Exposure to asbestiform amphiboles can trigger several diseases, including asbestosis, squamous-, small-, and large-cell carcinoma, adenocarcinoma, and mesothelioma. Currently, there is no clear epidemiological evidence that non-asbestiform amphiboles (e.g., cleavage fragments) may lead to these malignancies [1
], but the role of amphibole cleavage fragments needs to be determined and placed within the context of the toxicity of asbestos-like phases, even if it might play only a synergistic role with their fiber-shaped counterparts. Since “asbestos” is an inadequate and incomplete non-scientific term inherited from industrial and regulatory uses [2
], it is important to provide details of all the phases, including their shapes and dimensional parameters, present in complex amphibole-bearing rocks or powders in order to assess their potential effect in disrupting normal physiological functions and homeostasis (e.g., iron homeostasis in cells).
This approach is even more important after evidence has become available that asbestiform minerals may not only induce fibrosis and malignancies, but also other diseases. One example is autoimmune disease, where an association between exposure and autoantibodies has been shown to exist [3
]. While asbestosis and mesothelioma all over the world are immediately related to asbestos, the recognized cases of autoimmune disease are much rarer because of a lack of systematic investigation and sufficient available cohorts of study [4
]. One of the rare case studies is Libby, Montana, where detailed mineralogical investigations have revealed that six amphiboles occur, namely: winchite, richterite, tremolite, magnesio-riebeckite, magnesio-arfvedsonite, and edenite. However, only tremolite is regulated as asbestos, and the exact identification of the different fiber compositions can be dependent on the data reduction method used or can vary among different laboratories [5
Furthermore, recent studies show evidence that ingestion of asbestos fibers may lead to development of malignancies [6
]. Yet, there is no study on the eventual contribution of other non-asbestiform amphiboles or mixed-habit powders to this possible disease pathway. The description of these recognized dangerous minerals as well as of minerals that are similar crystallochemically or in terms of their habit is important for the identification of geographic areas of potential risk and for understanding the transformation that occurs to the minerals during the interaction with the biosphere, which would help the medical community in describing the generation and development of these diseases.
In this paper, we describe the mineralogical characteristics of a metamorphic rock sample from Passo di Caldenno (Western Valmalenco, Northern Italy), which contains naturally occurring asbestos (NOA). NOAs are typical secondary mineral assemblages in mafic and ultramafic rock sequences [10
], where a wide range of geological processes, including shear deformation and fluid-rock interaction, may produce a considerable rock alteration and serpentinization, with formation of asbestiform amphiboles. The particle population will be described through several statistical parameters and by applying a fractal model over different magnification ranges, in an attempt to describe its wide variability in terms of dimensions and habit. The sample characterization has been conducted through a multi-analytical approach, and we will provide details on the dissolution of the bulk mineral powder in a simulated interstitial lung fluid by tracking the release of select elements and by recording the dimensional re-distribution of the amphibole particles.
Passo di Caldenno (2517 m above sea level [m asl]) is located in the Central Alps (Sondrio province, Northern Italy), near the border of south-eastern Switzerland (Figure 1
a). It is the northern termination of Monte Caldenno (2669 m asl) and the western limit of Valmalenco. The study area is geologically rather complex, as it is located at the boundary between the Penninic and Austroalpine nappes [15
] (Figure 1
b). The Penninic Malenco unit, the largest outcrop of ultramafic rocks in the Alps [16
], is delimited by the Valtellina Valley to the south and the Engadine Valley to the north. These two valleys correspond to the regional structures of the Insubric Line (or Tonale Fault) and the Engadine Line, respectively (e.g., [16
]). The Malenco unit is overlain by the Austroalpine nappes, which consist of gneisses, schists and phyllites associated with marbles, calc-schists and quartzite [18
]. Despite the pervasive recrystallization phenomena resulting from the Alpine metamorphism, several lenses of pre-Alpine high-grade assemblages are preserved in the Malenco unit [17
]. Penninic and Austroalpine nappes were intruded by Oligocene felsic plutons, including the 30–32 Ma Bergell granitoids (Masino-Bregaglia) and the Tonalite di Triangia intrusion (Figure 1
b). Amphiboles have been reported to occur in all major mafic lithologies of the study area (e.g., [17
]), including the following varieties: (i) ‘orthoamphibole’, pargasite, tschermakite, ‘hornblende’, and tremolite in gabbros, and (ii) magnesio-hornblende, cummingtonite, and tremolite in spinel peridotites. Secondary amphibole-bearing assemblages, however, occur as well, mostly in mafic and ultramafic rocks, but also in carbonates, within the contact aureole surrounding the Bergell pluton (e.g., [15
2. Materials and Methods
The sample was characterized using a multi-analytical approach combining: Visible-Light Microscopy (VLM), X-Ray Powder Diffraction (XRPD), Electron Probe Micro-Analysis (EPMA), µ-Raman Spectroscopy, Fourier-transform Infrared Spectroscopy (FTIR), BET surface area determination, laser diffraction, Field Emission Environmental Scanning Electron Microscopy (FE-ESEM) coupled with Energy-Dispersive X-Ray Spectroscopy (EDXS). Mineral dissolution in Gamble’s solution was tracked using Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES). After the dissolution experiment, the remaining solid material was studied using FE-ESEM-EDXS.
The sample (≈15 × 5 cm) was examined as received (≈231 g of material) and broken into small pieces (1–3 cm across) using a hammer. These fragments were studied without further preparation using a Leica M165 C stereomicroscope (Leica Microsystems, Wetzlar, Germany) equipped with a Leica IC80 D camera (Leica Microsystems, Wetzlar, Germany) and an LED illumination system.
Powder diffraction data were collected using a Scintag X1 diffractometer (ThermoFisher, Waltham, MA, USA) with Cu-Kα1 radiation (λ = 1.54055 Å, 40 mA, 45 kV), fixed divergence slits, and a Peltier-cooled Si (Li) detector with a resolution of <200 eV. A divergent slit width of 2 mm and a scatter-slit width of 4 mm were employed for the incoming beam, whereas a receiving slit width of 0.5 mm and scatter-slit width of 0.2 mm were used for the diffracted beam. Data were collected in step-scan mode in the 2–70° 2θ range, with a step size of 0.05° 2θ, and a counting time of 3 s/step.
Amphibole compositions were obtained by combining wavelength-dispersive EDXS techniques using a CAMECA SX50 electron microprobe (CAMECA, Gennevilliers, France) at ‘‘Centro di Studi per il Quaternario e l’Evoluzione ambientale—CNR’’ (Rome). Analytical conditions were: 15 keV accelerating voltage, 20 nA beam current, 5 µm beam size and 100 s counting time. Data reduction was made using the ZAF4/FLS software (standard version) by Link Analytical (Oxford, UK). Compositions were determined relative to natural and synthetic standards.
2.4. µ-Raman and FTIR Spectroscopy
µ-Raman analyses on untreated samples were carried out employing a confocal Horiba-Jobin Yvon LabRam HRVIS (HORIBA, Ltd., Kyoto, Japan), equipped with a motorized x-y stage and connected to a BX41 Olympus microscope (Olympus Corporation, Tokyo, Japan). The backscattered Raman signal was collected with a 50× objective, and Raman spectra were obtained for non-oriented crystals. The 532 nm line of a He-Ne laser was used as excitation source, with the laser power (80 mW) controlled by a series of density filters. Spectra were collected in the region between 100 and 4000 cm−1
, with a hole aperture of 200 μm, diffraction gratings of 1800 g/mm, and integration times of 10 to 40 s, for 10 accumulation cycles. The lateral and depth resolutions were approximately 2 and 5 μm, respectively. The system was calibrated against the Stokes Raman signal of pure Si at 520 cm−1
before each experimental session. Spectral manipulation, such as baseline adjustment, smoothing and normalization, were performed with the Labspec 5 software package (version 5.64.15, HORIBA, Ltd., Kyoto, Japan) (Horiba Jobin Yvon GmbH 2004, 2005). Band-component analysis was undertaken using a Lorentzian function (Fityk software package (version 1.3.1), [20
The chemical composition of the same fragments characterized by µ-Raman was further determined using a JEOL JSM IT300LV (JEOL, Ltd., Akishima, Tokyo) scanning electron microscope (SEM) equipped with an Oxford Inca Energy 200 EDXS microanalysis system (Oxford Instruments plc, Abington, UK), a Pentafet detector and an ultrathin window (Z > 5). All spectra were obtained at an accelerating voltage of 15 kV, a working distance of 10 mm and an exposure time of 60 s. Primary standardization was performed on SPI Supplies and Polaron Equipment analytical standards, whereby daily standardization was performed against a high-purity metallic Co standard.
Powder FTIR spectra were collected using a Nicolet iS50 FTIR spectrometer (ThermoFisher Scientific, Waltham, MA, USA) equipped with a DTGS detector and a KBr beam splitter; the nominal resolution was 4 cm−1, and 64 scans were averaged for each sample and for the background. Samples were prepared using the conventional KBr pellet technique: one disk was prepared for the investigation of the OH-stretching region (4000–3000 cm−1) by mixing 7 mg of amphibole powder with 150 mg of KBr; a second disk was prepared by mixing 0.5 mg of amphibole powder with 150 mg of KBr for analysis of the framework modes region (<1200 cm−1). In addition, a single-crystal spectrum in the OH-stretching region was collected using a Bruker Hyperion 3000 FTIR microscope (Bruker Corporation, Billerica, MA, USA) with a beam dimension of 20 µm. The resulting pattern was virtually identical to the powder pattern.
For all the following analyses, the sample fragments were initially crushed into mm-wide crystals. This material was then ground by hand for 5 min in 2-propanol using an agate pestle and mortar to obtain a powder. As described in the following FE-ESEM-EDXS section, the sample material was further investigated on a larger scale, as mm-sized crystals and cm-sized rock fragments, to complement the morphological information obtained through VLM.
2.5. BET Surface Area Analysis
Once the powder was dry, 2.6476 g of the sample were loaded in a tube and conditioned under N2 flow at 100 °C. The BET specific surface area (m2·g−1) was then determined using a Micromeritics TriStar 2000 instrument (Micromeritics, Norcross, GA, USA) and N2 as probe gas.
2.6. Laser Diffraction
The particle-size distribution of the powdered material was analyzed with a Beckman-Coulter LS 13 320 particle-size analyzer (Beckman Coulter Inc., Brea, CA, USA). For this purpose, ~0.5 g of sample was dispersed in distilled water and inserted in the instrument.
A FEI Quanta 600 FEG Mark II Field Emission SEM (FEI Company, Hillsboro, OR, USA) was used to obtain detailed information on the starting material. We studied (i) the morphology of the sample subsequent to the hammering in a 3 cm-sized fragment, and (ii) the hand-ground sample to determine the habit of individual, fractured particles, their dimensional distribution, and track the chemical transformation before and after the interaction with the mimicked interstitial lung fluids. Each specimen was mounted on a 12.7 mm SEM stub by using conductive carbon tabs. The SEM was set to environmental mode with a voltage of 15 kV and a chamber pressure of 0.38 torr. Variable spot dimension was used to optimize imaging or chemical analyses (size 3 and 6, respectively). All images shown here were taken in secondary electron (SE) mode.
The size-distribution and morphological studies were conducted under the same conditions, but with a voltage of 30 kV and a magnification of 102×, 103× and 104× to cover three orders of magnitude in our measurements and to obtain three separate dimensional distributions. The descriptive statistics show the parameters that are commonly used for the regulatory descriptions of fiber populations, i.e., their length (L), width (w), and aspect ratio (), as well as additional parameters, such as the real area (measured), the calculated area (the area given by ), the dimension “s” which is given by , and the aerodynamic equivalent diameter (Dae). This dimensional study was performed on both the hand-ground starting material and the material after it had interacted for 168 h with Gamble’s solution.
The dimensional distributions obtained for the starting material were modelled by applying a fractal model. The rationale of our choices is explained in the corresponding section. In general, the applied procedure was as follows: (i) removal of outliers, (ii) definition of the optimal binning by qualitatively comparing the Scott’s and Freedman-Diaconis’s methods, (iii) discretization of the distribution using the K-means algorithm, (iv) power-law fitting of the experimental distribution, and (v) calculation of the fractal dimension “D”.
The powdered sample was morphologically and chemically characterized again after it had been exposed for 168 h to Gamble’s solution. In particular, we studied the , , and atomic ratios in the bulk of both the powdered sample before the interaction and the powdered sample after 168 h of interaction with Gamble’s solution, as well as in Gamble’s solution itself.
The release of elemental species from the powdered sample was investigated by dissolution experiments using Gamble’s solution (simulating lung fluid), following previous experimental approaches but with a few modifications [21
]. The modified Gamble’s solution was composed of (g·L−1
O, 0.203; NaCl, 6.019; KCl, 0.298; Na2
, 0.126; Na2
, 0.063; CaCl2
O, 0.368; NaC2
O, 0.952; NaHCO3
, 2.604; and Na3
O, 0.097, brought to a final pH of 7.4. Subsequent to the preparation of Gamble’s solution, the powdered sample was added to the solution to a final concentration of 1.25 mg·mL−1
. This suspension was mixed homogenously using a magnetic stir bar while dispensing into 10 mL batch reactors. Additionally, negative controls containing Gamble’s solution only were also prepared. All batch reactors were prepared in triplicate and subsequently incubated at 37 °C to mimic human body conditions. Ultimately, we prepared a total of 48 batch reactors, each containing 10 mL of suspension.
The entire contents of three batch reactors for each of our two experimental conditions (Gamble’s with powdered sample and the negative control, for a total of 6) were sacrificed per time point at 0, 0.5, 1, 12, 24, 48, 96, and 168 h after dispensing. At each time point, the 10 mL suspensions from all 6 batch reactors were centrifuged at 2000× g for 5 min to concentrate the powdered sample into a pellet. The supernatants containing just the Gamble’s solution were decanted into clean tubes and stored at 4 °C until further processing by ICP-OES as outlined below. The pellets containing the powdered samples were subsequently resuspended in milliQ water (18.2 Ω·cm, 22 °C), transferred into larger 50 mL Falcon tubes, then spun again to wash the mineral phases. This washing step was repeated once more before the water was carefully removed using a serological pipet. The pellets collected at 168 h were then dried before FE-SEM-EDXS analysis as described above.
The Gamble’s solutions from all 48 batch reactors were analyzed by ICP-OES for specific elements using a Spectro Genesis instrument equipped with a Modified Lichte nebulizer and 25 mm axial plasma torch. The Gamble’s solutions were diluted down 1:100 in clean milliQ water before analysis. The elemental concentrations were quantitatively determined for Si at a wavelength of 251 nm, Mg at 288 nm, Ca at 315 nm, Fe at 259 nm, and Al at 308 nm. The elemental dissolution rate is given by , where “El” is the specific element concentration in solution (mol·g−1), “t” the considered reaction time (s), and subsequently normalized to the surface area (m2·g−1).
The VLM and FE-ESEM morphological observations on the original sample revealed a fibrous morphology and the occurrence of asbestiform, well-elongated individual fibers arranged in prismatic millimetric and sub-millimetric aggregates. Whereas the asbestiform units are well aligned on undisturbed surfaces, fractured areas show randomly oriented fibers. After short grinding by hand (5 min and minimal mechanical stress), the observed morphology is clearly different, with a predominance of non-asbestiform characteristics.
The combination of XRPD and EPMA data identified the sample precisely at the boundary between the ideal compositional ranges of tremolite and actinolite.
The small surface area determined by BET analysis (0.4786 ± 0.0104 m2
) is compatible with our morphological observations as well as with literature data for bulk amphiboles, which were prepared in a very similar fashion (0.46 m2
]). The surface area determined here, however, is one order of magnitude lower than that of a nearly pure asbestiform tremolite (short fibers, surface area 9.2 m2
The dimensional analysis documents that the estimation of the average length and its distribution obtained through laser diffraction is in good agreement with the measurement performed by the FE-ESEM over three orders of magnitude. In detail, however, laser diffraction failed to properly describe the studied population, as it returned a flat edged bimodal, poorly sorted distribution (Table 2
). This is most likely a consequence of the fact that laser diffraction cannot take into account the rich variety of habits that comprise the investigated powder and the presence of particle agglomerates. The FE-ESEM investigation of the crystal habit at different magnifications, on the other hand, revealed that the predominant habit is different at different magnifications (102
×, and 104
×), and therefore, we conclude that the use of a single magnification when describing a sample can result in a different distribution of both dimension and habit.
Typically, microscopically-recorded dimensional distributions fail to correctly describe the abundance of the smaller particles due to instrumental limit or observation issues [38
], superposition of particles or the reciprocal screen effect, and insufficient contrast or poor resolution at a certain magnification range. Examination of all three power-law curves obtained at 102
×, and 104
× documents that the information regarding the lower end of the size distribution is enhanced, which results in a higher frequency of small particles compared to the one detected at lower magnifications. Inspection of the fractal model chart reveals that the observations made at 104
× complement those made at lower magnifications (102
×), including particles with L
< 6 µm. On the other hand, the 104
× data fail to describe the longest particles present in the population. The overlap between the 102
× and the 103
× data is larger than the overlap between the 103
× and 104
The fractal model yields a good fit (R2
≥ 0.90) for all three studied distributions, where the resulting fractal dimensions are similar and within the calculated standard error. The slight tendency of D
to decrease with increasing magnification can be due to the fact that we investigated a complex powder in which each individual habit contributes at different magnifications. A lower value of D
is compatible with a lower resistance to fracture, which is expected for particles that can easily be separated mostly along the elongation axis. The fractal dimension calculated at 104
x is very similar to the one calculated by Wylie [38
] for a tremolite from California with β
< 3:1, investigated by SEM (D
= 2.13), but slightly higher than the one of another Californian tremolite, which exhibits an aspect ratio >3:1 (D
= 1.70). Wylie [38
] further reported identical D
values of 1.42 for two Korean tremolite samples, one with β
< 3:1, the other with β
> 3:1. These results thus document that a given mineral from different outcrops may show a different fracture resistance, regardless of its morphology. In general, however, the fibrous variety of a given mineral has a smaller D
value than the non-fibrous counterpart [38
]. Overall, our data suggest that combining descriptive statistics with a fractal model over different magnification ranges is more effective in describing a complex particle population than observation and counting made at a fixed magnification for a given number of fields of view. The descriptive statistics and the fractal model are compatible, and they may be applied together.
Regulations from different countries commonly define asbestos fibers by a certain length, width, and β
. When we change magnification, both the average β
values and the Areal
values remain similar. However, individual L
values change according to the selected magnification. This observation can lead to problems with the application of different microscopy techniques or methods for evaluating the presence of regulated fibers (e.g., moving from VLM to SEM, or even to TEM). This is also true when the magnification is changed during the observation, as shown by our fractal model and in accordance with Wylie [38
], who compared VLM, SEM and TEM. The use of a fractal approach should help in closing the gap in information generated using a single instrument with a fixed magnification, where it is assumed that the observed particles are the product of a scale-invariant phenomenon.
The dimensional term “s
” is theoretically the best mono-dimensional descriptor of a particle’s projected area [34
]. A limitation exists, however, when an investigator wants to assess the potential toxicity or breathability of a powder using this dimension, since there is no published investigation on this topic. The term “s
” can be substituted by Dae
, which is a mono-dimensional term describing the 3-dimensionality of the particles, which takes into account the ratio between the major dimensions (L
). This term can be used to describe the aerodynamic behavior of the particles in the environment (e.g., air) as well as their deposition in the human body (e.g., lungs). After the interaction with Gamble’s solution, we recorded a general increase in the range of most dimensional parameters and their σn−1
values. The dissolution process, thus, not only generated new particles in the smallest dimensional categories, but also led to the formation of particles in the larger categories. In fact, large non-measurable particles (data not reported in the statistical summary) were visible at 102
× in the original starting material, and we believe that these particles may be responsible for the larger particles observed in the sample after its interaction with Gamble’s solution. This somewhat counterintuitive effect of dissolution on the particle population makes a differential study across different magnifications even more important.
Furthermore, after interaction with Gamble’s solution, the morphology of the particles is different. Overall, the dominant habit is the same before and after the interaction, but their percentages changed slightly: fibers increased by 1.33% at 102×, becoming the dominant habit; bladed particles increased by 4.00% at 103×, and remained the dominant habit; bladed particles decreased by 3.33% at 104×, but also remained the dominant habit. These modifications seem to be more intense for shorter particles, possibly as a consequence of a suspected higher number of defects along the c-axis of non-asbestiform amphiboles. Particles occasionally show deeply striated and stepped surfaces, probably indicating that the dissolution proceeded selectively between the I-beams of the amphibole structure and not just on planes intersecting the c-axis.
The last dimensionally related aspect of particular importance is related to the percentage of EMPs according to the NIOSH classification [37
]. We have shown here that changing magnification from 102
× to 104
× led to a reduction of the percentage of EMPs from 54.00% to 2.00% in the starting material, and from 67.33% to 12.00% in the sample after the dissolution experiment. In addition, the interaction of the particle population with Gamble’s solution led to an increase in the percentage of NIOSH-classified EMPs by 13.33% at 102
× and by 10.00% at 104
×, but to a 5.33% decrease at 103
×. Again, when considering all magnifications, there was an overall increase of EMPs by 18.00% when comparing the powdered materials before and after the experiment.
Our dissolution experiment showed that Si and Mg were released into Gamble’s solution at the same rate in the first 48 h. For longer exposures (96–168 h), a nearly congruent dissolution was observed. These two elements maintained a similar bulk ratio in the solid before and after 168 h of interaction, and their ratio in solution is only slightly lower, although the difference is statistically not significant. A similar dissolution study of an actinolite-rich sample, which contained considerable amounts of chlorite and other phases, was conducted by Critelli et al. [43
]. These authors showed that at near-neutral pH, chlorite had a dissolution rate that was slightly higher than that of actinolite (with partially overlapping error bars). Therefore, some of the Mg and Si detected in our own dissolution experiment might have been derived from the small amounts of chlorite (estimated at 2 wt. % by XRPD). However, the potential contribution from chlorite dissolution is probably very small in our case.
The calculated Si-release rate for the first 48 h (1.21 × 10−12
± 1.13 × 10−13
) is one order of magnitude smaller than the one calculated for tremolite in the experiment by Critelli et al. (3.6 × 10−11
]). The slower dissolution observed in our study is likely related to the smaller surface area (about one order of magnitude smaller) and to the fact that Gamble’s solution is likely to reach saturation faster than the buffer solution used in the cited work. Alternatively, this may be simply due to the different experimental conditions.
Another interesting result of our experiment is that the Ca concentration in solution increased strongly in the first 24 h, but decreased subsequently. This decrease could be due to recrystallization and settling of Ca-bearing particles, which we did indeed observe at a micrometer/sub-micrometer scale (data not shown). Similarly, Ca-phosphate precipitation on the surface of crocidolite fibers was observed by Pacella et al. [45
]. The Fe was barely released into the solution at the chosen experimental conditions. According to the literature, Fe most likely rearranges itself into new structures with lower coordination geometry while being oxidized at the surface of the particles (e.g., [45
Notably, in our experiment no Al release was detected. Therefore, and because our sample is almost entirely composed of tremolitic amphibole, with a very small amount of chlorite (2 wt. %), we conclude that the released elements are derived almost exclusively from the dissolving tremolitic amphibole.
The calculated biodurability values allow us to speculate that the studied mineral may be characterized by a long biopersistence in the human body. The particle population studied here, however, shows a shorter lifetime than a tremolite examined in similar conditions [42
]. Rozalen et al. [42
] have shown that a fiber of 0.1 µm in diameter dissolves at pH 7.4 in 92 years, whereas in our case a fiber of 0.1 µm in similar conditions has a lifetime of 72 ± 2 years.