Formation of Iron Oxide Nanoparticles in the Internal Cavity of Ferritin-Like Dps Protein: Studies by Anomalous X-Ray Scattering

DNA-binding protein from starved cells (Dps) takes a special place among dodecamer mini-ferritins. Its most important function is protection of bacterial genome from various types of destructive external factors via in cellulo Dps–DNA co-crystallization. This protective response results in the emergence of bacterial resistance to antibiotics and other drugs. The protective properties of Dps have attracted a significant attention of researchers. However, Dps has another equally important functional role. Being a ferritin-like protein, Dps acts as an iron depot and protects bacterial cells from the oxidative damage initiated by the excess of iron. Here we investigated formation of iron oxide nanoparticles in the internal cavity of the Dps dodecamer. We used anomalous small-angle X-ray scattering as the main research technique, which allows to examine the structure of metal-containing biological macromolecules and to analyze the size distribution of metal nanoparticles formed in them. The contributions of protein and metal components to total scattering were distinguished by varying the energy of the incident X-ray radiation near the edge of the metal atom absorption band (the K-band for iron). We examined Dps specimens containing 50, 500, and 2000 iron atoms per protein dodecamer. Analysis of the particle size distribution showed that, depending on the iron content in the solution, the size of the nanoparticles formed inside the protein molecule was 2 to 4 nm and the growth of metal nanoparticles was limited by the size of the protein inner cavity. We also found some amount of iron ions in the Dps surface layer. This layer is very important for the protein to perform its protective functions, since the surface-located N-terminal domains determine the nature of interactions between Dps and DNA. In general, the results obtained in this work can be useful for the next step in studying the Dps phenomenon, as well as in creating biocompatible and solution-stabilized metal nanoparticles.


INTRODUCTION
Approximately one-third of the well-characterized proteins and almost half of the enzymes contain from one to several metal ions [1,2]. Among those, one of the most vital and abundant chemical elements is iron. This metal is commonly found in the hemes o r iron-sulfur prosthetic groups in proteins [2,3]. The biological signifi cance of iron is determined by its ability to undergo reversible redox reactions. Divalent and trivalent iron is essential in vital processes in all eukaryotes and most prokaryotes.
Iron participates in various metabolic processes, such as oxygen transport, DNA synthesis, and electron transport for energy production. It is a universal microelement that ensures normal functioning of all body systems at the cellular level.
On the other hand, the excess of iron promotes the damage of DNA, proteins, and lipids, leading to the disruption of cellular homeostasis. The existing evolutionary developed mechanisms of detoxifi cation and iron removal from the cytosol include oxidation of excessive divalent iron to the trivalent iron in the Fenton reaction [4]: Fe 2+ + H 2 O 2 → Fe 3+ + OH + OH − In living organisms, trivalent iron is accumulated in the ferritin-like proteins. The superfamily of ferritin protein has evolved to provide iron sequestration in its soluble, non-toxic, and bioavailable form [5]. These proteins function as iron depots and store it in a non-reactive form, this limiting its involvement in cellular processes and protecting the cell from the oxidative damage initiated by this metal [6,7]. Ferritin is a protein complex that plays a role of major intracellular iron depot in humans and animals. It is found virtually in all organs and tissues. Mammalian apoferritin (protein lacking iron ions) is a 24-mer protein with the molecular mass (MM) of ~450 kDa; each of its polypeptide subunits has MM of ~20 kDa. Ferritin is a globular protein with the inner cavity that can store approximately 4500 ions of hydrated trivalent iron oxide (Fe 2 O 3 • H 2 O) together with a variable number of phosphate groups. The external diameter of the protein is 12-13 nm; the dimeter of its inner cavity is 7-8 nm [7,8]. The protein component of the ferritin molecule has numerous pores through which iron is transported [9].
There are at least three types of ferritin-like prokaryotic proteins -bacterial ferritin, bacterioferritin, and dodecamer ferritin (mini-ferritin) -that are related to eukaryotic ferritins. Similar to the mammalian ferritins, bacterial ferritin and bacterioferritin consist of 24 subunits and have the central cavity that can accommodate ~2500 iron atoms in bacterial ferritin [10] and ~1800 iron atoms in bacterioferritin [10,11]. The presence of heme (iron protoporphyrin IX) located between each two protein subunits and linked to methionine in each of these subunits is a characteristic feature of bacterioferritin that distinguishes it from other ferritin forms. As a result, the bacterioferritin molecule contains 12 heme groups, the role of which is yet to be elucidated [12][13][14].
Dps (DNA-binding protein from starved cells) has a special place among the dodecamer mini-ferritins. Its most important function is protection of genome from harmful factors such as starvation, high temperature, UV-and γ-radiation, toxins, chemical shock, and oxidative stress. Dps binds to DNA thus forming a stable Dps-DNA complex that protects DNA from damage [15][16][17].
One of the important consequences of this protective response is the emergence of bacterial resistance to antibiotics and other drugs. That is why formation of protective Dps-DNA complexes in the stress-induced bacterial cells has attracted a signifi cant attention of many research groups in recent two decades. The most known works in this area are the studies by Minsky and colleagues, who were the fi rst to prove this phenomenon experimentally [18][19][20].
In bacterial cells, Dps is present in minor amounts, and its synthesis is induced during the stationary phase of bacterial growth, during starvation, or under oxidative stress [21]. The Dps dodecamer lacks structural modules for the recognition of specifi c nucleotide sequences. It is assumed that its binding with the negatively charged sugar-phosphate DNA backbone occurs via electrostatic interactions with the lysine-enriched N-terminal domains of Dps monomers [15,[22][23][24]. However, the exact mechanism of Dps binding to DNA is still unknown [25].
Dps monomer is a polypeptide consisting of 167 amino acid residues (MM 18.7 kDa), that contains a highly conserved sequence of four α-helices [26]. The Dps dodecamer has MM 224.4 kDa and displays type 23 symmetry with the external diameter of 8-9 nm with the inner cavity diameter of 4-5 nm. As Dps is a ferritin-like protein, each monomer is capable of binding up to 40 iron atoms. Hence, Dps can accommodate approximately 500 Fe 3+ ions in its inner cavity [27].
Therefore, Dps simultaneously performs two vitally important functions in bacterial cells: (i) protects bacterial genome from unfavorable external factors by forming a crystal complex with DNA; (ii) serves as an iron depot and protects cells from the oxidative damage initiated by the iron excess. These two functions are interrelated. In particular, Zhao et al. [27], who studied DNA damage in vitro, demonstrated that Dps prevents the cleavage of Escherichia coli DNA simultaneously exposed to FeSO 4 and H 2 O 2 . That implies that by forming the protein-DNA complex and neutralizing hydrogen peroxide during interaction with iron, Dps preserves E. coli genome under stress conditions [28].
Despite the fact that E. coli Dps does not contain any canonical ferroxidase sites present in E. coli bacterioferritin or bacterial ferritin, Dps oxidizes Fe 2+ at some sites in the protein macromolecule followed by its accumulation as Fe 3+ in the protein central cavity [27]. It has been assumed that the ferroxidase sites of Dps are formed by the iron-binding amino acids, such as aspartate, glutamate, and histidine [28,29]. Thus, the N-terminal domain of the expressed Deinococcus radiodurans Dps (DrDps1) contains a region (residues 30-55) that includes the metal-binding site (Asp 36 x 2 His 39 x 10 His 50 x 4 Glu 55 motif) and is located at the external surface of the dodecamer. Disruption of this site aff ects the protein self-assembly, as well as reduces its DNA-binding capacity, i.e., decreases its protective capacity [30]. BIOCHEMISTRY  Unlike Dps, far from all bacterial dodecamer mini-ferritins can simultaneously protect an organism from the oxidative stress by forming a crystalline complex with DNA and serve as a source of iron in the case of its defi cit [31]. That is one of the reasons why the unique features of Dps attract so much attention. Moreover, there is also another aspect that makes investigating this protein so important. Accumulation of a large number of metal ions in the inner cavity of Dps protein presumes an existence of magnetic properties; such protein molecules could be considered as natural biosensors of electromagnetic radiation. Dps is capable of transmitting the received signal to DNA, which gives a promise for the development of a new generation of logic elements. That is why investigating the formation of iron nanoparticles inside the stabilizing Dps protein shell has not only scientifi c, but also practical signifi cance.
In order to use the small angle X-ray scattering (SAXS) for elucidating the structure of metal-containing specimens and size distribution of metal nanoparticles in the sample, the contribution of these components to the total scattering should be separated. Traditional SAXS techniques are used for studying the structure of original samples prior to the formation of metal particles (i.e., structure of the matrix) [32]. The standard approach based on the subtraction of the matrix scattering from the total scattering of the metal-containing sample is possible only in the cases, when the structure of the matrix does not change during formation of the metal particles in it. When structural changes in the original sample are presumed, the commonly used method is anomalous SAXS (ASAXS) [33][34][35][36]. In this method, experimental SAXS curves obtained for the metal-containing samples are recorded at diff erent energies of the X-ray radiation: (i) close to the edge of the absorption band of the metal (when only matrix scattering is recorded) and (ii) away from the band edge (when total scattering is recorded), so that the diff erence between these two curves could be attributed only to the scattering of nanoparticles containing the metal atoms. Analysis of the contribution of each component allows to calculate the size distribution of the metal nanoparticle and to evaluate the structure of the metal-containing sample during its interaction with the metal.
In this study, we investigated accumulation of iron oxide in the inner cavity of the ferritin-like Dps protein using ASAXS.

MATERIALS AND METHODS
Preparation of iron-containing Dps samples. Dps was expressed and purifi ed according as described earlier [37,38]. The purifi ed protein was concentrated with an Amicon® centrifugal concentrator (10 kDa cut-off ; Merck Millipore, USA) to the concentration of 3 mg/ml followed by dialysis against a buff er containing 50 mM NaCl, 0.5 mM EDTA, 50 mM Tris-HCl (pH 8.0). The purity of the obtained protein sample was evaluated by electrophoresis in a 15% polyacrylamide gel; protein concentration was determined from absorbance at 280 nm (A 280 ) using molar absorption coeffi cient from [39].
Freshly prepared solution of FeSO 4 ⋅ 7H 2 O was added to the purifi ed Dps (3 mg/ml) in a buff er containing 50 mM NaCl, 0.5 mM EDTA, and 50 mM Tris-HCl (pH 8.0) in the amounts corresponding to 50, 500, and 2000 iron atoms per dodecamer, and the mixture was incubated for 30 min at room temperature. The resulting samples are designated in the text as Dps-Fe50, Dps-Fe500, and Dps-Fe2000, respectively.
SAXS and analysis of the obtained data. Traditional SAXS. Dps structure was investigated by traditional SAXS with the Petra III synchrotron (DESY, Germany), beamline P12 [40]. The P12 beamline was equipped with an automatic sample changer and two-dimensional Pilatus 2M detector (DECTRIS, Switzerland). The scattering intensity, I(s), was measured in the range of momentum transfer values 0.08 < s < 3 nm -1 , where s is scattering vector (4πsinθ/λ), 2θ is the scattering angle, and λ is the radiation wavelength (0.124 nm). For each sample, 50 frames were recorded to evaluate possible radiation damage. No radiation damage was observed in the process.
The scattering curves were primarily processed with the PRIMUS program [41]. Analysis of the obtained data and structural modeling were performed using the ATSAS software package [42].
The radius of gyration (R g ) of the scattering particles was determined from the initial part of the scattering curve in the region of the smallest s values using the equation (1): (1) which is valid in the region ( sR g ) < 1.3. The scattering intensity at the zero angle, I(0), was determined from the slope of the linear part of the Guinier plot (ln I(s) versus s 2 ), which is proportional to molecular mass of the scattering object.
Molecular masses were calculated from the SAXS data by two diff erent methods: (i) using Bayesian approach (MM Bayesian ) [43] and (ii) based on the excluded volume V p (Porod volume) inaccessible to the solvent [44] using the empirical ratio between V p and MM (1.65 for proteins) [45].
The GNOM program [46] was used to calculate the distance distribution function p(r), which was required for the reconstruction of the Dps protein shape in solution based on the SAXS data. The distance distribution function, p(r), was determined as an indirect Fourier transform of the scattering intensity in accordance with the equation (2): where I(s) is the scattering intensity. The maximum particle size (D max ) was calculated under conditions p(r) = 0 at r > D max .
The ab initio method for the reconstruction of the low-resolution shape of Dps was based on the simulated annealing algorithm and was implemented using the DAMMIN program [47], which allows the construction of structural models with the minimization of residual χ 2 between the experimental scattering and scattering obtained for the models (3): where N is the number of experimental points, I exp (s j ) and σ(s j ) are experimental intensities and errors, I calc (s j ) is the calculated intensity for the model, c is the scaling coeffi cient.
The structure of the iron-containing protein was calculated using the multiphase ab initio modeling, which allowed to obtain not only the low-resolution structure of the protein component in the complex, but also to determine the location of iron atoms in the protein matrix. The two-component (two-phase) model was constructed with the MONSA program [47]. This program takes into account the diff erences in the electron densities of the protein and metal components of the complex, as well as the ratio between their volumes. SAXS curves from the original protein and protein containing metal atoms were used to produce the two-phase model of the protein-metal complex. The theoretical scattering intensity for the constructed models was calculated using the CRYSOL program [48].
ASAXS experiments were also carried out with a synchrotron Petra III, beamline P12, and involved recording of X-ray scattering curves at diff erent wavelengths (λ), i.e., at diff erent energies of the incident beam (E). The measurements were conducted for the original Dps protein and Dps samples containing 50 (Dps-Fe50), 500 (Dps-Fe500), and 2000 (Dps-Fe2000) iron atoms per dodecamer. The scattering data were recorded at several diff erent photon energies E k with the E 0 energy (10,000 keV, λ = 0.124 nm) being suffi ciently far from the edge of the Fe absorption band, and, hence, selected for investigating the structure of the original Dps molecule (traditional SAXS technique). The obtained SAXS and ASAXS data were corrected for the background scattering and fl uorescence, and were processed with the ATSAS software package [42] using the recently developed strategy for ASAXS data acquisition and processing [49].
The atomic scattering factor was determined from the following equation (4): (4) where dispersion correction factors f′(E) and f′′(E) become signifi cant in the vicinity of the edge of the resonance atom absorption band. In our case, the measurements were conducted close to the K-band (absorption band of Fe atom), i.e., at the photon energy E = 7.125 keV (λ = 0.174 nm). The changes in the dispersion corrections factors f′(E) and f′′(E) with changes in the energy of photons used in this study, E, are shown in Fig. 1.
For each sample, the scattering curves I(s, E k ) were recorded with 7 diff erent energies of the incident radiation E 1-7 : 7.100, 7.110, 7.118, 7.125, 7.128, 7.130, and 7.133 keV in the region at the edge of K-band (E = 7.125 keV). The anomalous correction factors for E = 7.125 keV were 8.13 and 0.48, respectively. The difference between the scattering curves produced at diff erent energies Δ k (s) = I(s, E 0 ) -I(s, E k ) is proportional to the scattering by the resonance atoms [33][34][35][36]. These difference curves were used for calculating the volume size distribution functions D V (R). The integral equation (5) was solved for D V (R) using the GNOM program [46] assuming a spherical shape of the formed nanoparticles.
In this equation, R is the sphere radius; R min and R max are the minimal and maximal sizes, respectively; i 0 (x) = {[sin(x) -x cos(x)]/x 3 } 2 and m(R) = (4π/3)R 3 are the sphere form factor and its volume, respectively. The scattering length density for anomalous atoms is defi ned as follows: Δρ = (N 0 2 -N k 2 )e/v at , where N 0 and N k are the number of electrons contributing to scattering far from the resonance and at E = E k , respectively; e is the electron charge; and v at is atomic volume.
We also used an alternative approach to analyze the size distribution of Fe nanoparticles formed in the Dps protein using the MIXTURE program [41]. In this approach, the scattering intensity I(s) of the mixture of k diff erent components with diff erent sizes is represented as a linear combination (6): (6) where ν k is the volume fraction of component k, I k is the scattering intensity for this component, K is the number of components. The MIXTURE program models the scattering of mixtures containing K number of scattering objects with varying shape and size using theoretical scattering from the simple bodies (sphere, hollow sphere, ellipsoids, cylinders, etc.). Each object is characterized by its own volume fraction, average size, polydispersity distribution width, contrast, and, optionally, by the possibility of interparticle interactions. An experimental BIOCHEMISTRY (Moscow) Vol. 87 No. 6 2022 scattering pattern is approximated by the weighted combination of the calculated individual (partial) scattering curves from the components to minimize residual χ 2 between the experimental and model data.

RESULTS AND DISCUSSION
SAXS studies of the Dps structure. We used traditional SAXS technique to investigate the Dps structure in solution and to determine the SAXS invariants, such as radius of gyration (R g ), Porod volume (V p ), distance distribution functio n p(r), the maximum size (D max ), and molecular mass (MM). Preliminary experiments have shown the absence of concentration dependence and aggregate formation in the solution at the Dps concentrations ranging from 1.0 to 3.0 mg/ml. Based on this observation, we selected the sample with the concentration of 3.0 mg/ml for further SAXS experiments and data processing, as it produced a suffi ciently informative SAXS curve with a low noise in the range of angle vectors 0.25 < s < 3.0 nm -1 and clearly pronounced shape factor (Fig. 2, curve 1).
The SAXS curve presented in Fig. 2 is typical for a solution of a monodisperse spherical protein. The distance distribution curve p(r) (Fig. 2, inset), which describes the shape of a scattering object [32] and was used in the DAMMIN program for reconstruction of the low-resolution structure from the SAXS data [47] also indicates a spherical shape of the protein. It can be concluded based on the p(r) function profi le that in this case, the spherical object is hollow, since the distance distribution function is asymmetric and its maximum is shifted to the right. Moreover, considering that the amplitude of the p(r) function is proportional to the electron density of individual parts of a scattering object, it can be suggested that the protein has a less dense surface layer, because there is a weakly scattering "tail" in the p(r) profi le in the size range of ~8-9.6 nm. Based on the published data and our previous research, such scattering corresponds to the fl exible N-terminal domains of the protein [24,50,51]. The low-resolution Dps structure (bead model) was reconstructed from the SAXS data using the DAMMIN program [47]. The shape reconstruction was based on the algorithm of annealing within the sphere with a diameter of the maximum size of the protein molecule, D max , which, in turn, was determined from the distance distribution function and was 9.6 nm. The results of the reconstruction are presented in Fig. 2b with grey beads. The residual χ 2 for the experimental data was 1.9, which indicates good agreement of the experimental SAXS curve with the scattering curve produced by the obtained low-resolution shape (Fig. 2a, curve 2). The bead model of the Dps structure has an inner cavity and corresponds to the known crystal structure of the protein (PDB ID: 1DPS) (Fig. 2b), although the bead model is slightly bigger in size due to the scattering by the N-terminal domains (which are absent in the 1DPS atomic-resolution structure) due to their fl exibility and impossibility of their crystallization [26].
The main structural characteristics of Dps determined directly from the SAXS curve without modeling (SAXS invariants) are presented in Table 1.
The data in Table 1 together with the reconstructed low-resolution shape of the protein, indicate its native state, correspond to the characteristics of this protein known from the literature [26,50,51], and make it possible to further use the protein in investigating the processes of accumulation of iron atoms in the inner cavity of the Dps dodecamer.
ASAXS studies of the iron nanoparticles formation in Dps. Experimental ASAXS curves recorded at diff erent energies are presented in Fig. 3.
Analysis of the scattering curves for the samples with diff erent Fe 3+ content and measured at diff erent radiation energies E 1-7 revealed the following features of the Dps interaction with iron ions.
1. The scattering curves recorded at the energy E 0 in the regions far from the iron absorption band (10,000 keV, λ = 0.124 nm) and at the energies E 1,2 (7.100 and 7.110 keV) were virtually identical for all Fe-containing samples. (Hence, we omitted SAXS curves recorded at E 0 in Fig. 3 to avoid plot overloading.) 2. The low iron content (50 atoms per Dps molecule) did not aff ect the general structural characteristics of the protein at the energies relatively far from the iron absorption band (K-band). It is important to emphasize once more that the preliminary structural analysis of the original Dps protein at the energy E 0 revealed that the protein was in the state suitable for its use as a matrix for the formation of iron nanoparticles.
3. In contrast to the specimens with the low iron content (Dps-Fe50), Dps-Fe500 and Dps-Fe2000 demonstrated a signifi cant increase in the scattering intensity at very small angles, i.e., exhibit high polydisperse due to the ability of iron as a transition metal to form stable complexes with, for example, protein amino groups, thus bringing closer together protein chains of  the neighboring macromolecules. The protein matrix, thus, changed signifi cantly upon interaction with the iron-containing compounds, which strictly required the use of ASAXS in this case. 4. In the samples with the high iron content (Dps-Fe500 and Dps-Fe2000), the presence of large metal nanoparticles was expected due to the formation of metal-containing protein aggregates.
5. The scattering curves recorded at diff erent photon energies E 1-7 demonstrated a certain dependence on the energy of incident beam in the range of angle vectors 0.5 < s < 1.3 nm -1 with the minimum at s = 0.95 nm -1 (Fig. 3). Although this dependence was most pronounced for Dps-Fe2000, nevertheless, for all iron-containing samples it was possible to calculate the diff erence between the scattering curves obtained at diff erent energies Δ k (s) = I(s, E 1 ) -I(s, E k ), which was proportional to the scattering by the resonance Fe atoms and could be used for the analysis of size distribution of metal nanoparticles formed in the protein.
6. The maximum of the anomalous signal for all Fe-containing samples was observed at the photon energy E 5 = 7.128 keV, which is close to the K-band of Fe absorption (Fig. 1). Accordingly, the scattering curve at this energy was used to evaluate the resonance signal Δ 5 (s) = I(s, E 1 ) -I(s, E 5 ) only from the iron nanoparticles formed in the Dps inner cavity. The criterion that the diff erence signal is indeed due to Fe particles is the absence of the signal from the protein matrix, i.e., from the Dps form factor, on the obtained anomalous scatter-ing curves. This criterion was met for all Fe-containing samples (Fig. 4a).
The obtained curves of the anomalous signal were then used to calculate the volume size distribution functions D V (R) of iron oxide nanoparticles formed in the internal cavities of the Dps dodecamer. (Fig. 4b). It is important to note that the ASAXS curves were well pronounced for Dps-Fe500 and Dps-Fe2000, but not for Dps-Fe50 (weak signal, high noise). Hence, the size distribution function calculated from this curve should be considered only as an estimate.
Analysis of the volume size distribution functions (Fig. 4b) revealed the following regularities. Predominantly small nanoparticles ~2-nm in size were formed at low iron ion concentrations. The increase in the iron content to 500 atoms per protein molecule resulted in the additional emergence of 4 to 5-nm particles (shoulder in the D V (R) function), as well as minor amounts of larger structures (up to 20 nm) associated with the iron oxide nanoparticles in protein aggregates. At the iron content of 2000 atoms per dodecamer, mostly 4 to 5-nm particles were formed, as well as large Fe-containing structures, while the 2-nm particles were absent. This implies that at the high iron concentration, the entire inner cavity of the protein is fi lled with metal nanoparticles. Their size is limited by the size of the Dps dodecamer inner cavity and determined by the iron concentration in the solution, which is important when using this protein matrix for the formation of solution-stabilized metal nanoparticles.
Similar results were obtained using an alternative approach for determining the fractional composition of iron nanoparticles in the protein dodecamer with the MIXTURE program, suggesting that the shape of the formed particles is spherical. For each specimen, three fractions with a broad size range (D min -D max ) were specifi ed. For each fraction, the average particle size and volume ratio of the fraction were calculated using the MIX-TURE program ( Table 2).
Similar to determining the volume size distribution D V (R) with the interactive GNOM program, the use of MIXTURE program revealed a predominant presence of the 2-4-nm nanoparticles in all samples together with a small amount of larger formations. Since both methods were used to determine the volume fractions Table 2  of the nanoparticles, in quantitative terms the number of large particles was low even for the high iron concentration in the initial solution, but these large particles signifi cantly contributed to SAXS due to their size (see Fig. 3, b and c). The SAXS curves calculated from the fractional composition of each sample presented in Table 2 are in good agreement with the obtained anomalous signals (Fig. 5).
Multiphase ab initio modeling based on the SAXS data. According to the published data, the iron-binding sites are located not only in the inner cavity of the Dps protein, but also at the surface of the Dps dodecamer [28][29][30]52]. Although the nanoparticles do not form on the dodecamer surface, each monomer contains up to four amino acid residues capable of metal binding in the region of the N-terminal domains, so the bound iron atoms contribute to the scattering of the entire macromolecule and to the size distribution, in particular, due to their high electron density. The ab initio multiphase modeling (MONSA program) is a SAXS technique that allows to use the diff erence between the electron density of diff erent parts of the scattering object to localize some structural features associated with this diff erence [47]. The calculations are based on the diff erences between the electron density of the protein and metal components of the complex, as well as the ratio between the volumes of these components. In order to obtain a two-phase protein-metal model, we used the original SAXS curves of the iron-free and those of the iron-containing protein.
We selected the Dps-Fe500 specimen for the modeling, which provided a reliable detection of the formed nanoparticles in the protein inner cavity, but at the same time, allowed localization of the signifi cantly smaller metal-containing groups on the dodecamer surface on the background of strong scattering by the large particles. The results of modeling are presented in Fig. 6.
The multiphase modeling has allowed for the fi rst time to visualize the location of iron atoms in the pro-tein surface layer, and, at the same time, to confi rm the presence of most metal ions in the central cavity of the Dps dodecamer. We observed a good agreement between the experimental data and calculated curves for the bead models with χ 2 = 2.1. This generalized model provides the most comprehensive picture of formation of iron oxide nanoparticles in the ferritin-like protein that also protects bacterial genome.

CONCLUSIONS
Not all mini-ferritins protect bacterial genomes during oxidative stress, and not all nucleotide-associated proteins are capable of detoxifi cation and accumulation of iron atoms. Hence, the multifunctionality of the ferritin-like protein Dps is unique, and its consequence is the stable resistance of bacteria to drugs and antibiotics. Undoubtfully, the main feature of Dps, which attracts a considerable interest of researchers, is its ability for the in cellulo biocrystallization with DNA. The persistent microbial cell formed via this mechanism are resistant to numerous adverse environmental factors; they can retain their viability for a long time and give rise to a new population with preserved pathogenic properties under favorable conditions [53]. That is why DNA "archiving" by in cellulo co-crystallization with Dps requires attention and detailed investigation. However, it is equally important to study the ability of Dps for detoxifi cation and trans-formation of toxic Fe 2+ into non-toxic Fe 3+ ions followed by the accumulation of the trivalent iron in the protein inner cavity that is protected from the external exposure. These both functions of Dps are closely interrelated [30]. In our previous studies, we have investigated the formation of the Dps-DNA co-crystals, in particular, those formed with the involvement of divalent metal ions [37,38,50]. In this work, we studied the second feature of the functional properties of Dps, namely, the accumulation of iron ions in the inner cavity of this mini-ferritin. Using ASAXS technique, we were able to demonstrate that 2 to 4-nm nanoparticles were basically formed within the protein molecule, indicating that the growth of metal nanoparticles was limited by the spatial characteristics of the protein inner cavity. However, a certain number of iron ions was found in the Dps surface layer. This layer is very important for the protective functions of this protein, because the N-terminal domains located in this layer defi ne the type of interactions of Dps with DNA in solution, since these fl exible lysine-enriched N-terminal domains are responsible for the interaction with the negatively charged sugar-phosphate backbone of deoxyribonucleic acid [15,[22][23][24]. Therefore, the N-terminal domains must be accessible to bind DNA [50,51,53]. However, iron atoms could bind to the negatively charged amino acids (Asp and Glu) in the same N-terminal regions of Dps. Iron is a transition metal and can form stable complexes, for example, with the amino group nitrogen (N) in amino acids. Therefore, divalent iron ions could form bonds with the neighboring protein chains, pulling them towards each other and stabilizing the surface regions of the dodecamer. In this case, the N-terminal domains are pressed to the protein surface and the Dps-DNA crystal complexes are not formed [50,54]. However, in the presence of chelating agent (e.g., EDTA), the bonds between the iron ions and amino acids on the dodecamer surface are disrupted, which can result in the restoration of the Dps interactions with DNA without aff ecting protein inner regions where the nanoparticles have formed. Hence, the charge of the iron ions in the composition of Dps (on the surface and in its core) is of great importance and requires detailed investigation.
SAXS structural studies do not allow to determine the valence of the iron atoms. By defi nition, the function of ferritin-like proteins, including Dps, is oxidation of Fe 2+ to the trivalent state followed by accumulation of this non-toxic form in the protein central cavity. However, it was demonstrated that the mechanism of removal of the divalent iron from the solution and its oxidation at the respective protein sites is more complicated. In particular, the use of the Mössbauer spectroscopy for determination of the iron ion charge in Dps demonstrated that this charge was non-uniform. The protein contained iron in a form of magnetite FeO · Fe 2 O 3 , i.e., a mixture of diand trivalent iron ions [55]. The authors concluded that this composition of the Dps inorganic core supports its dodecamer structure.
In general, formation of nanoparticles that contain metal atoms, and not only iron, but, for example, cobalt and possibly others, in the Dps inner cavity is of certain practical interest. Hence, the studies carried out in this work can be useful not only in terms of studying the unique properties of Dps, but can also contribute to the development of solution-stabilized biocompatible nanocapsules with magnetic properties. The plasticity and structural stability of the Dps protein matrix provides this possibility [56]. A. Gruzinov, European Molecular Biology Laboratory (EMBL), for conducting SAXS experiments and valuable discussion.
Funding. This work was supported by the Russian Science Foundation (project no. 18-74-10071).
Ethics declarations. The authors declare no confl icts of interest in fi nancial or any other sphere. This article does not contain description of studies with human participants or animals performed by any of the authors.
Open access. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution, and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and in-dicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.