For our experiments we used a synthetic amphibole powder prepared by Della Ventura et al. (1997). This compound, Ni-richterite [Na(NaCa)Ni₅Si₈O₂₂(OH)₂], does not exist in nature therefore it is not included in the regulated asbestos list. It has been chosen for the tests because it was available as a powder consisting of almost mono-dimensional, very thin and needle-like crystals (Figure 2) with length ranging between 5 and 15 µm and width <1 μm, thus corresponding dimensionally and morphologically to asbestos amphiboles (e.g., Belluso et al., 2017; Vigliaturo et al., 2021). One point to be stress here is that the terminology used in asbestos studies is often confusing because several disciplines (geology, mineralogy, industry, medicine, regulatory agencies etc.) have developed their own nomenclature somewhat independently. Accordingly, the term “asbestiform” is commonly used to describe minerals having longitudinal parting that can be split into individual fibres; the term “fibre” is used for materials that, besides having an elongated morphology, can be bent under force, while the terms “particle” and “fragment” are used for small amphibole crystals matching the above textural criteria (Gunter et al., 2007). In our work, we used a material that cannot be defines as “asbestos” being synthetic and chemically different from the regulated species, however it has all morphological and physical properties of asbestos, allowing it to be termed “asbestiform”.

Five milligrams of powder were suspended in 50 ml of Merck^® pure ethanol. Note that this solvent is not suitable for actual experiments with cells because it may induce cytotoxic effects; it was used here because of its fast evaporation time (evaporation enthalpy 42.4 kJ/mol at 303 K: Van Ness et al., 1967) that guarantees quick and efficient depositions for further image analysis.

The prepared suspension was transferred on silicon substrates, by using both the conventional pipette (hereafter “drop” method) and the “micro-drop” device. Silicon substrates were chosen because they proved to be efficient materials for cell-fibres interaction tests (Yao et al., 2017). They could be directly used in this study for both optical and SEM imagery without any further preparation. For the micro-drop depositions, we used the MD-K-130 instrument of Microdrop Technologies (GmbH Norderstedt, Germany, Figure 3), equipped with an inner nozzle diameter of 70 µm that generates single droplets with a volume of 180 pL (1% volume repeatability), with a variable drop rate from 1 to 2000 Hz. This device, designed for advanced micro-dispensing, inkjet printing and R&D applications (see https://www.microdrop.de) allows to control with high precision the pattern of droplets deposited on the substrate at well-defined and repeatable positions (see layout in Figure 3, right). The method can ensure also sterile working conditions once the metallic nozzle is efficiently cleaned and kept. During the deposition, the dispensing head remains fixed, while the substrate moves along X-Y trajectories thanks to two translation slides assembled perpendicularly, allowing the positioning control of the substrate with an accuracy of ±1 µm. The whole system is handled by a code written in LabView^® (a graphical programming environment), permitting the control of several parameters such as deposition area, deposition time, uniformity and amount of deposited liquid. Therefore, the fibres/liquid ratio to be used for cell-culture experiments depends on the parameters selected for the deposition. Just like an inkjet printer, we can start from a single deposition pattern where we release the minimum dose and may increase the amount of fibres with additional depositions on the same (or different) locations. In other words, incremental amounts of fibres on the substrates for different viability tests can be prepared by multiple depositions from the same suspension instead of preparing different suspensions.

The depositions were investigated by reflected-light optical microscopy and scanning electron microscopy (SEM): optical images were acquired by using a Leica optical system Z16 APO, while SEM images were acquired with a Zeiss FE-SEM Gemini III instrument operated at 15 kV accelerating voltage and 40 nA beam current.

Image processing and analysis were performed by using the LabView^® vision assistant software. Selected images were first converted to grayscale, applying both convolution and smoothing filters, then binarized by applying an adequate thresholding range. The converted files were analysed through the “Particle Analysis” tool build in the LabView^® environment to count particles in each generated binary image and to obtain the morphological features for each object. In our analysis, all extremely small particles (<2 µm in length) were included into a single class, because they could not be efficiently discriminated from the background. However, this simplification does not affect our conclusions regarding the suitability of the depositions for toxicological tests, because all dimensional classes with lengths <5 µm are not considered to be carcinogenic (Stanton et al., 1981; Davis and Jones, 1988), although the pathogenic potential of short fibres cannot be totally excluded (Pollard, 2020). They are included in the following plots for statistical purposes to describe the distributions obtained with the two methods, i.e., drop vs. microdrop.

Asbestos counting on both optical (collected with ×40 objective) and SEM images was performed manually following the NIOSH fibre counting Method 7400 (Ashley and O'Connor, 2017) that recommends, based on the Asbestos International Association (AIA) guidelines, that objects with these specific features are fibres: (i) longer than 5 µm along their elongation axes, (ii) width <3 μm, (iii) length/width ratio >3, (iv) have no sticking particle larger than 3 µm in diameter; in addition, (v) crossed fibres are counted individually, (vi) branched fibres are counted as a single fibre, (vii) entangled fibres are not counted.

The distribution of fibres resulting from the above-described methods was investigated through the analysis of optical images (Figure 5). The analysis aimed at quantifying the textural parameters such as homogeneity of the deposition, amount and types of entangling fibres, preferential orientation of the particles, number of single fibres etc. An example of phase segmentation and binarization of the images obtained by the drop and the micro-drop depositions are given in Figure 5. Full numerical data provided by the analytical routine are given in the Supplementary Table S1 while a summary of the number of counted objects and their total area with respect to the images, is given in Table 1.

The counting obtained by using the drop method shows an average surface coverage (Table 1) of 4.4% (1.1–10.4%) with a mean of 24 objects per image (5–37), while the sample deposited using the micro-drop method shows an average coverage of 5.0% (1.9–6.8%) with a mean of 212 objects per image (80–352) (Table 1). Therefore, although a very similar amount of material is deposed in the same area (4.4% vs. 5%), the micro-drop method yielded a much better distribution, clearly recognized in Figure 5 with a far larger (9 times higher) number of counted objects.

The elongation factor (EF), expressed as their length/diameter ratio is displayed in Figure 6 against the normalized number (to 1 mm²) of objects; 46 versus 930 particles with EF > 3 (a class including asbestiform particles) are counted for the depositions obtained with the drop and micro-drop methods, respectively. In more detail, for the classical drop method the fibrous particles are about 15% out of the total counted particles (169 over 1144), whereas for the micro drop they are 31% (930 over 2964) (Table 2). This large difference straightforwardly indicates that the depositions obtained via the conventional micropipette method may produce unrealistic results in toxicology test. Moreover, we note also that with the drop deposition ∼60% of the total objects (Figure 6; Table 2) have a very low elongation factor (EF < 2.2), meaning that the deposited particles are extremely entangled, i.e., forming very stubby objects. However, it must be underlined here that the category “objects with elongation factor >3” in Figure 6 includes, besides objects <3 µm in length also objects with a diameter >3 μm, that cannot be considered asbestos according to the AIA rules cited above.

The size distribution of objects identified through the image analysis is displayed in Figure 7, in terms of measured area in µm² (Figure 7A) or length (7b) for drop and micro-drop depositions, respectively. From the data in Table 2 and the images in Figure 7 it is evident that the number of small objects (i.e., those having an area <2 μm²) deposited by the drop method is ∼20 times lower than the number of similar objects deposited by the micro-drop method (77 vs. 1559, corresponding to 25% and 52% out of the total counts, respectively). However, the parameter “area” derived via the image processing and plotted in Figure 7 does not consider the shape of the objects. In the present case, since the used powder contains (see Figure 1) only very fibrous crystallites, any “small object” can be considered as a single asbestos fibre. The same kind of information is obtained by inspection of Figure 7B where the number of counts is plotted against their maximum length, obtained as the longest diameter of the rectangle contouring the object. Moreover, the number of objects with length <5 µm is more than seven times lower with the drop than with the micro-drop method, respectively (127 vs. 933, Table 2).

A very common source of error in the analysis of images is related to the classic workflow based on background subtraction and color thresholding (segmentation), particularly when objects to be identified share colors and hue comparable to the background. This problem is exacerbated when treating optical images due to their intrinsic relatively low resolution.

In environmental monitoring, the exposure risk to airborne asbestos is assessed by counting the fibres contained in a defined volume of sampled air; the technique commonly used for this task is the manual counting on phase contrast microscopy (PCM), particularly in indoor applications, or in SEM images. The important point here is that the asbestos counting needs to follow precise and standardized rules, such as the protocol recommended by the NIOSH Method 7400 (Ashley and O'Connor, 2017; see also https://www.cdc.gov/niosh/nmam/pdf/7400.pdf). Such possibility is not provided by standard image processing packages and for this reason asbestos counting for environmental purposes is usually done manually. To complement the statistical data discussed above, we used the same procedure on our depositions with the goal of verifying how many single particles on the substrates can be morphologically considered as asbestos fibres according to the NIOSH recommendations. This information is essential in providing robust additional evidence for the suitability of the obtained depositions for foreseen real toxicological tests. In the annotated examples (Figures 8, 9), the number and fraction of single asbestiform fibres in the micro-drop deposition is clearly much higher than in the drop deposition. Normalizing to the total studied areas, the counted fibres in the optical images are 92 vs. 1557 fibres/mm² for drop and micro-drop samples (∼18 times higher), respectively (Table 3). The same calculations for SEM images (Table 3) yielded a similar result: 65 vs. 1916 fibres/mm² (∼29 times higher). An intriguing point from Table 3 is that, for the drop deposition, the number of single fibres for mm² for the optical images is slightly higher in comparison to the SEM images, although reasonably similar (92 vs. 65, Table 3). This is explained by considering the larger inspected area (3.68 mm² for optical vs. 2.08 mm² for SEM, Table 3) for a sample that is extremely heterogeneous (Figure 8), hence the counting is underestimated because of the smaller area imaged by SEM. The numbers of Table 3 could also be affected by an error introduced by the operator that, during the manual counting, needs to avoid the entangled particles. This being the case, the number of fibres/mm² counted from the optical images (92, Table 3) is probably more accurate than the number from the SEM images (75). In the micro-drop case, where objects are more homogeneously distributed, although the surface covered by optical images is 5 times larger (0.98 vs. 0.19 mm²) the number of counted single fibres is lower (1547 vs. 2275, Table 3) and this clearly reflects the higher resolution of electron microscopy over optical microscopy. In addition, as it is evident from the images given in Figures 4b, 8b and 9b, for highly homogeneous distribution of particles even a small area is representative of the whole sample.

The distribution of the fibres lengths counted from SEM images (Figure 10) shows that the two classes 5–8 and 8–11 µm are almost equal and in total constitute 70% of the counted fibres (see cumulative trend over the histograms); their concentration gradually decreases towards longer dimensions for the micro-drop deposition. This result is in agreement with the strongly mono-dimensional nature of the used powder. For the drop samples very few particles with short length (5–11 µm classes) are counted, again stressing the superior reliability of the micro-drop deposition for in vitro tests.