Molecular speciation controls arsenic and lead bioaccessibility in fugitive dusts from sulfidic mine tailings

Communities nearby mine wastes in arid and semi-arid regions are potentially exposed to high concentrations of toxic metal(loid)s from fugitive dusts deriving from impoundments. To assess the relation between potentially lofted particles and human health risk, we studied the relationship between pharmacokinetic bioaccessibility and metal(loid) molecular speciation for mine tailings dust particulate matter (PM), with elevated levels of arsenic and lead (up to 59 and 34 mmol kg−1, respectively), by coupling in vitro bioassay (IVBA) with X-ray absorption spectroscopy (XAS). Mine tailing efflorescent salts (PMES) and PM from the surface crust (0–1 cm, PMSC) and near surface (0–25 cm) were isolated to <10 μm and <150 μm effective spherical diameter (PM10 and PM150) and reacted with synthetic gastric and lung fluid for 30 s to 100 h to investigate toxic metal(loid) release kinetics. Bioaccessible (BAc) fractions of arsenic and lead were about 10 and 100 times greater in gastric than in lung fluid simulant, respectively, and 10–100% of the maximum gastric BAc from PM10 and PM150 occurred within 30 s, with parabolic dissolution of fine, highly-reactive particles followed by slower release from less soluble sources. Evaporite salts were almost completely solubilized in gastric-fluid simulants. Arsenate within jarosite and sorbed to ferrihydrite, and lead from anglesite, were identified by XAS as the principal contaminant sources in the near surface tailings. In the synthetic lung fluid, arsenic was released continuously to 100 h, suggesting that residence time in vivo must be considered for risk determination. Analysis of pre- and post-IVBA PM indicated the release of arsenic in lung fluid was principally from arsenic-substituted jarosite, whereas in synthetic gastric fluid arsenic complexed on ferrihydrite surfaces was preferentially released and subsequently repartitioned to jarosite-like coordination at extended exposures. Lead dissolved at 30 s was subsequently repartitioned back to the solid phase as pyromorphite in phosphate rich lung fluid. The bioaccessibility of lead in surface tailings PM was limited due to robust sequestration in plumbojarosite. Kinetic release of toxic elements in both synthetic biofluids indicated that a single IVBA interval may not adequately describe release dynamics.


Figure S1
Location Map Figure S2 PMSC SLF SGF Zn Figure S3 PMES SLF SGF Zn Figure S4 SEM-EDS micrographs of PMES Figure S5 EDS SEM-micrographs of PM150 in SGF and SLF Figure S6 SEM micrographs of PM10 in SLF Figure S7 XRD from PM150 in SLF and SGF Figure S8 XRD from PM10 in SLF and SGF Figure S9 Iron XANES pre-edge Figure S10 Iron EXAFS fits PM10 Figure S11 Iron EXAFS fits PM150 Figure S12 Sulfur XANES with pyrite highlight Figure S13 Sulfur LCF fits Figure S14 Arsenic EXAFS fits PM10 Figure S15 Lead XANES and EXAFS Figure S16 PM10 v bulk tailings elemental enrichment plot Table S1 Synthetic lung fluid (SLF) and synthetic gastric fluid (SGF) formulations Table S2 Iron EXAFS PM10 fits Table S3 Iron EXAFS PM150, jarosite, and ferrihydrite fits Table S4 Iron EXAFS PMES and PMSC fits Table S5 Sulfur EXAFS PM10 fits Table S6 Arsenic EXAFS PM10 fits Table S7 Arsenic EXAFS PM150 fits Table S8 Arsenic EXAFS PMSC fits Table S9 Lead LCF EXAFS fits Table S10 Elemental concentrations of PM10 and bulk tailings  Samples were collected from the tailings pile (A). The town of Dewey-Humboldt (B) flanks the tailings. The site was designated a National Priorities Listed site in 2008 (C); contaminants of concern in the tailings (D) are arsenic and lead. Aerial photos from Google Earth®.  Scanning electron micrograph (SEM) and energy dispersive X-ray spectra (EDS) of efflorescent salt particulate matter (PMES) isolated from IKMHSS tailings before in vitro bioassay. EDS from the micrograph indicates Al, Ca, Mn, Fe and Zn sulfate salts. Scale bar indicates magnification. Energy dispersive X-ray spectroscopy (EDS) of <150 µm effective spherical diameter particulate matter (PM150) isolated from IKMHSS tailings unreacted (1-2) and after in vitro gastric simulant bioassay for 1 hour (3-4), and 48 hours (5-6); and after in vitro lung simulant bioassay for 1 hour (7-8), and 48 hours (9-10). See Fig. 3 for micrographs. Figure S6. Micrograph of PM10 unreacted and after 24 h and 7 days in synthetic lung fluid.
Electron micrograph of <10 µm effective spherical diameter particulate matter (PM10) isolated from IKMHSS tailings before (left 2 panels, unrx) and after in vitro lung simulant bioassay for 24 hours (top right, L 24h) and 7 days (bottom right, L 7d). X-ray transparent acicular needle laths were observed 24 hours post IVBA. Scale bar indicates magnification.   Figure S8. X-ray diffraction of PM10 pre and post IVBA in synthetic lung and gastric fluid.
Synchrotron X-ray diffractograms collected at 0.965 Å (12.6 keV) and converted to Cu Ka at 1.5406 Å for comparison. *Diffractograms are normalized to the quartz hkl(112) peak at 1.818 Å. The black center diffractogram is the unreacted PM10, reaction progress in SLF at 24 h is shown in red, reaction progress in SGF at 1 h is shown in blue. Reference mineral peaks fitting the data are shown for quartz, pyrite, jarosite, and gypsum.
Qtz (112) Int = 1 * Figure S9. Iron pre-edge XANES analysis of PM150 The Fe XANES pre-edge for PM150 before and after reaction in vitro with synthetic lung fluid and reference jarosite and ferrihydrite. Black lines show data, red dashed lines are the fits to the data. a. pre-edge peak at 7111.9 eV present with jarosite, b. and c. are jarosite pre-edge peaks at 7113.5 2.0 3.63 / Fe EXAFS fit (k=2-12) constrained by jarosite and ferrihydrite structures. The amplitude reduction factor was fixed at S 0 = 0.7. Due to the covariance of coordination numbers (N) and the Debye-Waller term (σ 2 ), σ 2 were fixed based on best agreement to reference fits from crystallographic values from (O'Day 2004;Savage 2005), interatomic distances (R) were iteratively varied, σ 2 was a linked parameter anchored to the Fe-O backscattering contribution and held constant. a Jarosite component: b Ferrihydrite component: c Degrees of freedom in the signal Nidp= 2ΔkΔR/π, k= 2 -12 (Δk = 10) and R= 1 -4.5 (ΔR = 3.5), the number of variables must not exceed the degrees of freedom, Nidp was 22, fits were constrained to 12 or fewer variables; / parameter linked in fit to the parameter directly above. Estimated errors for R ± 0.02 Å, N or σ2 ± 30% based on empirical fits to known reference compounds were for atoms beyond the first shell (see O'Day et al., 2004).   Table S2. S 0 = 0.7, σ 2 fixed based on best agreement to reference fits (O'Day et al., 2004;Savage et al., 2005), R values were iteratively varied, σ 2 was as a linked parameter, / parameter linked in fit to the parameter directly above. a Jarosite and b Ferrihydrite components described in the text. c Nipd= 2ΔkΔR/π, k= 2 -12 (Δk = 10) and R= 1 -4.5 (ΔR = 3.5), Nipd was 22, fits were constrained to 12 or fewer variables. . Estimated errors for R ± 0.02 Å, N or σ2 ± 30% based on empirical fits to known reference compounds were for atoms beyond the first shell (see O'Day et al., 2004).  Table S2 and S3. S 0 = 0.7, σ 2 fixed based on best agreement to reference fits (O'Day 2004;Savage 2005), R values were iteratively varied, σ 2 was as a linked parameter, / parameter linked in fit to the parameter directly above.. a Jarosite and b Ferrihydrite components described in the text. c Nipd= 2ΔkΔR/π, k= 2 -12 (Δk = 10) and R= 1 -4.5 (ΔR = 3.5), Nipd was 22, fits were constrained to 12 or fewer variables. . Estimated errors for R ± 0.02 Å, N or σ2 ± 30% based on empirical fits to known reference compounds were for atoms beyond the first shell (see O'Day et al., 2004).   Table S5). The spectra show that sulfate species (gypsum, jarosite and sulfate associated ferrihydrite) are dissolved relative to pyrite in the pH 1.5 gastric fluid. Fractional contribution of reference mineral to PM10 and PM150. Reference minerals a gypsum, b jarosite, c pyrite, and d sulfate sorbed to ferrihydrite; e sum of the fractional fit components not normalized to unity, f reduced R-factor, and g χ 2 are given as goodness-of-fit parameters; PM samples are unrx = unreacted, reacted in SLF for 1 hour, L1h; 24 hours, L24; 7 days L7d; and in SGLF for 1 hour, G1; 24 hours,GL24; 48 hours, G48. Fits shown in Fig. S13.  Table S6). Due to the covariance of coordination numbers (N) and the Debye-Waller term (σ 2 ), N were fixed based on crystallographic values of arsenate tetrahedra (O'Day 2004;Savage 2005), interatomic distances (R) was iteratively varied and the σ 2 was allowed to vary as a linked parameter anchored to the As-Fe backscattering contribution and held constant; / parameter linked in fit to the parameter directly above. Degrees of freedom in the signal Nidp= 2ΔkΔR/π, k= 1 -12 (Δk= 11) and R= 1 -4.5 (ΔR = 3.5); the number of variables must not exceed the degrees of freedom, Nidp was 24, fits were constrained to 6 or fewer variables. Estimated errors for R ± 0.02 Å, N or σ2 ± 30% based on empirical fits to known reference compounds were for atoms beyond the first shell (see O'Day et al., 2004).Spectral fits shown in Fig. S14. .001 Due to the covariance of coordination numbers (N) and the Debye-Waller term (σ 2 ), N were fixed based on crystallographic values of arsenate tetrahedra (O'Day 2004;Savage 2005), interatomic distances (R) was iteratively varied and the σ 2 was allowed to vary as a linked parameter anchored to the As-Fe backscattering contribution and held constant. / parameter linked in fit to the parameter directly above. Degrees of freedom in the signal Nidp= 2ΔkΔR/π, k= 1 -12 (Δk= 11) and R= 1 -4.5 (ΔR = 3.5); the number of variables must not exceed the degrees of freedom, Nidp was 24, fits were constrained to 6 or fewer variables. Estimated errors for R ± 0.02 Å, N or σ2 ± 30% based on empirical fits to known reference compounds were for atoms beyond the first shell (see O'Day et al., 2004). .009 Due to the covariance of coordination numbers (N) and the Debye-Waller term (σ 2 ), N were fixed based on crystallographic values of arsenate tetrahedra (O'Day 2004;Savage 2005), interatomic distances (R) was iteratively varied and the σ 2 was allowed to vary as a linked parameter anchored to the As-Fe backscattering contribution and held constant; / parameter linked in fit to the parameter directly above. c Degrees of freedom in the signal Nidp= 2ΔkΔR/π, k= 1 -12 (Δk= 11) and R= 1 -4.5 (ΔR = 3.5); the number of variables must not exceed the degrees of freedom, Nidp was 24, fits were constrained to 6 or fewer variables. Estimated errors for R ± 0.02 Å, N or σ2 ± 30% based on empirical fits to known reference compounds were for atoms beyond the first shell (see (1)).

Figure S15. Lead LIII XANES and EXAFS of Iron King mine tailings.
Unreacted and IVBA reacted PM10 and PM150. Left panel: lead XANES references: 1. anglesite, 2. plumbojarosite, 3. beaverite, 4. pyromorphite, 5. cerussite, 6. lead adsorbed to ferrihydrite, 7. galena. Center panel: lead XANES of PM10 and PM150 A. PM10 SGF 1 h, B. PM10 SGF 7 day, C. PM10 unreacted, D. PM150 SGF 1 h, E. PM150 SGF 7 day, F. PM150 unreacted, reference anglesite and plumbojarosite shown. Right panel: lead EXAFS fit of PM10 and PM150 A. PM10 SGF 1 h, B. PM10 SGF 7 day, C. PM10 unreacted, D. PM150 SGF 1 h, E. PM150 SGF 7 day, F. PM150 unreacted, fit components anglesite and plumbojarosite shown. Fits to χ(k)·k 2 EXAFS given in Table S9.  Theoretical phase-shift and amplitude functions were calculated with the program FEFF6 (REHR, 1993) using atomic clutters taken from the crystal structures of ferrihydrite and jarosite for Fe and scorodite and angelellite for As, known Fe and As(V) minerals with geometries similar to those expected for absorber backscatterer interactions of Fe and As in the tailings. Multiple scattering paths (MS) from As (V)-O tetrahedra were included as they have been shown to improve EXAFS fits beyond the first shell for arsenate compounds (BEAULIEU and SAVAGE, 2005;ONA-NGUEMA et al., 2005). During Fe EXAFS fitting, the values of interatomic distance (R, Å) of the Fe-O, Fe-S, and Fe-Fe shells were allowed to vary. The photoelectron threshold energy shift, ΔE0 (eV), was allowed to float as a common parameter during fit iterations, i.e., a single ΔE0 parameter was used for all backscatterer paths in a fit. There is a strong correlation between Debye-Waller (σ 2 ) and coordination number (N); therefore N was initially held constant during EXAFS fitting based on crystallographic values. Based on empirical fits to known As reference compounds, estimated errors were R ± 0.02 Å, N or σ 2 ± 30% for atoms beyond the first shell (see O'DAY et al., 2004b).
Fe-O was fit to 2-shells; 1.96 A for apical oxygen and 2.08 A for equatorial oxygen and the sum of N for the two oxygen paths was constrained to 6, but distribution and R were allowed to vary in the fit. This allowed the addition of multiple paths without exceeding the Nyquist criterion of independent fit parameters (Nidp) or increasing the degrees of freedom: Ninp = 2ΔkΔR/π (eq. 1)