Main

There has been a recent shift in the study of infectious-disease mechanisms from the use of two-dimensional (2D) monolayers to the use of organotypic, 3D cell culture models that mimic the morphological and functional features of their in vivo parental tissues1,2,3,4,5,6,7. These 3D systems have tremendous potential for bridging the gap between cell-based discovery research and animal models for studying both host–pathogen interactions and human disease progression, as well as for the development of novel drugs and therapeutics.

Conventional 2D cell culture involves growing cells as monolayers on solid, impermeable surfaces (plastic or glass) or in uniform suspension. Indeed, 2D monolayers have contributed greatly to our understanding of infectious-disease processes, including the host immune and physiological mechanisms used to defend against viral, bacterial, fungal and parasitic infections8,9,10,11,12, as well as microbial virulence strategies13,14. However, it is becoming increasingly evident that this 'flat biology' approach, in which key phenotypic and functional characteristics are often lost, is not predictive of in vivo tissue responses1,3,7,14,15,16,17,18,19,20,21,22. One key reason for the loss of differentiation that occurs in monolayers is the dissociation from the native in vivo 3D structure to 2D propagation on flat, impermeable substrates in vitro, which also prevents cells responding to chemical and molecular gradients in three dimensions (reflecting the apical, basal and lateral cell surfaces)15,18,21,23. Because 2D monolayers lack the complexity, and often the physiological relevance, of the tissues that are encountered by a pathogen during the natural course of infection in vivo, they are often unfaithful predictors of the infection process. Thus, to overcome some of the inherent limitations associated with 2D monolayers, as well as the high cost, availability and variability of animal models, attention has moved to 3D cell culture models.

The shift towards using 3D tissue models as high-fidelity tools to facilitate the transition from basic cellular research to clinical applications is logical, given that tissues and organs are 3D structures. In contrast to 2D monolayers, 3D cell culture models are modular, adaptable biomedical systems that range in complexity from a single cell type (monotypic), representing the minimum unit of the differentiated tissue in vivo16,21, to complex co-culture models that recapitulate both the 3D architecture and the multicellular complexity of the parental tissue. This continuum of 3D cell culture systems can be used in a hierarchical approach (that is, results can be compared across a collection of increasingly complex experimental models) to provide a new generation of models that offers more reliable and more predictive correlations between in vitro studies and in vivo outcomes, and may ultimately streamline the discovery and evaluation of new drugs and therapeutics3,5,6,7,14,17,21,22,24,25,26,27,28,29. The establishment of complex 3D tissue equivalents from cell lines, primary cells and organ cultures can be achieved through various methodologies, including spontaneous aggregation in a suspension culture, implantation into 3D matrix scaffolds and culture in transwell sytems or in rotating culture bioreactors21,27,28,30,31,32,33,34,35,36. Each of these systems ranges in complexity and carries distinct advantages and disadvantages that must be considered, depending on the experimental question being addressed (Box 1). These considerations include the increasing time and cost of experimental set-up, the need for model optimization and validation when different cell types are incorporated, and the level of expertise required to successfully adapt the system to the specific experimental question.

Although a range of 3D cell culture systems exist, this Review highlights how one of these systems, the NASA (National Aeronautics and Space Administration)-designed rotating wall vessel (RWV) bioreactor, has been used as a platform for studying the cellular and molecular responses of both hosts and pathogens. Given the extensive literature on 3D cell culture and the diversity of cell types used, it is beyond the scope of this Review to cover this literature in its entirety, but readers are referred to several recent reviews3,6,7,14,17,21,22,27,28,29,37,38. Here, we focus on a subset of studies that have used RWV-derived 3D organotypic models to provide novel insights into host–pathogen interactions, and we discuss how this knowledge can advance our understanding of the mechanisms of pathogenesis and may lead to the development of novel therapeutics.

The rotating wall vessel bioreactor

The design of the RWV bioreactor (Fig. 1) is based on the principle that organs and tissues function in a 3D environment. This optimized form of suspension culture is used for growing 3D cell cultures that maintain many specialized features of in vivo tissues36,39. Moreover, the RWV provides a low-fluid-shear growth environment similar to that encountered by pathogens in certain regions of the body (including between the brush border microvilli of epithelial cells and in utero); fluid shear is a biomechanical force known to influence cellular differentiation and development in mammals40,41,42,43,44,45,46. The dynamic culture conditions in the RWV allow cells to grow in three dimensions, to aggregate based on natural cellular affinities (facilitating co-culture of multiple cell types) and to differentiate into 3D tissue-like assemblies19,36. Although cells cultured in the RWV require more time to grow and are initially more costly than monolayers, the inherent flexibility of these systems is ideal for the design of hierarchical models with a modular functionality and multicellular complexity, allowing different cell types to be mixed and matched to explore fundamental biological questions.

Figure 1: Operational principles of rotating wall vessel (RWV) technology.
figure 1

Confluent monolayers are trypsinized and introduced into the slow-turning lateral vessel (STLV) together with porous microcarrier beads coated with extracellular matrix (ECM). Rotation of the STLV along a horizontal axis offsets sedimentation, causing gentle falling of the beads and cells through the culture medium within a restricted orbit, resulting in the establishment of three-dimensional tissue models. A central core consisting of a gas-permeable membrane provides adequate gas exchange. Enlarged representations of epithelial cells grown on a porous microcarrier bead using the RWV, and a three-dimensional aggregate of lung epithelium removed from the RWV and analyzed by light microscopy, are shown.

Full size image

There are two different RWV designs, the high-aspect rotating vessel (HARV) and the slow-turning lateral vessel (STLV), which differ in their aeration source. However, the operational principle is the same for both types of reactors. The RWV is a cylindrical, rotating bioreactor that is completely filled with culture medium; the sedimentation of cells in the vessel is offset by the rotating fluid, creating a constant, gentle fall of cells through the medium under conditions of physiologically relevant fluid shear38. To initiate 3D cell culture in the RWV, cells are first grown as conventional monolayers in standard tissue culture flasks (Fig. 1). At the appropriate density, the cells are removed from the flask, re-suspended in medium and incubated with porous extracellular matrix (ECM)-coated microcarrier beads (or other scaffolding) for attachment. The porosity of the beads allows the cells to respond to chemical and molecular gradients in three dimensions, in a manner akin to the response of the tissue in vivo. The cell–bead complexes are then introduced to the RWV and rotation is initiated. As mentioned, other types of scaffolding can be used in the RWV to support cellular aggregation and differentiation. For example, a recent report used novel hyaluronan hydrogel-coated microcarriers as scaffolds for RWV-derived 3D intestinal models, which allowed for enzyme-free cell detachment, with applications in tissue regeneration and transplantation47. In addition, cells can also spontaneously aggregate under scaffold-free conditions during RWV culture. The choice of scaffold material, and whether to use scaffolds at all, depends on the nature of the study.

After RWV-derived 3D cell cultures are established, they can easily be exposed to infectious agents and external compounds. This can be accomplished by removing the 3D cells from the RWV and distributing them evenly in multi-well plates or other convenient formats for testing — cells recovered from the RWV retain their differentiated state long enough to be amenable to a wide range of experimental manipulations, as determined empirically for each cell type by immunohistochemical, histological and functionality assessments — or by introducing the pathogen or compound into the RWV simultaneously with the 3D cell cultures. The RWV design also allows easy manipulation of culture conditions, including the addition of cells for co-culture purposes, removal of cellular aggregates for analysis and replacement of media at various time points. RWV-derived 3D models can therefore easily be incorporated into existing assays used for infection studies, such as: assays testing microbial adherence, invasion and intracellular survival; microscopic examination; transcriptomic, proteomic and metabolomic analyses; expression profiling for cytokines and other inflammatory mediators; and flow cytometry3. For studies that require homogeneous cell suspensions, such as flow cytometry and cell viability assays48,49, 3D cell cultures can be removed from the microcarrier beads using conventional enzymatic and non-enzymatic treatments (for example, trypsin or EDTA, respectively). Although the treatments for flow cytometry analysis will disrupt the delicate 3D architecture, the expression of cellular markers such as those for differentiation and apoptosis can still be quantified at the single-cell level. Finally, the inherent flexibility of this system is ideal for the design of modular models that can be mixed and modified to explore fundamental questions in biology and that can also be implemented in drug, adjuvant and vaccine screening.

A wide range of RWV-derived 3D models have been extensively characterized, and many different cell and tissue types, derived from both cell lines and primary cells, have been developed. This includes, but is not limited to, 3D models of the human small intestine, colon, bladder, lung, liver, placenta, neuronal tissue, tonsil, lymphoma and vaginal epithelium, which are being used to investigate mechanisms of cellular differentiation and tissue morphogenesis as well as host–pathogen interactions that lead to infection and disease35,36,39,44,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66 (Figs 2,3).

Figure 2: Rotating wall vessel (RWV)-derived three-dimensional (3D) cell culture models of human tissue used for infection studies.
figure 2

A range of RWV-derived 3D models are currently being used for infectious-disease research. This includes, but is not limited to, the models shown here. For neuronal tissue, a scanning electron micrograph (SEM) is shown of 3D SH-SY5Y cell aggregates with well-developed neurite extensions (shown at ×2,000). This model is currently being used to study HIV-associated dementia66. For the lung, an SEM of 3D A549 lung cell aggregates (×2,000) is shown on the left, and an SEM of 3D A549 lung cell aggregates infected with Pseudomonas aeruginosa (×7,500) is shown on the immediate right. For the liver, an SEM of primary human liver cells (×1,500) is shown; RWV liver models have been applied to the study of hepatitis C virus infections60. For the bladder, a transmission electron micrograph is shown of 3D 5637 bladder cell aggregates showing close association of uropathogenic Escherichia coli (labelled 'b') with superficial urothelial cells (asterisks indicate regions displaying a loss of structural integrity). For the vaginal epithelium, an SEM of 3D V19I vaginal epithelial cells is shown (×300). This model is currently being used to study several sexually transmitted infectious agents64. For the colon, SEMs are shown of 3D HT-29 aggregates (×2,000) on the left and of 3D HT-29 aggregates infected with Salmonella enterica subsp. enterica serovar Typhimurium (×2,000) on the immediate right. For the small intestine, an SEM is shown of the 3D Int-407 aggregates that have been used to study the early stages of human enteric salmonellosis (×500). For tonsillar tissue, a light micrograph is shown of primary tonsillar cells 10 days after infection with Borrelia burgdorferi. The 3D HT-29 and Int-407 model development and imaging was carried out by the C.A.N. group. The neuronal tissue SEM is reproduced, with permission, from Ref. 66 © (2008) Elsevier. The left lung SEM is reproduced, with permission, from Ref. 19 © (2004) American Society for Microbiology (ASM), and the right lung SEM is reproduced, with permission, from Ref. 48 © (2005) ASM. The liver SEM is reproduced, with permission, from Ref. 115 © (1999) Springer. The bladder image is reproduced, with permission, from Ref. 61 © (2006) ASM. The vaginal epithelium SEM is reproduced, with permission, from Ref. 64 © (2010) Society for the Study of Reproduction. The colon SEMs are reproduced, with permission, from Ref. 63 © (2006) Elsevier. The tonsil image is reproduced, with permission, from Ref. 52 © (2005) University of Chicago Press.

Full size image
Figure 3: Rotating wall vessel (RWV)-derived three-dimensional (3D) cell culture models display in vivo-like differentiation properties that are crucial for pathogenesis.
figure 3

A | Immunohistochemical profiling of RWV-derived 3D models from various human cell lines shows expression and organization of biomarkers, as well as cellular organization, that are relevant to those found in normal tissues in vivo. 3D models have been established from cell lines that have been or are currently being used in infection studies, as indicated below. Nuclei are stained with 4′,6-diamidino-2-phenylindole (DAPI; blue), and the actin cytoskeleton is labelled with phalloidin (red) (except in part Ag); other labels are protein-specific antigens, as indicated. Aa | A549 lung epithelium co-cultured with U937 monocytes with labelled MUC5AC, a mucin marker (yellow). Ab | A549 lung epithelium co-cultured with U937 monocytes, with labelled CD45, a mononuclear-phagocytic-cell marker (white-yellow). Ac | HT-29 colonic epithelium labelled with an E-cadherin tight junctional marker (orange). Ad | Int-407 small-intestinal epithelial cells exhibit a unilayer morphology. Ae | Int-407 small-intestinal epithelial cells with labelled sialyl Lewis A antigen, a microfold cell marker (green). Af | HTB9-5637 bladder epithelium labelled with a uroplakin- and urothelial cell-specific antigen (green). Ag | Huh7 liver cells with labelled hepatitis C virus receptor SR-B1 (red). Ah | SH-SY5Y neuronal cells co-cultured with HTB-14 astrocytes and THP-1 monocytes, with labelled glial fibrillary acidic protein (GFAP), an intermediate filament protein marker (green). Development and imaging of the 3D models in parts Aa–Af and Ah were carried out by the C.A.N. group. B | Scanning electron micrographs of Int-407 monolayers and 3D aggregates infected with Salmonella enterica subsp. enterica serovar Typhimurium (×2,000). Ba | An uninfected monolayer control, and monolayers at 2 h post-infection (p.i.). Bb | An uninfected 3D control, and 3D cells at 2 h p.i. Part Ag image is reproduced, with permission, from Ref. 60 © (2009) BioMed Central. Part B images are reproduced, with permission, from Ref. 49 © (2001) American Society for Microbiology.

Full size image

The ability to successfully use these models as surrogates for human tissues depends on the extent to which they mimic the dynamic cellular interactions that exist in vivo. For example, normal tissue growth and maintenance relies on a complex 3D network of interactions involving multiple cell types and molecules, and these interactions are actively stabilized by tensional forces between the cell cytoskeleton, adhesion molecules on the cell surface and an extracellular environment comprising the ECM, growth factors, cytokines and hormones17,21,27,46,55,67,68. These spatial interactions are crucial for regulating diverse cellular processes related to proliferation, differentiation, survival and immune function46. Studies have shown that cells cultured in the RWV can recapitulate many of the fundamental aspects of the parental tissue in vivo, including the 3D spatial organization and polarity, cellular differentiation, multicellular complexity and functionality48,49,51,53,55,56,60,61,63,64,65,66 (Table 1). Moreover, cells cultured in the RWV can spontaneously differentiate into multiple cell types that are normally found in parental tissues48,49,51,63.

Table 1 Select three-dimensional human organotypic models
Full size table

Figures 2,3 and Table 1 give an overview of the key in vivo-like architectural and phenotypic characteristics associated with RWV-derived 3D cell culture models that are crucial for mimicking the host response to infection and that are essential for advancing our understanding of the fundamental mechanisms governing pathogenesis. Such characteristics include the induction of apical and basal polarity, the enhanced formation and physiological distribution of tight junctions, the production of mucus, the in vivo-like cellular organization (the formation of unilayers or stacking) and differentiation into the multiple cell types that are normally found in tissues in vivo. Although RWV-derived 3D models mimic many key structural and functional characteristics of parental tissues, they do not completely reconstruct the in vivo situation, and thus there are limitations to their ability to fully predict the complex behaviours of whole tissues. For example, current models do not recapitulate all of the cell types found in parental tissues, nor are they innervated or vascularized. These are important features for studying host–pathogen interactions and the polymicrobial interactions between microorganisms and the host. Adding these functionalities in a stepwise manner to existing 3D cell culture platforms would allow the development of a series of increasingly complex 3D model systems, each of which represents a physiologically relevant assay system in its own right.

Several other 3D model systems have been used for infection studies (Box 1), including static transwell (permeable support) cell cultures69,70,71,72,73, propagation in 3D polyethylene glycol hydrogels74, cultures in spinner bottles75 and cultures in 3D collagen gels76. Although many key findings have been made using these systems, only those infection studies using RWV-generated 3D models are discussed here.

Host–pathogen interactions

Mechanisms of microbial pathogenesis. There are many pathogens that lack practical and representative cell culture or animal models which accurately reflect the host response to infection, and the RWV has enabled researchers to study such pathogens50,52,58,59,62. In this regard, 3D models of the intestine have been shown to have in vivo-like expression levels and distribution patterns of key biological surface markers that are directly accessible to pathogens49,51,63 (Table 1), and this could contribute to the ability of these models to support productive pathogen infection and replication. Moreover, 3D intestinal epithelial-cell aggregates initiated from a single cell type (enterocytes) have been shown to differentiate into heterogeneous populations of epithelial cells relevant to those normally found in the parental tissue from which they were derived, including mucin-secreting cells (goblet cells), and to express sialyl Lewis A antigen, a marker for microfold cells (M cells)49,77 (Fig. 3Ae). RWV-derived 3D intestinal epithelial cells also exhibit unilayer formation on the surface of microcarrier beads, and this is relevant to the morphology of normal human intestinal tissue49,63 (Fig. 3Ad).

3D models of small-intestinal epithelium were used in infection studies with human norovirus (a human-specific pathogen that currently cannot be propagated in vitro or in animal models) and were shown to support virus uptake, which was accompanied by the induction of cytopathic effects. Using limiting-dilution PCR, this study also suggested the potential for this model to support minimal to modest viral replication62. Additional studies are ongoing in numerous laboratories to validate and advance this finding. Recently, Cryptosporidium parvum, a protozoan parasite that previously lacked physiologically relevant in vitro and in vivo models, was shown to induce morphological changes indicative of a successful infection in RWV-derived 3D models of large-intestinal epithelium50. Similarly, there was no ex vivo tissue system to study the pathogenesis of Borrelia burgdorferi, the aetiological agent of Lyme disease. However, B. burgdorferi can invade and mount a productive intracellular infection in an RWV-generated 3D tonsillar-tissue model52 (Fig. 2). Recently, an RWV-derived 3D liver model was shown to mimic the differentiated and polarized state of hepatocytes in vivo more closely than monolayers, and this model was used to assess whether hepatocyte polarity affects the cellular entry of hepatitis C virus, a pathogen that has lacked robust cell culture and small-animal infection models60. This model was shown to be highly permissive for infection with hepatitis C virus and is currently being used to study viral entry and replication. For these pathogens, a physiological in vivo-like 3D cell culture model has provided an enabling tool for researchers to investigate in vitro those host–pathogen interactions that were previously difficult or impossible to study.

The 3D small-intestinal- and colonic-epithelium models have also proved to be useful tools in investigating the well-studied and easily culturable enteric bacterial pathogen, Salmonella enterica subsp. enterica serovar Typhimurium49,63. Unlike the pathogens discussed above, S. Typhimurium can mount a productive infection in numerous cell lines grown as monolayers8. The prevailing paradigm of how S. Typhimurium initiates enteric disease is that bacterial invasion of the intestinal epithelium is essential for virulence and that this process requires the virulence-associated, Salmonella pathogenicity island 1 (SPI-1)-encoded type III secretion system (T3SS), a 'molecular syringe' that injects virulence-associated effector proteins into host cells78. These effector proteins manipulate host signalling pathways and elicit cellular responses, including changes in signal transduction pathways and cellular functions, such as rearrangements of the actin cytoskeleton, induction of pro-inflammatory cytokines and apoptosis, and regulation of epithelial-cell proliferation and differentiation78,79,80,81. Much of this information was derived from studies using S. Typhimurium-infected epithelial cells grown as monolayers. However, the degree to which the SPI-1 T3SS and its secreted effectors contribute to intestinal disease in humans remains unknown8,82.

Studies of S. Typhimurium infection in animal models have shown that gastrointestinal disease can occur in the absence of SPI-1, in part because the organism can enter the intestinal epithelium through epithelial cell types other than enterocytes, including M cells83,84. Interestingly, the first human outbreak was recently reported for diarrhoeal disease caused by clinical strains of Salmonella enterica subsp. enterica serovar Senftenberg lacking SPI-1 (Refs 85,86). In agreement with this finding, recent work has shown that a strain of S. Typhimurium carrying a mutation in invA, and thus lacking a functional SPI-1 system, could invade into 3D models of human colonic intestinal epithelium to the same extent as the wild-type parental strain63 (see Supplementary information S1 (figure)). To our knowledge, this is the first time that an in vitro model of intestinal epithelium has paralleled the SPI-1 T3SS-independent intestinal invasion response observed in vivo; these results suggest that the SPI-1 T3SS might not be the main determinant for invasion of S. Typhimurium into human intestinal tissue. One possible reason for the observed invasion of the invA mutant into 3D intestinal epithelial cells could be the presence of M-like cells in this model. Moreover, we recently used 3D models of small- and large-intestinal epithelium to help elucidate the role of the S. Typhimurium T3SSs encoded by both SPI-1 and SPI-2 (SPI-2 encodes virulence-associated proteins important for intracellular survival) in the early stages of human enteric salmonellosis. The results of these studies further support the fact that a more in vivo-like response is obtained with the 3D intestinal models than with monolayers (C.A.N, unpublished observations) and further emphasize the value of these systems for ascertaining the molecular mechanisms of human enteric salmonellosis.

A 3D model of human bladder urothelium was developed to investigate the molecular mechanisms by which another well-characterized pathogen, uropathogenic Escherichia coli (UPEC), causes urinary tract infections61. Specifically, human bladder 5637 cells cultured in the RWV displayed a stacked organization and a heterogeneous cellular morphology that was shown to closely resemble normal human urothelial tissues. When these models were infected with wild-type UPEC (Fig. 2), the pathogen successfully colonized and induced exfoliation (an innate host defence mechanism) of the organoids with much less damage than was observed in monolayer cultures infected with the same strain. Mutational analysis showed that UPEC alpha-haemolysin mediated the exfoliation damage. The results obtained with the 3D cultures were consistent with observations from animal infection studies using a mouse model of cystitis.

Collectively, these studies highlight how crucial aspects of disease and pathogenesis can be missed or misunderstood by using model systems that do not closely mimic the host, and they could provide an explanation, in part, as to why there have been many failed attempts to develop effective vaccines and antimicrobials. Indeed, several laboratories are currently exploring the potential of RWV-derived 3D models as a platform technology to streamline vaccine discovery and development.

Host responses to infection. The complexity and in vivo-like characteristics of 3D cell culture models make them not only useful systems to investigate the mechanisms involved in microbial pathogenesis, but also valuable tools to study the host response to microbial infection. The intestinal and lung mucosa are two major portals of entry for microorganisms, and both tissues have been modelled using the RWV and subsequently used to study enteric and respiratory pathogens, respectively48,49,63,87.

The intestinal epithelium serves two major functions in protecting the host against disease: providing a defensive physical barrier, and inducing host innate immunity through pathogen recognition, resulting in inflammatory responses88,89,90,91. As architectural integrity and controlled inflammation are crucial protective properties of the intestine, a representative cell culture model must display similar characteristics. Infection of 3D small-intestinal epithelial cells with S. Typhimurium (Fig. 3b) and of 3D large-intestinal epithelial cells with C. parvum oocysts resulted in less severe host structural damage than was seen in infected monolayers49,50. Indeed, SEM analysis of 3D small-intestinal epithelium after infection for 2 hours with wild-type S. Typhimurium revealed a minimal loss of structural integrity, no induction of apoptosis and a faster recovery of cell structure than was seen for monolayers, which exhibited a major loss of structural integrity and a robust induction of apoptosis49. This response of 3D small-intestinal epithelium to S. Typhimurium infection is relevant to the scenario that is thought to occur in vivo, in which S. Typhimurium-induced damage to the small-intestinal epithelium is short-lived and followed by fast cell recovery and repair, and restoration of the intact mucosal surface.

Interestingly, the differences observed in the 3D intestinal models in the response to infection with S. Typhimurium and C. parvum are more reflective of the pathology of the in vivo infections49,50. As previously mentioned, the increased formation and physiological localization of tight junctions and ECM proteins observed in 3D intestinal epithelial models could potentially provide a more protective barrier against invading pathogens, in turn limiting the infection and preserving the structural integrity of host cells49,63. In addition, 3D intestinal epithelial cells have been shown to secrete in vivo-like levels and have in vivo-like distribution patterns of mucus, which contains antimicrobial enzymes that may also reduce bacterial adherence, invasion and extracellular survival49,92, and mucosal-protective prostaglandins, which could prevent injury to mucus-coated cells and help maintain the structure of the protective epithelial barrier during infection49,93,94.

A protective and regulated inflammatory response is also crucial for host defence and survival during infection9,10,11,12,88,95. In response to S. Typhimurium infection, both monolayers and 3D cultures of small-intestinal epithelial cells secreted pro-inflammatory cytokines, with monolayers exhibiting significantly higher expression levels of these immunomodulatory factors (which included tumour necrosis factor (TNF), interleukin-6 (IL-6), IL-1α and IL-1β) than 3D cells49. This is in agreement with the increased level of apoptotic cell death that is associated with monolayers at early post-infection time points and is not predictive of the in vivo infection. Conversely, the 3D small-intestinal epithelial cells expressed higher levels of anti-inflammatory cytokines (such as IL-1-receptor antagonist) than monolayers and, correspondingly, apoptosis was not induced in these cells at the same time points following S. Typhimurium infection, which reflects the situation in vivo49. Furthermore, the enhanced basal level of transformation growth factor-β expression in uninfected 3D small-intestinal epithelial cells mimics that of the normal parental tissue, compared with basal levels in uninfected monolayers49. Interestingly, in a separate study, 3D colon epithelial cells expressed lower levels of the pro-inflammatory cytokine IL-8 than monolayers following S. Typhimurium infection, but the 3D culture exhibited greater structural damage63. Although results from the 3D colon studies might initially seem to contradict those from the 3D small-intestine studies, they are indeed consistent with in vivo infection profiles, which indicate that the colon is the primary site of intestinal infection by non-typhoidal Salmonella spp.88, and it would not be advantageous for the colon to respond to S. Typhimurium infection with robust IL-8 secretion, given the large amount of lipopolysaccharide present in the normal colonic microbial flora96. Likewise, the application of 3D models of human lung epithelium to the study of Pseudomonas aeruginosa infection showed a controlled and regulated balance of pro-inflammatory (IL-6, IL-12 and TNF) and anti-inflammatory (IL-10) cytokine responses to infection that is also relevant to the in vivo infection process48,97.

Collectively, these findings further demonstrate the usefulness of 3D models as meaningful predictors of the outcomes of, and host responses to, in vivo infections49,63,98. Taken together, these architectural and inflammatory differences between monolayers and 3D aggregates in response to pathogen challenge resulted in notable differences in infection outcomes, with the 3D models responding in important ways that are more reflective of an in vivo infection. Moreover, although mucosal epithelial cells are excellent sentinels of innate immunity, the further development of these models to include immune cells such as macrophages and dendritic cells will provide a hierarchy of models that range in cellular complexity and can be used interchangeably to study host–pathogen interactions and inflammatory responses. In this regard, we and our collaborators have initiated the development of advanced 3D immunocompetent co-culture models of lung (Fig. 3), intestinal and vaginal epithelium that contain functional immune cell populations for use in modelling the infectious-disease process. Each of these 3D models, ranging from monotypic cultures to immunocompetent co-cultures, represents a physiologically relevant assay system in its own right for use in such studies.

Future clinical applications

Infectious diseases continue to rank among the leading causes of death worldwide, especially in developing countries, where diarrhoeal diseases and respiratory and sexually transmitted infections are major problems99. There is an urgent need to develop novel therapeutics and drugs that target emerging and re-emerging infectious agents, particularly to counteract the increasing problems of antibiotic and antiviral resistance. In this regard, it has become increasingly evident that polymicrobial infections (in which the presence of one microorganism predisposes the host to subsequent infections with other pathogens), including those associated with biofilms, have an important role in human health and disease100. A more complete understanding of the physical and chemical interactions that occur between microorganisms and host tissues and that can increase the susceptibility of the host to secondary infection will lead to improved strategies for preventing and treating infectious diseases100.

As discussed previously, several human 3D organotypic models are currently being used to study infectious diseases (Table 1). These models provide novel platforms to explore the complex nature of host–pathogen interactions and to investigate the drug resistance strategies that are used by pathogens to circumvent current treatments. The incorporation of these models as surrogates for human tissues in the early stages of the drug design process can potentially aid in reducing the number of inadequate drug candidates that enter into clinical trials, by providing in vitro models that are more predictive of in vivo responses6,101. Despite considerable scientific and technological advancements in the pharmaceutical industry and in basic research over the past 30 years, approximately 90% of drug candidates in the developmental pipeline still fail to make it to market102. Most of these failures are associated with problems with toxicity or ADME (absorption, distribution, metabolism and excretion), or simply a failure to translate preclinical success in cell culture and animal models to human clinical trials102. Several 3D models have been shown to more reliably mimic the drug sensitivity patterns observed in vivo, including penetration into target tissue and drug resistance, and to also more closely parallel patient responses to treatment6,14,27,28,64,67,101,103,104,105, which may reduce the number of drug candidate failures.

One preclinical area that may benefit from the use of more in vivo-like 3D cell culture models is the development of microbicides, which are compounds that protect against sexually transmitted infections. Currently, no licensed microbicides are available, and the failure of the anti-HIV-1 microbicide nonoxynol-9 (N-9) in clinical trials further emphasizes the need for more predictive models of success. N-9, the active ingredient in the over-the-counter spermicide Conceptrol (McNeil), displayed promise in both in vitro and animal models as a candidate microbicide; however, it was ultimately found to be ineffective in clinical trials and, in some cases, increased the incidence of HIV-1 transmission106,107. Recently, a model of human 3D vaginal epithelium was developed (see Supplementary information S1 (figure)) that more closely paralleled the toxicity associated with N-9 in vivo than did cervical explants64. This model could potentially be used in the future to screen candidate microbicides for toxicity and efficacy before advancing to clinical trials. Moreover, the incorporation of patient-derived primary cells, including immune cells, into these models could help predict personalized responses to a panel of drug candidates or to predict the potential toxicities of certain drug combinations.

Conclusions

Three-dimensional cell culture models are valuable tools for studying the mechanisms of infectious disease and promise to facilitate the translation of basic science to the clinic. These physiologically relevant model systems are reproducible, experimentally flexible and offer targeted high-throughput platforms. These models have advanced our understanding of the molecular and cellular mechanisms that underlie pathogenic processes in both the host and pathogen, have enabled the study of difficult-to-culture pathogens, and have exciting potential for use as powerful screening tools for therapeutics and for drug and vaccine target validation and discovery.

Although no single in vitro model system will provide a complete understanding of the fundamental mechanisms governing pathogenesis, a hierarchical modelling approach that uses a range of advanced cell culture models, including immunocompetent co-culture models, in combination with animal models will advance our mechanistic understanding of normal versus diseased cellular processes. To this end, 3D models are modular, tractable biomedical systems that are currently being engineered to integrate features such as multiple cell types (including immune cells), cell-instructive scaffolds, vascularization, innervation and biomechanical forces. Each of these systems could be adapted to high-throughput screening platforms and should therefore make a major contribution to preclinical drug and therapeutic discovery.