New insights on the biomineralisation process developing in human lungs around inhaled asbestos fibres

Once penetrated into the lungs of exposed people, asbestos induces an in vivo biomineralisation process that leads to the formation of a ferruginous coating embedding the fibres. The ensemble of the fibre and the coating is referred to as asbestos body and is believed to be responsible for the high toxicological outcome of asbestos. Lung tissue of two individuals subjected to prolonged occupational exposure to crocidolite asbestos was investigated using synchrotron radiation micro-probe tools. The distribution of K and of elements heavier than Fe (Zn, Cu, As, and Ba) in the asbestos bodies was observed for the first time. Elemental quantification, also reported for the first time, confirmed that the coating is highly enriched in Fe (~20% w/w), and x-ray absorption spectroscopy indicated that Fe is in the 3+ oxidation state and that it is present in the form of ferritin or hemosiderin. Comparison of the results obtained studying the asbestos bodies upon removing the biological tissue by chemical digestion and those embedded in histological sections, allowed unambiguously distinguishing the composition of the asbestos bodies, and understanding to what extent the digestion procedure altered their chemical composition. A speculative model is proposed to explain the observed distribution of Fe.


Results
At the optical microscope the AB appear as optically transparent fibres surrounded by a birifrangent golden-brown coat (Fig. 1c, and S1 in the Supporting Information (SI)). Secondary electron SEM images show that their overall diameter typically ranges between 1 and 4 μ m, and their average length falls in the range between 20 and 120 μ m. According to previous studies 7, 10 , both optical microscopy (OM) and SEM images confirmed the typical features of the AB: their coating was often segmented along the fibre length into spaced spherical or ellipsoidal units, and often knobbed at the extremities 7,16 . SEM micrographs also show the inner crocidolite fibre in parts of the AB where the coating is interrupted ( Figure S2 in the SI). The diameter of the fibres was measured to fall in the range between 0.3-0.4 μ m, in agreement with that reported for crocidolite fibres 18 . Several uncoated fibres, were also detected ( Fig. 1a and S1 in the SI). In the following, the abbreviation dAB will be used to refer to AB recovered from lung tissue after its chemical digestion and filtration on porous membranes, and hAB to the ones embedded in histological sections of the original lung tissue.
Elemental quantification. Semi-quantitative EDS microanalysis was performed for preliminary characterization and to support XRF elemental quantification. The concentration of Fe measured by EDS was similar to that calculated by XRF (~27% wt. EDS vs. ~23% w/w XRF). Nevertheless, with the exception of Fe, all other elements detected by XRF were not detected by EDS. This can be explained by the fact that XRF analysis can reveal the presence of buried matter thanks to the higher probing depth of the X-rays compared to electrons 17 (tens of μ m vs. few μ m, depending on the element considered and on the sample matrix). Spatially resolved elemental quantification with high lateral resolution was performed using XRF maps acquired on dAB at beamline ID18F (in air, at 14.4 keV), and maps acquired on dAB and hAB at beamline ID21 (in vacuum, at 7.3 keV). Representative XRF spectra acquired at the two beamlines are reported in Figure S3 in the SI. The elemental concentrations, reported in Table 1, were calculated by averaging the signal from pixels in selected areas of the AB, the lung tissue, or the background (see Methods and SI sections for more details). Since there were no significant differences between the results of Case A and Case B, elemental concentrations reported in Table 1 correspond to the weighted average of Cases A and Case B. The elemental composition of the hAB and dAB is compared with that of the lung tissue (TS), and with the ferritin (FR) and bovine liver (BL) references (Table 1). Elemental quantification of an empty membrane filtered with the same NaClO solution used to digest the lung tissue was performed to check for possible external contamination in the dAB (see spectrum Bkg2 in Figure S3 in the SI).
Iron levels in the dAB and hAB. XRF elemental quantification indicated that Fe is by far the most concentrated element in the AB, exceeding the concentration of the other elements by up to three orders of magnitude (Table 1). Iron was found to be highly enriched in both dAB and hAB with respect to the Fe detected in the surrounding lung tissue (TS) and in the biological reference sample (BL) (~300x and ~1000x, respectively), in line with the qualitative results reported by Pascolo et al. 17 . As can be seen in Fig. 2 and Table 1, in all the investigated AB a significant difference in the concentration of Fe in the inner and outer areas was observed (Fig. 3). This is particularly evident in the higher resolution maps reported in Fig. 4 Elemental quantification of dAB and standards was obtained from measurements performed at incident photon energy of 14.4 keV at ID18F, while that of hAB was performed on data acquired at 7.3 keV at ID21, therefore no information on Cu, Zn, and As is available for the latter. The errors represent the standard deviations on N XRF measurements (N = 9 for the dAB (5 for Case A, 4 for Case B), and N = 8 for the hAB and the TS (4 for Case A, and 4 for Case B). See SI for details about the estimation of the absolute uncertainties. "nd" stands for not detected or below the 0.001% w/w detection limit. a AB deposited on cellulose esters porous membranes upon filtering the chemically digested lung tissue; b AB in paraffin-embedded histological lung sections; c Horsespleen ferritin reference (F7879 from Sigma-Aldrich); d Bovine liver certified standard (SRM1577b from NIST); e average of the XRF signal of the pixels corresponding to the inner AB area; f average of the XRF signal of the pixels corresponding to the outer AB area; g measurements performed on areas of the histological sections without AB.  AB were similar to that measured for the ferritin standard (~14% w/w), and in agreement with the Fe concentration range reported for ferritin (10-30 wt.% ref. 19). This finding supports the presence of Fe-storage proteins with high Fe-loading, as was already proposed 7,17,20,21 . On the other hand, the inner part of the AB contains from two to almost three times Fe (~27 to ~39% w/w) compared to the outer part. Similar trend for the Fe concentration was found for both the dAB and the hAB ( Table 1). The average Fe concentration detected in the lung tissue (TS ~0.06% w/w) is of the same order of magnitude of that measured in the bovine liver reference (BL ~0.02% w/w), indicating that the tissue surrounding the AB is only slightly enriched in Fe with respect to the biological tissue reference. The concentration of Fe in the samples and references is summarized in the histogram shown in Fig. 2.
Trace elements. Copper, Zn, As, and Ba were detected in the dAB in the 0.02-0.3% w/w concentration range ( Table 1). The presence of Cu and Zn is not surprising because they are well known essential physiological metals in human and animal organisms, and are present in several metallo-proteins and enzymes 22 . Arsenic, on the other hand, is widespread in the environment, in particular as a contaminant in groundwater 23,24 , and moderate As-contamination is common in the region from where the samples originate 25,26 . On the other hand, Ba may originate from barite (BaSO 4 ), which is a common addictive in cement (the presence of Ba was previously reported on digested AB by means of ICP-MS analysis 22 ). The presence of Cu, Zn, and As in the hAB cannot be confirmed due to the lower excitation energy used (7.3 keV), which was due to the technical features of beamline ID21. Quantification was not performed for elements lighter than Fe because large shifts were observed between nominal and experimental concentrations of the calibration standard below 5 keV. In maps acquired at 7.3 keV on hAB, quantification of K, P, and Ca was prevented also because of significant impurities of the same elements detected in the polyethylene-naphtalate (PEN) membrane used to support the histological sections (see spectrum Bkg2 in Figure S3 in the SI). The maps acquired at 14.4 keV (Fig. 3, and S4 and S5 in the SI) show that the spatial distribution of Fe, Cu, Zn, and As clearly mimics the morphology of the dAB, indicating either that those elements were involved in the biomineralisation process, such as Fe, or that they were adsorbed on the AB at some stage, which is probably the case of Cu, Zn, and As. In some dAB, lower elemental concentrations were observed inside the spherical lobes at both ends of the AB, suggesting a hollow structure ( Figure S5 in the SI). Elemental distribution maps of dAB and hAB acquired at higher resolution at ID21 beamline further confirm how Fe precisely reproduces the AB features, including the symmetric segmented sections perpendicular to the longitudinal axis and the rounded lobes at the extremities (Fig. 4). In vacuum experimental setup allowed for the detection of lighter elements, such as Si, P, and S. Maps acquired on the dAB revealed the presence of P, K, and Ca in small spots randomly distributed on the porous membranes ( Figures S6 and S7 in the SI). In the hAB, those elements were instead observed co-localized with Fe ( Fig. 4, S8, and S9 in the SI). Sulphur was only detected in the lung tissue surrounding the hAB (Figures S8 and S9 in the SI), where it is rather uniformly distributed. The distribution of Si and Fe in areas where the coating is missing or interrupted reveals the embedded asbestos fibres (Fig. 4), in agreement with the elemental composition of crocidolite asbestos, which is mainly made of Si, but also contain ~27 wt.% of Fe. Accordingly to SEM images, higher resolution maps confirmed that the fibres are  Figure S4 in the SI. The maps were acquired in air at 14.4 keV, with a pixel size is 2.5 × 2.0 μ m 2 , and a dwell time of 1s. about ten times thinner than their coating. It is interesting to note that single fibres seem to favour the deposition of the Fe-coating from the host organism with respect to bundled fibres (Fig. 4). Silicon is more concentrated in the inner part of the AB, matching the embedded asbestos fibre, but is also rather uniformly distributed at lower concentration on the entire AB area ( Figures S6-S9 in the SI).

Elemental Distribution.
It is also interesting to note that the distribution of Ba differs from that of the other elements detected on the AB. Fluorescence maps acquired on dAB show that it is present in higher concentrations in the inner part of the AB, matching the area where the concentration of Fe is highest (Figs 4h, S6 and S7 in the SI). This is particularly evident in the map shown in Fig. 4h, in which the co-localization of Si, Fe, and Ba is highlighted by RGB colour combination.
X-ray Absorption Spectroscopy. X-ray absorption near edge structure (XANES) spectra of selected hAB and of several Fe reference compounds were acquired in an energy range between 7.0 and 7.3 keV, which includes the K absorption edge of Fe (7112 eV). Following the work of Wilke et al. 27 , after subtraction of an edge-shaped baseline, the pre-edge peaks of the XANES spectrum of a hAB were fitted with two pseudo-Voigt functions peaked at 7114.0 and 7115.1 eV, and a Gaussian function peaked at 7116.8 eV (Fig. 5, inset). The average energy position of the two lower energy peaks weighted for their areas is equal to 7114.6 eV, which, according to Wilke et al., indicates that Fe is in the 3+ oxidation state, compatibly with the presence of ferritin (or ferrihydrite). The unique features of the reference spectra ( Figure S10 in the SI) indicate that it is possible to unambiguously identify the Fe speciation in the AB by fingerprint analysis of the XANES spectra (i.e. by comparison with the Fe reference spectra). The spectral features of the hAB spectrum match very well with those of the ferritin standard spectrum (Fig. 5). Linear combination fitting (LCF) using the spectra of Fe(III)-compounds was attempted to improve the match with the hAB spectrum. The match did not improved indicating that contributions from other Fe compounds, such as elemental iron or hematite, which were previously claimed to contribute up to ~20% to the XANES spectra of the AB 17 , can be excluded, at least in amounts larger than 5% (it is also worth noting that the presence of 1% of metallic iron, claimed in the same study, cannot be confirmed because the error associated with LCF procedure is in the range between 5 and 15% [28][29][30] ). The XANES spectra of the hAB and of the ferritin standard are both very similar to the spectrum of the ferrihydrite standard ( Figure S10 in the SI). This is not surprising, considering that the ferritin protein stores Fe at its interior in the form of ferrihydrite, and that, being XANES only sensitive to the local structure around the absorber atom (Fe, in the present case), it cannot distinguish between ferritin and ferrihydrite. On the other hand, the lack of a significant (> 5%) contribution from the embedded crocidolite fibre, which was reported to contribute ~20% in a previous study 17 , can be explained by considering that the diameter of the fibre is about one-tenth that of its Fe-coating, and that the fluorescence signal scales accordingly.

Discussion
Pulmonary alveolar macrophages are able to convert Fe into forms that can be retained indefinitely in the tissues. The iron proteins specifically dedicated to iron storage in human and animal organisms are ferritin and hemosiderin. Ferritin has several functions: it uptakes Fe 2+ , catalyses its oxidation to Fe 3+ , and limits its bioavailability to cell constituents 31 . Theoretically, a single ferritin molecule can contain up to 4500 iron atoms (i.e. 27 wt. %) 19 . The protein component of Fe-free ferritin, apoferritin, has a hollow spherical protein shell of outer diameter 12-13 nm and inner diameter of 7-8 nm. The cavity communicates with the surface by eight channels, which Fe can enter and leave. Iron enters in the form of Fe 2+ and it is oxidized to Fe 3+ as it is transferred into the core, where nucleation of ferrihydrite takes place. For this reason, pioneering studies exploiting x-ray and electron diffraction techniques 7,32 identified the ferruginous coating as made of ferrihydrite. Hemosiderin, on the other hand, is formed by incomplete degradation of ferritin and conglomeration of iron and ferritin proteins, and differs from ferritin in having a higher iron-to-protein ratio. In addition, being less soluble in aqueous solutions, it represents a more stable and less available form of iron storage than ferritin. Under conditions of high Fe-excess, some of it may be stored in hemosiderin, whose formation may be favoured by the oxidative lung conditions 10,19 . Ferrihydrite, a poorly crystalline ferric oxide-hydroxide 33,34 , is abundant in a variety of aqueous geochemical environments and, because of its reactivity and large specific area, it has well known uptake ability towards As 35,36 and Se 37,38 , Cr 39 , and U 40,41 , Ra and Ba 42 . It has also been reported that, at physiological pH and under oxidizing conditions, as in the lungs, ferrihydrite could preferentially uptake large ionic radius species, such Ra and Ba, metals, such as Pb, Cd, and Zn, and semi-metals such As and Se 22 . Elemental distribution maps acquired on dAB (Figs 3, and S4-S5 in the SI) confirmed that they are efficient scavengers for metals (Cu and Zn), semi-metals (As), and rare earths (Ba), in agreement with the affinity of ferrihydrite toward these species. Although Ba has been reported as a trace element in some asbestos samples 43 , and although its presence could be explained by the barite (BaSO 4 ) used in production process of fibre-reinforced cement 44 , the lack of Ba in the hAB samples would suggest that it was introduced at some step of the digestion procedure (although Ba was not detected in the NaClO solution used to digest the lung tissue samples as can be seen from the corresponding XRF spectrum, Bkg1, in Figure S3 in the SI). Similarly, since the presence of Cu, Zn, and As in the hAB could not be confirmed, due to the lower excitation , is also shown to highlight the similarity of the two spectra. The hAB spectrum is the average of eight spectra acquired in different points of two different hAB. The inset shows the fit of the pre-edge peak performed using two pseudo-Voigt and one Gaussian functions, after subtraction of a baseline simulating the absorption edge. energy used (7.3 keV), it cannot be excluded that their presence could also be due to external contamination occurred during the digestion step, and to their subsequent enrichment on the AB during the filtration through the porous membranes. Further measurements on hAB at higher energies would help settling this point.
In dAB maps, K, P, and Ca were found in small agglomerates randomly distributed on the porous membranes, where residual lung tissue accumulated ( Figures S4-S7 in the SI), while in hAB maps those elements were found associated with the AB, co-localized with Fe. This would suggest that they were present as soluble species on the AB and then removed during the digestion with NaClO. While P and Ca have already been observed in the AB in previous works, claiming their active role in the biomineralization process 16,17 , the presence of K associated with the fibres' coating has never been directly observed before.
High resolution fluorescence maps (Fig. 4) show that Si is more concentrated in the inner part of the AB, revealing the inner asbestos fibre, but it is also spread at lower concentration over the entire AB area, suggesting the incipient dissolution of the fibre itself. The possible Degradation of asbestos fibres after long residence time in the lungs is a matter of a long standing debate. This subject is of interest because it is widely accepted that among the reasons for the high carcinogenicity of asbestos fibres there are, on one hand, their high biopersistence in biological tissues, and, on the other hand, the possibility that the elemental constituents of the fibres are released to the surrounding tissues upon degradation of the fibres itself 45 . The elemental distribution observed in this work is in agreement with recent observation by high resolution TEM of the formation of silica-rich amorphous coatings around asbestos tremolite-actinolite fibres 46 . The authors suggested that these coatings are associated with the dissolution of the amphiboles, and observed that silicon is the last to dissolve as the coatings progressively accumulates. They therefore conclude that solid state diffusion of Ca, Fe, and Mg ions out of the crystal lattice of the fibres results in a silica residue that eventually replaces the fibre by alteration.
Altogether, the observed distribution of Si and Fe suggests a possible model for the evolution of the AB during prolonged residence in the lung tissue. As demonstrated by in vitro studies 47,48 , alveolar macrophages start to deposit endogenous Fe on the asbestos fibres soon after their injection in the host tissue, and XANES results confirmed that Fe is deposited in the form of ferrihydrite, the mineral core of ferritin and hemosiderin. The concentration of Fe in the inner part of the AB is higher with respect to the more recently deposited external layer. This may be attributed to the gradual conversion of the primarily deposited and Fe-overloaded ferritin in the inner part of the AB, into hemosiderin with higher Fe content, and to subsequent precipitation of its ferrihydrite core (the presence of hemosiderin in the AB has been proposed already in 1965 49 , but its actual presence could not be confirmed due to difficulty of distinguishing between ferritin and hemosiderin). An alternative explanation for the observed distribution of Fe may arise from the release of exogenous Fe from the asbestos fibre itself, which would undergo to gradual dissolution, as the observed distribution of Si would suggest. The Fe distribution may also be determined by the combination of the above-mentioned mechanisms, one being a consequence of the other: Fe-excess originating from the dissolution of the fibre may induce the conversion of ferritin into hemosiderin. In the framework of this model, the observed peculiar distribution of Ba could be determined by possible differences in the uptake abilities of ferritin and hemosiderin. The model describing the observed data is depicted in the sketch shown in Fig. 6. It is worth noting that the model remains speculative, as it is not possible to confirm it in laboratory time-scale experiments.
The comparison of the results obtained on the dAB and on the hAB, allowed unambiguously distinguishing the composition of the AB from that of the lung tissue, and understanding to what extent the digestion procedure altered their chemical composition. In particular, such comparison suggested that species previously reported to be associated with the AB 22 (such as Ba), may have been introduced by external contamination during the digestion of the biological tissue, and subsequently migrated in the AB, which were proven to be excellent scavengers for elements having high affinity with ferrihydrite. Phosphorus, K, and Ca, on the other hand, were found to be associated with the AB coating, but were dissolved and removed by chemical digestion with NaClO. Finally, the comparison between the XANES spectra of the AB and of ferritin, indicated that hematite and metallic iron, whose presence in the AB was claimed in a previous study 17 , are absent or present in amounts well below 5%, and thus that the AB are mainly composed by ferritin and/or hemosiderin.

Samples.
Human lung samples were collected after forensic autopsy from two former workers of an asbestos plant in North-West Italy. Both cases were affected by pulmonary asbestosis; case A had also pleural mesothelioma, while case B had also lung cancer. The grade of asbestosis (Table 2), was established according to Craighead et al. 50 . Accordingly to the type of asbestos used in the plant, EDS microanalysis of uncoated parts of the fibres ( Figure S2 in the SI) allowed identifying them as crocidolite asbestos, which has ideal chemical formula NaFe 2 (Fe, Mg) 3 Si 8 O 22 (OH) 2 and contain ~27% Fe 18 . More detailed information on the nature of asbestos fibres from samples with the same origin can be found in a previous work 51 . Lung samples were preserved in formalin (10%) until non-neoplastic portions (0.25 g) of lung tissue were digested in 30 mL sodium hypochlorite solution (NaClO, reagent grade, chlorine content 10-15%, Merck) to dissolve the organic matrix. The suspension of inorganic material was filtered through mixed cellulose esters porous membranes (pore size 0.45 μ m, Millipore) to recover the AB. The membranes were then thoroughly washed with warm (40 °C) deionized water to dissolve NaCl crystals formed during digestion, and finally air-dried. SEM and light microscope images of these samples (dAB) are shown in Fig. 1, and S1 in the SI. Further portions of non-neoplastic lung tissues were collected to prepare 3 and 10 μ m-thick histological sections with a microtome. To avoid interference during XRF experiments and increase the probability to locate the AB, 10 μ m-thick non-stained histological sections were fixed on polyethylene-naphtalate (PEN) membranes (from MMI). Due to the extreme difficulty experienced in locating the AB embedded in the lung tissue, a laser micro-dissector (Nikon) coupled with an optical microscope was used to cut 100 μ m-diameter areas of the histological sections centred on the AB. Light microscope images of the AB in the cut sections are shown in Fig. 4a and d.
Three μ m-thick sections were embedded in paraffin, fixed on standard microscope slides, and then stained with haematoxylin and eosin, according to the standard protocol 52 , for histological examination. Lung samples were examined to estimate the number of AB by optical microscope (Leica DMLB) and SEM (Cambridge Stereoscan S360). According to the international standard 53 , the concentration of AB was expressed as their number per gram of dry weight (g dw ). This quantity was derived following the procedure described in Belluso et al. 54 : (i) the whole membrane was observed by optical microscopy at 400x magnifications, and (ii) a portion of the membrane representing 0.7% of its total area (about 2 mm 2 ) was observed by SEM at 2000x magnifications. The equivalent dry weight was calculated by dehydrating 2.5 g of wet tissue of each specimen at 60 °C for 3 days. In both specimens the burden of AB largely exceeded the amount established by the European Respiratory Society guidelines (10 3 /g dw ) to indicate a high level of occupational exposure to asbestos 53 (Table 2). Synchrotron experiments. Several AB from each human subject were analysed during each of the three synchrotron experiments performed at two different beamlines at the ESRF synchrotron facility (Table S3). Different experimental conditions (namely, in-vacuum or in-air acquisitions, 7.3 or 14.4 keV incident photon energy, and different beam spot-size, Table S3 in the SI) were dictated by the technical features of the beamlines, and allowed to obtain complementary information on the samples (i.e. elements lighter or heavier than Fe). dAB samples supported on cellulose esters membranes and hAB samples supported on PEN membranes were mounted on dedicated sample holders, covered with 4 μ m-thick Ultralene ® polymer film, and measured using solid state fluorescence detectors in 45°/45° geometry. Detailed information on the beamlines technical features and on the experimental setups is reported in the SI.
Elemental quantification. Semi-quantitative EDS measurements were performed with a Si(Li) detector (Link-Oxford Pentafet ATW2) coupled with the SEM, in vacuum (~10 −5 mbar) and at electron accelerating voltage of 15 kV. A Co standard was used to check the stability of the incident beam. The electron beam excitation area and penetration depth were estimated to be of ~1 μ m 2 and < 1 μ m, respectively. Absolute elemental quantification by XRF was performed on the fluorescence maps acquired on dAB at the ID18F beamline and on hAB at the ID21 beamline at the ESRF, at incident photon energies of 14.4 and 7.3 keV, respectively. Fluorescence signal of each detected element was fitted and deconvoluted using the PyMCA software package 55 , which also allows defining the absorbing matrix and taking into account absorbers between the sample and the detector (air, windows, filters, and others). Quantification was performed by manually selecting the pixels of the areas of interest (inner or outer AB area, whole AB area, and whole map area for the references), and averaging the fluorescence signals of each pixel. Great attention was paid to the choice of the calibration standards and to the definition of the absorbing matrix (see SI). Certified reference standards included SRM1832-1833 and bovine liver (SRM1577b) from NIST, and a reference standard composed by seven ultra-thin layers of different metals (Pd, Ca, La, Fe, Cu, Pb, and Mo) sputtered on a 0.2 μ m-thick silicon nitride membrane from AXO-Dresden, were used. A ferritin reference (F7879, Sigma-Aldrich) was prepared by dropping and drying a small amount of the ferritin solution on a Kapton film. The working energies and experimental conditions allowed to detect elements from Si to Br (K-edges), and from Ru to Pb (L-edges), but elemental quantification was considered reliable only for elements heavier than Ca. Further details on the quantification procedure (definition of the absorbing matrix, data processing, and analysis) can be found in the SI.
Elemental distribution. High resolution fluorescence maps (pixel size 0.5 × 0.5 μ m 2 ) were acquired on the dAB and hAB in vacuum (10 −5 mbar), at incident photon energy of 7.3 keV at the ID21 beamline at the ESRF. The excitation energy and the in vacuum setup allowed for the detection of elements from Na to Fe (K-edges) and from Rb to Nd (L-edges). Fluorescence maps acquired on dAB at ID18F have lower resolution (2.5 × 2.0 μ m 2 ), but allowed for the detection of elements heavier than Fe (K-edges from Co to Br).
X-ray Absorption Spectroscopy. XAS is a chemical selective tool informative of the oxidation state and of the chemical and crystallographic neighbourhood of a selected element 56 . Several XANES spectra were acquired at the Fe K-edge at the ID21 beamline on different points of selected hAB using beam spot sizes of 0.5 × 0.5 μ m 2 .
Reference XANES spectra of commercial horse spleen ferritin (Sigma-Aldrich) and of other relevant Fe commercial compounds (magnetite, hematite, hemin, haematin, and Fe(II)-L-ascorbate by Sigma-Aldrich, and Fe metallic foil by MaTeck) were measured along with the samples ( Figure S10 in the SI). The 2-lines ferrihydrite and crocidolite UICC standard reference spectra were kindly granted by Dr. A. Voegelin and Prof. A. Gualtieri, respectively. XANES spectra were energy calibrated with respect to the spectrum of an iron foil, and background subtracted and normalized using the IFEFFIT software package 57 . Further details on the Linear Combination Fitting (LCF) and pre-edge peaks analyses of the XANES spectra are reported in the SI.
All methods were carried out in accordance with relevant guidelines and regulations, and all experimental protocols were approved by the Bioethical committees of the Martini hospital (Turin, Italy) and of the University of Torino (Turin, Italy). Informed consent was obtained from the subjects at the moment of their hospitalization for generic research purposes.