Cellular Imaging of Intracellular Bacterial Pathogens

ABSTRACT

The spatial dimensions of host cells and bacterial microbes are perfectly suited to being studied by microscopy techniques. Therefore, cellular imaging has been instrumental in uncovering many paradigms of the intracellular lifestyle of microbes. Initially, microscopy was used as a qualitative, descriptive tool. However, with the onset of specific markers and the power of computer-assisted image analysis, imaging can now be used to gather quantitative data on biological processes. This makes imaging a driving force for the study of cellular phenomena. One particular imaging modality stands out, which is based on the physical principles of fluorescence. Fluorescence is highly specific and therefore can be exploited to label biomolecules of choice. It is also very sensitive, making it possible to follow individual molecules with this approach. Also, microscopy hardware has played an important role in putting microscopy in the spotlight for host-pathogen investigations. For example, microscopes have been automated for microscopy-based screenings. A new generation of microscopes and molecular probes are being used to image events below the resolution limit of light. Finally, workflows are being developed to link light microscopy with electron microscopy methods via correlative light electron microscopy. We are witnessing a golden age of cellular imaging in cellular microbiology.

INTRODUCTION

Cellular imaging encompasses a large variety of methods that can be considered to be among the most powerful tools for investigating the molecular and cellular details of host-pathogen interactions, particularly when microbes display an intracellular lifestyle. This field has been expanding rapidly during the last 20 years and will likely gain further relevance as it continues to develop. Optical, fluorescence-based approaches are especially popular among infection biologists. In addition, we are witnessing a renaissance of electron microscopy (EM) due to numerous novel ultrastructural approaches changing the perception of EM from that of a confirmatory approach to that of a tool that drives new biological questions.

The explosion of optical, imaging-based studies is based on the following. First, a new generation of microscopes has been brought to the market during the last 2 decades. These microscopes are more easily accessible to biologists, the hardware—especially light sources and detectors—has been dramatically improved, and microscopes are now typically fully automated. Second, molecular and cell biologists have developed a huge catalogue of genetically encoded probes, which started in 1994 with the application of green fluorescent protein (GFP) for cellular imaging (1). Simultaneously, chemists have worked with cell biologists to synthesize a plethora of low-molecular-weight, fluorescent probes or sensors to investigate specific cellular and functional features. Third, computer-assisted analysis of imaging data has moved cellular imaging away from a descriptive and qualitative discipline towards the realm of quantitative studies. Unbiased algorithms are able to extract “high-content” information from the obtained images. These developments are enhanced by open-source and user-friendly software, such as FIJI (http://fiji.sc/Fiji), Cell Profiler (http://www.cellprofiler.org), and ICY (http://icy.bioimageanalysis.org). In addition, analytical pipelines based on artificial intelligence, including deep learning and neural networks, have recently emerged, underlining the interdisciplinary character of this research field.

Cellular imaging for infection research is highly complex, often requiring expertise not only in biology but also in physics (particularly optics), engineering, mathematics, computer science, and chemistry. Therefore, delivering exhaustive information on cellular imaging in a short article on the interaction between pathogens and their hosts is elusive. Here, we provide a short description of some important principles of optical imaging, including methods that breach the diffraction limit of light. In addition, we address some developments commonly considered crucial for the study of host-pathogen interactions, such as labeling, “functionalized” probes, and optogenetic assays. Then we describe screening approaches focusing on the statistical methods used to quantify image-based screening data. Finally, we highlight methods that combine optical imaging with ultrastructural approaches, in particular the game-changing character of large-volume EM methods.

MICROSCOPY AND FLUORESCENCE

Microscopy Principles: Diffraction Limit, Contrast, and Fluorescence

While most microbes are visible by optical microscopy, viruses and some very small bacteria require ultrastructural imaging. Therefore, using imaging techniques to study the interaction of microbes with their host cells is common sense. In doing so, the scholar Antonie Van Leeuwenhoek succeed in discovering the first microbes, or “animalcules,” in the late 17th century. Since then, microscopy techniques have continued to be developed, pushing the boundary of what biologists can see. The human eye can distinguish objects down to 100 micrometers, and microscopy is required for anything smaller (Fig. 1). There are numerous excellent textbooks on the fundamental principles of optics and its use in imaging; we recommend Principles of Fluorescence Spectroscopy by J. R. Lakowicz for its comprehensive character (2).

FIGURE 1.

Wavelengths, objects, and their recognition. The sizes of different objects and the tools that can be used for their visualization are shown.

Light can be characterized either as a wave or as particles. Its wave characteristics are fully described by the Maxwell equations. It is visible in the range from purple (about 400 nm) to red (800 nm). The wave character of visible light determines the resolution limit of an optical microscope; it is about 250 nm in the x and y directions and about 450 to 700 nm in the z direction. This was formulated by Ernst Abbe in 1870 with the famous equation d = λ/2NA (where d is the resolution limit, λ is the wavelength, and NA is the numerical aperture of the objective). As many biological samples do not provide sufficient contrast to be visualized in a light microscope, researchers have developed different contrasting methods throughout the last 100 years. Contrast can be enhanced by using simple bright light illumination, for example, through the phase-contrast procedure (developed by Frits Zernicke in the 1930s), by differential interference contrast (also called Nomarski contrast), or by dark-field contrasting. These techniques do not provide molecular specificity for the imaged objects; therefore, other methods—mainly based on fluorescence—have been established to provide labeling specificity (see “Fluorescent Probes” below for specific examples).

Fluorescence is a widely present phenomenon of light-matter interactions. It describes how molecules interact with light through a defined sequence of light absorption and emission at specific wavelengths at defined energy states. This phenomenon is commonly depicted in a Jablonski diagram (Fig. 2). The light absorption leads to the excitation of the electron cloud of a given molecule to an active state (also called S₁) from its ground state (S₀). The energy from the excited state is emitted within the range of a couple of nanoseconds. Before, energy is lost through collisions or vibrations or crossing to other energy states. The difference between the excitation and emission energy becomes “visible” through the Stokes shift, which is defined as the difference between the excitation peak wavelength and the emission peak wavelength. Fluorescence imaging has become one of the most widely employed techniques in cell biology and infection biology due to its specificity (biomolecules of interest can be labeled) and its sensitivity (single fluorescent molecules can be detected).

FIGURE 2.

The Jablonski diagram explains the basic principles of fluorescence. Molecules are excited by incoming light to reach a higher energy level (S₁). After vibrational loss of energy (among other losses), the molecule falls back to its low energy level (S₀), emitting fluorescent light. An excitation-and-emission spectrum for a hypothetical fluorophore that could be similar to GFP is shown on the right. Image courtesy of Gael Moneron (Institut Pasteur).

Superresolution Microscopy

Objects that are closer to each other than the resolution limit cannot be distinguished as separated. Superresolution techniques break the diffraction limit by temporally or spatially modulating the excitation, or activation, light or by specific postprocessing of the obtained imaging data, exploiting the blinking properties of fluorophores and their point spread functions (3).

Stimulated emission depletion (STED) is a technique that increases the resolution by shaping the excitation light and optically modifying the diffraction limit to reduce its effective diameter. This is achieved by overlaying two excitation beams, one with a doughnut shape and another that is superimposed. The doughnut shape can be modified to enable precise illumination in the subdiffraction range. The two-dimensional resolution for STED is between 30 and 70 nm, with no increase in axial resolution.

Photo-activated localization microscopy (PALM) and stochastic optical reconstruction microscopy (STORM) techniques are based on single-molecule blinking. They overcome the diffraction limit by using light to turn on only a sparse subset of the fluorescent molecules of interest. Using a mathematical postprocessing step, they estimate the centroids of the localized points to create a reconstructed superresolution image. The difference between PALM and STORM lies in the nature of the fluorescent labels, which are either photoswitchable organic dyes such as Alexa Fluor 647 or Cy5 for STORM or photoswitchable fluorescent proteins for PALM. Currently, these techniques allow localization of the position of the molecule of interest with a precision of 20 nm.

Finally, structured illumination microscopy (SIM) illuminates the entire field with striped patterns of light (4). These excitation patterns mix with the spatial pattern of the sample, and they produce interference patterns, called moiré fringes. To reconstruct a superresolution image, multiple images are taken with different pattern orientations, the information from the frequency space outside the observable region is collected, and the illumination pattern is removed by postprocessing. SIM provides a resolution of about 100 nm in the x-y direction and 400 nm axially (5).

Despite a significant improvement of the spatial resolution, standard superresolution techniques provide a lower temporal resolution, and they are further limited because of photobleaching and the phototoxicity, as they require longer time and light exposure to acquire the sample. In addition, most superresolution techniques require special sample preparations, which can be limiting for many experiments.

FLUORESCENT PROBES

The steady development of novel fluorescent probes for the labeling of specific molecular and cellular structures has been a driving force to make fluorescence microscopy a workhorse for cell biologists and infection biologists (6). Fluorescent probes are typically designed to localize low numbers of labeled biomolecules down to individual molecules inside fixed or living cells. While the majority of published studies have employed protein labeling, it is also possible to label all kinds of biomolecules, including nucleic acids, lipids, and sugars. Labeling is achieved through a specific affinity between a small fluorophore and the molecule of interest, such as fluorescently labeled phalloidin intercalating into F-actin structures (Fig. 3).

FIGURE 3.

Fixed confocal acquisition of ruffle formations and Salmonella entry in HeLa cells. All salmonellae expressed a fluorescent plasmid (in red). Salmonella lipopolysaccharide is immunolabeled before cell permeabilization (in green). The host actin cytoskeleton and nuclei are stained with phalloidin (in grey) and DAPI (4′,6-diamidino-2-phenylindole; in blue). Bars, 10 μm.

Additionally, genetically encoded fluorophores allow the expression of fluorescent protein chimeras of molecules of interest or genetically fluorescent sensors (see “Genetically Encoded Probes for Cellular Compartments” below). Numerous engineered GFPs and their derivatives are available, and they span the entire visible spectrum of light. Among them are molecules like mCherry (7), plant-derived fluorescent proteins, such as miniSOG or iLov (8), and others, such as dsRed (6). The physical properties of these fluorophores are highly variable and need to be taken into account by the experimentalist. For example, subtleties of the spectra, bleaching properties, and quantum yield need to be assessed. Also, in some cases, the fluorescent proteins impede the function of the tagged proteins; hence, alternative approaches have been developed to overcome issues arising through the bulkiness of GFP derivatives.

Genetically Encoded Probes for Cellular Compartments

Tagging of host factors is particularly useful for the study of intracellular pathogen trafficking. Within cells, many pathogens can remain within a membrane-enclosed vacuole and avoid fusion with lysosomes to create a unique replicative niche. Alternatively, others escape from this niche, rupturing their vacuole to access the host cytosol for subsequent spread (9). Vacuolar maturation is characterized by the interplay of diverse bacterial effectors and host factors that modulate the lipid and protein composition of the vacuolar membrane. During these processes, small GTPases such as Rabs, ARFs, and Rhos are key regulators of host vesicular trafficking and cytoskeleton dynamics. Tracking their localization by fluorescence microscopy has provided a large amount of information on the host pathways subverted by each pathogen during invasion.

These trafficking events are commonly studied using time-lapse optical microscopy with cells transformed or transfected with DNA coding for fluorescently tagged proteins (Fig. 4). Many pathogens induce membrane ruffles at their entry site or macropinocytic cup formation that can be observed using fluorescently tagged actin or Rac1. Macropinosomes can be followed using Rab5, SNX1/5, or EEA1. Maturing endosomes are monitored with Rab7 as classical probes for late endocytic compartments and with LAMP1 for their endolysosomal identity. The phosphatidylinositol phosphates (PIPs) are lipids that provide a specific identity to each endosomal compartment. For example, PI(4,5)P2 can be used to investigate early membrane ruffling, while PI(3,4,5)P3 localizes at the macropinocytic cup, and PI(3)P is enriched on early endosomes after scission from the plasma membrane. Fluorescent biosensors are used to detect the PIPs, such as the lipid-binding PX domain of p40phox or the tandem dimer of the FYVE domain of EEA1 (2xFYVE), which are most frequently used to detect PI(3)P (10, 11). However, considering that pathogens subvert the host in multiple ways to establish unique intracellular pathogen niches, the pathogen-containing endosomal compartment behaves differently from physiological compartments. Therefore, these trafficking markers may behave in unexpected ways in infected cells compared to noninfected cells. Clarity can be obtained only through in-depth studies of the involved trafficking molecules.

FIGURE 4.

Time-lapse imaging of epithelial cells infected with Salmonella using bacterial and host genetically encoded probes. HeLa cells were transfected with GFP-actin (top) or GFP-Rab5 (bottom) to follow the ruffle formation and the phagosomal trafficking upon infection with dsRed-expressing Salmonella (yellow arrowheads indicate ruffles). White bars, 10 μm; yellow bars, 2 μm.

Fluorescent reporters have also been developed to differentiate between vacuole-bound and cytosolic bacteria. For example, galectin-3 is a host lectin that can be used as a dynamic marker for vacuolar escape simply by expressing a fluorescently tagged version of the protein within the host cell. As long as the vacuolar integrity is maintained, the galectin-3 fluorescent signal is uniformly distributed within the entire host cell. Upon vacuolar rupture, the signal accumulates in the vicinity of the damaged membrane of bacterium-containing vacuoles, forming a readily visible galectin-3 “ghost.” This has revealed that Shigella escapes into the cytoplasm about 10 minutes after bacterial entry (12, 13). In addition, the galectin-3 marker and other galectins have been successfully used to follow vacuolar rupture in the context of various bacterial pathogens (e.g., Listeria, Salmonella, Legionella, etc.) (13–15).

Functionalized Probes

Molecular switches, such as small GTPases or kinases, regulate the host cytoskeletal and trafficking organization. Upon infection, these proteins are commonly targeted by pathogens to reorganize the host environment. They exist in an active form and an inactive form. To differentiate between these conditions and to delineate the molecular switches, functional fluorescent probes have been developed to measure their activity status through fluorescence energy transfer (FRET). A large catalogue of such probes is now available, and a few of them have been used for bacterial entry. For example, it was shown via ratiometric FRET image analysis that Rac1 is activated at the site of Yersinia entry in typically nonphagocytic cells using the interaction of the FRET pair Rac1-CFP with a yellow fluorescent protein fusion of a domain of its downstream adaptor, PAK1 (16). In the case of Listeria entry, the same principle was used to monitor Rac1 activation and the implication of small signaling lipids of the inositol-phosphatidyl family (17). Furthermore, FRET sensors were used to investigate vinculin activation during the early entry steps of Shigella (18). There are two hurdles that prevent the broad usage of the FRET probes. One issue is that most of them display a limited functionality; for example, it is not clear whether the probes localize in a manner identical to that of the endogenous proteins. Secondly, FRET measurements are often done through relatively slow ratiometric procedures. This can be improved via fluorescence lifetime imaging for measurements in real time, which are important to trace the highly dynamic events during the rapid entry of different pathogens.

We have developed functionalized probes to track vacuolar rupture under both fixed-cell and living-cell conditions in a robust way. This method is based on a cephalosporin-derived FRET probe, CCF4-AM, that can be cleaved by beta-lactamase. Using this assay in real time, we revealed the rapid escape of Shigella from the vacuole within less than 10 minutes after cellular entry, which was in line with the results obtained with fluorescently labeled galectins (12). This technique has been successfully adapted for other pathogens, including Mycobacterium tuberculosis, Francisella, and Listeria (19–21). It is highly versatile for high-throughput screening, as described in more detail below. Another fluorescence imaging approach relies on the use of fluorescent dextran, a fluid-phase marker for endocytic compartments that accumulates within the Salmonella-containing vacuole. By preloading cells with Alexa488-dextran overnight and challenging them with mCherry-expressing Salmonella, the integrity of the Salmonella-containing vacuole can be surveyed by time-lapse imaging assessing the colocalization of mCherry-Salmonella with Alexa488-dextran a few hours postinfection. This method has been used to demonstrate that the Salmonella pathogenicity island 2 type III secretion system is required for bacterial replication within the vacuole but not for cytosolic replication (22). This technique can differentiate between vacuolar and cytosolic bacteria, but unlike the galectin-3 and FRET probe methods, it does not yield precise quantitative information on the timing of vacuolar rupture.

Optogenetics

Optogenetics is a relatively new technique to modulate the localization or the activity status of biomolecules in a spatially and temporally defined way (23). Used in conjunction with regulatory proteins, this technique makes it possible to turn them on and off simply by shining light on the observed specimen.

Notably, optogenetics can be used to induce ruffling and macropinosome formation within chosen regions of a given cell. The pathway associated with receptor tyrosine kinase-induced macropinocytosis can be activated downstream of the receptor tyrosine kinase using optogenetics on modified small GTPases. Fujii and colleagues characterized the activation and deactivation of the small GTPase Rac1 using microscopic photomanipulation (24). Expression of the genetically encoded photoactivatable-Rac1 (PA-Rac1) in RAW264 macrophages enables the local and reversible control of macropinocytosis using blue laser irradiation. The irradiated region of macrophages under the persistent activation of PA-Rac1 displays PI(4,5)P2 accumulation, actin enrichment, membrane ruffling, and unclosed macropinosomes. Deactivation of PA-Rac1 by ceasing irradiation is needed to stop membrane ruffling and lead to the acquisition of maturation markers such as PI(3)P and Rab21 by the preformed macropinosomes. Thus, using an optogenetic technique shows that the targeted activation of Rac1 is sufficient to induce ruffles and macropinosomes. Through this approach, pathogens could be forced to be internalized at specific sites within challenged host cells.

IMAGING-BASED SCREENING

Loss-of-function screens have been instrumental in uncovering numerous features of the host-pathogen cross talk. Loss of function means a targeted depletion or reduction of expression of a specific gene. The most widely used approach still uses RNA interference (RNAi) as a suitable tool for gene function knockdown. The reason for this is the accessibility and the ease of use of small interfering RNA (siRNA) libraries against host genes. siRNA gene knockdowns can be combined with recent advances in imaging technologies (see above), including automated microscopy and image processing software that have been milestones in the development of large-scale screening approaches. Over the past 15 years, image-based RNAi screening has emerged as a powerful technique to unravel the molecular mechanism of intricate cellular processes such as cell division, membrane trafficking, and host-pathogen interactions (25, 26).

RNAi is a conserved posttranscriptional gene silencing process mediated by double-stranded RNAs (dsRNAs). In brief, the double-stranded RNAs are cleaved by the endoribonuclease Dicer, resulting in short fragments, typically of 21 to 23 nucleotides, referred to as siRNAs. These siRNAs are then incorporated into the RNA-induced silencing complex, where one strand of the siRNA, the antisense strand, serves as a guide to specifically recognize and pair with the complementary target mRNA, resulting in mRNA cleavage. siRNA can be chemically synthesized and delivered into the cells, commonly by liposome-based transfection or electroporation. Libraries of siRNAs targeting the entire genome or only subsets of genes involved in specific processes (e.g., membrane trafficking and apoptosis) or family members (kinases and molecular motors) are commercially available for medium- to high-throughput screening applications.

More recently, loss-of-function screens have also been performed via alternative gene-targeting approaches. CRISPR (clustered regularly interspaced short palindromic repeat)-Cas-based gene targeting in particular has become very popular. Through this method, genes can be targeted via small guide RNAs against specific host genes. The targeted genes are deactivated by the Cas9 nuclease, which forms a complex with the guide RNA. Despite the increased usage of CRISPR-based screening methods, they are still less well characterized than siRNA-based approaches, and obtained hits require a careful follow-up characterization with regard to their specificity.

Undoubtedly, image-based loss-of-function screening can yield valuable insight; however, its success requires careful assay design and data analysis. Practically speaking, the experimental pipeline of a screening assay can be divided into the following steps: (i) experimental planning, (ii) assay development and validation, (iii) primary screening, (iv) screen analysis and confirmation, and (v) hit characterization (25, 26). The initial step consists of clearly defining the biological question and appropriate design of the screening assay, including exercising particular care in choosing the model and the knockdown/knockout library (i.e., using a genome-wide library versus a selected subset). One of the most challenging steps (and usually the most time-consuming) is assay development and validation. Often, this step is underestimated by the experimentalists. Each individual experimental parameter (e.g., the knockdown or knockout procedures, fluorescence biosensor/stain, imaging setup, etc.) needs to be established and optimized, yielding a sensitive, robust, and reproducible assay. In particular, microscopy-based screening requires the development of an automated image analysis algorithm for reliable qualitative and/or quantitative measurements of large data sets.

The reliability of the assay (or quality control) is assessed using statistical tests such as the Z factor (or Z′) and the strictly standardized mean difference, which measure the magnitude of the difference between negative and positive controls, ensuring that the selection of effective positive and negative controls is maximal and that the assay can identify hits with a wide range of phenotypic effects (Fig. 5A) (27, 28). Once the assay is statistically validated, the actual screen is usually conducted in plate duplicates, typically taking a few days to weeks depending on the throughput.

FIGURE 5.

Statistical tools for imaging-based screens. (A) Statistical analysis tests of knockdown screens to assess quality control. (B) Statistical analysis tests for data normalization and hit identification from the screens. Mathematical formulas and interpretation are shown. Adapted from references 27 and 28.

Post-primary-screening analysis relies on a number of statistical methods for data normalization and hit selection (Fig. 5B). Data normalization is required to compare and combine data from different plates by removing systematic errors from the raw data. Widely used methods for data normalization are the z score and its variant the robust z score, which basically determine the number of standard deviations from the mean and the median, respectively, for the control population. Because the robust z score is based on the median, it is insensitive to outliers and thus more suitable for knockdown screens. Typically, hit identification relies on a standard deviation threshold, which commonly follow the empirical rule of ±3 standard deviations, that is directly linked to the P value in normally distributed data (i.e., Gaussian distribution).

Alternatively, a more sophisticated method for data normalization and hit identification is the robust strictly standardized mean difference, which has been shown to better measure the effect sizes across experiments. It is noteworthy that this statistical technique grants control of both the false-positive and false-negative rates and provides a valuable classification of the hit effects based on a rigorous probability interpretation (28) (Fig. 5B). Ultimately, the statistical analysis provides a list of target gene candidates that will be further validated and characterized.

Hit characterization usually involves a multidisciplinary approach combining cell biology (e.g., live-cell imaging), bioinformatics (e.g., protein network analysis), and/or biochemistry (e.g., a pull-down assay). Altogether, although they are challenging, image-based loss-of-function screening approaches open exciting avenues for the study of host-pathogen cross talk at the cellular level.

CORRELATIVE LIGHT AND EM OF LARGE VOLUMES

Many of the molecular players and general mechanisms that have roles during host-pathogen cross talk have been identified and characterized in some detail at the level of optical diffraction-limited microscopy. The above-described multidimensional fluorescence microscopy approaches have been instrumental in this, providing dynamic insights into host and pathogen factors during infection, as well as functional information through the development of a multitude of cellular reporters. Nevertheless, how these pathogen and host factors function in the three-dimensional cellular environment at subdiffraction resolution cannot be resolved by optical approaches. One way to tackle this issue has been the development of superresolution microscopy (see above). Most superresolution approaches rely on fluorescence and can provide increasingly high-resolution information (down to about 10 nm by PALM, STORM, or STED). However, with these methods, one visualizes only the labeled elements, not the complete cellular environment.

This global cellular environment can be imaged at molecular resolution by EM. Nevertheless, EM-based approaches are limited in their ability to fully describe complicated three-dimensional cellular interfaces between the pathogen and host, due to an important and often overlooked factor—their limits of acquisition volume. Only in the past 10 years have we seen the emergence of new EM techniques which have been developed to not be constrained by these limits. One of these techniques is focused ion beam scanning EM (FIB/SEM), in which large-volume tomograms are acquired (29). Briefly, the principle of large-volume FIB/SEM is that an embedded biological sample is exposed to a focused ion beam capable of removing a thin layer of material (5 to 10 nm thick) in a highly precise manner termed “milling.” Between sample millings, a scanning electron beam is used to image the newly exposed surface. By repeating this process hundreds or even thousands of times, a large sample volume can be acquired with up to 5-nm resolution in all axes. Data sets produced by FIB/SEM can be thoroughly examined from any orientation and subjected to detailed quantitative analysis. This allows the high-resolution three-dimensional visualization of intracellular pathogens and their hosts within a large sample volume, providing an unprecedented level of structural detail.

In recent years, FIB/SEM has been combined increasingly frequently with three-dimensional fluorescence microscopy using correlative light EM (CLEM) to address biological questions. CLEM enables the study of a single site of interest using two approaches, each offering unique advantages. Labs routinely employing fluorescence microscopy for the study of pathogens often have a wide arsenal of fluorescent labels (for example, GFP-labeled proteins and organelle-specific stains) that provide precise biological information about the localization of pathogens, proteins, organelles, and compartments. By initially examining samples with fluorescence microcopy, transient or rare biological events as well as specific regions of interest within a pathogen or host cell can be identified. By using a finder grid or another reference system during light microscopy acquisition, samples can be processed for FIB/SEM so that the region of interest that was initially identified can be located again precisely and acquired as part of a large volume.

In CLEM-FIB/SEM, a biological sample is initially imaged by light microscopy (typically three-dimensional confocal or epifluorescence microscopy plus deconvolution) followed by FIB/SEM acquisition at the same location (Fig. 6). Transient or rare biological events can be pinpointed by fluorescent labels prior to FIB/SEM acquisition, allowing imaging of precise stages of pathogen invasion not easily accessible by classic EM investigation. Furthermore, the fluorescent signals can later be correlated to details within the large ultrastructural volume, providing molecular labeling of the observed structures. Therefore, the defining feature of FIB/SEM is its ability to provide access to intricate structural detail placed within a much broader cellular context. Generally, fluorescently labeled pathogens (viruses, bacteria, or parasites) represent objects that can be readily exploited as alignment fiducials for CLEM. The FIB/SEM experimental workflow is relatively fast and simple compared to other advanced EM approaches, allowing quick turnaround times between biological experiments and structural insights. Thanks to this, CLEM-FIB/SEM constitutes a major investigative tool to facilitate ongoing research in a lab.

FIGURE 6.

Large-volume CLEM workflow via multidimensional confocal and FIB/SEM imaging. Biological samples are prepared on gridded glass-bottom slides for time-lapse imaging (1), and events are tracked dynamically at high resolution (2). After site identification under the light microscope (3), locations are retrieved in the electron microscope (4), and three-dimensional volumes are obtained by milling and scanning of the prepared specimen (5). Afterwards, both image data sets are correlated and segmented (6). A typical data set spans 10 μm by 10 μm by 10 μm. Invading Shigella organisms are depicted, the forming entry foci are segmented (gold), and macropinosomes (orange) in the vicinity of the entering bacteria (blue) are identified. Images were taken by Allon Weiner (Institut Pasteur).

Correlative FIB/SEM offers several important features: CLEM allows rare or transient biological events to be subjected to ultrastructural investigation. Acquisition of large cellular volumes significantly increases the chance that the event imaged by light microscopy will be fully contained within the ultrastructural volume and therefore unambiguously identified at high resolution. The three-dimensional information obtained by fluorescence microscopy (e.g., confocal) can be directly correlated with the FIB/SEM data, labeling pathogens, compartments, and organelles of interest. As the acquisition volume of FIB/SEM is much larger than that of conventional tomography, significantly more fluorescent signal can be correlated with ultrastructural data, enhancing the information content. CLEM-FIB/SEM can also be performed in conjunction with even larger-volume micro-computed tomography studies or even in vivo imaging (30).

THE NEXT CHALLENGES

“The sky is the limit” describes the feeling that one has in contemplating the ever-increasing number of imaging techniques for the investigation of infection at the cellular level. In this overview, we focus on techniques that we consider will be of key relevance in the coming years. A number of issues need to be tackled in a systematic way.

First, we need to bridge dimensions in imaging. Despite improved CLEM procedures, it is still challenging to obtain molecular resolution in the cellular context. This could be overcome through the consequent implementation of superresolution imaging in the CLEM pipelines. Alternatively, single-molecule methods need to become more accessible. This requires probe development and more sensitive detectors. Another issue is the integration of cellular imaging with tissue imaging that could be achieved either by developing multiphoton imaging or by using more accessible tissue models, for example, organoids.

A second challenge is the need for precise quantification. Cellular reporters need to be fine-tuned to depict the precise process that one wants to study. Also, consequent development of analytical pipelines, including algorithms that use neuronal networks for the analysis of subtle phenotypes, should become routine.

Finally, as imaging is based mainly on probes, one needs to take into account that they interfere with biological processes. In particular, the overexpression of fluorescently tagged proteins often interferes with the processes the protein of interest is involved in. This can be tackled through the labeling of the endogenous proteins using approaches as CRISPR-Cas. Another approach is the consequent development of organic probes with minimal interference, for example, coupling click chemistry with fluorescence imaging.

The consequent integration of the different microscopy-related fields will keep cellular imaging at the forefront of the understanding of the cellular processes taking place during infection.