Enhanced darkfield microscopy and hyperspectral imaging with spectral mapping enable screening, localization, and identification of nanoscale materials in histological samples with improved speed and accuracy over traditional methods. The goal of this paper is to provide methods for darkfield imaging and hyperspectral mapping of metal oxide nanoparticles in histological samples.
Nanomaterials are increasingly prevalent throughout industry, manufacturing, and biomedical research. The need for tools and techniques that aid in the identification, localization, and characterization of nanoscale materials in biological samples is on the rise. Currently available methods, such as electron microscopy, tend to be resource-intensive, making their use prohibitive for much of the research community. Enhanced darkfield microscopy complemented with a hyperspectral imaging system may provide a solution to this bottleneck by enabling rapid and less expensive characterization of nanoparticles in histological samples. This method allows for high-contrast nanoscale imaging as well as nanomaterial identification. For this technique, histological tissue samples are prepared as they would be for light-based microscopy. First, positive control samples are analyzed to generate the reference spectra that will enable the detection of a material of interest in the sample. Negative controls without the material of interest are also analyzed in order to improve specificity (reduce false positives). Samples can then be imaged and analyzed using methods and software for hyperspectral microscopy or matched against these reference spectra in order to provide maps of the location of materials of interest in a sample. The technique is particularly well-suited for materials with highly unique reflectance spectra, such as noble metals, but is also applicable to other materials, such as semi-metallic oxides. This technique provides information that is difficult to acquire from histological samples without the use of electron microscopy techniques, which may provide higher sensitivity and resolution, but are vastly more resource-intensive and time-consuming than light microscopy.
As nanomaterials are increasingly used in a variety of industries and applications, there is a need for nanoscale imaging and characterization methods that are more rapid, affordable, and convenient than traditional modalities, such as electron microscopy. To visualize nanoparticle (NP) interactions with cells, tissues, and living systems, many optical techniques have been employed, including differential interference contrast (DIC) microscopy1 and evanescent field-based approaches, such as total internal reflection (TIR) or near-field scanning optical microscopy (NSOM)2,3. However, these are high-end analytical approaches, out of reach of most non-specialist labs4. Electron microscopy, including transmission electron microscopy (TEM), has also been used to study NP interaction with cells5,6,7,8. High angle annular darkfield (HAADF) scanning transmission electron microscopy has been used to study the interaction of NPs with viruses9. Confocal microscopy is another popular technique used to study NP-cell interactions10.
In recent years, darkfield-based hyperspectral imaging (HSI) techniques have been used as a promising analytical tool for studying NPs in biological matrices11. HSI systems generate a three-dimensional representation of spatial and spectral data, known as a hypercube or datacube12. The spatial map of spectral variation is used for material identification. Spectral profiles of known materials can be generated and used as reference libraries for comparison to unknown samples. One of the major advantages with HSI systems is its ability to combine imaging with spectroscopy, thereby establishing the location and distribution of unknown NPs in vivo or ex vivo as well as linking them to another reference particle of known, similar composition.
There are several advantages of using HSI systems over conventional imaging techniques: minimal sample preparation is required; sample preparation is typically non-destructive in nature; image acquisition and analysis is faster; the technique is cost-effective13; and the spatial distribution and analysis of compounds of mixed composition and/or in complex matrices is more easily accomplished14.
For nanomaterials research involving precious samples, one of the most important considerations is the availability of a non-destructive imaging method, which allows for the potential to repeatedly examine samples by one or more methods. Repeated or multiple analyses may be desired to develop comprehensive datasets that would not be available from a single method. To this regard, studying its optical properties is the safest way to analyze the sample. By using an enhanced darkfield microscope (EDFM) and HSI system to study the optical response of the sample – namely reflectance, but also absorbance and transmittance – feature identification and characterization can be performed15. Potential characterization endpoints include an assessment of relative size and shape of nanoparticles or agglomerates and distribution of nanoparticles within a sample.
In this paper, we describe mapping methods specifically for metal oxide nanoparticles in post-mortem tissue using a HSI system based on a pixel-spectral match algorithm referred to as a spectral angle mapper (SAM). We chose this particular application because it has the potential to complement current and future in vivo nanotoxicology research, wherein animal models are used to evaluate the health implications of exposure to engineered nanomaterials. Application of this method could also inform nanoscale drug delivery research that utilizes tissue or animal models. In particular, nanoparticle absorption, distribution, metabolism, and excretion throughout organs and tissues could be investigated with this system. A wide variety of applications are being investigated for use in biomedical research11.
This method could be utilized for assessment of different biological samples (such as various tissues types, bronchoalveolar lavage samples, and blood smears) that have been exposed to nanoparticles of a variety of elemental compositions16-19. Furthermore, this method is useful for studying nanoparticle biodistribution in vivo and in vitro, which is pertinent for nanoscale drug delivery studies11. Beyond biological samples, EDFM and HSI can be used to evaluate nanoparticles in environmental samples, such as wastewater20. Occupational exposure assessment can be facilitated by the use of this technique as well, since it may be used to evaluate the efficacy of personal protective equipment in preventing nanoparticle penetration. Furthermore, the research team is currently developing a similar EDFM and HSI protocol for the evaluation of filter media samples of nanoparticles collected from occupational exposure assessments. While preparation of these different sample types for EDFM and HSI may vary, it is important that they are prepared in such a way that they may be easily visualized by the optical system. Typically, the sample should be prepared as if it would be visualized via traditional brightfield microscopy. There are several hyperspectral imaging systems commercially available11.
For tissue samples that have undergone conventional histological staining, the identification and analysis of metal oxide nanoparticles may be achieved through a combination of EDFM, HSI mapping, and image analysis techniques. While there is flexibility in sample preparation for histology or immunohistochemistry (e.g., using fixed or frozen tissues; type of stain), it is important that samples are sectioned to a thickness of 5-10 µm for optimal visualization. Samples used here were formalin-fixed and paraffin-embedded prior to sectioning with a rotary microtome to 6 µm thickness, mounted on glass microscope slides, stained with hematoxylin and eosin and coverslipped. Porcine skin tissues from an ex vivo cutaneous penetration toxicology collaboration were used for this study. The tissues were exposed to metal oxide nanoparticles (alumina, silica, ceria) in aqueous suspensions. Detection of the region(s) of interest (high contrast elements) with EDFM is a critical first step that facilitates subsequent HSI mapping and analysis. Positive and negative control samples must be imaged and analyzed first in order to create a spectral library for reference. The collected spectra from the positive control are exported to a positive control spectral library. Then, all the spectra from the negative control images are subtracted from the positive control’s spectral library in order to improve specificity (reduce false positives). The resulting filtered spectral library is considered the RSL that serves for the analysis of materials of interest. All tissue samples undergo the same imaging process and are mapped against the RSL. The resulting image will contain only areas with elements of interest over a black background. This image could then be analyzed with ImageJ using its threshold and particle analysis functions to obtain the area of mapped particles per field of view. Numerical data obtained from ImageJ can be exported to a spreadsheet for further analysis.
It is important to consider that as biological samples are inherently different from one another, and staining methods may affect visualization through EDFM and HSI, the exposure settings should be determined in accordance to what produces the best high contrast image for a specific type of sample. Although reduction of false positives can be achieved through the filtration of spectral libraries, it is crucial to obtain reliable negative controls that have avoided contamination with the element of interest, as the spectra corresponding to this contamination could potentially be filtered from the positive control’s spectral library, increasing false negative rates. Also, the spectrum intensity range that is detectable with the hyperspectral imaging software cannot surpass the limit of the particular software (for this study, that is 16,000 units): areas with a high number of accumulated particles that produce spectral intensities above the intensity limit are left out of the spectral library, due to the risk of increasing the number of false negatives.
While the HSI system confers many advantages over traditional methods, there are some drawbacks and limitations to consider. One is that the large amount of optical data collected may require substantial computing power. Another is that HSI can be time-intensive, particularly at the beginning stages when reference spectral libraries are being created. Also, imaging time may require several minutes per image capture, making it slower than simple darkfield imaging; however, it is still faster than performing sampling preparation and visualization by electron microscopy. Additionally, complex systems may result in multiple characteristic spectra, which require the development of highly specialized controls and make the establishment of standardized, universal reference libraries difficult26. Finally, the technique results in lower resolution than probe- or electron-microscopy techniques, such as atomic force microscopy or transmission electron microscopy, which can resolve individual atoms. This technique’s resolution is limited due to its photonic nature, which prevents it from being a high-precision tool for measuring particle sizes at the nanometer level or for precisely locating materials at a sub-micron level. While the technique may be able to pinpoint particles within certain tissue compartments or cellular organelles greater than 1 µm (such as cell nuclei), smaller organelles or features are challenging to visualize accurately with this method. Also of note, given its spatial resolution, this method cannot differentiate between single nanoparticles and agglomerates11.
Other considerations include: certain materials (such as noble metals) have much higher reflectance and distinct spectral profiles, which can make them easier to analyze and spectrally map with this tool. Others, such as the semi-metallic oxides investigated in this study and carbon-based nanomaterials24, 27, may be more challenging due to their elemental composition, shape, and depending on the matrix. In two murine inhalation studies by Mercer et al., a similar system to the one employed in this study was used in order to locate carbon nanotubes in the lung and in secondary organs based on their remarkably high contrast with surrounding tissues. However, hyperspectral mapping was not demonstrated in either study, likely because the unique shape of carbon fibers was a sufficient trait for identification. Another consideration pertains specifically to tissues: since depositing nanoparticles of interest to specific organs through normal biophysical processes is unpredictable (and often itself the subject of study), determination of an applicable positive control can be difficult and requires consideration of how generation of the control might affect the state of a material of interest. For example, if a spectral library is created from pristine nanoparticles of interest, it may be difficult to map the library to those same nanoparticles in tissues or cells due to changes in the spectra resulting from alteration of the particle (e.g., due to change in pH, dissolution, agglomeration, protein binding) and the overall microenvironment or matrix. Finally, the technique is limited in its semi-quantitative nature: it can only be as quantitative as other two-dimensional microscopy techniques of similar resolution, which means it cannot be easily used to perform tasks such as characterizing total organ burden of a material.
Overall, EDFM and HSI provide several advantages over conventional nanomaterial imaging and characterization techniques, such as TEM, HAADF and DIC. EDFM/HSI allows for faster image acquisition and analysis, which saves time and cost in comparison to more intensive conventional techniques. Additionally, sample preparation for EDFM/HSI is typically both minimal and non-destructive, which saves time and also allows for more flexible analysis of a given sample since it may then be used for other techniques. Furthermore, HSI is versatile, allowing for analysis of nanoscale materials of many compositions and in a variety of matrices. The research team is working to refine the method described here for other materials and sample types, including an in-depth assessment of the specificity of the technology. A critical next step under investigation by the research team is validation of these techniques against traditional gold standards (e.g., Raman, TEM, SEM) for the materials and tissue types of interest.
The authors have nothing to disclose.
The authors thank Günter Oberdörster, DVM, PhD and Alison Elder, PhD (University of Rochester) and Mary Frame, PhD (Stony Brook University) for animal research collaborations resulting in tissue samples for analysis. Additionally, the authors thank: Christina Rotondi (Albany Medical College Histology Core); Rani Sellers, DVM, PhD and Barbara Cannella, PhD (Albert Einstein College of Medicine Histology and Comparative Pathology Facility); Leonardo Bezerra and Ahlam Abuawad (Brenner research team members); and Leslie Krauss, Byron Cheatham and Elyse Johnson (CytoViva). This work was supported in part by CDC-NIOSH grant OH-009990-01A1 and the NanoHealth and Safety Center, New York State, awarded to S.B.
CytoViva 150 Unit (condenser) | CytoViva (Auburn, AL) | mounted to Olympus BX43 microscope | |
Olympus BX43 Microscope – Analyzer Slot – HSI with 10x and 40x air objectives and 100x oil immersion objective | obtained through CytoViva (Auburn, AL) | for use with CytoViva 150 Unit condenser | |
Dagexcel-M Digital Firewire Camera – Cooled; includes Exponent 7 software | obtained through CytoViva (Auburn, AL) | enhanced darkfield camera and software | |
CytoViva Hyperspectral Imaging System 1.4; includes Pixelfly hyperspectral camera, XY stage controller, ENVI hyperspectral imaging software | obtained through CytoViva (Auburn, AL) | hyperspectral camera and software | |
cleanroom cleaned glass microscope slides (glass B slides) | Schott NEXTERION | 1025087 | reduced debris and artifacts compared to conventional glass microscope slides for optimal imaging |
cleanroom cleaned glass microscope coverslips (#1.0; 22mm x 22mm x 1.45mm) | Schott NEXTERION | custom | reduced debris and artifacts compared to conventional glass coverslips for optimal imaging |
type A microscopy immersion oil | Fisher Scientific | 12368B | multiple suppliers |
70% isopropanol in water | multiple suppliers | ||
ImageJ software | National Institutes of Health (NIH) | free open-source software online download | |
metal oxide nanoparticles | supplied to the research team by industrial partners | alumina, silica, and ceria nanoparticles in aqueous suspensions. Due to a Non-Disclosure Agreement between the authors and industry partners, further product information cannot be disclosed. |