Identifying the reactive sites of hydrogen peroxide decomposition and hydroxyl radical formation on chrysotile asbestos surfaces

Background Fibrous chrysotile has been the most commonly applied asbestos mineral in a range of technical applications. However, it is toxic and carcinogenic upon inhalation. The chemical reactivity of chrysotile fiber surfaces contributes to its adverse health effects by catalyzing the formation of highly reactive hydroxyl radicals (HO•) from H2O2. In this Haber-Weiss cycle, Fe on the fiber surface acts as a catalyst: Fe3+ decomposes H2O2 to reductants that reduce surface Fe3+ to Fe2+, which is back-oxidized by H2O2 (Fenton-oxidation) to yield HO•. Chrysotile contains three structural Fe species: ferrous and ferric octahedral Fe and ferric tetrahedral Fe (Fe3+tet). Also, external Fe may adsorb or precipitate onto fiber surfaces. The goal of this study was to identify the Fe species on chrysotile surfaces that catalyze H2O2 decomposition and HO• generation. Results We demonstrate that at the physiological pH 7.4 Fe3+tet on chrysotile surfaces substantially contributes to H2O2 decomposition and is the key structural Fe species catalyzing HO• generation. After depleting Fe from fiber surfaces, a remnant fiber-related H2O2 decomposition mode was identified, which may involve magnetite impurities, remnant Fe or substituted redox-active transition metals other than Fe. Fe (hydr)oxide precipitates on chrysotile surfaces also contributed to H2O2 decomposition, but were per mole Fe substantially less efficient than surface Fe3+tet. Fe added to chrysotile fibers increased HO• generation only when it became incorporated and tetrahedrally coordinated into vacancy sites in the Si layer. Conclusions Our results suggest that at the physiological pH 7.4, oxidative stress caused by chrysotile fibers largely results from radicals produced in the Haber-Weiss cycle that is catalyzed by Fe3+tet. The catalytic role of Fe3+tet in radical generation may also apply to other pathogenic silicates in which Fe3+tet is substituted, e.g. quartz, amphiboles and zeolites. However, even if these pathogenic minerals do not contain Fe, our results suggest that the mere presence of vacancy sites may pose a risk, as incorporation of external Fe into a tetrahedral coordination environment can lead to HO• generation.


Background
The term asbestos refers to a heterogeneous group of five fibrous amphiboles and one fibrous serpentine mineral (chrysotile) [1,2]. Due to its favorable properties such as a large tensile strength, heat resistance and noncombustibility, asbestos has been used in a variety of industrial applications [3], e.g. in thermal and electrical insulation, roofing, cement pipes and sheets, flooring and coatings [4,5]. However, respiratory exposure to asbestos minerals causes adverse health effects like pneumoconiosis, fibrosis of the lung, pleural plaques and effusions, carcinomas predominantly in the lung (but also in the larynx and ovaries) and mesotheliomas in the pleura and peritoneum [2,4,6,7]. Because of their carcinogenic potential, the WHO-IARC has classified all asbestos minerals as group 1 carcinogens [8]. More than 100,000 people die each year because of asbestos-related illnesses, mostly following occupational exposure [9]. Because of the intrinsic health hazard of asbestos, its use has been banned in European countries from the late 1980s onwards [10]. In northern American countries its use has not yet been banned [10] and in some Asian countries it even increases [11,12].
Chrysotile [Mg 3 Si 2 O 5 (OH) 4 ] accounts for more than 95% of all historically used asbestos [13]. As a result, exposure to asbestos predominantly concerns chrysotile. Therefore, we have focused on this mineral in this study. Chrysotile asbestos consists of octahedral Mg hydroxide layers and tetrahedral Si layers which bundle together to a fiber with a Mg hydroxide layer at the surface [14,15]. During petrogenesis, Fe is substituted into the crystal lattice (usually up to 2-4 wt%) [16]. Ferrous and ferric Fe are found in the Mg layers (Fe 3+ oct and Fe 2+ oct , respectively), whereas in the Si layers, exclusively ferric Fe is found (Fe 3+ tet ) [17,18]. Fe is by far the most abundant redox-active metal in chrysotile [16].
Weathering of chrysotile at circumneutral pH is commonly described as a layer-by-layer dissolution of alternating Mg and Si layers. Mg layers at the fiber surface dissolve within hours, whereas exposed Si layers dissolve much slower and therefore determine the overall dissolution rate [19,20]. However, enhanced dissolution of Fe 3+ tet from the Si layer by ligands like the siderophore desferrioxamine-B (DFOB) increases the Si dissolution rate, presumably through the formation of vacancy sites in the Si layer that labilize it [21].
Asbestos-induced pathologies can be linked to its high persistence in vivo [7,22,23], its fibrous morphology and the surface chemistry of the fibers [2,7]. Asbestos fibers lodged in lung or pleural tissue induce continuous, yet unsuccessful attempts of macrophages and neutrophils to phagocytose the fibers -a process called frustrated phagocytosis. During this process, enzymatically formed reactive oxygen species (ROS) like hydrogen peroxide (H 2 O 2 ) and superoxide anions (O 2

•-
) are released into the immediate extracellular environment [7]. Both exhibit a low potency for cellular damage under homeostasis [24] and can be enzymatically detoxified. At elevated concentrations H 2 O 2 and O 2 •may, however, interact with Fe on the fiber surface. This interaction induces cyclical redox reactions generating hydroxyl radicals (HO • ), which have a high potency to damage DNA, proteins and lipids [2,[24][25][26][27]. In this Haber-Weiss cycle, Fe acts as a catalyst: Fe 3+ is reduced by O 2 •to Fe 2+ , which is back-oxidized by H 2 O 2 in the so-called Fenton reaction, yielding Fe 3+ and HO • [2,28]. In the presence of Fe 3+ , H 2 O 2 may decompose to hydroperoxyl (HO 2 • ), which can either directly reduce Fe 3+ to Fe 2+ or decompose to the even stronger reductant, O 2 •- [29]. Despite the important role of H 2 O 2 and its degradation products in Fe redox cycling at the chrysotile fiber surface, H 2 O 2 decomposition by asbestos has only been assessed in a limited number of studies [30][31][32]. An involvement of the Fenton and Haber-Weiss pathways in H 2 O 2 decomposition by asbestos was demonstrated by Eberhardt et al. (1985) [30]. Furthermore,   [31] assessed H 2 O 2 decomposition for various Fe containing minerals. They found that H 2 O 2 decomposition rates by chrysotile and crocidolite were comparable, yet smaller than by magnetite and substantially larger than by hematite. H 2 O 2 decomposition by chrysotile partly occurs through Fenton reactions involving Fe surface species [2,25,33]. However, not all Fe surface species are equally Fenton-active or have an equal potential to form hydroxyl radicals.   [31] demonstrated that Fe 2+ oct on chrysotile surfaces does not play a substantial role in HO • generation. Recently, Walter et al. (2019) suggested that per mole Fe the potential to generate HO • is substantially larger for surface exposed Fe 3+ tet than for Fe oct [21]. Furthermore, Fe 3+ tet is the only Fe surface species in chrysotile that remains Fenton-active during long-term dissolution (weeks) at circumneutral pH, because the Si layer in which it is incorporated dissolves slowly, whereas Fe oct in the readily dissolving Mg layers rapidly precipitates to Fenton-inactive Fe (hydr)oxide minerals [21]. Depletion of all Fe surface species (including Fe 3+ tet ) from chrysotile surfaces by ligands like DFOB decreased the radical yield of the fibers, almost to background values [21,33]. Apart from structural Fe, also external Fe that associates with surfaces of asbestos (or other silicates) may generate ROS and increase oxidative stress in vivo and in vitro [2,[34][35][36][37].
To our knowledge, the relation between Fe speciation at chrysotile fiber surfaces and H 2 O 2 decomposition rates has not yet been established. Also, the relation between the speciation of external Fe after associating with the chrysotile fiber surface and the change in radical yield and H 2 O 2 decomposition rate of the fibers has not been explored previously. Hence, the current understanding of which Fe species at the chrysotile surface participate in the prerequisite step of the first stage (H 2 O 2 decomposition to reductants), and in the second stage (Fenton oxidation) of the Haber-Weiss cycle is incomplete. Establishing the reactive sites of H 2 O 2 decomposition and HO • generation on chrysotile surfaces is important in assessing the overall redox reactivity of chrysotile asbestos, which is a major determinant in its pathogenicity [25,38]. In this study we addressed this knowledge gap.
We hypothesize that H 2 O 2 is decomposed, either by structural Fe 3+ tet in exposed Si layers of the dissolving fibers, or by secondary Fe minerals precipitated on the fiber surface. The precipitated Fe may originate from external sources or from fiber dissolution during which structural Fe is released. Furthermore, we hypothesize that external Fe only substantially contributes to the HO • yield of chrysotile fibers when it becomes tetrahedrally coordinated by incorporation into a Si layer. The rationale for this hypothesis is the high potential of surface Fe 3+ tet for generating HO • [21], compared to the low potential of Fe (hydr)oxides [37] precipitated on chrysotile surfaces. Finally, we hypothesize that chrysotile fibers with surfaces depleted in Fe (e.g. due to preconditioning with a ligand) may still pose a health hazard if external Fe is incorporated into vacant surface sites in the Si layer.
The hypotheses were tested in batch incubation experiments. Samples were analyzed by ICP-OES (inductively coupled plasma optical emission spectrometry), UV-VISphotospectrometry, Mössbauer spectroscopy and EPR (electron paramagnetic resonance) spectroscopy.

Chemical reagents and asbestos characterization
All chemical reagents used in this study were at least pro analysis grade and were ordered from VWR (unless otherwise mentioned). Chrysotile asbestos was purchased from Shijiazhuang Mining IMP&EXP Trade Co, China. The material was characterized by XRD-Rietveld phase analysis, Raman spectroscopy, BET specific surface area measurement, Mössbauer spectroscopy, fusion digestion and neutron activation analysis [21]. The BET specific surface area (SSA) of Shijiazhuang chrysotile fibers was 20.3 m 2 g − 1 (with a standard deviation of 0.9 m 2 g − 1 , [21]), and phase impurities were established by XRD-Rietveld analysis: Shijiazhuang chrysotile contains 86.4 ± 4.6% chrysotile fibers, whereas phase impurities in the fiber material are brucite, talc, chlorite, magnetite, quartz and calcite [21]. Key results on the bulk of Shijiazhuang chrysotile are presented in Table 1: Shijiazhuang chrysotile asbestos contains ≈249 g kg − 1 Mg and ≈188 g kg − 1 Si; the stoichiometric Mg/Si ratio is close to 1.5. Fe (≈20 g kg − 1 ) and Al (≈8 g kg − 1 ) are the major substituents. Mössbauer analyses demonstrated that in pristine Shijiazhuang chrysotile asbestos, almost all Fe is substituted into the octahedral Mg layer (≈ 55% Fe 3+ oct and ≈ 38% Fe 2+ oct ), whereas only 7% is substituted into the tetrahedral Si layer (Table 1). Magnetite (1.5 ± 0.2% in Shijiazhuang chrysotile) hosts approximately 32% of the total bulk Fe ( Table 1).

Preparation of fiber suspensions
All experiments were carried out in fiber suspensions with a fiber to solution ratio of 1 g L − 1 . The non-metalcomplexing tertiary amine ("Better") buffer [39] MOPS (3-(N-morpholino)propanesulfonic-acid) was used at a concentration of 50 mmol L − 1 to maintain the pH of experimental solutions at 7.4 ± 0.3. The ionic strength of the buffer solutions was adjusted to 300 mmol L − 1 by addition of NaCl. Solutions in blank treatments contained only pH-buffer and electrolyte, while DFOB (Novartis) treatments additionally contained 1 mmol L − 1 DFOB. In H 2 O 2 decomposition experiments DFOB was used to quench the redox-activity of Fe. This method has been used previously, e.g. in refs [40,41]. Finally, H 2 O 2 decomposition was also studied in 0.1 mol L − 1 NaOH solutions in which chrysotile fibers are practically insoluble [21].

Preconditioning of chrysotile fibers
Fibers were preconditioned to obtain fiber types with different specific surface chemistry. The preconditioning involved incubation of the fibers in blank solutions  [21,37]. Moreover, in DFOB-altered fibers the Fe content of the dissolved outermost Mg layer as well as the Fe content of the slowly dissolving Si layer was complexed and mobilized by DFOB. Fe mobilization from the Si layer presumably leads to the formation of vacancy sites, which promote Si dissolution [21]. During preconditioning up to 4% of the fiber mass dissolved; assuming a cylindrical fiber geometry with constant length, this corresponds with a 2% decrease in SSA, which is smaller than the standard deviation on the BET-SSA analysis and was considered negligible.
To test whether external Fe can be incorporated into vacancy sites in the Si layer and whether this incorporated Fe participates in H 2 O 2 decomposition and HO • generation, DFOB-altered fibers were suspended in solutions buffered at pH 7.4 containing 0, 3, 30 and 300 μmol L − 1 of Fe 2+ under anoxic conditions in a N 2filled anoxic chamber (Brown box). The suspensions were then immediately oxygenated outside the anoxic chamber by air bubbling for 24 h, while magnetically stirring them at 500 rotations per minute. The Fe 2+ rapidly oxidized and Fe not incorporated into vacancy sites precipitated onto fiber surfaces as Fe (hydr)oxide minerals, coloring the fibers beige to yellow (see Fig. 1). As a negative control, the same concentrations of Fe were precipitated onto blank-altered fibers (which presumably lack vacancy sites in the Si layer) following the same procedure. The obtained altered fiber types are referred to as "DFOB-altered fibers + 0, 3, 30 or 300 μmol g − 1 Fe" and "blank-altered fibers + 0, 3, 30 or 300 μmol g − 1 Fe". Preconditioned fibers were collected in Büchner funnels on 0.47 μm Nylon membranes (Magna) and dried by vacuum filtration. To remove potentially adsorbed DFOB ligand or metal-DFOB complexes, fibers were washed with ultra-pure water and then vacuumdried and stored in an evacuated desiccator until they were used in follow-up experiments. Metal and Si concentrations mobilized during the fiber preparations are presented in Additional file 1: Table S1. 57 Fe addition and Mössbauer analyses 57 Fe Mössbauer spectroscopy was performed at room temperature in standard constant acceleration mode with a 57 CoRh source, relative to which all center shift data are given. The analyzed fiber types were DFOBaltered and blank-altered fibers + 0 μmol g − 1 Fe, and DFOB-altered and blank-altered fibers + 3 μmol g − 1 Fe. These fiber types were prepared following the procedure described above, except that isotopically enriched 57 Fe (Sigma Aldrich, > 95 atom % isotopic purity) was used. The isotopically enriched metallic 57 Fe-powder was dissolved over night at 70°C in a 2 mol L − 1 HCl solution, according to Arrigo et al. (2017) [42]. This procedure yielded a 57 Fe 2+ solution, which was purged with N 2 for 2 h and then put into the anoxic glove box. The isotopic composition of Fe in the stock solution was verified by ICP-MS ( 57 Fe accounted for 99.2% of the total Fe), and the Fe 2+ concentration was verified spectrophotometrically with a ferrozine assay [43]. Aliquots of the 57 Fe 2+ stock solution were added to DFOB-altered and blankaltered fiber suspensions to obtain an added concentration of 3 μmol g − 1 57 Fe 2+ .
After vacuum filtration and drying of the fibers, 700 mg of each fiber type were ground in a tungsten carbide ball mill (Resch Schwingmühle MM 400) for 30 s (a duration that does not affect Fe 2+ /Fe 3+ ratios in minerals [44]) at 30 strokes per minute in order to avoid spatial anisotropy of fibers in specimens. 500 mg of the milled fibers were pressed between Teflon foils (Zuma). Mössbauer measurements required up to 2 weeks per samples (Fig. 2). The spectroscopic data were analyzed by solving the full Hamiltonian. Thickness of the samples was taken into account after Mørup and Both (1975) [45]. A ferrihydrite sub-spectrum (based on data from Murad and Schwertmann, 1980, [46]) was used to account for Fe precipitation on blank-altered fibers + 0 μmol g − 1 57 Fe (precipitation of Fe from the dissolved Mg layer) and DFOB-altered and blank-altered fibers + 3 μmol g − 1 57 Fe (precipitation of added 57 Fe). Ferrihydrite was selected, because under the experimental conditions such a poorly crystalline Fe 3+ (hydr)oxide mineral is most likely to precipitate. Fits involving DFOB-altered fibers + 0 μmol g − 1 57 Fe were done with and without ferrihydrite sub-spectrum; including the ferrihydrite sub- Fig. 1 Preconditioned fibers. From left to right: DFOB-altered fibers with 0, 3 and 30 μmol g − 1 Fe, respectively, blank-altered fibers with 0 μmol g − 1 Fe and DFOB-altered fibers with 300 μmol g − 1 Fe spectrum did not significantly improve the fit. Because presumably precipitation of ferrihydrite was prevented by addition of DFOB, the fit without the ferrihydrite sub-spectrum was used for comparison with the other treatments. Each sample was measured two times: first in a wider velocity range (± 10.6 mm s − 1 ) to cover the full magnetically split-spectrum of magnetite impurities, which allowed to obtain the amount of magnetite in the samples, and second in a narrow velocity range (± 4.6 mm s − 1 ) to better resolve the chrysotile and Fe 3+ (hydr)oxide contributions. The obtained hyperfine parameters for both velocity ranges are presented in Additional file 1: Table S2, the spectra of the narrow velocity range are presented in Fig. 2 and the spectra of the wide velocity range in Additional file 1: Figure S1. The magnetite contents were calculated based on the wide velocity range data. After determination of the percentage of magnetite, the percentages of remaining Fe species were determined using the narrow velocity range data by multiplying the narrow velocity range percentages of these Fe species with (100% -magnetite% (wvr))/(100% -magnetite% (nvr)). Finally, all percentages were multiplied by the total amount of 57 Fe in each treatment. For blank-altered fibers the Fe content equaled the average content in pristine fibers measured by neutron activation analysis (NAA , Table 1), the amount of Fe removed by DFOB in DFOB-altered fibers was determined from the dissolved Fe concentration after reaction with DFOB. The amount of 57 Fe added was known.

Experimental procedure for H 2 O 2 decomposition experiments
In the H 2 O 2 decomposition experiments, metal mobilization from, and decomposition of H 2 O 2 by, differently preconditioned fibers were assessed. Included fiber types were: pristine fibers, blank-altered fibers, DFOB-altered fibers and both blank-altered and DFOBaltered fibers + 0, 3, 30 or 300 μmol g − 1 Fe. Experiments were carried out at pH 7.4 in MOPS buffer, and additionally for pristine and blank-altered fibers in a 0.1 mol  57 Fe. Spectra for blank-altered fibers and DFOB-altered fibers + 3 μmol g − 1 57 Fe were fitted with a ferrihydrite sub-spectrum to account for Fe (hydr)oxide precipitation L − 1 NaOH solution. The initial experimental H 2 O 2 concentration was 3.3 g L − 1 (≈ 0.3%), which was prepared by diluting a 30% stock solution (Sigma Aldrich, for trace analysis) a hundred times. The H 2 O 2 concentration of the stock was determined by redox titration with KMnO 4 : 334 ± 2 g L − 1 H 2 O 2 . Experiments were carried out in duplicates in 15 ml PP tubes (VWR) that were shaken in an end-over-end shaker at 15 rounds per minutes (RPM) at 20 ± 2°C in the dark. Samples were taken destructively after 0. 5 [47]. One ml of a 1.9-2.1% titanium (IV) oxysulfate solution (Sigma Aldrich) was added to 0.5 ml of the diluted filtrate and light absorption by the resulting peroxytitanyl-ion was measured at 410 nm by a Varian Cary 50 UV/VIS spectrophotometer (ɛ = 689 L mol − 1 cm − 1 ). H 2 O 2 concentrations in the samples were quantified by an external linear calibration method (7 to 42 mg L − 1 H 2 O 2 ); filtrates were diluted down to fit the calibration range. Because H 2 O 2 also reacts with MOPS buffer [48], a control treatment to determine the H 2 O 2 decomposition rate in absence of fibers was also included. Also for experiments in 0.1 mol L − 1 NaOH a control treatment without fibers was included. In an additional experiment, H 2 O 2 decomposition by pristine, blank-altered and DFOB-altered fibers was examined at pH 7.4 in the presence of 1 mmol L − 1 DFOB using the same experimental procedure. The absorption maximum of the FeDFOB complex (425 nm; ɛ = 2460 L mol − 1 cm − 1 , [49]) and the peroxytitanyl-ion (vide supra) [47,50] are in close proximity. However, FeDFOB concentrations were orders of magnitude smaller and the molar absorption coefficients of the complexes are less than one order of magnitude different. Therefore, the contribution of FeDFOB to overall light absorption at 410 nm could be neglected.

EPR spin trapping analyses of hydroxyl radicals generated by Fe on chrysotile fibers surfaces
The HO • yield of fiber specimens in the presence of H 2 O 2 was quantified with 5-5-dimethyl-1-pyrroline N-oxide (DMPO) as spin trapping agent using a X-band EPRspectrometer (Bruker EMX) and a split ring resonator (Bruker MD5). This spin trapping technique has frequently been used for this purpose before [26,31,37,51,52]. Eleven mg of fibers were incubated for 0.5 h in 0.5 ml of a 125 mmol L − 1 H 2 O 2 and 12.5 mmol L − 1 DMPO solution buffered at pH 7.3 with a 250 mmol L − 1 chelextreated phosphate buffer. After 25 min of incubation at room temperature and 5 min of centrifugation (14,000 RPM), 50 μl of the supernatant were pipetted into a glass capillary (intraMark Blaubrand), which was then sealed with Critoseal. Subsequently, the capillary was transferred into the resonator. The instrumental settings for the EPR measurements are described in Walter et al. (2019) [21]. EPR measurements were performed on four subsamples from each type of preconditioned fibers (quadruplicates). To quantify the change in HO • yield, the signal intensity (Intensity peak-to-peak (Ipp)) of the second peak from the left in the DMPO/HO • quadruplet of altered fibers was determined and expressed as a percentage of the Ipp of pristine fibers, which was measured as a reference in each measurement session. For comparison, also the HO • yield of the poorly crystalline Fe (oxy)hydroxide 2-line ferrihydrite (3 ± 0.2 mg, synthesized according to Schwertmann and Cornell (2000), [53]) was measured following the same procedure. An amorphous Fe (hydr)oxide like 2-line ferrihydrite may precipitate upon Fe addition to the fibers and subsequent oxygenation [53,54].

Statistical analysis and supplementary data
Statistical analysis of the EPR spin trapping data was performed with the program SPSS Version 25. A square root transformation of the data was carried out to reduce skewness. Homogeneity of the transformed data was tested with the Levene's test (α = 0.05). Differences among treatments were established by applying the univariate general linear model procedure and the Tukey post-hoc test (α = 0.05). A statistical test was employed to answer a) if the HO • yield increased with the amount of Fe applied to DFOB-altered fibers and b) if, through addition of Fe to DFOB-altered fibers, the HO • yield of blank-altered fibers could be reached.
The data included in Figs. 2, 3, 4 and 5 are reported in Additional file 1: Table S2 to Table S5, respectively. The p-values from the statistical analyses of the EPR data are reported in Additional file 1: Table S6.

Results
Color changes related to Fe at chrysotile surfaces Complexation and mobilization of Fe from the beige pristine chrysotile fibers by DFOB resulted in the whitish color of DFOB-altered fibers (Fig. 1). Interaction of DFOB-altered fibers with 3 μmol g − 1 Fe changed the whitish color to greyish after exposure to oxygen; interaction with 30 μmol g − 1 Fe reversed the color to beige, comparable to the color of pristine (not shown) and blank-altered fibers, and interaction with 300 μmol g − 1 Fe changed the color to yellow (Fig. 1). Interaction of pristine fibers with 0, 3 and 30 μmol g − 1 Fe did not lead to a clear change in the beige fiber color of blank-altered fibers, whereas addition of 300 μmol g − 1 Fe again changed the color of the fibers to yellow (Additional file 1: Figure S2).

Coordination environment of 57 Fe after interaction with chrysotile surfaces
The contribution of the subspectra to the overall Mössbauer spectrum differed per fiber type ( Fig. 2 Table S2c) and lowered the Fe 3+ tet content by half, from 15.0 to 7.7 μmol g − 1 Fe (Additional file 1: Table S7). As only 4% of the fiber mass had dissolved during preconditioning, this suggests that Fe 3+ tet sites are not homogenously distributed over the chrysotile bulk, but are strongly enriched in Si layers near the fiber surface. Addition of 3 μmol of 57 Fe per gram of DFOB-altered fibers increased the 57 Fe 3+ tet fraction ( Fig. 2 panel a and c) from 2.2 to 3.8% (Additional file 1: Table S2c). This corresponds with an increase in total Fe 3+ tet bulk content by 3% (Additional file 1: Table  S7), suggesting a marginal recovery of Fe 3+ tet sites at chrysotile surfaces by addition of 57 Fe. Most added 57 Fe (1.8 μmol g − 1 Fe ≈ 60%) was retrieved as Fe 3+ oct . In contrast, addition of 3 μmol of 57 Fe per gram of blankaltered fibers resulted in a decrease in 57 Fe 3+ tet fraction (Fig. 2 panel b and d) from 4.0 to 3.3% (Additional file 1: Table S2c) and a negligible change (0.3%) in total Fe 3+ tet bulk content (14.9 μmol g − 1 Fe, Additional file 1: Table  S7). Most added 57 Fe (1.8 μmol g − 1 Fe ≈ 60%) was retrieved as magnetite.

Dissolution of metals and Si from pristine and preconditioned chrysotile fibers in the presence of H 2 O 2
In Fig. 3, Mg and Si concentrations mobilized from pristine and preconditioned chrysotile fibers in the presence of H 2 O 2 (initial concentration: 3.34 g L − 1 ) are reported as a function of time. Fe concentrations were in the submicromolar range throughout all these experiments (data no shown). In contrast to our results, Ghio et al. (1998) reported somewhat higher mobilized Fe concentrations from a different chrysotile preparation by H 2 O 2 [32].
For Mg mobilization from pristine chrysotile fibers, two stages could be distinguished (Fig. 3, Panel a1): a fast first stage during which the outermost Mg layer of the fibers rapidly dissolved (first few days), and a slower second stage during which the outermost Mg layer had been completely dissolved, Si dissolution had become rate limiting and mobilized Mg concentrations reached a plateau at approximately 500 to 600 μmol L − 1 (between 96 and 336 h). Mg mobilization from DFOB-altered and blank-altered fibers could not be divided into two dissolution stages, but increased linearly throughout the experiment. Si concentrations mobilized from all three fiber types increased linearly throughout the experiment (Fig. 3, Panel a2). Mobilized Si concentrations were consistently higher for DFOB-altered fibers than for blankaltered and pristine fibers. Adsorption and precipitation of added Fe onto DFOB-altered fiber surfaces decreased the rate of Mg (Fig. 3, Panel b1) and Si (Fig. 3, Panel b2) mobilization throughout the experiment. The decrease in Mg and Si dissolution rates were however not proportional to the amount of Fe applied and reached a maximum of approximately 50% with addition of 30 μmol g − 1 Fe. Adsorption and precipitation of added Fe onto blank-altered fiber surfaces did not decrease Mg and Si mobilization as strongly as for DFOB-altered fibers (Fig.  3, Panel c1 and in which k tot is the overall decomposition constant. Chrysotile fibers accelerated H 2 O 2 decomposition relative to the MOPS-buffer control treatment by a factor 2 to 5, depending on the pretreatment (Fig. 4a, Table 2). H 2 O 2 decomposition was fastest with pristine fibers and slowest with DFOB-altered fibers. The addition of DFOB as a redox quencher for Fe largely inhibited differences in the H 2 O 2 decomposition rate between pristine, blank-altered and DFOB-altered fibers (Fig. 4b).
In the treatment with DFOB-altered fibers, the application of DFOB as redox quencher had no effect on H 2 O 2 decomposition; for the treatments with pristine and blank-altered fibers, H 2 O 2 decomposition decreased as a result of DFOB addition ( Fig. 4a and b). For the treatments with DFOB-altered fibers and DFOB-altered fibers + 0 μmol g − 1 Fe, k tot values did not differ (2.5*10 − 3 h − 1 ), demonstrating that the fiber preparation procedure without Fe addition did not affect the H 2 O 2 decomposition rate. Fe addition to DFOBaltered fibers increased k tot values by up to over a factor 2 in the DFOB-altered fibers + 300 μmol g − 1 Fe (5.3*10 − 3 h − 1 ); the increase in k tot was non-proportional to the amount of Fe added (Fig. 4c, Table 2). A non-proportional increase in k tot values was also found for treatments in which Fe had been added to blank-altered fibers. The relative increase was however smaller, amounting a factor 1.4 (from 4.2*10 − 3 h − 1 to 6.0*10 − 3 h − 1 ; Fig. 4d, Table 2). The contributions from different reactive sites on chrysotile surfaces to overall H 2 O 2 decomposition can be estimated in a tiered approach (Table 3), under the assumption that the various degradation mechanisms are independent, and their decomposition constants add up to the k tot of the reaction. Equation 1 can then be rewritten to equation 2: in which k 1 to n represent the contributions of the individual H 2 O 2 decomposition pathways to the overall decomposition constant k tot . In addition to contributions from tetrahedral Fe and Fe (hydr)oxide precipitates, the difference in decomposition rate between the MOPS buffer control and the DFOB-altered fiber treatment suggests a contribution from a remnant H 2 O 2 decomposition pathway (Fig. 4b, Table 3 [21]), that this Fe largely precipitated in the blank treatment as only a small fraction of the Fe is located in the slowly dissolving Si layer (Table 1), and that precipitation of an additional 30 μmol g − 1 Fe had the same effect size on the k tot value as the Fe that precipitated from the outer layer. The contribution of Fe (hydr)oxide precipitates to the k tot value can then be calculated by subtracting the k tot value of the blank-altered treatment from the blank-altered + 30 μmol g − 1 Fe treatment. Finally, the contribution from tetrahedral Fe was calculated by subtracting the contributions from the MOPS buffer, Fe (hydr)oxide precipitates and the H 2 O 2 decomposition pathway from the k tot value of the blank treatment (Table 3). Following this approach, the k tot value of the blankaltered fiber treatment (4.2*10 − 3 h − 1 ; Table 2, treatment 2) was broken down to contributions from the three types of active surface sites and the MOPS buffer (equation 2). The contributions of the active surface sites to  Table 3 Contributions from different reactive surface sites and the MOPS buffer to the overall H 2 O 2 decomposition rate constant (k tot ) for the blank-altered fiber treatment. The fitted constants (k (Exp.x)) for the treatments reported in Table 2 were used and linear additivity was assumed k tot were comparable, varying within a factor 2, and also the contribution from the MOPS buffer fell within this range ( Table 3). The solution pH had a strong effect on the H 2 O 2 decomposition rate: in 0.1 mol L − 1 NaOH (pH 12-13) the decomposition rate by pristine and preconditioned fibers was approximately an order of magnitude faster than at pH 7.4 ( Table 2).
Effect of Fe addition to preconditioned chrysotile fibers on HO • generation Pretreatment of Shijiazhuang chrysotile asbestos decreased the HO • yield relative to pristine fibers to 50 ± 10% for blank-altered fibers and to 9% for DFOB-altered fibers [21]. The HO • yield of blank-altered and DFOBaltered fibers + 0 μmol g − 1 Fe (Fig. 5) corresponded with these values. For all treatments with Fe addition to DFOB-altered fibers, the HO • yield was larger than for the + 0 μmol g − 1 Fe treatment. The HO • yield increased non-proportionally with the amount of Fe added, from 7% (+ 0 μmol g − 1 Fe) to 36% (+ 300 μmol g − 1 Fe) ( Fig.  5a; Additional file 1: Table S5). Although a factor 10 more Fe had been added in the DFOB-altered + 300 μmol g − 1 Fe treatment than was extracted in the DFOB-pretreatment, the HO • yield remained lower than in the blank-altered + 0 μmol g − 1 Fe fiber treatment ( Fig.  5; Additional file 1: Table S5), suggesting that the HO • yield could be largely, but not fully recovered. The addition of Fe to blank-altered fibers did not consistently increase the HO • yield of chrysotile (Fig. 5b). Furthermore, the HO • yield of 3 mg of 2-line ferrihydrite was 11% (relative to the HO • yield of 11 mg pristine chrysotile fibers). The total amount of Fe in 3 mg of 2-line ferrihydrite (≈1.8 mg Fe) is a thousand times larger than the 1.8 μg Fe on the fiber surface of the aliquots of DFOB-altered fibers + 3 μmol g − 1 Fe. Despite this large difference, the increase in HO • yield (an increase from 7 to 17%) due to the 3 μmol g − 1 Fe addition was comparable with the overall HO • yield of 3 mg of ferrihydrite (11%).

Speciation of added Fe and implications for fiber dissolution
Si dissolution from DFOB-altered fibers was over a factor three faster than from pristine fibers, whereas Si dissolution from blank-altered and pristine fibers were comparably fast (Fig. 3, Panel a2). The faster Si mobilization from DFOB-altered fibers is a consequence of the complexation of Fe 3+ tet by DFOB during pretreatment. Presumably this led to the formation of vacancy sites in the Si layer resulting in Si labilization which enhanced Si dissolution rates [21]. Si mobilization from blank-altered fibers was considerably slower, because no Fe 3+ tet had been removed from the Si layers during pretreatment. The larger Mg mobilization rate from DFOBaltered fibers compared to blank-altered fibers presumably resulted from the larger rate-controlling Si mobilization rate, allowing segments of deeper Mg layers to dissolve more rapidly; in both treatments the outer Mg layer had been dissolved during pretreatment.
Mössbauer spectroscopy analyses of DFOB-altered and blank-altered fibers + 3 μmol g − 1 57 Fe demonstrated that the absolute increase in tetrahedrally coordinated 57 Fe content was more than 5 times larger when added to DFOB-altered fibers compared to blank-altered fibers (Additional file 1: Table S7). However, assuming that no isotope exchange occurred, the data imply that only a small fraction of the Fe 3+ tet sites depleted by DFOB were recovered by 57 Fe additions.
Despite the apparently low recovery of depleted vacancy sites as observed by Mössbauer spectroscopy, the interaction of Fe with DFOB-altered fibers re-stabilized the labilized Si layer, reduced the Si dissolution rate, and consequently also reduced the Mg dissolution rate (Fig.  3, panel b1 and b2). The 25% reduction in Si and Mg dissolution rate by addition of only 3 μmol g − 1 Fe and the fact that Fe addition beyond 30 μmol g − 1 did not lead to a further decrease in dissolution rates, supports that the effect of Fe addition originates from the stabilization of the Si layer rather than from surface coverage by precipitated Fe (hydr)oxide minerals that prevent dissolution. The latter observation also suggests that between addition of 3 and 30 μmol g − 1 Fe, all vacancy sites became occupied with tetrahedrally coordinated Fe and further Fe addition did not affect dissolution rates. The absence of similar trends in Si and Mg dissolution for Fe addition to blank-altered fibers further indicates that external Fe only becomes tetrahedrally coordinated if there are vacancy sites present in the surface Si layer (Fig. 3, Panel c1 and c2).

Active sites of H 2 O 2 decomposition on chrysotile surfaces
At pH 7.4, the H 2 O 2 decomposition rate (Fig. 4a) (as well as the HO • yield (Fig. 5)) was highest in the treatment with pristine fibers. This is presumably related to a (transient) contribution from Fe in the outermost Mg layer, which dissolves within a few days at this pH. In the treatment with NaOH the Mg layer did not dissolve at all (Additional file 1: Table S8) and the lasting contribution from Fe in this layer may in part explain the higher H 2 O 2 decomposition rate.
In addition to two Fe-related modes of H 2 O 2 decomposition by chrysotile, a third, remnant mode was identified ( Fig. 4a and b), which, to our knowledge, had not yet been described for asbestos. It made the largest contribution to the k tot of blank-altered fibers in our experiments (Table 3) and may also be relevant in vivo.
Magnetite impurities in the Shijiazhuang chrysotile that do not dissolve during the DFOB pretreatment may contribute to the remnant H 2 O 2 decomposition mode. H 2 O 2 is more rapidly decomposed by magnetite than by asbestos per unit of mass [31], but magnetite is only a phase contaminant in Shijiazhuang chrysotile asbestos (1.5 ± 0.2%), whereas chrysotile is the predominant phase (86.4 ± 4.6%) [21]. Therefore, we assume that the contribution of magnetite to H 2 O 2 decomposition rates is small. Furthermore, other substituted metal ions (e.g. Cr, Mn, Ni) that are not or only slowly mobilized by DFOB might have contributed to the remnant H 2 O 2 decomposition mode. And finally, the contribution from small amounts of remnant Fe that were either not mobilized by DFOB during the pretreatment or that became exposed during the H 2 O 2 decomposition experiments as a result of Mg and Si dissolution is counted towards the remnant decomposition mode.
In spite of the smaller surface concentration of Fe 3+ tet in blank-altered fibers relative to octahedral Fe which had precipitated as Fe (hydr)oxide minerals, their contributions to H 2 O 2 decomposition were comparable (Table  3). Several factors may contribute to the comparatively large contribution of Fe 3+ tet per mole Fe. First, only a fraction of the Fe in Fe precipitates resides at the mineral surface and is able to react with H 2 O 2 , whereas all tetrahedral Fe substituted into the exposed Si layer can contribute to H 2 O 2 decomposition. Secondly, in other silicate minerals like nontronites, it has been shown that Fe 3+ tet is preferentially reduced over octahedral Fe [55][56][57] suggesting a lower redox potential of Fe 3+ tet in silicate minerals. This lower redox potential may contribute to a higher reactivity of Fe 3+ tet with regard to H 2 O 2 decomposition. Also for Fe (hydr)oxide minerals, it has been demonstrated that for equal masses the H 2 O 2 decomposition rates were larger for minerals containing Fe 3+ tet , like magnetite (even higher than chrysotile), than for minerals that do not contain Fe 3+ tet , like hematite [31].
Similarly to H 2 O 2 decomposition, a much higher reactivity with respect to the HO • yield was observed for tetrahedral Fe than for octahedral Fe precipitates. The reason for the high redox reactivity of Fe 3+ tet in silicates (and potentially Fe (hydr)oxide minerals) has, to our knowledge, not yet been examined.

Active sites of HO • generation by structural and external Fe on chrysotile surfaces
Contrary to Fe addition to blank-altered fibers, Fe addition to DFOB-altered fibers clearly increased HO • generation by chrysotile (Fig. 5a). Since DFOB-altered fibers were significantly depleted in Fe 3+ tet sites, this suggests that upon Fe addition, Fe was incorporated into vacancy sites in the Si layer of DFOB-altered fiber surfaces where it became tetrahedrally coordinated and particularly active in HO • generation. It should be noted, however, that Mössbauer data seem to indicate that addition of 3 μmol g − 1 57 Fe only resulted in a small increase of Fe 3+ tet ; only 3.0% of the Fe 3+ tet removed in the DFOB pretreatment was recovered through Fe addition. HO • generation, however, recovered to a substantially larger extent by 3 μmol g − 1 Fe addition: 10 percentage points relative to untreated fibers, corresponding to 19% of the difference between the blank-altered + 0 μmol g − 1 Fe treatment (with the pristine Fe 3+ tet content) and the DFOB-altered + 0 μmol g − 1 Fe treatment (with Fe 3+ tet mostly depleted).
The limited recovery of Fe 3+ tet by 3 μmol g − 1 57 Fe addition may be related to the observed enhanced Si dissolution, possibly creating vacancy sites in the Si layer and subsequent edge pit formation. Edge pit formation would make the sites unsuitable for accommodating Fe 3+ tet coordination when the 57 Fe was added. However, the factor six discrepancy between recovered Fe 3+ tet and recovered HO • generation seems to indicate that recovered Fe 3+ tet sites may be underestimated by Mössbauer data. Indeed, the recovery of Fe 3+ tet sites was calculated under the assumption that no Fe-isotope exchange occurred in these sites over the timescales of the experiment. Considering the significant time gap between 57 Fe addition and Mössbauer spectroscopy, it is conceivable that isotope exchange reactions did occur over the timescales of the experiment. In this case, we may have under-estimated the increase of Fe 3+ tet sites after addition of 57 Fe. Finally, the reactivity of Fe 3+ tet sites regarding HO • generation may be heterogeneous as a result of differences in local coordination environment and the recovery of such sites may not be linearly related to the recovery of reactivity.
Blank-altered fibers do not have vacancy sites in the Si layer, and therefore addition of Fe did not lead to a clear increase in HO • yield. However, the HO • yield of the blank-altered + 0 μmol g − 1 Fe treatment, in which surface Fe 3+ tet was preserved, was still higher than the HO • yield of fibers from the DFOB-altered + 300 μmol g − 1 Fe treatment. This suggests fewer exposed Fe 3+ tet surface sites in the latter treatment, potentially as a result of a loss of vacancy sites due to ongoing dissolution of the Si layer.
Assuming that Fe addition to blank-altered fibers mainly lead to precipitation of Fe (hydr)oxide minerals, the lack of differences in HO • yield between blankaltered fiber treatments with different amounts of added Fe suggests that these Fe (hydr)oxide minerals do not contribute to HO • generation. This corresponds with results from previous studies: the HO • yield of hematite, which contains no Fe 3+ tet [54], was below the LOD in a study by   [37], while the HO • yield of magnetite, which does contain structural Fe 3+ tet [54], corresponded with 60% of the HO • yield of chrysotile asbestos on a per mass basis [37]. The difference in reactivity between Fe 3+ tet and octahedral Fe was larger for HO • generation than for H 2 O 2 decomposition.
The high Fenton reactivity of Fe 3+ tet in chrysotile may, analogously to H 2 O 2 decomposition, be explained by the lower redox potential of Fe 3+ tet compared to octahedral Fe, as observed in nontronites [55][56][57], and a potentially rapid back-oxidation of the Fenton-active Fe 2+ tet to Fe 3+ tet by H 2 O 2 , yielding HO • . In contrast to Mg and Si mobilization and H 2 O 2 decomposition, addition of 30 μmol g − 1 Fe to DFOB-altered fibers did not recover the HO • yield to the level of blank-altered fibers. For DFOB-altered fibers + 300 μmol g − 1 Fe the HO • yield (37 ± 14%) was still significantly lower than for blankaltered fiber + 0 μmol L − 1 treatment (58 ± 6%) (Fig. 5, Additional file 1: Table S5). This incomplete recovery of the Fenton reactivity when adding an excess of Fe may suggest a loss of vacancy sites during preconditioning e.g. due to edge pit formation, leading to a smaller number of Fe 3+ tet surface sites than in the blank-altered fiber treatment.

Conclusions
The results from this study demonstrate that both Fe 3+ oct in Fe (hydr)oxide precipitates and Fe 3+ tet contribute to H 2 O 2 decomposition by chrysotile asbestos; for asbestos fibers incubated at pH 7.4 in absence of a ligand (blank-altered) the contributions of both Fe species were comparable (within a factor 1.5), despite the excess of octahedral sites. A remnant mode of H 2 O 2 decomposition by chrysotile was identified, which may be related to magnetite impurities, redox active substituted trace metals not removed by DFOB during pretreatment and remnant Fe. HO • generation by chrysotile asbestos is likely governed by Fe 3+ tet ; the contribution from Fe precipitates is negligible.
The occurrence of Fe 3+ tet in Fe (hydr)oxide minerals may also be correlated with their HO • yield and their H 2 O 2 decomposition capacity. However, whereas Fe (hydr)oxide minerals are not pathogenic [58], many silicate minerals other than chrysotile are. In many pathogenic silicates Fe 3+ tet has been detected, e.g. in quartz, in amphiboles and in zeolites [59][60][61][62][63][64]. Even if these minerals do not contain Fe, our results demonstrate that the presence of vacancy sites in their Si lattice can pose a risk, because incorporation of external Fe into the tetrahedral coordination environment can lead to HO • generation. This may be particularly relevant for zeolites (e.g. erionite), which often have a non-detectable bulk Fe content, but a higher potential to induce mesothelioma than asbestos [2]. The dissolution of tetrahedral Al (which is a stoichiometric constituent of framework silicates) may create abundant vacancy sites in the Si lattice of zeolite fibers, available for the incorporation of Fenton-active tetrahedrally coordinated Fe.
To conclude, our results suggest that Fe 3+ tet governs HO • generation by chrysotile at circumneutral pH, and that Fe 3+ tet may also contribute to the hazard of other pathogenic silicates.
Additional file1: Figure S1. Wide velocity range Mössbauer spectra of DFOB-altered fibers (Panel a and c) and blank-altered fibers (Panel b and d) with 0 or 3 μmol g − 1 added 57 Fe. Figure S2. Preconditioned fibers. Panel a.) From left to right: blank-altered fibers + 0 μmol g − 1 Fe, blankaltered fibers + 3 μmol g − 1 Fe, blank-altered fibers + 30 μmol g − 1 Fe and blank-altered fibers + 300 μmol g − 1 Fe; Panel b.) Fiber preparation for Mössbauer analyses, from left to right: blank-altered fibers + 3 μmol g − 1 57 Fe, blank-altered fibers + 0 μmol g − 1 57 Fe, DFOB-altered fibers + 0 μmol g − 1 57 Fe, blank-altered fibers + 3 μmol g − 1 57 Fe. Table S1. Mobilized Mg, Si and Fe concentrations in μmol L − 1 during pretreatment (no duplicates available). Table S2. Mössbauer hyperfine parameters of DFOB-altered and blank-altered fibers + 0 or 3 μmol g − 1 57 Fe, analyzed in the narrow (Table a) and the wide (Table b) velocity range. The Fe species distributions were calculated from both the wide and the narrow velocity range data combined (see materials and methods) and are presented in Table  c. Table S3. Mobilized Mg and Si concentrations from 1 g L − 1 pristine, DFOB-altered and blank-altered fibers incubated at pH 7.4 (50 mM MOPS) with addition of 3.34 g L − 1 H 2 O 2 . Table S4. Residual H 2 O 2 concentrations during H 2 O 2 decomposition by pristine fibers, DFOB-altered fibers, blankaltered fibers and the MOPS buffer as a function of time. Table S5. HO • yield of DFOB-altered and blank-altered fibers relative to pristine fibers (i.e. 100%). Table S6. Results from the statistical analysis of the EPR spin trapping data presented in Fig. 5: p-values from the univariate general linear model and Tukey post-hoc test procedure. Table S7. Changes in 57 Fe and total Fe speciation upon addition of 3 μmol g − 1 57 Fe to blank-altered and DFOB-altered chrysotile at pH 7.4, as determined by Mössbauer spectroscopy. Table S8