Modeling Innate Antiviral Immunity in Physiological Context

Abstract

An effective innate antiviral response is critical for the mitigation of severe disease and host survival following infection. In vivo, the innate antiviral response is triggered by cells that detect the invading pathogen and then communicate through autocrine and paracrine signaling to stimulate the expression of genes that inhibit viral replication, curtail cell proliferation, or modulate the immune response. In other words, the innate antiviral response is complex and dynamic. Notably, in the laboratory, culturing viruses and assaying viral life cycles frequently utilizes cells that are derived from tissues other than those that support viral replication during natural infection, while the study of viral pathogenesis often employs animal models. In recapitulating the human antiviral response, it is important to consider that variation in the expression and function of innate immune sensors and antiviral effectors exists across species, cell types, and cell differentiation states, as well as when cells are placed in different contexts. Thus, to gain novel insight into the dynamics of the host response and how specific sensors and effectors impact infection kinetics by a particular virus, the model system must be selected carefully. In this review, we briefly introduce key signaling pathways involved in the innate antiviral response and highlight how these differ between systems. We then review the application of tissue-engineered or 3D models for studying the antiviral response, and suggest how these in vitro culture systems could be further utilized to assay physiologically-relevant host responses and reveal novel insight into virus-host interactions.

Graphical Abstract

I. Introduction

The innate immune system is an essential component of host defense against viral infection. An effective innate antiviral response requires successful pathogen detection followed by signal transduction and expression of interferons (IFNs), as well as an array of pro-inflammatory cytokines and chemokines. Together these molecules promote the establishment of a local antiviral state and help recruit immune cells to the site of infection. The innate immune response also facilitates cross-talk with the adaptive immune system to orchestrate resolution of the ongoing infection and ensure a more rapid response upon subsequent challenge. Without an intact innate antiviral response, viral replication proceeds unchecked, potentially leading to disseminated infection and life-threatening consequences for the infected individual (reviewed in [1]). This was recently evidenced in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-infected individuals that suffered severe outcomes as a result of genetic defects in the IFN pathway or autoantibody-mediated neutralization of IFN [2–4].

The discovery of key innate immune sensors and effectors – beginning with the description of interferon activity by Isaacs and Lindenmann in 1957 [5,6] – has led to an increased understanding of the molecular basis for innate antiviral immunity. Subsequently, an immense amount of research has been conducted to define which innate immune sensors and effectors are critical for defense against specific viruses and their underlying mechanisms of action. Since standard cell lines in monolayer culture are readily scalable and generally amenable to overexpression or knockout approaches, they remain the ‘go to’ system for the identification of novel pro- and antiviral host factors and initial phenotyping of their impact on viral replication. However, antiviral signaling pathways are defective in some cell lines. Furthermore, viruses exhibit tropism for specific cell types in vivo; not only does sensor and effector expression vary between cell types, but the same cell may also respond differently (e.g., in type and magnitude) depending on cell differentiation state, cell-cell interactions, and other microenvironmental cues. The spatiotemporal patterns of the response will also depend on the location of infected and uninfected cells within the tissue and other physical dynamics within the system. Therefore, both cell type and its context are important factors when studying innate antiviral immune responses. In this regard, animal models are useful, as the complexity of these interactions is certainly captured in vivo; however, species-specific differences in immunity may not allow for direct extrapolation of results to infections in the human population, and human challenge studies are not always feasible.

Cell culture systems that capture relevant cell types, tissue architecture, and biomechanics to recapitulate the initiation, dissemination, and impact of antiviral signals have the potential to yield further insight. Tissue-engineered and 3D cell culture platforms are capable of meeting these requirements by incorporating relevant homo- and heterotypic cell-cell interactions, extracellular matrix (ECM) components, and multi-organ compartments. Additionally, these cultures can be derived from primary cells which have fewer defects in antiviral pathways that often result from mutations accumulated through continuous passaging or immortalization methods. Here we provide a brief overview of pathogen detection and signaling pathways that are central to the interferon response, review how these sensors and effectors can vary between model systems, and highlight the application of more physiologically-relevant in vitro systems to probe novel concepts in intrinsic and innate antiviral immunity.

II. Key players in innate antiviral defense

Activation of the innate antiviral response is based on the ability of a cell to distinguish ‘self’ from ‘non-self’. This is achieved, in part, through cellular expression of pattern recognition receptors (PRRs) that sense conserved structural features associated with bacterial, fungal, or viral pathogens, collectively termed pathogen-associated molecular patterns (PAMPs), as well as molecular signatures associated with damaged or dying cells referred to as damage-associated molecular patterns (DAMPs). Several classes of PRRs have been defined since Charles Janeway first hypothesized their existence [7], and individual receptors can be grouped by localization, ligand specificity, and function. Membrane-bound PRRs include the Toll-like receptors (TLRs) and C-type lectin receptors (CLRs), while cytosolic receptors include the RNA-sensing RIG-like receptors (RLRs), the absent in melanoma 2 (AIM2)-like receptors (ALRs), and nucleotide-binding oligomerization domain (NOD)-like receptors (NLRs). The importance of these receptors in the antiviral response is underscored by the vast array of viral strategies that have evolved to antagonize PRR activation and downstream signaling pathways (reviewed in [8–10]).

TLRs were the first PRRs described in humans [11–13], and to date, seven (TLR2, −3, −4, −7, −8, −9, and −10) of the ten TLRs identified in humans have been shown to play an important role in mediating host antiviral responses [reviewed in [14]]. TLRs are found on cell-surface and intracellular (e.g., endosomal) membranes where they are well-positioned to detect viral glycoproteins (e.g., TLR2 [15]; TLR4 [16], also reviewed in [14]) or nucleic acid moieties (e.g., TLR3 [17], TLR7 [18]; TLR8 [19]; TLR9 [20]) during initial virus uptake into the cell (Figure 1, left). Indeed, virus-associated PAMPs are often linked to nucleic acid structures found in viral genomes, replication intermediates, or viral transcripts such as uncapped or double-stranded (ds) RNA as well as CpG-DNA that are either absent or sequestered in host cells. Activated TLRs signal via their Toll-Interleukin-1 resistance (TIR) domain through adapter molecules such as myeloid differentiation factor 88 (MyD88) and TlR domain-containing adaptor-inducing interferon-β (TRIF), ultimately leading to the nuclear translocation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), activator protein 1 (AP-1), or interferon regulatory factors (IRF3, IRF7). Collectively, these transcription factors drive the production of both pro-inflammatory cytokines and interferons. Similar to TLRs, CLRs are also membrane-associated and found at the cell surface. CLRs, however, bind carbohydrates and are most well-known for their role in mediating fungal immunity. Activated CLRs may signal through immunoreceptor tyrosine-based activation motif (ITAM), hemi-ITAM, or immunoreceptor tyrosine-based inhibitory motifs (ITIM); alternatively, CLRs such as dendritic cell-specific intercellular adhesion molecule-3 grabbing non-integrin (DC-SIGN) and liver/lymph node-specific intercellular adhesion molecule-3 grabbing non-integrin (L-SIGN) mediate internalization of the receptor-ligand complex (reviewed in [21]). CLR-ligand endocytosis can promote antigen processing and major histocompatibility complex (MHC) presentation; however, in some cases, viruses such as human immunodeficiency virus (HIV) [22] and dengue virus (DENV) [23] have hijacked this interaction to promote trans-infection or virion uptake into target cells.

Figure 1: Diagram of pattern recognition receptor-initiated signaling events, interferon production, and subsequent interferon signaling leading to interferon-stimulated gene expression.

Left) Pattern recognition receptors (blue), including membrane-associated TLRs, and cytosolic RNA (RIG-I, MDA5), and DNA sensors (cGAS, IFI16, DAI, DDX41), signal through adaptor proteins and downstream kinases (red), resulting in transcription factor activation (yellow; NF-κB, IRF3) and subsequent expression of IFN and pro-inflammatory cytokines. Right) Following secretion, IFN-I and IFN-III bind to their respective receptors on the cell surface. Receptor dimerization triggers the JAK/STAT pathway leading to phosphorylation and homo- or heterodimerization of STAT proteins. Both STAT1/STAT2 heterodimers (in association with IRF9 to form ISGF3) and STAT1 homodimers can drive ISG expression via interaction with ISRE and GAS elements, respectively, in target promoters. Created with BioRender.com.

RLRs include retinoic acid-inducible gene I protein (RIG-I), melanoma differentiation-associated protein 5 (MDA5), and laboratory of genetics and physiology 2 (LGP2). Both RIG-I and MDA5 sense dsRNA, although RIG-I associates preferentially with short 5′ di- and triphosphorylated short dsRNA [24,25] and MDA5, with long dsRNA ([26]; reviewed in [27]). Notably, although known for its role in defense against RNA viral infections, RIG-I can also recognize DNA viral infections via RNA polymerase III-mediated conversion of AT-rich dsDNA into 5’-triphosphate (ppp) RNA [28,29]. Upon PAMP interaction, RIG-I and MDA5 interact with the downstream adaptor mitochondrial antiviral-signaling protein (MAVS) (also known as IFN-β promoter stimulator I (IPS-1), virus induced signaling adaptor (VISA), or caspase activation recruitment domain (CARD) adaptor inducing IFN-β (Cardif)) via their CARD domain to trigger interferon expression. LGP2, on the other hand, lacks a CARD domain, indicating LGP2 cannot initiate downstream signaling but may instead regulate RIG-I and MDA5 [30]. Cytosolic DNA receptors have also been identified and include DNA-dependent activator of IFN-regulatory factors (DAI) [31], DExD/H box helicases (e.g., DEAD-box helicase 41 (DDX41) [32]), interferon gamma inducible protein 16 (IFI16) [33], AIM2 [34], and cyclic GMP-AMP synthase (cGAS) [35], among others. Many of these receptors signal through stimulator of interferon genes (STING) and TANK-binding kinase 1 (TBK1) leading to IRF3 activation, although other pathways have also been reported (e.g., through β-catenin [36]) that also result in IFN gene transcription. In contrast, AIM2 engages apoptosis-associated speck-like protein containing a caspase recruitment domain (ASC) leading to inflammasome activation and caspase-1-mediated proteolytic processing of immature IL-1β and IL-18 [34], despite a lack of structural similarity to the NLR class of PRRs typically associated with inflammasome activation. While NLRs play an important role in host immunity, their role in antiviral defense is less well understood and both positive and negative effects on the interferon response have been described, often through intersection with the antiviral signaling pathways noted above (reviewed in [37]).

To date, three types of interferon have been defined according to shared structural features, receptor usage and biological effects: type I IFN (IFN-I; including 13 IFN-α subtypes, −β, −ε, κ, −ω) [5,38–41], type II (IFN-II; IFN-γ) [42], and type III (IFN-III; comprising 4 IFN-λ subtypes) [43–45], with IFN-I and -III having well-established antiviral activity. This activity is mediated through autocrine and paracrine signaling where IFN-I and -III bind unique heterodimeric receptors – IFNα/β receptor subunit 1 (IFNAR1) / IFNα/β receptor subunit 2 (IFNAR2) in the case of IFN-I [46,47]; IFNLR1 / IL-10R2 in the case of IFN-III [43,44] (Figure 1, right). Receptor ligation by either IFN-I or -III leads to receptor dimerization and Janus kinase 1 (JAK1) phosphorylation which then triggers phosphorylation of the receptor and subsequent recruitment of signal transducer and activator of transcription (STAT) proteins (reviewed in [48]). Phosphorylated STAT1 and STAT2 heterodimerize and associate with IRF9 to form IFN-stimulated gene factor 3 (ISGF3) which can then translocate to the nucleus and stimulate antiviral gene transcription through binding of interferon-stimulated response elements (ISREs). Beyond this canonical signaling pathway, other kinases (e.g., JAK2, tyrosine kinase 2 (TYK2)) and STAT proteins can be phosphorylated, or STAT-independent pathways (e.g., mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K)) activated after IFN treatment (reviewed in [49]). Notably, STAT1 homodimerization, or heterodimerization with other STATs (e.g., STAT3), can promote expression of additional interferon-stimulated genes (ISGs) through interaction with a gamma interferon activation site (GAS). Although hundreds of ISGs and non-coding RNAs have been identified ([50]; reviewed in [51]), the mechanism of action of relatively few have been described in detail. Nonetheless, it is clear from these studies, that ISGs antagonize the viral life cycle at each stage, helping to clear infection through elimination of viral components, induction of apoptosis, or by conferring resistance to uninfected cells. IFN-I can also have antiproliferative effects, modulate cell differentiation, and further sculpt the immune response through activation or suppression of specific immune cell subsets (e.g., T cells and regulatory T cells) [52,53]. These differential effects may be the result of IFN concentration, variation in receptor expression levels, or time of IFN exposure (reviewed in [54]).

Finally, negative regulators of the IFN response play an essential role in resolving inflammation after viral clearance to avoid excessive toxicity and immunopathology associated with interferonopathies. Resolution of interferon expression and signaling occurs through a variety of mechanisms mediated by direct protein-protein interaction, post-translational modification, or non-coding RNAs that result in the degradation or inactivation of the PRRs, signaling intermediates, transcription factors, IFNs, and IFN receptors mentioned above. For example, JAK / STAT signaling is curtailed by IFN receptor endocytosis [55] as well as through the binding of ubiquitin-specific peptidase 18 (USP18) to IFNAR2, thereby blocking JAK1 interaction [56]. Similarly, suppressor of cytokine signaling (SOCS) proteins bind phosphorylated tyrosine residues to inhibit JAK activity and further target signaling components for ubiquitination and subsequent degradation via their Src homology 2 (SH2) and SOCS box domains, respectively (reviewed in [57]).

III. Important considerations in modeling authentic antiviral responses

IIIa. Variation in the expression and function of innate immune sensors and antiviral effectors

Interspecies variation

Faithful recapitulation of human antiviral responses in a model system requires matched expression and functional conservation of relevant PRRs, signaling molecules, IFN receptors, and the full repertoire of ISGs. As noted above, cell lines from different species are often used as an in vitro “host,” while animal models are a frequently-employed tool for assessing viral pathogenesis. Mice in particular have been extensively utilized for analyzing various aspects of the host response to infection, and are therefore our focus in this section, although certainly additional variation exists across other model species (reviewed in [58–61]). While the overall architecture of the murine immune system is remarkably similar to humans given our drastically different lifestyles, key differences do exist [62–65], including in the phenotype and function of specific leukocyte populations (reviewed in [66]) and in a multitude of host factors that contribute to the innate antiviral response [67].

As species lineages radiated, each was subjected to their own set of diverse pathogens which drove variation in innate immune genes and led to significant changes in immune profiles over time [68]. For example, among the 84 rapidly-evolving, higher-primate, immunity-related genes as identified in Barreiro and Quintana-Murci [69], remarkably 30 of these have been linked to HIV. While this highlights the impact of just a single virus, the authors stipulate these virus-driven alterations in immunity are a product of selection not solely targeting HIV, but rather against a multitude of ancestral retroviral pathogens. Thus, species-specific aspects of innate antiviral immunity must be considered when using a non-human model system. This is true for both in vitro or in vivo systems since the level of conservation between humans and animals impacts the utility of both small animal models and cell lines from other species in dissecting the parameters of innate antiviral immunity for specific pathogens.

Not surprisingly, proteins involved in regulating the immune response, especially transcription factors and kinases have been shown to be more highly conserved between species, constrained by their multi-functional roles, whereas cytokines display strong specifies-specific diversity, attributed to maintaining a singular role in combating host-specific pathogens [67]. TLRs, for example, are incredibly well conserved and are estimated to have originated with the diversification of vertebrates. Only TLR10, expressed in humans not mice, and TLR11–13 in mice and not humans, represent the main discrepancy to the overall TLR conservation [70]. Similarly, some NLRs have undergone expansion in mice (e.g., Nlrp1, Nlrp4, Nlrp9), while others have been lost (e.g., NLRP8, NLRP13) [71]. However, despite similar representation of highly-conserved orthologous proteins, the cellular expression pattern and tissue distribution of some PRRs is not uniform between species. For example, TLR4 in humans is highly expressed in spleen and peripheral blood lymphocytes (PBLs) [12,72]; while in mice, expression is most pronounced in the lung, heart, and spleen [73]. Expression profile differences are predicted to also exist in TLR2, TLR3, and TLR9 (reviewed in [74]). The molecular basis for these differences in basal TLR expression as well as differences in TLR upregulation in response to stimuli may be related to poorly conserved promoter sequences [75].

Evolution across species can also impact PRR ligand affinity and specificity, leading to differences at the functional level even in cases where expression profiles and the signal-transmitting domain (e.g., TIR, in the case of TLRs) is well-conserved. For example, the optimal CpG motif for human TLR9 activation (GTCGTT) differs from murine TLR9 (GACGTT) [76]. Further, there are species-specific differences in recognition of single-stranded RNA, with TLR7 and TLR8 acting as the primary sensors for this PAMP in humans, and TLR7 and TLR13 acting in mice [18,19,77,78]. Notably, murine TLR8 was originally thought to be non-functional, but subsequent work demonstrated activity after imidazoquinoline stimulation in the presence of DNA oligonucleotides [79] and a more recent study by Zhang and colleagues suggests a role for TLR8 outside of immunity [80]. Further, even in cases where human and mouse innate immune proteins function similarly in their respective hosts, relatively minor differences in sequence can have significant impacts on virus-host interactions and resulting antiviral response. For example, viruses such as hepatitis A virus (HAV) and hepatitis C virus (HCV) are known to abrogate IFN production via cleavage of the upstream adaptor molecule, MAVS; however, a lack of sequence conservation at the HAV cleavage site in mice curtails the ability of HAV to block signaling via this mechanism, leading to an enhanced host response and failure of the virus to establish infection [81]. Similar barriers exist for other viruses such as DENV and Zika virus (ZIKV), both of which fail to cleave mouse STING [82,83] and degrade mouse STAT2 [84,85], restricting infection.

Differences in interferon gene composition also exist between species, illustrated by the expression of only two functional IFN-III genes, Ifnl2 and Ifnl3, in mice, compared to the four expressed in humans. The implications of this in modeling the human antiviral response, however, remain unclear since the role of specific interferon subtypes in humans remains elusive despite purifying (in the case of IFN-α) or positive (as observed for IFN-λ) selection to suggest essential, unique functions [86]. Additionally, while IFN-I signaling is conserved across vertebrates, IFNAR1 and IFNAR2 proteins differ substantially, with only 50–51% sequence identity between humans and mice, making it difficult to relate IFN receptor activation and the precise nature of the interferon-stimulated response in murine models back to a human context [87]. Finally, many inbred mice used in laboratory experiments lack one or more interferon-stimulated genes (ISGs). CBA/J and BALB/c strains, for example, contain defective Mx1 genes [88]. While these defects may facilitate the use of otherwise non-permissive hosts to model pathogenesis, (e.g., influenza (IAV) [89]), researchers focusing on specific aspects of intrinsic antiviral immunity at the molecular level should carefully select the host background to ensure accurate modeling.

Intrahost variation

In addition to the differences in innate antiviral host factor expression patterns between species described above, variation exists within a host as a function of tissue and cell type [90–92]. This variation can impact the initiation, magnitude, and even primary mechanism of an antiviral defense; thus, even when using human cells to model antiviral responses, the tissue source and identity of the cells must be considered. For example, stem cells – a useful tool in a wide range of tissue engineering and regenerative medicine applications – are known to be resistant to viral infection (reviewed in [93]). This suggests they employ different antiviral defense mechanisms from their differentiated counterparts. Supporting this hypothesis, early observations noted that stem cells do not produce interferon and are insensitive to IFN treatment [94,95]. More recently, Maillard and colleagues demonstrated antiviral RNA interference (RNAi) in undifferentiated cells [96] and Wu et al. revealed that stem cells are characterized by intrinsic expression of canonical ISGs [97] providing additional insight into the underlying mechanistic basis for stem cell resistance to viral infection.

PRR expression profiles across fully differentiated cells are unique to individual receptors with RLRs being found in most cell types and NLRs, CLRs, and TLRs frequently detected in myeloid cells such as dendritic cells and macrophages in addition to endothelial and epithelial cells, suggesting a more restricted expression profile (reviewed in [98–102]). Furthermore, expression of specific TLRs across all epithelial cells is not uniform. Ioannidis et al. found that human tracheal epithelial cells lacked the TLR8 protein, and only expressed TLR6 and TLR2 in basal cells while other TLRs (TLR4, 5, 7, 9, 10) were confined to the luminal surface. In contrast, TLR1 and TLR3 were detected throughout the pseudostratified epithelium [92] (Figure 2A). Similar data have been reported in the gut, where Price et al. utilized TLR reporter mice to visualize TLR expression in intestinal epithelial cells, revealing temporal differences and spatial compartmentalization that correlated with ligand-induced responses [91]. Most notably, TLR7 and TLR9 were not detected in epithelial cells, but found in the underlying lamina propria, and TLR5 expression in the small intestine was only found in Paneth cells in the crypt (Figure 2B). Huhta and colleagues took an immunohistochemical approach to similarly characterize TLR expression in human and mouse alimentary tracts, noting higher expression of TLRs in the small intestine [99], while Faure-Dupuy et al. clearly show distinct TLR expression profiles in primary liver cell subsets [103].

Figure 2: Variation in Pattern Recognition Receptor Expression Profiles Across Cell Types.

A) Immunofluorescence staining using TLR-specific antibodies (red; nuclei, blue) in histological sections of human tracheal epithelium reveals differences in TLR expression and distribution across cell types. Images are oriented with basal cells along the bottom and the airway lumen on the top (reproduced from [92]). B) Green-fluorescent protein (GFP), yellow-fluorescent protein (YFP), or tdTomato (TOM) was inserted in the 3’ end of the endogenous murine TLR gene of interest to create TLR reporter mice. Sections of TLR reporter mouse intestinal epithelium were then stained with antibodies targeting GFP/YFP (green) or tdTomato (red) to identify specific TLR expression across the small intestine, proximal and distal colon regions. Sections were also probed with anti-Epcam antibodies to identify epithelial cells (gray). Scale bars = 50 μm (reproduced from [91]).

In addition to PRRs, the basal level of downstream transcription factors, such as IRFs, also contribute to variation in IFN production. Plasmacytoid dendritic cells (pDCs), for example, have high constitutive levels of IRF7 [104] that enable them to produce very high levels of IFN-I, on the order of 200–1000 times more IFN-α than other blood cells, upon stimulation [105]. Differences in the type(s) of IFN a particular cell produces and the type(s) of IFN it can respond to are also critical. Nearly all cells can produce IFN-I, although subtype expression varies, with IFN-β being ubiquitous, IFN-α being produced primarily by leukocytes [106], and other IFN-Is (e.g., IFN-κ, IFN-ε) being expressed in a tissue-specific manner [39,107,108]. Similar to IFN-I, IFN-III is produced by many cell types; however, while essentially all nucleated cells can respond to IFN-I, IFN-III activity is observed primarily at epithelial surfaces as a result of receptor distribution [43,109]. Lastly, the pattern of ISG induction varies by cell type [110,111] – an observation that will likely be reinforced through the ongoing application of single cell technologies.

Primary cells vs. continuous cell lines

Cells from a given species, tissue, or of a certain type may be cultured as primary cells, or may be cancer-derived or immortalized, and thus propagated continuously. These cell lines have been essential for a multitude of seminal work in understanding the antiviral response to infection since they are amenable to genetic knockout and overexpression studies, and can be readily scalable to accommodate high-throughput screens. Continuous cell lines are chosen for their fast growth rates which, when combined with their inability to undergo senescence, make for an incredibly robust experimental pipeline. However, as IFN can drive anti-proliferative and pro-apoptotic responses, it is not surprising that antiviral signaling pathways intersect with those involved in tumor suppression. Cells lacking antiviral protein expression are more readily immortalized [112], while conversely, cells lacking tumor suppressors (e.g., p53) or expressing oncogenes (e.g., SV40 large T-antigen, papillomavirus E6) are impaired in their ability to make and respond to IFN ([113,114]; reviewed in [115]) and are thus more susceptible to viral infection [116,117]. Further supporting the link between antiviral and antitumor immunity, Critchley-Thorne and colleagues [118] demonstrated that IFN signaling and subsequent ISG expression is impaired in breast cancer, and Hare et al. note that antiviral genes are often silenced or selected against during tumorigenesis [119]. Therefore, it is not surprising that many cancer-derived cell lines have defects in their antiviral components. For example, Huh7.5 cells contain a mutation in the RIG-I gene (T55I) that impairs IFN signaling and has been linked to enhanced permissiveness for HCV [120] and Vero cells have lost the ability to produce IFN due to spontaneous gene deletions [121,122]. While these defects have made them useful for cultivating certain difficult-to-grow viruses or for amplifying attenuated vaccine candidates, they are not optimal for assessing authentic antiviral responses.

While continuous cell lines are usually still capable of responding to IFN-I and potent IFN-I inducers, such as the dsRNA mimic polyinosinic-polycytidylic acid (Poly I:C) [123], implementing studies using primary cells allows for a more authentic recapitulation of the cell’s progenitor tissue, including its innate immune response (Table 1). However, these cells are not without their own drawbacks. Passaging of primary cells rapidly approaches their senescent end-point, resulting in significant differences from earlier passages including cell enlargement, impeding apoptosis, and alterations to their normal cell cycle. Most significantly, older primary cells have heightened basal-level expression of IFN-β and mount an abnormally stronger interferon response ([124]; reviewed in [115,119]). Therefore, when performing innate immunity studies in primary cells it is critical to carefully manage over-passaging and to match passages for all experiments to ensure reproducibility.

Table 1.

Comparison of primary and continuous cell lines.

	Primary Cells	Continuous Cell Lines
Origin	Direct from Donor	Tumor-derived or immortalized
Ease of Use	Difficult	Easy
Genetic Manipulability	Low	High
Transfectability	Low	High
Growth Rate	Low	High
Passaging	Very Limited	Indefinite
Intact Signaling Pathways	Well preserved	Often defective
Representative of Progenitor Tissue	High	Low
Reproducibility	Inter-donor variation	Often clonal
Differentiable	High	Few cases
Preparation Time	Days to weeks	Immediate
Polarity	High	Low
Cost	High	Low

To overcome many of the challenges of primary cell culture studies, methods to immortalize these cells have been applied. Transduction of primary cells containing human telomerase reverse transcriptase (hTERT) instead of SV40 large T-antigen best preserves the complex innate immune framework of the non-immortalized primary cell, thus allowing for creation and implementation of immortal primary cell lines [125]. Through this method, cells are uniformly immortalized which lessens the bottleneck effect during spontaneous immortalization and also does not apply selective pressures which can silence antiviral pathways. Granted, even when primary cells are immortalized by hTERT, they too will eventually garner enough mutations to distinguish them from their primary progenitors and ultimately behave more like a cancerous cell line. Still, utilization of these cells can address the passaging difficulties of non-immortalized primary cell lines, thus allowing for more robust studies requiring cell modifications and selection, including overexpression and knockout using CRISPR technologies [119].

IIIb -. The impact of the microenvironment on innate antiviral responses

Cell-cell interactions

After identifying the appropriate cell type(s) in which to assay innate antiviral responses, consideration of the cellular context is critical since it can impact cellular phenotype, function, and ability to respond to stimuli. In vivo, cells are frequently in contact with one another and typically more than one cell type is present in the microenvironment. Thus, even in cases where cells are not in direct contact with one another, the inherent heterogeneity in antiviral response across component cell types can impact the system as a whole (Figure 3). This is especially relevant when striving to understand the effects of innate immune responses on a multicellular or tissue level.

Figure 3: Impact of cell-cell and cell-extracellular matrix interactions on antiviral responses.

Infected cells secrete cytokines (e.g., IFNs) that amplify antiviral and pro-inflammatory responses through autocrine and paracrine signaling. Notably, the type and magnitude of the host response can vary across different cell types (compare brown and blue cells), each of which contribute to the antiviral and inflammatory state at the culture-wide (i.e., tissue) level. Additionally, transport of the second messenger cGAMP across gap junctions can activate intrinsic immunity in neighboring cells, while viral components (e.g., nucleic acids, proteins, virions) can be transferred directly to other cells via nanotubes, contributing to the spread of infection along with other modes of virus cell-to-cell spread [129,133]. In the extracellular space, ECM components can impact local cytokine concentration and activation, while ECM disruption following release of proteases during infection produces DAMPs that can further stimulate inflammation. Created with BioRender.com.

For adjacent cells, intercellular contacts can contribute to virus spread and propagation of antiviral signals. For example, a variety of reports have described direct cell-cell transfer of viral components through cell protrusions such as tunneling nanotubes [126–128]. This mechanism of virus spread is not only more efficient than the assembly, release, and receptor-mediated uptake of nascent viral particles, but also shields the virus from antibody-mediated neutralization (reviewed in [129]). Furthermore, delivery of viral components (i.e., PAMPS) into neighboring cells via non-canonical pathways may lead to the activation of other cellular sensors or evasion of PRR-mediated detection altogether. While additional studies are needed to address this specific hypothesis, existing reports have shown that viruses being transmitted via cell-cell spread are less sensitive to the activity of intrinsic antiviral restriction factors. In particular, in the case of HIV, neither tetherin [130] nor rhesus TRIM5-α [131] is effective at restricting cell-cell transmission of virus between T cells at the virological synapse. Murrell and colleagues similarly described the ability of IFN-α treatment to efficiently block cell-free, but not cell-cell mediated infection of dendritic cells (DCs) and Langerhans cells (LCs) by human cytomegalovirus (CMV) [132]. Furthermore, the direct transfer of viral proteins that act to antagonize antiviral responses, such as the IAV NS1 protein, is hypothesized to suppress innate antiviral pathways in recipient cells [126].

Antiviral signals can also be directly transferred between adjacent cells. Specifically, the second messenger cyclic GMP-AMP (cGAMP), which is produced following cGAS activation, can pass through gap junctions due to its small molecular weight [134,135]. Furthering these observations, Pepin et al. demonstrated that connexin-dependent cGAMP transfer from epithelial cells to phagocytes leads to STING-dependent transactivation and contributes to the propagation of antiviral responses, impacting IAV titers [136]. In some cases, cell-cell contact, or short-range exosomal transfer may even be required for activation of the antiviral response (reviewed in [137]). For example, Decembre and colleagues studied pDC responses during DENV infection, noting that pDCs only produced IFN-α when co-cultured with infected cells [138]. Thus, in sum, the component cell types comprising the model, as well as the direct cell-cell contacts influence the magnitude and dynamics of the antiviral response.

Cell-ECM interactions

Microenvironmental impacts on the host response are not limited to cell-cell interactions but are also influenced by extracellular matrix (ECM) components (Figure 3). The ECM comprises a diverse set of ~300 proteins arranged in a tissue-specific manner with varying concentrations and patterns [139]. Historically, the extracellular matrix was thought of as simply a scaffold to allow for cell adhesion and tissue development. However, more recent work has shown that the ECM is both a dynamic compartment – being degraded and remodeled during infection and subsequent tissue repair – and an active player in modulating host immune responses. Specifically, ECM can regulate cell proliferation, differentiation, and migration through interaction with cell surface adhesion molecules and receptors, and influence cellular responses to soluble immunomodulatory factors by controlling their spatial distribution and activation state (reviewed in [140–142]). Indeed, the dissemination of cytokines that are secreted during infection in vivo is influenced by both surrounding cells and the structure and porosity of the ECM, which limits their simple diffusion ([143,144]; reviewed in [145]). As a result, local concentrations of cytokines can vary, yielding differential effects on cells throughout the microenvironment. Notably, this is in contrast to cytokine release in vitro in traditional monolayer cell culture models where gradients cannot be maintained in the liquid medium.

ECM components, such as proteoglycans and fibrillar proteins, can also directly bind growth factors and cytokines, including IFNs (reviewed in [146]). For example, Yoshida et al. [76] described the ability of fibronectin associated with the extracellular matrix to directly bind IFN-α and further demonstrated that ECM-bound IFN-α can induce STAT1 expression and cytotoxic effects in Panc-1 cells. Cellular exposure to different types of matrices can also impact their responsiveness to the same input signal. Kuwashiro et al. [77] demonstrated that fibrotic collagen associated with liver fibrosis can attenuate ISRE-driven luciferase activity in Huh7 cells after IFN-α treatment. This phenotype was linked to an increase in β1-integrin expression in cells cultured on type I collagen and was also shown to limit the effect of IFN on HCV replication [77].

Beyond ECM component interactions with host-derived soluble molecules, ECM components can also influence the antiviral response through direct pathogen interactions or stimulation of PRRs. The former can result in viral particle neutralization, as shown for tenascin-C – HIV interaction [147], or enhanced clearance, as described in the case of mindin and IAV [148]. Notably, the release of proteases and hydrolases such as matrix metalloproteinases (MMPs), a disintegrin and metalloproteinases (ADAMs), and a disintegrin and metalloproteinases with thrombospondin motifs (ADAMTs) triggered during infection or released from dying cells can fragment the ECM. These degradation products (e.g., heparan sulphate; hyaluronic acid) can then act as DAMPs to trigger pro-inflammatory cytokine expression and leukocyte recruitment through TLRs or other cellular receptors (reviewed in [149]). Furthermore, ECM remodeling enzymes can also influence the antiviral response directly through other mechanisms. Here, Marchant et al. showed that MMP-12 is taken up by virus-infected cells where it drives transcription of NFKB1A, facilitating IFN-α secretion [150]. While these data define a direct role for MMP-12 in gene transcription, the authors also demonstrated MMP-12 can cleave IFN-α, preventing its ability to signal through its receptor, and thereby attenuating the antiviral response.

Cellular polarity

In vivo, individual cells adopt diverse morphologies through adhesive interactions with ECM and by the formation of junctions with neighboring cells, giving rise to distinct biological interfaces and in some cases, apical-basolateral polarity. It is well known that such cellular polarization can impact modes of viral entry; human parainfluenza virus 3, for example, can only infect the airway epithelium via the apical side [151] while measles virus preferentially enters through the basolateral surface [152]. However, in addition to viral entry factors, PRRs and IFN receptors also exhibit polarized expression which impacts how cells sense viral pathogens and respond to antiviral signals [153,154] (Figure 4A and 4C). For example, Muir et al. showed apical localization of TLR2 in human airway epithelial cultures at air-liquid interface (ALI) derived from individuals with cystic fibrosis, while TLR4 and TLR5 were predominantly basolateral [155]. Using primary human airway epithelial cells, Cienwicki et al. demonstrated that IFNAR2 localization changes as these cells undergo differentiation into a pseudostratified epithelium [153]. Notably, the predominantly basolateral detection of IFNAR2 in mature ALI airway cultures in vitro recapitulated the pattern of expression in human tissue ex vivo (Figure 4C), suggesting these cells are well-poised to respond to IFN-I produced by other cell types in the underlying tissue and may also limit receptor access when the epithelial barrier is intact. Similarly, in the GI tract, TLRs are concentrated on the basolateral surface of mucosal epithelial cells to minimize stimulation by commensal microbiota, but trigger host responses to more invasive pathogens [156]. Specifically, the asymmetrical distribution of TLR3 in the gut was linked to elevated IFN production following basolateral inoculation of intestinal organoids with mammalian reovirus, despite the fact that similar levels of infection were achieved after apical delivery of virus [154] (Figure 4B). Polarized cells may also release interferon and other cytokines asymmetrically following infection. This has been observed following infection by parainfluenza viruses and SARS-CoV-2 [157,158] (Figure 4D). The release of different amounts of cytokines to distinct extracellular compartments over different timescales may impact the cell types that receive the signal and the spatiotemporal dynamics of the response that follows which is discussed further in the section below.

Figure 4: Polarized receptor expression and cytokine release impacts the dynamics of the host response to infection.

A) Basolateral localization of TLR3 (green) in human colon organoids co-stained for the cellular tight junction protein ZO-1 (red), cytoskeletal protein actin (magenta), and nuclei (blue). Scale bar = 20 μm. B) Intestinal organoids shown in (A) were inoculated with mammalian reovirus via the apical (MRV A) or basolateral (MRV B) surface and expression level of both the MRV μ2 genome segment and IFN-β was determined by quantitative reverse transcription PCR. The data show that despite similar levels of infection (top panel), basolateral inoculation resulted in significantly stronger induction of intrinsic antiviral immunity (bottom panel), likely due to the localization of TLR3. Mean +/− st dev; p values were calculated with two tailed unpaired t-test (reproduced from [154]). C) Immunohistochemical detection of IFNAR2 in ciliated human trachea ex vivo (top; staining shown in brown) and in human airway epithelial cultures over the course of differentiation in vitro (at day 1, 6, and 29; bottom). Scale bar = 10 μm. (reproduced from [153]). Airway lumen is positioned at the top in all images. D) Polarized cytokine release (apical vs. basolateral) following infection of an in vitro model of human airway epithelium with SARS-CoV-2 (MOI = 0.5; 72 hpi). Cytokines were quantified by Luminex assay and values below the limit of detection are designated nd. All data were analyzed using one-way or two-way ANOVA. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 (reproduced from [158]).

Biomechanics

In addition to the impacts of cell-cell and cell-ECM interactions on cellular polarity and the antiviral response summarized above, cellular contacts and ECM rigidity further impact cellular responses as mechanical stimuli. Indeed, cells are subject to a variety of mechanical forces in vivo which can regulate intracellular signaling through the activation of mechanosensory ion channels, yielding effects on gene expression and overall cell phenotype.

In the lung, for example, airway cells (including epithelial, endothelial, and mesenchymal cells) are subjected to continuous cyclical stretch and compression during the respiratory cycle. While the impact of these forces on the IFN response specifically is unclear, it has been shown in alveolar cells to impact surfactant secretion and even contribute to a pro-inflammatory phenotype during mechanical ventilation (reviewed in [159]). Recently, Kilic et al. studied the effect of compression on normal airway epithelial cells, demonstrating changes in gene expression as early as 3 hours after mechanical stimulation and the induction of an asthma-like transcriptional profile [160]. Similarly, Solis et al. reported the expression of pro-inflammatory and chemoattractant mediators by macrophages and monocytes exposed to in vitro cyclical hydrostatic pressure was dependent on the ion channel, PIEZO1 [161]. PIEZO1 drives calcium-mediated activation of transcription factors AP-1 and endothelin 1 (EDN1) in response to mechanical stimuli - elegantly demonstrating how these types of extracellular cues can impact the innate immune response even in the absence of PRR activation [161]. Undoubtedly, immune cells sense and respond to a variety of mechanical stimuli during extravasation, transmigration, and eventual infiltration to specific tissues during the response to viral infection, and recent advances in technology facilitate further work in this area (reviewed in [162]).

In blood vessels, shear stress induced by non-laminar flow impacts NF-κB activity [163] and gene expression profiles in endothelial cells [164,165], while laminar flow was observed to suppress TLR2 levels [132] in addition to IFN-γ-mediated STAT1 activation and CXC chemokine expression [166]. Beyond the impact of biomechanics on gene expression, fluid flow within a tissue can impact the local concentration of soluble mediators and thus the spatiotemporal dynamics of the host response. Conversely, when flow is slowed, such as at arterial bifurcations, it can lead to high concentrations of cytokines due to a ‘pooling effect’ yielding to inflammation [167]. While the impact of flow on endothelial cell phenotypes has been investigated primarily through the lens of atherosclerosis, these findings have implications for modeling antiviral responses in endothelial cells under static vs. dynamic conditions in vitro.

IV. Utility of tissue-engineered and 3D cell culture systems in modeling innate antiviral responses

Historically, traditional monolayer cell culture systems have provided the foundation for the discovery of many of the innate immune pathways discussed in section II. However, as detailed in section III, cellular context impacts antiviral responses. Notably, a variety of in vitro model systems, including precision-cut tissue slices, epithelial cultures at ALI, organoids, and ‘on-chip’ technologies, among others, can incorporate human primary cells, recapitulate aspects of tissue architecture and function, and integrate biological forces present in vivo such as cyclical stretch and fluidic shear (Figure 5). Since many of these systems are susceptible and permissive for viral infection, they provide a more robust platform to model physiologically-relevant antiviral responses. Detailed descriptions of these model systems, including organoids of the brain [168,169], airway [170,171], gut [172,173], placenta [174,175], and liver [176,177], have been reviewed elsewhere [178–180]; thus, in this section, we limit our discussion to the application of these models to studying innate antiviral responses.

Figure 5: In vitro model systems with emergent properties.

Schematics are shown for a variety of tissue culture platforms that recapitulate aspects of organ-specific, in vivo physiology. For tissue slice cultures, organotypic tissue is excised and cultured ex vivo for a short duration prior to infection. Air-liquid interface (ALI), organoid (also spheroid and enteroid), and on-chip cultures all are initiated from dissociated tissue before an expansion phase. For ALI cultures, cells are seeded on transwells for several days before media is removed from the apical chamber establishing the ALI system. Organoids, spheroid, and enteroid cultures all result in a recapitulation of the progenitor organ system, yet differ in how the culture is initiated. On-chip cultures have cells seeded into microfluidic chips specialized for a specific experimental goal. Created with BioRender.com.

As noted above, pseudostratified and organotypic models of the airway epithelium and intestinal mucosa recapitulate the cell type-specific expression and polarized distribution of PRRs and IFN receptors observed in vivo [91,92]. Thus, it is not surprising that these models, as well as other in vitro tissue culture systems with emergent properties, have been shown to respond to immunomodulatory stimuli and mimic in vivo innate immune responses. For example, Henjakovic et al. characterized cytokine production in precision-cut ex vivo lung slices in response to LPS, IFNγ, and dexamethasone [181] while Temann et al. determined that these ex vivo cultures can respond to LPS stimulation even after long-term culture, indicating this ex vivo system is a robust platform that can predict inflammatory responses to various stimuli [182]. Similarly, ex vivo lung slices were shown to mount a pro-inflammatory response following exposure to a nanoparticle-based IAV vaccine candidate and were subsequently used to provide evidence of antigen-primed T cells in the lung after initial in vivo vaccine challenge in mice [183,184]. Regarding other airway models, microarray analysis of tracheal bronchial cells obtained through brush biopsies compared to unstimulated primary human airway epithelial (HAE) cells cultured at either ALI or in submerged conditions and Calu-3 cells, found that differentiated HAE cultures at ALI best mimic the gene expression profile of these cells in vivo [185]. Hui et al. later compared IAV replication kinetics and tissue tropism preferences between human airway organoid cultures and ex vivo bronchus cultures from the same donor and found airway organoids to be a suitable mimic that was also capable of eliciting potent immune responses [186]. More recently, Cheemarla and colleagues compared SARS-CoV-2-infected organoids with serially swabbed nasopharyngeal samples and demonstrated comparative viral kinetics and innate antiviral responses including high levels of CXCL10 [187]. Furthering this analysis, the authors utilized these cultures to assess the ability of a preceding rhinovirus 1A (RV-1A) infection to inhibit SARS-CoV-2 through the induction of protective ISGs culture-wide, including in bystander cells [187]. This was similar to previous work by the same group where RV-1A-induced immunity protected against 2009 pandemic H1N1 IAV in ALI HAE cultures [188].

Outside the airway, 3D models of brain microglia grown in the presence of a hydrogel matrix have also been shown by multiple groups [189–192] to respond to LPS stimulation, while Abreu et. al [193] established a differentiated stem cell-derived micro brain sphere capturing multiple cell types capable of phenocopying in vivo innate immune responses including production of CCL-2, TNF-α, IL-1β, IL-6, and IL-10. Importantly, Watanabe et. al were able to develop a cerebral organoid system, inclusive of cortical and basal ganglia regions, that improved upon previous models by mounting more reproducible innate antiviral signaling events following viral challenge by pathogens such as ZIKV [194]. Likewise in the liver, the development of a 3D microfluidic primary human hepatocyte model by Ortega-Prieto and colleagues that could garner hepatitis B virus (HBV) replication upon low MOI infection and produce an in vivo-like innate response including IL-8, macrophage-inflammatory protein-3ɑ (MIP-3ɑ), SerpinE1, and monocyte-chemoattractant protein-1 (MCP-1) has proven to be a major development for comparing HBV strains and their potentially differing activation of host immunity [195].

Beyond their utility in profiling in vivo-like host responses, cell culture models with emergent properties have also been employed to uncover novel aspects of pathogen detection and antiviral response that would otherwise be absent in conventional monolayer cultures or cell lines (Table 2). Some examples of this have been highlighted earlier in this review (e.g., the polarity of antiviral responses). Additionally, these models are useful in understanding antiviral defense in complex or maturing tissues, for instance, in ZIKV transmission to the developing fetus during pregnancy and subsequently, in the developing brain. Here, Bayer et al. [196] demonstrate the inability of ZIKV (and DENV) to infect full-term primary human trophoblasts due to constitutive release of IFN-III which – through autocrine and paracrine signaling – protect trophoblast and adjacent non-trophoblast tissue from viral invasion. To further understand the mechanism of antiviral protection across the maternal-fetal barrier, Corry et. al were able to recapitulate the protective effects of IFN-λ1 and IFN-λ2 signaling in 3D primary trophoblast and organotypic chorionic villous explant models to highlight the antiviral properties of differentiated trophoblasts [197], and thus that maternal-fetal viral transmission is largely limited to early-stage, post-conception trophoblasts [198]. Through the use of human embryonic stem cell-derived cerebral organoids, Dang et al. were then able to confirm the activation of TLR3 upon ZIKV infection leading to a pro-apoptotic cascade, and ultimately a decrease in organoid volume characteristic of microcephaly [199]. Thus, without the use of these embryonic organoids, this connection between TLR3 signaling, apoptosis, and microcephaly would be unrealized. Likewise, additional studies utilizing brain organoids were able to link both neonatal herpes simplex virus I and CMV with microcephaly [200,201]. Moreover, while primary keratinocyte organoid models have become the gold-standard for studying and assessing the antiviral innate immune response against human papillomavirus (HPV), Jackson et. al recently reported an enhanced organoid model which incorporates Langerhans cells that can be utilized for studying early-stage infection and subsequent innate responses [202].

Table 2.

Current applications and description of ongoing advances in in vitro model technology that will enable further interrogation of antiviral responses in physiological context.

Furthermore, advanced culture techniques have yielded improved models of the blood-brain barrier, allowing for a better understanding of how antiviral defense impacts the ability of viruses to invade the brain. These techniques include transwell and ‘on-chip’ technology [203], the latter of which have been designed to incorporate constant shear stress which promotes in vivo endothelial physiology [204]. Still, blood-brain barrier on-chip systems are often inadequate for investigating the antiviral response given the large volume of virus often needed in addition to the low cell density of these platforms. However, Bramley et. al [205] recently established a 3D model system recapitulating the blood-brain barrier after growing human brain microvascular endothelial cells within a bioreactor to provide constant shear force. Using this system, they determined that the physical vascular barrier provides greater antiviral defense compared to the ability of these 3D cultures to mount an antiviral innate immunological response through ISG-mediated signaling. They further demonstrate that disruption of the tight junctions as a result of stimulation with proinflammatory cytokines including TNF-α, lead to enhanced viral susceptibility.

V. Additional applications and current challenges of assessing innate antiviral immunity in 3D model systems.

Innate antiviral immunity is complex and multifaceted. Understanding the mechanisms that regulate the activation, propagation, and resolution of antiviral and pro-inflammatory responses requires a model system that recapitulates the in vivo setting where infection takes place. Traditional monolayer cell culture methods featuring cell lines have – and still provide – immense value, due to their simplicity, ease of use and genetic manipulability. However, where possible, efforts should be made to validate findings and further study innate antiviral immunity through the lens of more specialized cell culture systems with emergent properties. These systems allow for better representation of the natural host and therefore preserve more biologically relevant responses.

Current 3D and tissue-engineered models have been successfully employed to define the polarity of viral infection and host response, as well as cell type-specific responses, and the impact of infection on developing tissues, as reviewed above. However, while these model systems clearly represent more physiologically-relevant platforms in which to interrogate virus-host interactions, their utility in assessing the antiviral response has not been completely exploited. For example, organoids and ‘on-chip’ platforms could facilitate additional investigation into the role of extracellular influences (e.g., ECM; fluid flow) on ISG and cytokine expression, in addition to the spatiotemporal dynamics of IFN signaling. Notably, transgenic mice expressing IFN-sensitive, cell-based reporters have been developed to visualize and track the host response in vivo [206,207]; the translation of these approaches to 3D culture models in vitro will enable analysis in human systems with tunable parameters.

Improving existing cell culture systems with emergent properties will also further their application in understanding the antiviral response. For instance, polarized intestinal epithelial models have difficulty replicating the crypt-villus topography, with proliferating cells resident in the crypts and differentiated cells occupying the villi [174]; thus, utilizing intestinal organoids have been heavily relied upon for mimicking intestinal processes including comparisons across viruses [208]. Recently, Drummond and colleagues were able to grow enteroid cultures from isolated primary intestinal crypts which will be a useful tool for assessing cell type-specific antiviral responses due to their ability to induce differentiation into all four classic cell types of the gut: enterocytes, Paneth, goblet, and enteroendocrine [209]. Indeed, using a similar enteroid model and through the application of single-cell RNAseq, Triana et. al demonstrated a pro-inflammatory response in SARS-CoV-2-infected cells while bystander cells upregulated ISGs in a cell type-specific manner [210].

‘On chip’ systems are often specially designed to address a specific biological question and are thus incredibly focused in their utility, often making them less ideal for broad investigations into the innate response (reviewed in [211]). Nonetheless, Villenave, Wales, and Hamkins-Indik et. al developed a gut-on-a-chip capable of supporting coxsackievirus B1 infection and apical cytokine release [212], establishing a foundational on-chip device for investigating innate antiviral immune responses. Finally, recapitulating virus-induced hemorrhaging, caused by Ebola and Lassa fever viruses among others, are difficult to model in vitro, and thus nonhuman primate models have been heavily relied upon in understanding pathogenesis, antiviral responses, and therapeutic development. However recently, an on-chip method was developed where upon use of a phase-guide, two parallel channels can be produced which are sensitive to infection and vascular leakage [213]. While these methods remain in their infancy, certain on-chip technologies have provided the basis for in vitro interrogation of host-responses against these destructive pathogens.

Since viral infections often trigger more severe outcomes in those with underlying conditions, it is important to note that researchers have also made advancements in establishing ALI and organoid models derived from primary cells sourced from diseased donors, as well as through the manipulation of cells to either induce or correct the diseased phenotype. Inclusive among these models are ulcerative colitis, asthma, chronic obstructive pulmonary disease, cystic fibrosis, and other fibrotic lung syndromes [154,214–216]. Importantly, this latter technology now enables virus challenge experiments and analysis of antiviral host responses across healthy and diseased states in an isogenic setting.

Still, despite these advancements, challenges remain in order to fully implement studies interrogating the antiviral response in tissue culture systems with emergent properties (Table 2). For example, unlike their 2D counterparts, 3D or tissue-engineered model systems remain difficult to genetically modify or utilize in high-throughput platforms. Advancements in cell culture techniques and the establishment of stem cell-derived models have removed some of these constraints by enabling more extensive passaging of primary cells or providing a renewable source of cells, respectively. As a result, researchers have a larger window for cell manipulation and selection, as well as greater numbers for downstream applications [217,218]. However, while cultures lacking specific PRR or ISGs expression would be a powerful tool to determine their impact during infection in a relevant setting, examples of 3D systems in which any host factor has been genetically altered remain scarce [219–222]. Cell culture models with emergent properties could also be further improved through the routine incorporation of additional cell types, such as endothelial and immune cells, and tissue-relevant mechanical forces (reviewed in [171]) to better capture microenvironmental influences on cellular responses and immune modulation. In sum, through greater utilization and continued improvement of these 3D culture systems, a heightened understanding of the initial antiviral immune response can be achieved, while potentially discovering novel mechanisms which can only be observed in these more relevant, in vivo-like model systems.

HIGHLIGHTS.

• The innate antiviral immune response is essential for controlling viral infection.

• Antiviral responses vary according to species, cell type, and microenvironment.

• 3D models allow interrogation of antiviral immunity in a physiological context.

• Further advancing 3D models will enable additional insight into the host response.

ABBREVIATIONS

ALI

air-liquid interface

AIM2

absent in melanoma 2

AP-1

activator protein 1

CARD

caspase activation and recruitment domain

cGAS

cyclic GMP-AMP synthase

CLR

C-type lectin receptor

CMV

cytomegalovirus

DAMP

damage-associated molecular pattern

DC

dendritic cell

DENV

Dengue virus

dsRNA

double-stranded RNA

ECM

extracellular matrix

HCV

hepatitis C virus

HIV

human immunodeficiency virus

hTERT

human telomerase reverse transcriptase

IAV

influenza virus

IFN

interferon

IFNAR2

interferon alpha and beta receptor subunit 2

IRF

interferon regulatory factor

ISG

interferon-stimulated gene

ISRE

interferon-stimulated response elements

JAK

Janus kinase

LGP2

laboratory of genetics and physiology 2

LPS

lipopolysaccharide

MAVS

mitochondrial antiviral signaling protein

MDA5

melanoma differentiation-associated protein

MMP-12

matrix metalloproteinase 12

NF-κB

nuclear factor kappa-light-chain-enhancer of activated B cells

NLR

nucleotide-binding oligomerization domain (NOD)-like receptor

PAMP

pathogen-associated molecular pattern

pDC

plasmacytoid dendritic cell

PRR

pattern recognition receptor

RIG-I

retinoic acid-inducible gene I protein

RLR

RIG-like receptor

RV

rhinovirus

SARS-CoV-2

severe acute respiratory syndrome coronavirus 2

SOCS

suppressor of cytokine signaling

STAT

signal transducer and activator of transcription

STING

stimulator of interferon genes

TIR

Toll-Interleukin-1 resistance

TLR

Toll-like receptor

ZIKV

Zika Virus