Integrated molecular imaging technologies for investigation of metals in biological systems: a brief review

Abstract

Metals play an essential role in biological systems and are required as structural or catalytic co-factors in many proteins. Disruption of the homeostatic control and/or spatial distributions of metals can lead to disease. Imaging technologies have been developed to visualize elemental distributions across a biological sample. Measurement of elemental distributions by imaging mass spectrometry and imaging X-ray fluorescence are increasingly employed with technologies that can assess histological features and molecular compositions. Data from several modalities can be interrogated as multimodal images to correlate morphological, elemental, and molecular properties. Elemental and molecular distributions have also been axially resolved to achieve three-dimensional volumes, dramatically increasing the biological information. In this review, we provide an overview of recent developments in the field of metal imaging with an emphasis on multimodal studies in two and three dimensions. We highlight studies that present technological advancements and biological applications of how metal homeostasis affects human health.

Graphical Abstract

Introduction

Metals play a pivotal role in biology and are required by approximately 40% of enzymes for proper function [1]. This dependency is largely due to the unique inorganic and electronic properties of transition metals, supporting a myriad of functions in their proteinaceous clients. As such, transition metals have been studied extensively and ‘metallomics’ has emerged as a rapidly expanding field [2].

Metals must be obtained externally to maintain sufficient cellular or organismal pools. However, metal levels must also be strictly regulated to prevent toxicity at high concentrations. Disruption of metal homeostasis can have detrimental effects on an organism, and various neurodegenerative and cardiovascular diseases as well as cancers have been associated with alterations in metal levels [3–6]. Analogous to eukaryotic systems, bacterial pathogens rely on metals to replicate [7]. During infection, vertebrate immune systems take advantage of this dependency and actively restrict pathogen access to transition metals in processes termed nutritional immunity [8, 9]. Consequently, the field of host-pathogen interactions has defined metal competition as a critical step dictating the outcome of infection [7]. In response to the broad importance of metals in biologic systems, analytical technologies have emerged for metal analysis and imaging.

Elemental imaging technologies allow visualization of both distributions and relative abundances of metals within tissue samples [10]. Imaging studies have focused on many aspects of metal biology, ranging from analysis of individual metalloproteins to complex pathways in mammalian organ systems. The development of multimodal imaging strategies has been aided by the utility of various in vivo imaging technologies in diagnostic and clinical applications [11]. Despite the outstanding contributions imaging has made to diagnostics, we will focus in this brief review on recent advances of metal imaging in biomolecular research. We discuss current studies that integrate multiple imaging modalities with a specific focus on metal distributions and correlative biomolecular distributions. Resultant information highlighting two-dimensional (2D) and three-dimensional (3D) localization of analytes within a tissue sample can provide insight into the chemical and elemental dynamics of a biological system or disease state. By combining the information derived from multiple imaging modalities, insight into complex biological interactions and mechanistic processes can be achieved.

Imaging Modalities Provide Insight into Biological Metal Acquisition

In the following two sections, we will give a brief overview of methods used for multimodal elemental imaging as well as complementary technologies providing morphological and molecular sample characterization. A summary of the techniques and technologies described in this review can be found in table 1. Discussion of technologies to discern metal distributions will focus on both native and xenobiotic metals, and exclude diagnostic techniques such as radioisotopic and contrast imaging, except when complementary or necessary for integration. Excellent recent reviews on diagnostic imaging have been published [11, 12].

Table 1.

Techniques and technologies commonly employed for molecular or elemental imaging.

Name	Abbreviation	Brief Definition
X-ray fluorescence	XRF	Excitation of an atom by an electron, proton, or photon beam to be identified and quantified by the light emitted upon relaxation.
Micro X-ray fluorescence	μXRF	Subcellular resolution XRF
Electron probe x-ray microanalysis	EPXMA or EPMA	Specific adaptation of XRF principles to allow for μXRF
Particle induced x-ray emission	PIXE	Specific adaptation of XRF principles to allow for μXRF
Synchrotron x-ray fluorescence microscopy	SXRF	Specific adaptation of XRF principles to allow for μXRF
X-ray fluorescence imaging	XFI	Application of XRF or μXRF to the entirety of a sample, acquiring data at discrete x,y locations to represent data as false colored images.
Mass spectrometry	MS	Measurement of ions in the gas phase that can be produced through a variety of ionization processes
Inductively coupled plasma	ICP	Ionization strategy in which high energy plasma is used to produce elemental ions from a sample while simultaneously eliminating molecular structures and most chemical interferences.
Laser ablation	LA	A stage can that sample specific x,y locations across a surface typically using an ultraviolet laser.
Secondary ion mass spectrometry	SIMS	A technology where high energy primary ions generated by an ion gun are targeted to specific x,y locations on the sample surface. Impact of the primary ions both eject and facilitate the formation of secondary ions to be analyzed by MS.
Imaging mass spectrometry	IMS	A technology allowing for the visualization of distribution and abundance of ions from mass spectra as colored intensity maps.
Hematoxylin and eosin tissue stain	H&E	The most common tissue stain that allows for the determination of gross tissue morphologies and cellular differentiation.
Immunohistochemistry	IHC	An antibody-based staining technique, targeting specific antigens within a tissue section.
Imaging mass cytometry	IMC	An expansion of immunostaining that incorporates heavy metal conjugated antibodies to be analyzed by LA-ICP IMS.
Bioluminescent imaging	BLI	Use of light producing proteins such as luciferase to track or localize biological responses.
Fourier transform infrared spectroscopy	FTIR	A technology that results in a spectral ‘fingerprint’ of a sample without providing specific biochemical information. Infrared radiation directed on a sample is absorbed through transitions in molecular vibrational energy states and resulting spectra are read out on a pixel by pixel basis.
Matrix-assisted laser desorption/ionization	MALDI	Ionization strategy in which a small, organic chemical matrix that aids in analyte desorption and ionization when irradiated by an ultraviolet laser.
Magnetic resonance imaging	MRI	A common biomedical imaging modality that uses magnetism and radio waves to allow for 3-dimensional contrast images.
Autofluorescence microscopy	AF	Microscopy of fluorescence observed from endogenous chemical within tissues.
Fourier transform ion cyclotron resonance	FT-ICR	A high-resolution mass analyzer for MS that uses the cyclotron frequency of trapped ions in a magnetic field to determine the mass-to-charge ratio.
Atomic force microscopy	AFM	A very high resolution microscopy technique that uses piezoelectric elements to ‘touch’ or ‘feel’ a surface with a mechanical probe.
Absorption microcomputed tomography	μCT	A technology that uses X-ray cross sections to create virtual 3-dimesional models of objects.

1.1. Elemental Imaging Modalities for the Investigation of Metal Content and Distributions

A variety of strategies can be used to visualize metals within a thin tissue section [10]. Stains and chemical probes are the most basic and accessible reagents to enable visualization of endogenous metals within a tissue section. Here, introduction of chemicals to the surface of a tissue enables metal observation by either a visual or fluorescent signature. Chemicals required for this type of visualization typically select for only one elemental species and vary in sensitivity [13].

An unbiased and more sensitive method to simultaneously image multiple metals is X-ray fluorescence (XRF). This technology relies on atomic excitation and relaxation principles. In XRF, an atom is excited by an electron, proton, or photon (x-ray) beam. Multiple elements can then be identified and quantified by the light emitted upon relaxation. Depending on the specific adaptation of these principles, distinct modalities can allow for subcellular resolution (e.g. micro X-ray fluorescence microscopy (μXRF), electron probe x-ray microanalysis (EPXMA or EPMA), particle induced x-ray emission (PIXE), and synchrotron x-ray fluorescence microscopy (SXRF)). Resultant XRF datasets can be represented as false colored images, termed x-ray fluorescence imaging (XFI) [14–16].

Mass spectrometry (MS) allows the measurement of ions in the gas phase that can be produced through a variety of ionization processes. Metal ion analysis is routinely accomplished using inductively coupled plasma (ICP)-MS in which high energy plasma is used to produce elemental ions from a sample while simultaneously eliminating molecular structures and most chemical interferences [17]. A coupled laser ablation (LA) stage can sample specific x,y locations across a surface [18, 19].

Secondary ion mass spectrometry (SIMS) is an MS-based technology that enables imaging of metals from a tissue section with subcellular spatial resolution. High energy primary ions generated by an ion gun are targeted to specific x,y locations on the sample surface. Impact of the primary ions both eject and facilitate the formation of secondary ions to be analyzed by MS [20, 21]. Although the spatial resolution of SIMS surpasses that achievable with LA-ICP-IMS (e.g. ~1 μm / pixel with LA-ICP IMS and ~ 500 nm / pixel with SIMS), few metallic species ionize using SIMS [20]. The distribution and abundance of any ion recorded as mass spectra from both types of MS-based technologies can be visualized as colored intensity maps. This process of the production of images of specific ions is termed imaging mass spectrometry (IMS) [18, 22].

1.2. Insight into Biological Processes through Application of Molecular Imaging Modalities

While images of metal distributions within a tissue section can be insightful, complementary imaging strategies can provide morphological or molecular information, resulting in biological context of metals within complex systems. Among the strategies that can be interfaced with elemental imaging, histology using color stains (e.g. H&E) is the most basic and allows for the determination of gross tissue morphologies and cellular differentiation [13]. Immunohistochemistry (IHC) relies on antibody-based staining, targeting specific antigens within a tissue section [23]. Visualization of fluorophore conjugated antibodies using fluorescence microscopy offers advantages over traditional IHC including the potential for multiplexing through use of fluorophores with different emission/excitation spectra. A further expansion of immunostaining incorporates heavy metal conjugated antibodies to be analyzed by LA-ICP IMS, termed imaging mass cytometry (IMC). This experimental strategy allows the application of up to 30 antibody tags to be analyzed in a single experiment, although with limited throughput due to extensive analysis times [24]. In addition to using histological or antibody-based methods to characterize a tissue of interest, fluorescent or bioluminescent reporters can be employed and detected via fluorescent microscopy or in vivo (3D) bioluminescent imaging (BLI). Transcriptional or translational reporter constructs can be designed to address an array of biological questions and, in context of this review, can help delineate the cellular, organellar, or molecular makeup of a sample in question.

Additional methods are available and frequently employed to allow unbiased and unlabeled analysis of tissues. For example, Fourier transform infrared spectroscopy (FTIR) results in a spectral ‘fingerprint’ of a sample without providing specific biochemical information. Here, infrared radiation directed on a sample is absorbed through transitions in molecular vibrational energy states and resulting spectra are read out on a pixel by pixel basis [25].

IMS is not limited to the analysis of elements. The technology can be leveraged to map the distributions of thousands of molecular analytes directly from the surface of a tissue section without a priori knowledge of the analytes present [26]. Amongst a variety of molecular IMS sampling/ionization techniques, matrix-assisted laser desorption/ionization (MALDI) IMS is most common and enables analysis of multiple biomolecular classes including metabolites, lipids, peptides, and intact proteins [26–28]. In MALDI IMS, a tissue section is homogenously coated with a matrix, and an ultraviolet laser performs a raster at discrete x,y locations across a sample’s surface, resulting in coordinate mapped mass spectra. Acquired spectral data can be visualized as ion images [22]. Recent studies have leveraged MALDI IMS as a discovery approach in tandem with methods assessing metallic distributions to probe biological systems [29**–32].

1.3. Integrating multiple imaging modalities

1.3.1. 2D modalities

Multimodal imaging studies can incorporate a variety of acquisition technologies resulting in multiple data dimensionalities, image spatial resolutions, and data sizes. However, a common goal of image integration is to synergistically combine information derived from multiple technological sources to elucidate complex biological processes or interactions in situ. The key challenges in any multimodal study are image registration and data mining of the resultant datasets.

Verbeeck et al. recently advanced data mining strategies for multimodal data by integrating spatial molecular data with conventional models or scaffolds. In an automated approach, structural context provided by magnetic resonance imaging (MRI) and the publically accessible Allen Mouse Brain Atlas (AMBA) was correlated to IMS data [33, 34]. Specifically, IMS-atlas integration was used in one example to perform automated anatomical interpretation to isolate differential ions in specific tissue sub regions in normal rat brain and a Parkinson’s disease rat model [34]. However, this strategy has yet to be applied to elemental imaging data.

Evaluable, accurate registration between modalities was recently accomplished by Patterson et al. By leveraging the nondestructive nature of autofluorescence microscopy (AF), tissue structures can be identified and ultimately allow for computational image registration. IMS and AF data were aligned to specific morphological information from an H&E stain within a single tissue section. IMS datasets of different spatial resolutions were correlated based on this AF workflow to isolate data from the same physical locations with a computational metric to determine pixel overlap [35–37]. A consideration when using the AF registration workflow is the potential elemental, molecular, or morphological differences when analyzing and combining information from serial tissue sections.

In cases where common modalities and advanced image registration are not employed, effective multimodal imaging can be performed by acquiring multiple assay types from a single tissue section. A study by Holzlechner et al. overcame potential differences in serial tissues sections by analyzing the molecular and elemental constituents of the same tissue section in subsequent imaging experiments. Specifically, this study achieved quantitative LA-ICP IMS measurements of platinum from individual tissue sections previously subjected to MALDI IMS. Statistical analyses isolated molecular species that correlated with platinum distributions [29**]. Analogous to multimodal studies of metal and chemical distributions, Svirkova et al. mapped elemental and lipid distributions within a single bone tissue section by utilizing the nondestructive capabilities of μXRF imaging (μXFI) in tandem with MALDI IMS to (Fig 1A) [38].

Figure 1: Overview of recent studies employing elemental imaging in 2D.

A) Bright field microscopy images of sectioned bone (left) from two sample preparations (top and bottom). Images depicting the abundance of two lipids (m/z 760 in red, m/z 725 in green) as well as calcium distributions (blue) as detected by MALDI IMS and μXFI respectively are shown on the right. Adapted from Svirkova et al [38] B) NanoSIMS ion image overlay of copper (blue) and phosphorous (red) from zebrafish retina (left) and an electron micrograph of similar region (right) with false colored labels highlighting nuclei (red) and megamitochondria (blue). Inset, corresponds to one megamitochondrion and indicates co-localization of copper puncta and megamitochondria. Adapted from Ackerman et al [44***]

Another active area of research often employing elemental imaging by XFI in combination with a variety of additional modalities is in the various sub-fields of neuroscience, as reviewed elsewhere [39]. Within this area of neurological research, XFI has been applied to many multimodal imaging studies, often in combination with FTIR, with biological interests including Alzheimer’s disease, neurological effects of diabetes, intracerebral hemorrhage, and brain aging [40–43].

While previously mentioned studies successfully resolved elemental distributions in tissues at or above cellular resolutions, Ackerman and colleagues combined LA-ICP-MS and nanoSIMS with confocal and electron microscopy to quantify subcellular copper accumulation in a Zebrafish model of Menkes disease (Fig 1B) [44***]. This experimental setup highlights the power of combining metal imaging with other high-resolution modalities.

Beyond image registration, machine learning methods have been developed to computationally fuse image data from modalities with disparate spatial resolution and information content. This approach, termed data-driven image fusion, enables a number of predictive applications including increased effective spatial resolution, non-biological noise removal, and out-of-sample prediction [45]. Data-driven image fusion allows for correlation between data from high information volume imaging technologies such as IMS with more fine-textured, but lower information content imaging such as stained microscopy. As a proof-of-concept, Van de Plas et al. computationally combined stained microscopy with MALDI IMS to predictively increase the spatial resolution of MALDI IMS data beyond the capabilities of current instrumental platforms [46]. This approach is intriguing for the investigation of specific cell types in tissues and as a way to comprehensively mine molecular relationships as they relate to fine tissue structures. Building on these preliminary studies, our group has successfully employed fusion to combine high-performance MALDI Fourier transform ion cyclotron resonance (FT-ICR) IMS and H&E microscopy generated from renal tissue of mice systemically infected with Staphylococcus aureus. We show the predictive distributions of two intact protein distributions in and around individual bacterial abscesses at 15 μm spatial resolution (Fig 2). It is noted that to acquire the raw FT-ICR IMS data at 75 μm spatial resolution (10,000 total pixels) required ~9 hours. If the experiment was performed at the predicted spatial resolution of 15 μm it would have resulted in >250,000 pixels and would have required >9 days of continuous data acquisition. This is an example of the necessity for computational predictions where generation of the data exclusively through experimentation would be impractical or impossible. In addition to these examples from our research group, fusion has been applied in other studies combining other modalities such as SIMS, FTIR, scanning electron microscopy (SEM), and Raman spectroscopy [47–49]. However, to our knowledge, fusion has yet to be applied to elemental imaging data.

Figure 2: Data-driven image fusion applied to MALDI FT-ICR IMS of intact proteins from a S. aureus infected murine renal tissue section.

Multivariate linear regression is used to construct a cross modality model from H&E stained microscopy (1 μm spatial resolution) and MALDI FT-ICR IMS of intact proteins (75 μm spatial resolution, unknown protein at m/z 9,183.46, S100A8 at m/z 10,164.07) acquired from the same 7 day post S. aureus infected murine renal tissue section. The fused IMS-Microscopy data set enables prediction of protein ion images to higher spatial resolution (15 μm spatial resolution). Statistical measures of confidence are provided for both predictions (Reconstruction Scores) [46].

1.3.2. Visualizing multimodal imaging data in 3D

Thus far, this review has focused on studies that highlight 2D multimodal imaging approaches; however, biological systems inherently occupy 3D spaces. Therefore, the field of metal imaging has witnessed a trend towards investigations attempting to axially resolve elemental distributions. Because many elemental and molecular imaging experiments are inherently 2D, most approaches require sectioning and analysis of many serial tissue sections to generate 3D data. This strategy presents several challenges, including sequential image registration from tissue that can be damaged during sectioning, large data files from combining many multimodal 2D imaging experiments into a single 3D volume, and finally, computational demands required for 3D data volume reconstruction and mining [50]. Several studies have overcome these challenges to build 3D multimodal molecular images.

To entirely circumvent the need for tissue sectioning, Ghosal et al. used ablation depths generated by nanoSIMS to acquire 3D volumes of spores [51]. However, calculating the depth from SIMS ablation is challenging due to sample density variations. To address this challenge, Moreno et al. implemented depth probing and monitoring by atomic force microscopy (AFM) to allow for more accurate 3D analysis [52]. While this strategy is successful for smaller objects, visualization of analytes within large structures typically requires data to be collected from multiple tissue planes.

In an early example of 3D metal imaging, Hare et al. performed LA-ICP-IMS on 46 tissue slices of murine brains. Subsequently, 3D modeling was performed based on alignment of common morphological features. Although not strictly multimodal, this approach was an important step towards evaluating metal distribution throughout complex tissues [53].

To investigate multi-organism systems in 3D, Cassat et al. used iterative blockface imaging in combination with MRI, 3D bioluminescent imaging (BLI), and IMS to create a 3D model of S. aureus infected murine tissues (Fig 3A). In vivo MRI and BLI measurements allowed the creation of 3D volumes that depict anatomical, elemental, and molecular changes resulting from staphylococcal infection and host nutritional immunity. Notably, the achieved spatial resolution enabled investigation of individual abscesses and allowed for the characterization of heterogeneous bacterial gene expression and metallic distributions within a single organ. Due to the 3D structure and distributions of abscesses, multimodal volumes captured bacteria as well as immune cell distributions in different planes [32***].

Figure 3: Overview of recent studies employing elemental imaging in 3D.

A) Multimodal imaging of a murine kidney infected with S. aureus. Iron distribution (yellow, determined by LA-ICP-MS) and bacterial Fe-starvation dependent reporter activity (red sphere, assessed by BLI) are co-registered to blockface images. This image depicts one representative view of an entire 3D volume. Adapted from Cassat et al [32***] B) μCT 3D renderings of C. dubia (a-b) and distribution of selected elements (calcium: red, manganese: green, zinc: blue) as determined by μ-XRF (c-d). Adapted from Van Malderen et al [54**] C) μCT 3D image of Triticum aestivum. L. grain and elemental distribution throughout the sample as determined by LA-ICP-MS. Adapted from Van Malderen et al [55]

Modern imaging technologies allow for small animals to be visualized in their entirety. Using computational tomography guided slice registration, Van Malderen and colleagues combined absorption microcomputed tomography (μCT), 3D confocal μXFI, and LA-ICP-IMS to create 3D volumes of the metal distribution throughout the entire Ceriodaphnia dubia (water flea) body (Fig. 3B) [54**].

Most current 3D approaches rely on the sectioning of biological material. While various microtome technologies are commercially available, the sectioning of hard or brittle material remains challenging. In order to analyze the metal content of wheat and rye seeds, specimens were epoxy-embedded, stepwise polished to achieve lateral increments of 350 μM, and analyzed via LA-ICP-MS and complementary XFI. 3D volumes were generated by feature-based registration and compared to μCT renderings (Fig 3C) [55]. The studies above exemplify multimodal investigations of molecular and elemental analytes while perserving spatial fidelity. Three-dimensionality is achieved by creative application of various sectioning and depth-probing methods as well as registration strategies in order to reflect the true spatial properties of biological samples.

Perspectives

To fully characterize complex biological interactions or processes within tissue environments, multiple levels of information must be obtained and combined. Despite many advances in imaging, there is no universal imaging technique that captures elemental, molecular, and morphological information in a single experiment. Many critical developments have been made towards correlative multimodal imaging. Recent efforts to apply elemental imaging approaches in tandem with anatomical, morphological, or molecular imaging modalities have placed a precedent on how multimodal imaging strategies provide insight to biological systems. Specifically, increasing spatial resolutions without the sacrifice of molecular or elemental information have allowed for subcellular localization of analytes, spatially verified and complemented by other imaging modalities. Increasing acquisition speeds of molecular imaging technologies have allowed feasible acquisition of large 3D data volumes in tandem with elemental information. The utility of these approaches is clearly highlighted by the broad nature of studies that employ multimodal imaging strategies. Studies reviewed here reach from basic physiology and plant biology to various human diseases including bacterial infections. However, the combination of imaging information still faces many challenges including registration, alignment, 3D volume constructions, and the cost of data analysis. While these constraints can be surmounted, there is currently a limited number of published 3D multimodal studies. Implementation of accurate registration techniques, data fusion, and data analysis automation to 2D and 3D multimodal approaches will not only lessen the time of analysis, but also increase the use of multimodal imaging strategies throughout the scientific community. Nevertheless, the wealth of elemental, molecular, and morphological information acquired through the combination of modalities in 3D volumes, outweighs the technical challenges and allows unparalleled insight into biological systems. Therefore, the development, application, and integration of multimodal imaging technologies will be highly valuable for future studies addressing the importance of metals for nearly every aspect of life.