or viruses have been used successfully to prepare these nanostructures, nano-cage architectures including ferritins are currently the most advanced studies by far [150–152]. Apo-ferritin (empty ferritin structures, e.g. from horse spleen) is the biological matrix readily available at low costs and can be prepared by simple dialysis of ferritin under reducing conditions to displace the iron-oxide matrix by different metal oxides or sulfides. These ferritins (containing light and heavy chains) have dimensions of 12 nm as the outer diameter and 7 nm as the inner diameter (Figure 7A).
Size matters: the miniaturization of surface-adsorbed nanodots
Although apo-ferritin allows for the deposition of nanodots and high resolution, Dps enzymes with dimensions of 8 nm outer and 5 nm inner diameters with similar functional properties, but smaller size, allow for an even higher lateral density (Figure 7C). Both cage-like proteins can be considered as spatially highly restricted chemical reaction chambers for the homogenous biomineralization of inorganic material. A variety of techniques to reproducibly charge and achieve two-dimensional crystals of ferritin and Dps molecules has been developed by Yamashita and colleagues at NAIST (Nara Institute of Science and Technology) [53,149]. Using these templates, this group also succeeded to produce semiconductor nanodots on the basis of CdSe and ZnSe. The nanoparticles on the basis of CdSe have been produced in template spheres, the ferritin and Dps (from L. innocua) cages [53,149]. The size of the CdSe nanoparticle was estimated by TEM (transmission electron microscopy) to be 7 nm ±10%, as expected from the inner dimension of the ferritin shell. Mutant proteins of recombinant apo-ferritins were used to study the deposition of ZnSe. Mutants were prepared as (i) N-terminally truncated versions, (ii) mutations in the uptake channels and (iii) charge mutations introduced in the inner sphere protein cages. All mutations differentially influenced the formation of ZnSe particles and directly allowed insights into the course of biomineralization yielding nanoparticles of significantly different properties (size and photoluminescence) . Taken together, these advances resulted in the construction of FNGM arrays produced by bottom-up and top-down techniques with high mechanic persistence (100000 times read and write) .
To increase the density of memory being stored on a given surface, further miniaturization is required. These arrays of nanoparticles have superior properties compared with ferritin-based arrays in that they enable wider memory windows and have technical advantages exceeding previous developments. On the basis of L. innocua Dps, Yamashita and colleagues synthesized 4 nm CdSe particles arranged in a large hexagonal two-dimensional packing with a significantly higher nanodot density . The rapid crystallization of Dps enzymes in the presence of DNA has already been observed; therefore these particles seem to be prone to crystallization . Two-dimensional arrays of these particles at high lateral order were achieved by standard techniques and analysed by TEM methods. The lateral resolution of the two-dimensional crystals was high and the density of particles exceeded 2.8×1012 cm−2. Further improvement of these arrays on the basis of Dps proteins may lead to wider memory windows and higher persistence, and may be used as light-emitting devices and photodetectors and ultimately for highly dense memory storage [53,153].
FUTURE PROSPECTS AND OUTLOOK
Since the first report on Dps enzymes, more than 20 years have passed. Many protein species from archaea and Gram-positive and -negative bacteria have been studied by different techniques aiming to elucidate their specific function in a biological context. Several functions of Dps proteins were reported, some of which are recurring and archetypical, e.g. as iron storage and detoxification of ROS. Other (enzymatic) functions are rather unique, presumably for a small class of Dps proteins. Given this body of results and information, the structural and functional principles in vitro are relatively clear; however, in vivo many questions remain largely unanswered.
The physiological role of Dps proteins in vivo is still unclear, since a deletion of the protein has no visible effect on cell growth and survival, unless knockout strains are grown in the presence of enhanced iron concentrations. Although data have been collected on the iron distribution in E. coli, these reports are in part contradictory to each other, and a reliable iron proteome using an E. coli cell would advance the perspective on iron flux. Such a study would particularly improve the understanding of Dps/ferritin/Bfr functions.
Recent reports indicate the plasticity in uptake of various divalent ions such as Zn, Co and Cu under in vitro and in vivo conditions. Although it is unlikely that Zn forms ZnO or mixed Zn/Fe oxides in vivo, it may be envisioned that Cu+, Mn2+, Co2+ or Ni2+ form part of the nanocluster together with an excess of iron. The element distribution of such metal clusters isolated from living cells may give hints as to more general functions of Dps in the detoxification of divalent metal ions. This approach could possibly be expanded to study toxic heavy elements such as Hg, Cd or Pb after previous enzymatic removal of organo-alkyl donors. Uptake of these metals into Dps proteins is likely to occur according to the general uptake of divalent ions. Irreversible storage may be enhanced by overexpression of the protein and the introduction of cysteine residues into the inner sphere. The ability of, for example, E. coli to live under increased heavy atom concentrations could be easily tested. Similar in vitro studies may shed light on the plasticity of Dps to be used as a template for the formation of nanodots as a basis for nanotechnological studies. Nanodots of defined size, but various functional properties, may be engineered after manipulation of the Dps enzyme or the composition of the metal-oxide/sulfide/selenide clusters.
From the structural biology point of view, there are several interesting questions addressing Dps–cofactor–DNA interactions. So far, two major uptake pathways of Dps enzymes have been discovered in atomic detail; subsequent steps including the ferroxidase and nucleation centres, however, are less well understood. So far only the FOC of DpsA from H. salinarum shows additional iron-binding sites, but oxidation states and geometries in related Dps enzymes remain unclear. Nucleation of iron oxide on the inner surface is a general phenomenon of Dps enzymes, but the evolution of these initial small nuclei was studied only for DpsA. Furthermore, the plasticity of Dps complexes has not been characterized and additional metal elements may be used to study their oxidation at FOC centres and further nucleation. Finally, the structural basis of Dps–DNA complexes remains largely enigmatic. Crystal structure analysis of such protein–DNA complexes would not only allow determination of the specific influence of charged termini, but also additional determinants on the surface of Dps cages.
Work in the author's laboratory is supported by the Human Frontiers Science Program (HFSP) [grant number RGP61/2007], the Max Planck Society, the University of Tübingen, and the German Science Foundation [grant number DFG-ZE522/4-1].