Infections of the central nervous system (CNS) continue to be a major threat to human health. Many pathogens including viruses, bacteria, fungi and parasites can cause CNS infections (Kim, 2008). Among the fungi, Cryptococcus neoformans, Candida albicans, Histoplasma capsulatum, Coccidioides immitis, Aspergillus spp. and zygomycetes are among the most common causes of brain or meningeal infections (Gottfredsson and Perfect, 2000; Kleinschmidt-DeMasters, 2002; Chakrabarti, 2007; Murthy, 2007; Scully et al., 2008; Bariola et al., 2010; Liu et al., 2012). Among the fungal pathogens that cause meningeal and cerebral infection, C. neoformans is the most prevalent (Gottfredsson and Perfect, 2000; Kim, 2008; Liu et al., 2012).
Although cryptococcal cells enter the body through the respiratory tract and initially cause pneumonia, they can eventually disseminate into the brain via the bloodstream, causing meningoencephalitis (Mitchell and Perfect, 1995; Gottfredsson and Perfect, 2000; Kwon-Chung et al., 2000). To cause meningoencephalitis, blood-borne C. neoformans that is either free within the blood or contained within phagocytic cells (Chang et al., 2004; Kechichian et al., 2007; Charlier et al., 2009; Casadevall, 2010) must first become arrested in the brain vasculature, making contact with brain endothelial cells, and then transmigrate into the brain parenchyma. Therefore, arrest of free C. neoformans or C. neoformans containing phagocytes in the blood vessels, the interactions between organism and endothelium, and subsequent crossing of the blood–brain barrier (BBB) are critical stages leading to meningoencephalitis. Understanding the mechanisms involved in these processes are fundamental to our knowledge of the pathogenesis of this disease.
The BBB separates the brain parenchyma from the vascular compartment and is composed of specialized endothelial cells that are surrounded by the basal lamina, astrocytic end-feet, pericytes and microglial cells. Model systems have been established to study the interaction of fungi, especially C. neoformans with the BBB and subsequent invasion of the CNS. In this review, we first present a brief overview of the techniques and model systems that have been used to study CNS fungal infections. We will then focus on the recent application of intravital imaging to examine fungal migration to the CNS in real-time in a living mouse and will discuss the technical challenges and limitations of each approach.
Monolayers of brain endothelium in vitro under static and flow conditions; strengths and limitations
Fungal interaction with the endothelium of the BBB is a prerequisite to establishing CNS infection. To understand the mechanisms involved in the initial fungal and endothelial interactions and subsequent invasion, monolayers of primary and immortalized human brain endothelial cells (HBMEC) have been used as a model of the BBB in vitro (Chen et al., 2003; Chang et al., 2004; Jong et al., 2007; 2008a,2008b; Vu et al., 2009; Huang et al., 2011; Maruvada et al., 2012; Sabiiti and May, 2012). This in vitro model has been used to examine the C. neoformans and HBMEC interactions using a variety of techniques. Several useful assays have been developed that rely on microscopic observations, biochemical analysis and assessing the burden of organisms under various experimental conditions. Typically, C. neoformans is allowed to settle onto HBMEC monolayers, which are then washed to remove unattached fungal cells and adhesion to, and invasion of HBMEC by C. neoformans are observed by electron or optical microscopy (Chen et al., 2003; Chang et al., 2004; Jong et al., 2008a,b; Huang et al., 2011; Sabiiti and May, 2012) (Fig. 1A). The HBMEC are lysed and analysed for colony-forming units (cfu) to evaluate the frequency of adhesion to, and invasion of HBMEC (Chen et al., 2003; Chang et al., 2004; Jong et al., 2007; 2008a,2008b; Sabiiti and May, 2012). These assays suggest that C. neoformans invades HBMEC via transcytosis, the process by which the yeast cell moves through the interior of the endothelial cell to gain access to the brain (Chang et al., 2004) with reorganization of HBMEC cytoskeletal structures (Chen et al., 2003). It has been further demonstrated that invasion is mediated by the interaction between CD44 (the receptor for hyaluronic acid) expressed on HBMEC and cryptococcal hyaluronic acid (Jong et al., 2007; 2008b). By contrast, more recent results suggest that the invasion is independent of cryptococcal capsule (Sabiiti and May, 2012). These models also suggest that HBMEC protein kinase C-alpha and dual specificity tyrosine-phosphorylation-regulated kinase 3 are crucial for HBMEC invasion (Jong et al., 2008a; Huang et al., 2011). In addition, the transmigration of C. neoformans across the HBMEC monolayers has been studied with HBMEC grown in a transwell chamber (mimicking the luminal side of the vasculature) that allows transmigration through a collagen-coated microporous membrane (mimicking migration to the abluminal side of the BBB). Yeast cells are added to the top chamber and transmigration is determined by counting the cfu in the bottom chamber (Chen et al., 2003; Chang et al., 2004; Vu et al., 2009; Huang et al., 2011; Maruvada et al., 2012). This approach can determine the frequency of transmigration of C. neoformans across the HBMEC under various experimental conditions. More recently, C. neoformans phospholipase B1 has been shown to activate host cell Rac1 for invasion of HBMEC by means of this assay (Maruvada et al., 2012).
To extend the in vitro models, we recently developed an in vitro model of the cerebral vasculature under flow conditions (Shi et al., 2010) (Fig. 1B). This model has the advantage of mimicking physiological blood flow. The importance of vascular flow conditions and the resultant shear stress have been demonstrated in endothelial physiology and in leucocyte rolling and firm adherence (Uematsu et al., 1995; Albuquerque et al., 2000). Briefly, HBMEC are grown on a glass coverslip, placed in a flow chamber, and the flow chamber is placed on an inverted microscope stage equipped with a heated enclosure that keeps the chamber at 37°C. A syringe pump is used to draw buffer containing C. neoformans over the HBMEC monolayers at various shear stress that can be precisely controlled by modifying the flow. The interactions of the yeast with HBMEC (e.g. rolling and adhesion) are visualized by phase contrast microscopy and recorded for later analysis. We tested whether fungal–endothelial interactions showed any similarities to the leucocyte recruitment cascade (Dang et al., 2002; Kerfoot et al., 2008). Our experiments showed that C. neoformans did not appear to interact while flowing across a resting monolayer even when the endothelium was activated with lipopolysaccharide and very low shear stress was used (Shi et al., 2010). However, this does not preclude a role for shear stress for other organisms. For example, it has been reported that Neisseria meningitidis adhered in brain capillaries in which the shear stress was reduced (approximately 0.5 dynes cm−2) rather than the postcapillary venule in which the shear stress is higher (3.0 ± 1.5 dynes cm−2) (Mairey et al., 2006). Moreover, dynamic models using endothelial monolayers within hollow microporous fibres may provide additional insights (Cucullo et al., 2008).
In vitro models for fungal–endothelial interactions have advantages as well as drawbacks. One great advantage is that the timing of interactions (adhesion and transmigration) of C. neoformans with HBMEC can be precisely controlled. More importantly, the in vitro reductionist model has great power to identify molecular mechanisms that govern the invasion of yeast cells into HBMEC by drawing on a wealth of genetic, immunological and pharmacologic approaches (Jong et al., 2008a,b; Huang et al., 2011; Maruvada et al., 2012). In vitro models are amenable to visualizing physiological processes at high spatial resolution, and are likely to provide important information about the morphologic changes of the organism, the specific site of interaction, and the dynamic morphologic changes of the endothelium. In addition, optical microscopy is readily combined with a battery of genetic manipulations in which timed gain, or loss of function can be studied, as well as analysis of photoactivatable fluorescent proteins that may provide important new understandings of the pathogenesis.
However, in vitro models have some limitations. The BBB is not composed solely of endothelial cells, but is rather a complex tissue that consists of HBMEC, pericytes and astrocytes in a precise organization that is difficult to recreate in vitro. Additionally, C. neoformans is stationary and in prolonged contact with the surface of endothelial cells in static models, while in vivo, C. neoformans is circulating in the blood under flow conditions, driven by vascular pressure. Furthermore, even with flow chambers that apply shear stress across endothelial monolayers, it is difficult to mimic the complex set of haemodynamic forces near the BBB.
Monolayers of brain endothelium under circumferential wall stress; strengths and limitations
The role of circumferential wall stress is an unexplored aspect of vascular invasion by fungi. Wall stress is the lateral stress due to transmural pulse pressure across the vessel wall, and is a very important constituent of the physiologic responses of endothelial cells (Prado and Rossi, 2006). Lateral wall stress is important in inducing physiologic responses such as expression of ICAM-1 and VCAM-1 (Golledge et al., 1997), and is important in leucocyte adhesion to the vascular endothelium (Riou et al., 2007; Michell et al., 2011). Although lateral forces are somewhat analogous to the turgor pressure used by filamentous fungi when invading plants (Bastmeyer et al., 2002), models have not been developed to examine the role of lateral wall stress on the endothelial response and subsequent transmigration of fungi. Biomechanical events in which spherical yeast exert force on the vascular wall under the driving force of vascular pressure may play an important role in endothelial responses that facilitate transmigration.
Histopathology of endothelium and surrounding brain parenchyma; strengths and limitations
In addition to in vitro BBB models, traditional histology and cfu determination have also been used to study the brain invasion by C. neoformans. Studies using these approaches have shown C. neoformans can be transported by phagocytes (Chretien et al., 2002; Santangelo et al., 2004; Charlier et al., 2009). It has been shown that capsule structure changes are associated with C. neoformans crossing of the BBB (Charlier et al., 2005). In addition, it has been demonstrated using these approaches that urease contributes to brain invasion by C. neoformans (Olszewski et al., 2004; Shi et al., 2010). Traditional histology is very useful to determine the location of the yeast cells in the brain and to obtain detailed structural information with high resolution. Moreover, histopathology can augment cfu determination, which fails to distinguish between organisms that are replicating within the vascular bed and those replicating after transmigration into the tissue. However, brain invasion by C. neoformans is a transient and dynamic process and therefore histological approaches are limited by examining only one point in time during the invasion process.
Intravital microscopy (IVM) in vivo; strengths and limitations
As early as the 19th century, IVM was being used to image blood flow in living animals (Dutrochet, 1824). However, major breakthroughs in this field occurred in the 1990s, when fluorescence microscopy was improved and transgenic mice that expressed fluorescent proteins became available. Today, IVM enables us to investigate the location, motility, adhesion, and interactions of individual cells at high resolution in three physical dimensions over time in living animals, and is an important technique in cell biology, immunology, tumour biology, and neurobiology (Pittet and Weissleder, 2011). IVM has been recently used to observe filamentous growth of Candida albicans in the ears of mice (Mitra et al., 2010) and phagocytosis of C. albicans in zebrafish (Brothers et al., 2011). Recently, we established a real-time experimental model system in which we used a variety of IVM techniques to directly visualize the early dynamic interactions of C. neoformans with the brain microvasculature as well as the subsequent crossing of the brain microvasculature by the organisms in mice (Shi et al., 2010) (Fig. 2). In the following paragraphs, we will review how this technique works.
IVM: the choice of microscope
Wide-field microscopy, confocal microscopy and multiphoton microscopy can all be used to visualize fungus–host interactions in vitro and in vivo. There are many comprehensive guides to choosing the appropriate technique for a given application (Frigault et al., 2009; Murray, 2011). As a general rule, wide-field fluorescent microscopy is a technique best suited for imaging thin and transparent cells, tissues or organisms while confocal and multiphoton techniques are useful for imaging thicker specimens.
In a typical wide-field configuration, a sample labelled with a fluorochrome is illuminated by an intense light source such as arc lamps or light-emitting diodes (LEDS). The resulting fluorescence emission is imaged by sensitive cameras. Wide-field fluorescence microscopy is versatile and sensitive, but is plagued by out-of-focus signal that degrades image clarity and contrast; indeed, the out-of-focus signal can contribute the majority of the total detected signal for thick specimens (Conchello and Lichtman, 2005; Murray, 2011).
Numerous techniques have been devised to reduce the blur inherent within the wide-field fluorescence image. Dramatic improvements in axial (z) resolution can be obtained over standard wide-field techniques, a result known as optical sectioning. Fluorescence confocal microscopy techniques are based on using physical apertures (often pinholes) that reject the out-of-focus blur associated with wide-field microscopy (Conchello and Lichtman, 2005; Murray, 2011). Compared with wide-field microscopy, which is generally limited to specimens up to 30 microns in thickness, confocal microscopy uses visible lasers as light sources, which can penetrate most tissues up to 100 microns and have been used for imaging cell trafficking in living animals (Geissmann et al., 2005; Baer et al., 2007).
Confocal microscopy can be broadly divided into laser scanning and spinning disk techniques. Although both techniques rely on pinholes or apertures to reject out-of-focus light, their implementation varies. Laser scanning confocal microscopes illuminate and collect signal from samples point by point sequentially (Conchello and Lichtman, 2005) while spinning disk systems use rotating disks to illuminate and collect signal from thousands of points simultaneously (Graf et al., 2005; Stehbens et al., 2012). Laser scanning confocal microscopy offers better axial resolution and more flexibility in experimental design such as in the ability to adjust the pinhole size when compared with spinning disk techniques. Conversely, spinning disk microscopy offers faster image acquisition, higher sensitivity, and reduced photobleaching and phototoxicity (Graf et al., 2005). Note that an analogous technique known as swept-field microscopy offers similar advantages to spinning disk microscopy and has been used in many types of biological studies (Bembenek et al., 2007; Linley et al., 2008).
Multiphoton microscopy is another standard optical sectioning technique. The basis of this technique is the almost simultaneous absorption of two red or near-infrared photons that can excite fluorescence in the visible range (Helmchen and Denk, 2005; Amornphimoltham et al., 2011; Ustione and Piston, 2011). High-powered pulsed lasers are required to drive the two-photon (or three-photon) absorption. This excitation occurs within a restricted focal volume, in which the laser intensity is the highest. For this reason, multiphoton microscopy does not require pinholes because the images do not contain the appreciable out-of-focus blur inherent in wide-field or confocal techniques. Compared with visible light used in confocal microscopy, the near-infrared illumination wavelengths used in multiphoton microscopy can penetrate biological tissues deeper, up to 1 mm, making multiphoton imaging the preferred choice for deep tissue imaging (Helmchen and Denk, 2005; Oheim et al., 2001). The restricted focal volume also means that phototoxicity is usually reduced compared with single photon techniques, although photobleaching is increased (Ustione and Piston, 2011). Multiphoton microscopy, like laser scanning confocal microscopy, illuminates and collects image data point by point as the laser scans across the sample. Multiphoton imaging is slower than spinning disk and related techniques, but generally achieves greater signal-to-noise and thus higher-quality images when the thickness of tissue or sample is 100 microns or thicker (Helmchen and Denk, 2005). Multiphoton microscopy has two main disadvantages: the high cost of the pulsed laser used to excite the two- or three-photon fluorescence and the relative paucity of well-characterized probes compared with single-photon techniques (Oheim et al., 2001; Helmchen and Denk, 2005).
A computational technique known as full iterative deconvolution can be used to reduce the image blur in most optical microscopy techniques (Swedlow, 2007). However, full iterative deconvolution is usually applied to wide-field data sets. Image stacks are acquired by focusing through the sample at regular z intervals in order to collect as much of the total fluorescence emission as possible, and then an algorithm is applied to remove the blur based on the optical characteristics of the microscope. It is critical that the sample remains stationary while the images are acquired by focusing through the sample. The fastest deconvolution implementations acquire data sets at about 0.5–2 s intervals although in practice it can take several minutes, thus limiting its utility for living systems (Swedlow, 2007).
When choosing among the different techniques for studying fungal–CNS interactions, it is vital to consider the sample and time scales involved. Wide-field microscopy is ideally suited to study the interactions of fungal pathogens with cellular monolayers as well as for imaging thin fixed cells or tissue sections. Improvements in image clarity can be obtained by applying full iterative deconvolution if the samples are stationary over the time required to acquire all of the image stacks through the sample.
When no other technique is available, wide-field fluorescence microscopy can be used to visualize the superficial layers of the BBB in the rat or mouse. Such imaging may be adequate for applications that do not need high spatial resolution, such as tracking fungal cells (Shi et al., 2010), but is not sufficient for applications requiring high spatial resolution, such as those that require monitoring the fine details of the fungal–BBB contacts. For in vivo imaging, either laser scanning or spinning disk microscopy (Shi et al., 2010) can be used to visualize fungal–CNS interactions within superficial layers of the BBB to about 100 microns in depth. However, the spinning disk has the advantage of higher speeds and lower photobleaching and phototoxicity. Multiphoton microscopy may not be as fast as spinning disk or related techniques, but it offers the ability to penetrate deeper into the BBB; beyond the 100 micron limit generally observed with other techniques.
IVM: the choice of labelling of C. neoformans and brain vessels
For intravital imaging of the migration of C. neoformans to the CNS, the yeast cells can be labelled with fluorescent dyes. C. neoformans can be labelled with fluorescein isothiocyanate (FITC) or tetramethylrhodamine isothiocyanate (TRITC) (Shi et al., 2010). Additionally, other fluorochromes such as Alexa Fluor 488 or Alexa Fluor 594 (Okagaki et al., 2010) may provide better emission spectra, brighter fluorescence, and greater photostability as compared with other fluorophores. The labelling does not affect the trapping, invasion, and growth in the brain as determined by cfu (M. Shi et al., unpubl. data). We have used different fluorochromes to label two different fungal strains with different characteristics such as virulence and injected them simultaneously into the same mouse. In this way, the interactions of two different strains of the fungus with the brain microvasculature were compared simultaneously (Shi et al., 2010).
Despite the important advantages of these techniques, there are some limitations. The yeast cell loses the fluorescent label if it multiplies. Consequently, it may be difficult to track the fungus once it begins to replicate. However, this disadvantage might be overcome by using C. neoformans expressing green fluorescent protein (GFP) (Voelz et al., 2010; Huang et al., 2011), and it has great potential to be used for imaging the brain migration in living animals.
Fluorescence labelling can also be used to visualize the brain microvasculature as well as the yeast cells. For example, mice expressing GFP on an endothelial-specific promoter, or labelling endothelial molecules with reagents that incorporate fluorochromes can be used. Tie-2 GFP mice express GFP under the direction of the endothelial-specific receptor tyrosine kinase (Tek, formerly, Tie2) promoter. Consequently, endothelial cells expressing GFP can be visualized via fluorescent microscopy (Motoike et al., 2000) and have been used to examine the interaction of Cryptococcus with the endothelium (Shi et al., 2010). Alternatively, mice can be injected by the i.v. route with fluorescently labelled antibodies or lectins (Phillipson et al., 2006; Moriarty et al., 2008). For example, PECAM-1 is expressed on endothelial cells and concentrates at endothelial cell junctions (Albelda et al., 1991), making it useful for visualizing both the lumenal endothelial surface of vessels, and the more PECAM-1-intensive intercellular junctions. The microvasculature can be visualized via fluorescence microscopy within minutes after i.v. injection of the antibodies. This antibody has been used previously to study junctional extravasation of leucocytes (Phillipson et al., 2006), as well as interactions of Borrelia burgdorferi with endothelial cells in vivo (Moriarty et al., 2008). In addition to antibodies, mice can be injected i.v. with fluorochrome-conjugated bovine serum albumin, so that the vascular lumen can be visualized by fluorescence microscopy (McDonald et al., 2010).
IVM: the choice of surgery
The most critical step for intravital imaging of fungal migration to the brain is to make the brain microvasculature visible under fluorescence microscopy. Two major surgical methods have been described, i.e. a thinned-skull cranial window and an open-skull cranial window (Carvalho-Tavares et al., 2000; Holtmaat et al., 2009; Shi et al., 2010; Yang et al., 2010). For a thinned-skull cranial window, a square (about 2 mm × 2 mm) of the skull is thinned using a high-speed drill under a dissecting microscope. After removing the outer compact and spongy bone, the internal compact bone layer is exposed. Thinning is continued until the skull thickness is reduced to approximately 20 μm and until the cortical surface vasculature can be clearly visualized (especially when damp with saline). For an open-skull window, a square groove is cut and an island of skull (4 mm × 4 mm) is left intact in the centre. The central island of skull bone is slowly lifted, exposing the dura. The dura is gently removed to expose the underlying pia and cortical vasculature.
There are advantages and limitations to each technique. When a thinned-skull window is used, the brain tissue is still covered with the skull and it is not required to superfuse the brain with artificial cerebrospinal fluid. This technique is well suited for observations over long periods. However, the skull thickness is critical for image quality and achieving optimal and uniform skull thickness requires significant surgical proficiency. Additionally, the size of the window is limited by the challenges of achieving a very thin uniform layer of skull with breaking through the underlying cortex (Yang et al., 2010). By contrast, the open-skull window provides direct visualization of the surface of the brain with no skull over the window. Therefore, the cortical vasculature is much clearer under the microscope as compared with that exposed by the thinned-skull technique. In addition, the observation area can be larger as compared with the thinned-skull technique (Holtmaat et al., 2009). However, the brain tissue is not covered by the skull and dura, and superfusion with an artificial cerebrospinal fluid becomes essential. In addition, some superficial capillaries might tear during removal of the cranial bone, resulting in bleeding.
IVM: the strengths and limitations
Intravital microscopy allows us to directly visualize the dynamic interactions of the yeast cells with brain endothelial cells, leading to arrest of the yeast cells in brain vasculature and subsequent transmigration into the brain parenchyma in a living animal. This is crucial because the behaviour of C. neoformans and its interactions with brain endothelial cells are influenced by many factors including interactions with other cellular and extracellular components, anatomical compartmentalization, and shear stress in the vessel. This type of analysis is the only in vivo approach to study cryptococcal interactions with the vascular endothelium under the physiological and shear flow conditions normally encountered in a living animal. However, IVM also has limitations for studying the migration of fungal cells to the CNS. To visualize the fungal cells in the brain a bolus of the organisms is infused by the i.v. route. However, during a natural infection fungal cells are likely gradually released from the lung and enter the blood stream in limited numbers. It is not known to what extent the two are dissimilar. Another limitation of IVM in the current model is that the injected yeast cells are derived from in vitro culture. By contrast, in a natural infection the fungal cells invading brain are derived from the lung before accessing the blood, raising the possibility that these fungal cells have activated different metabolic pathways or differ in virulence or morphology (Okagaki et al., 2010; Zaragoza et al., 2010).
Future perspectives and concluding remarks
There are a number of valuable models available to image the interaction between fungi and the vasculature and parenchyma of the brain. Each has its advantages and limitations. Additionally, these tools may be used in conjunction with techniques to explore the cell biology and pathophysiology of the interaction between fungi and endothelium. There are a number of outstanding questions. For example, what is the role of environmental and epigenetic modulation in the lung on subsequent brain invasion? A number of virulence factors and receptor ligand interactions have been identified. What is the stage at which these factors play a role during arrest and transmigration into the brain? There is evidence that yeast cells leave the lung within macrophages. Do the macrophages carry the yeast to the site of arrest, through the BBB or into the brain parenchyma (Chretien et al., 2002; Santangelo et al., 2004; Shea et al., 2006; Kechichian et al., 2007; Charlier et al., 2009; Casadevall, 2010)? Additionally, how does the microbe respond to signals provided during contact with the endothelium and how does the endothelium respond to signals provided during contact with the fungi? Is there a stress response that is important in transmigration? These and other questions await further study.
Taken together, IVM enables dynamic studies with high resolution of cellular and subcellular events in real time and its application to examine fugal migration to the CNS has been considered as an important advance in this field. It is believed that IVM will become an increasingly valuable research tool in the field and make greater contributions to our understanding of fungal migration in the CNS, with continuous improvements of fluorescent and bioluminescent probes, microscopes and associated filters, cameras and image acquisition, analysis software, as well as transgenic mice expressing fluorescently labelled proteins.