Pathogens are typically detected by their pathogen-associated molecular patterns (PAMPs) (Alper et al., 2018), which, in turn, are recognized by pattern recognition receptors (PRRs) on cells in tissues. PRR activation induces an intracellular signaling cascade that triggers pro-inflammatory actions, such as cytokine production, to attract immune cells and activate blood vessel endothelial cells (BECs). The activated endothelium attracts peripheral immune cells, which transmigrate from the blood vessel to the infection site. This transmigration cascade is a multistep process (Figure 1), which includes the capture, rolling, arrest, crawling, and transmigration of immune cells at the level of post-capillary venules (Ley et al., 2007; Abbas et al., 2012; Sturtzel et al., 2017). After crossing the microvascular endothelium, immune cells migrate through the extracellular tissue space to the site of infection.

To initiate the process of capture and rolling, passively flowing immune cells must come into contact with the endothelium, which occurs when BECs are activated. BEC activation arises through cytokine signaling but also through altered blood flow. In normal laminar flow conditions, the BECs are tightly aligned, and the cells are in a quiescent state. However, when the flow is disturbed and oscillatory, BECs are less aligned because they sense the flow variations, leading to BEC activation (Heo et al., 2011). In both disturbed flow and cytokine activation, activated BECs upregulate adhesion molecules like E- and P-selectins and members of the IgCAM family such as ICAM-1 amd VCAM-1, the main molecules involved in immune cell capture and rolling (Ley et al., 2007; Abbas et al., 2012; Muller, 2013; Maas et al., 2018).

When the immune cells arrest after rolling, the cells flatten to decrease their exposure to shear stress and other circulating cells (Figure 1). (Maas et al., 2018) Chemoattractant exposure, as well as binding to integrins, is essential in regulating arrest with different signals inducing firmer adhesion of the immune cell to the endothelium. Firm adhesion to the endothelium allows the cells to crawl along the endothelium to find the optimal location for transmigration across the endothelial cell wall. The majority of cells follow the paracellular transmigration route, where the cell migrates between BECs, via the cell-to-cell junctions (Muller, 2011). A smaller percentage of immune cells opt for transcellular diapedesis, in which the immune cell crosses the barrier by going through a BEC.

After reaching the extracellular tissue space, the immune cells’ ability to migrate towards the site of infection is crucial. To accomplish this, the cell tracks various chemokines through the dense extracellular matrix (ECM) network. Tissue-resident macrophages and the epithelium produce chemokines to stimulate migration. To ensure that chemokines are correctly positioned, they are bound to the ECM, often to proteoglycans in the ECM (Gill et al., 2010; O’Dwyer et al., 2018). Immune cells respond to chemokines based on factors such as chemoattractant type, concentration, and exposure time (Amulic et al., 2012). Resident cells interact with the migrating immune cells and can also produce matrix-modifying enzymes, such as matrix metalloproteinases (MMPs) and collagenases, which specifically degrade the ECM, thus potentially facilitating movement of immune cells through the tissue (Parks et al., 2004; O’Dwyer et al., 2018; Miskolci et al., 2021). The combined action of MMP production, chemokine signaling, and cell-cell interaction facilitates immune cell migration towards the site of infection in the extracellular tissue space.

In vivo studies on innate immune cell extravasation across the vascular wall have primarily focused on easily accessible structures, such as muscle, skin, and mesentery (Halin et al., 2005). Research on these tissues benefits from better imaging quality than internal tissues that are challenging to reach. Within these tissues, immune cells mostly transmigrate from the postcapillary venule (Hickey and Westhorpe, 2013; Weirather and Frantz, 2015). The specific transmigration mechanisms used by immune cells vary considerably depending on the organ of interest, the immune cell subset, and the time during inflammation caused by infection. For example, in the skin, immune cells migrate from the postcapillary venule through a large extracellular space to reach the site of infection, whereas in the lung, the air-blood barrier is very narrow, resulting in the immune cells mostly transmigrating from the capillaries into the airspace (Maas et al., 2018). Thus, the molecular mechanisms by which immune cells transmigrate depend on their microenvironment, specifically the vascular characteristics of the respective tissue. Unfortunately, live imaging of extravasation in internal organ vessels can only be carried out after a complex surgical procedure or with expensive equipment such as a two-photon microscope to perform intravital imaging (Kim et al., 2019a). Surgery not only causes stress for the animals, which may activate immune cells, but anesthesia also influences vascular dilation and immune cell activation, making it crucial to identify alternative methods to study the mechanisms of inflammation and infection in vitro (Hickey and Westhorpe, 2013).

A microfluidic chip to model immune cell extravasation and migration during infection should be multicompartmental, to enable the creation of an endothelium-hydrogel-epithelium barrier, and perfusable, to perfuse immune cells and create a chemotactic gradient (Figure 2). Several commercialized microfluidic chips are available, such as from Mimetas, AimBiotech, and BEOnChip (Zhang and Radisic, 2017; Zommiti et al., 2022). However, a customized microfluidic chip can be developed if a specific design is desired. Figure 3 gives an overview of possible chip designs. Information concerning chip fabrication, typically using soft lithography, can be found elsewhere. (Cameron et al., 2022; Cho et al., 2022; Leung et al., 2022). Briefly, a mold is created by 3D printing or stereolithography and filled with polydimethylsiloxane (PDMS) mixed with a curing agent. Cured PDMS is post-processed and bonded to another PDMS part or a glass slide to close channel structures. The chip’s design is based on the abovementioned requirements: multi-compartmentalization with a hydrogel ECM and perfusability. Generally, designs can be categorized into two groups: chips with channels created inside the hydrogel structure and chips with a hydrogel barrier separating the microfluidic channels.

Hydrogels, as their name suggests, mainly consist of water, with a 3D polymer network forming the structure. The wide variety of available polymers with different properties enables the recapitulation of different ECM types (Liu et al., 2019). Importantly, a hydrogel’s mechanical properties can be tuned through various methods to be similar to those of human tissues. Hydrogel chemical composition and topographies can be adjusted to mimic the tissue of interest, and specific ECM components can be included if desired. An organ-specific hydrogel resembles the instructive native environment of the cells, to emulate a functional tissue environment.

The most commonly used hydrogels for on-chip culture are created from collagen, fibrinogen, Matrigel, gelatin, polyacrylamide (PA), polyethylene glycol (PEG), or hyaluronic acid (HA) (Caliari and Burdick, 2016). More recently, hydrogels comprising native tissue ECM have also been created to recapitulate the ECM of the organ of interest more accurately (Marhuenda et al., 2022). In most of the infection-on-chip models, hydrogels are the central environment between the epithelium and endothelium (Figure 2).

Culturing BECs on chip utilizes two main methods to model blood vessels: (1) generating a vessel-like three-dimensional structure and culturing monolayers of BECs on this defined structure (Figure 2) or (2) mixing BECs with mural cells inside a hydrogel to create self-assembled microvessels (Jeon et al., 2015; Zeinali et al., 2018; McMinn et al., 2019; Zeinali et al., 2021).

In barrier organs, such as the lung, intestine, and skin, the epithelium is the first physical barrier pathogens encounter upon entry. In the infection-on-chip model, the epithelial barrier is located next to the hydrogel ECM environment and acts as the site of infection (Figure 2).

In the lung, the alveolar epithelium comprises two main cell types: alveolar type I and II cells. The thin type I cells line the alveolar sac and allow gas exchange, covering around 95% of the alveolus (Weibel, 2015). Type II cells, on the other hand, are more stem cell-like, have regenerative potential, and produce surfactant. During infection, alveolar epithelial cells secrete cytokines to activate the endothelial barrier and attract immune cells from the bloodstream (Manicone, 2009). Throughout the pathogen removal process, the epithelium is severely damaged, leading to the accumulation of liquid and cell debris in the alveolar space (Yamada et al., 2016). Multiple lung-on-chip models with an epithelial barrier have been used to study lung infection (Deinhardt-Emmer et al., 2020; Deinhardt-Emmer et al., 2021; Bai et al., 2022).

The intestinal epithelial barrier consists of a multitude of cell types. During infection, the intestinal barrier is disrupted and immune cells can migrate into the underlying tissue. This was shown in a model of T Cell migration across the intestinal epithelial barrier (de Haan et al., 2021).

The skin has a complex, multilayered cellular barrier, that is often modelled in cell culture inserts (Kosten et al., 2016). To model skin toxicity from oral exposure to metals, a multi-organ on chip with gingiva and skin tissue was developed (Koning et al., 2022). In another model, T Cell migration upon skin inflammation was investigated (Ren et al., 2021).

For modeling the innate immune response on chip, macrophages, monocytes or neutrophils can be cultured inside the OOC (Figure 2; Table 1). The organ of interest plays a significant role in immune cell action, as immune cells are very responsive to the other cells in their environment. Table 1 gives an overview of immune cells used on chip, their source, and isolation method. Both the innate and adaptive immune response have been modeled on chip, or a mixed population of peripheral immune cells was utilized.

Even though the most physiological model would contain only freshly isolated human immune cells, the choice of cells also highly depends on the availability of specific cell types, the research question, and the organ of interest (Granton et al., 2018). Generally, combining multiple immune cell types to investigate their role during infection would be interesting, for example, by combining tissue-resident macrophages in the ECM hydrogel compartment with monocytes in the vascular compartment (Chen et al., 2023). In the infection-on-chip model, the main focus is the extravasation and migration of peripheral immune cells perfused through the blood vessel (Figure 2).

To attract innate immune cells, a chemical signal to direct them is necessary. This chapter focuses on solutions to generate a chemotactic gradient for immune cell migration. First, it should be mentioned that chemotaxis and haptotaxis are different processes that both take place in-vivo. Chemotaxis is defined as the directional migration of immune cells following a gradient of soluble chemokines in the environment, released from the site of infection (Petri and Sanz, 2018), while haptotaxis describes the migration of immune cells induced by a gradient of ECM-bound chemokines within the microenvironment. Here, the focus will be on chemotaxis, however, with ECM-based hydrogels, haptotaxis can also occur.

Many chemotactic agents are available, such as bacteria or viruses, factors secreted by other cells cultured in the system, or factors added to the cell culture media. For example, in an OOC, neutrophil extravasation can be induced by the presence of bacteria, pre-activated BECs, or a gradient of the bacterial peptide fMLP (Han et al., 2012; Hind et al., 2021). Here, we discuss forming a chemotactic gradient in a microfluidic device and some common chemotactic agents used for attracting immune cells.

To assess immune cell migration during infection, microscopy, specifically live cell imaging, is an essential tool. Accurate visualization of immune cell migration can be obtained with live cell imaging, where timelapse intervals and total imaging time are crucial to capture the process of interest (Figure 6A). However, there must be a balance between avoiding phototoxicity and increasing imaging frequency and length to capture all processes of interest.

Not only the microscopy itself but also the chip design influences the quality of the imaging and subsequent data analysis. For example, in the first lung on chip published by Huh et al. (Huh et al., 2010), neutrophils were migrating towards the site of infection, but migration dynamics could not be quantified, as it was occurring vertically across a porous membrane, and the immune cell migrated out of focus. Thus, to fully image the extravasation process, generating a model where immune cells migrate in a horizontal plane instead of vertically across a membrane is essential, as exemplified by multiple experimental designs (de Haan et al., 2021; Pérez-Rodríguez et al., 2022; Riddle et al., 2022). Live imaging of one focal plane can be carried out, and immune cell migration can be tracked.

After a timelapse video has been created, analysis of migration distance, speed, and direction is the next challenge (Figure 6). The authors refer to the following review for a more detailed overview of which parameters can be measured from a live imaging dataset of immune cell migration (Mattei et al., 2021). Traditional image analysis software such as ImageJ/FIJI or Imaris can aid in cell tracking, but these semi-automated solutions do not work for all imaged time-lapses. Thus, a tracking algorithm to track immune cell migration on chip can be developed in-house, for example, with Matlab (Pérez-Rodríguez et al., 2022). To simplify the analysis, an alternative to tracking individual immune cells is quantifying immune cells inside the hydrogel and their distance from the endothelial barrier over time (Figure 6B). With this method, no decisions on cell tracking are made, making the analysis less detailed but more robust. Next to imaging and tracking immune cells migrating horizontally in one focal plane, recent advances in spinning disk confocal microscopy have allowed for 3D timelapse imaging, to track cellular migration in all directions (Wen et al., 2021). This technique has not been applied in OOC models of infection yet, but would be of interest.

Advances in artificial intelligence (AI) have enhanced analysis methods, with the potential for more automation of (tracking) data analysis in the future (Ren et al., 2022). The first microfluidic experiments with deep learning algorithms to track cells have already been performed, such as an analysis pipeline to track bacteria in a microfluidic chip (Lugagne et al., 2020) and an algorithm to track tumor cell migration on chip (Zhang et al., 2018). In the second study, the migration direction and speed of the cells could be analyzed. So far, this technology is mainly used on relatively simple, single-cell OOC models and relies on manual adaptation during training (Ren et al., 2022). However, with more and more data becoming available, deep learning algorithms provide an immense opportunity for future analysis of immune cell migration on chip.

Current models of infection on chip have the common drawbacks of most OOC models, including only low to medium throughput, using the highly absorbing material PDMS for molding structures, and a lack of translation to clinical data. These issues must be addressed to engineer the next-generation of OOC models. To increase fabrication throughput and replace PDMS, OOCs have been fabricated with injection molding of polystyrene or similar transparent polymers, for example, to fabricate a high throughput OOC of cancer spheroid vascularization (Kim et al., 2022). Comparison of the data generated with OOCs to previously acquired data from animal studies and clinical trials is ongoing. For example, in the field of infection, many studies of immune cell extravasation have been carried out in mouse, rat, and rabbit models. With the appearance of new OOC models, it is important to investigate how these data can be compared and how both animal and OOC models translate to human diseases.

Additional complexity can be added to increase the relevance of OOC models of infection even further, for instance by mimicking both the innate and adaptive immune responses. This and other improvements may include replacing cell culture media with whole blood, adding tissue-resident immune cells, and integrating the lymphatic system.

Currently, only a subset of immune cells is perfused into a culture media or a buffer, but whole blood could be perfused to increase relevance. Not only does whole blood contain all cell types of interest, but blood also has a different viscosity and, therefore a different effect on the endothelium. Preliminary tests with whole blood perfusion in OOCs have been carried out and appear promising (Menon et al., 2017; Golomingi et al., 2022). Microfluidic devices can be modified to directly use whole blood for selective migration of neutrophils toward a site of infection. For example, whole blood was added in a microfluidic device with infected skin tissues, and neutrophil migration towards the infected skin tissues was observed (Kim et al., 2019b).

Another important part of immune surveillance is the tissue-resident immune cell population. These cells, mainly tissue-resident macrophages and T Cells, have a distinct gene expression profile and monitor the tissue microenvironment. Since these cells are present inside the tissue and their numbers are low, isolation is challenging, and source material is scarce. Further research is necessary to understand how these cells can be best isolated and cultured in vitro.

Lastly, a crucial part of the immune system, the lymphatics, is often overlooked. Generally, the main focus of infection-on-chip models is the interaction between immune cells circulating in blood vessels and the infected tissue, but for activation of the adaptive immune response, migration of dendritic cells from the site of infection to the lymph nodes to activate antigen-specific T Cells is necessary. Individual systems modeling lymph nodes or vessels on chip have been developed (Selahi et al., 2021; Shanti et al., 2021), however, to our knowledge, lymph vessels have not been incorporated into an OOC model of vasculature or other organs.

Next to infection, immune cells are also involved in other conditions, such as cancer, which adds the opportunity to utilize immunocompetent OOC models not only to model infection but also to model cancer immunotherapy. Using the human body’s natural defense system, cancer immunotherapy targets tumor tissue with the immune system, especially T Cells. Multiple studies are emerging with microfluidic models of immune cell migration towards a tumor, with known and unknown immunotherapy drugs enhancing tumor suppression by T Cells (Pavesi et al., 2017; Deng et al., 2018).

This review has discussed the generation of an infection-on-chip model to investigate infectious processes in a 3D in vitro model. By incorporating various cell types, cultured on or in hydrogel ECM environments, and perfusing the endothelial vessels, OOC models provide a more sophisticated alternative to basic 2D in vitro cultures. The modularity and versatility of OOC models enable the study of (disease) mechanisms in the presence or absence of particular aspects of the system. For example, research has demonstrated that the lack of endothelial cells prevents the migration of immune cells into a hydrogel, whereas the presence of endothelial cells significantly increases the number of immune cells that migrate (Wu et al., 2015; McMinn et al., 2019). It is impossible to study the impact of endothelium on immune cell migration in vivo as it is not feasible to completely eliminate endothelial cells from an animal. Consequently, OOC models offer an opportunity to supplement the findings from animal studies.

However, the challenge remains in translating the data obtained from these complex models to human disease. Initially, OOC models aim to replicate the disease symptoms observed in the clinic, such as drug-induced pulmonary oedema (Huh et al., 2012). Moreover, during the COVID-19 pandemic, a list of antivirals were tested in a lung-on-chip, showing that hydroxychloroquine, a drug that demonstrates efficacy against SARS-CoV2 in cell lines, did not exhibit antiviral effect on chip (Si et al., 2021). This finding translated to the clinic, where this drug proved ineffective against SARS-CoV2. Thus, to establish a correlation between research findings on infection-on-chip models and clinical data, clinical observations and readouts must be shown on chip.