HAEC are obtained from human respiratory tissue, which can originate from different anatomical sites of the RT (41). Respiratory tissue is now obtained in different ways: during lung transplantation, during tissue resection in cancer patients, during other surgical procedures (e.g. turbinoplasty/-ectomy or nasal polypectomy) or from cadaveric explants (41). In addition, nasal and bronchial brushings, which are less invasive, can be performed to obtain HAEC (42). HAEC from healthy donors are now commercially available (43). Additionally, material from, for instance, COPD patients and/or smokers is also available (44, 45). Acquiring HAEC from various donors allows us to study and compare the RT of both healthy and diseased individuals.

Primary undifferentiated HAEC (HAEC_un) are relatively easy to culture, but can only be passaged a few times. After obtaining tissue, the primary airway epithelial cells are directly isolated and cultured. For the isolation of primary airway epithelial cells from lung transplants or biopsies, the tissue is cut into smaller pieces and then dissociated via the addition of a protease-containing digestion cocktail, followed by generation of uniform single-cell suspensions (46, 47). Tissue obtained from brushings can be cultured directly (47). After isolation, HAEC can be used in this undifferentiated form for experiments. In addition to normal culture flasks or plates, HAEC can also be seeded on a collagen-coated semi-permeable membrane (transwell) (48). In this transwell system, medium is present on both the apical and the basolateral side. When a 100% confluent monolayer has formed to separate the apical and basolateral compartments, the HAEC_un can be used for experiments. These HAEC_un are not polarized and these cultures do not have ciliated cells or goblet cells, and therefore lack important characteristics of the airways.

Although more challenging to culture, differentiated HAEC (HAEC_dif) represent the RT more faithfully than their undifferentiated counterparts. To differentiate primary respiratory epithelial cells in the transwell system, medium is removed from the apical compartment after the confluent monolayer has formed, creating an air-liquid interface (ALI). This, in combination with specific growth factors, induces differentiation of these cells over a timeframe of 3-4 weeks (48). Eventually, a polarized, pseudostratified respiratory epithelium is formed, containing basal cells, ciliated cells, and goblet cells or club cells (depending on the anatomical location) (48, 49). These HAEC_dif resemble the human RT anatomically (50, 51) but a continuous airflow, blood flow, and the presence of immune cells are still lacking.

Comparing studies using primary HAEC is difficult, because of differences in donor variability (1) , anatomic source of the cells (2) , culture methods (e.g. medium and growth factors) (3) and the use of undifferentiated versus differentiated cells (4) ( Figure 3 ).

HAEC_un can be passaged approximately three times prior to differentiation, after which loss of epithelial integrity, differential gene expression, or senescence occurs (70–72). This restricts the number of experiments that can be performed. That is why immortalized variants of HAEC have been created. Jonsdottir and colleagues (2019) described that insertion of the Rho-associated protein kinase (ROCK) inhibitor Y-27632 via a lentiviral vector increased the longevity of the primary respiratory epithelial cells (73). Alternatively, exogenous induction of human telomerase reverse transcriptase (hTERT) can be used to prolong primary cell life (74). Immortalization resolves the issue of the restricted number of experiments, but it remains unknown if all the characteristics of the respiratory epithelial cells remain the same. And, even in these immortalized variants, senescence can occur after a few passages (75, 76). Genetic engineering of HAEC is challenging and may compromise translational aspects of the model.

In conclusion, primary HAEC can be extracted from all parts of the airways and differentiated into a pseudostratified monolayer resembling the in vivo situation. Although several variables and parameters must be considered, primary HAEC are a promising culture system to study RTIs.

SC-based models have the potential to overcome the problems associated with the limited lifespan of primary HAEC. SCs can be used to make organoids, which are three dimensional cultures that can self-organize and renew (77). Organoid culture is a novel and innovative technology and was first described in 2009 with organoids from the gut (78). In 2012, the technique to culture AO was developed (79). Until now, human AO are mostly made from SCs obtained from adult lungs. Adults tissues are more accessible than embryonic SCs or induced pluripotent (iPSCs) and have less ethical restraints. Nevertheless, AO from embryonic SCs are also used (80, 81). Although most studies describe AO obtained from the lung, AO can be made from almost all parts of the RT, and are thus a suitable alternative to primary HAEC (82). Additionally, genetic modifications are also more feasible in AO than in primary HAEC.

Different SCs can be used to generate AO. First, embryonic SCs obtained from the inner cell mass of an early staged embryo have the potential to form every cell of the human body (83). However, the use of these cells remains ethically controversial. Second, iPSCs can also differentiate into almost every cell type of the human body (84). To obtain these iPSCs, somatic cells (for example skin cells) are reprogrammed to the progenitor state by addition of several transcription factors. However, specific factors, including growth factors and cytokines, are needed for differentiation into respiratory epithelial cells (85), which has so far not been successful. The challenges to find the optimal factors to differentiate these iPSCs into airway epithelial cells can be overcome by using organ-specific adult SCs. These cells can differentiate into the cell types of the respective organ (86). In this case, airway epithelial SCs can be isolated from tissue or URT brushings (87). The isolated basal cells can subsequently be used for the formation of 3D undifferentiated AO, using Matrigel and medium with specific growth factors. These AO, cultured either in 3D in matrigel or on transwell filters at ALI, can then be differentiated into cultures that represent the cells of the RT, such as basal cells, goblet cells, ciliated cells or alveolar cells (81, 87–92) ( Figure 4A ). In a differentiated spheroid AO, the basal cells are on the outside of the sphere, the goblet cells excrete their mucus into the lumen while the cilia move the mucus around (see Figure 4B ). Similar to primary cell cultures, there are indications that culture medium influences organoid morphology and behavior, which is something that has to be studied in more detail (92, 93).

AO appear to be a promising model mimicking the human RT and once established can be maintained for a long period of time. Limited studies have been performed with AO and respiratory viruses so far due to the novelty of the technique. However, the studies that have been performed show great promise for studying virus-host interaction and drug screening studies in this model.

One of the mechanisms by which viruses enter a host cell is receptor-mediated entry; the appropriate receptor needs to be present on the cell for the virus to enter. Immortalized cell lines are often used to identify such a cellular receptor, for example by genomic- and interactome- based approaches (141, 142). Primary HAEC mimic the natural targets cells of respiratory viruses better than continuous cell lines and are thus important for receptor identification, as illustrated in the following examples.

For most HCoVs, receptors were initially identified in cell lines, followed by confirmation and verification of these receptors in primary HAEC (58, 66, 91, 105–107, 143–148). An example of the importance of using primary HAEC was seen in the search for a model for HCoV-HKU1. This virus only replicates in primary differentiated tracheobronchial epithelial cells. Through studies in these cells, the HCoV-HKU1 receptor (O-acetylated sialic acid) was ultimately identified (108, 149). Differential binding of avian and human influenza viruses to their receptors was investigated in primary HAEC_dif (bronchial), which express both α-2,3- and α-2,6-linked oligosaccharides. Infections with avian and human influenza viruses confirmed that viruses with avian origin preferably bind to α-2,3-linked oligosaccharides, whereas viruses from human origin bind to α-2,6-linked oligosaccharides (133). For most HRV subtypes their receptor was identified in immortalized cell lines and validated in HAEC, except for HRV-C (96–99, 150–153). For HRV-C the receptor was found in primary differentiated bronchial epithelial cells and in ex vivo organ transplants, since this HRV subtype does not replicate in continuous cell lines (101, 154).

Many HRSV receptor candidates have been proposed in literature, but arguably the relevant in vivo receptor has not yet been identified. Ciliated epithelial cells are the main target cell of HRSV (155), therefore primary HAEC_dif are the most suitable model to perform receptor studies. CXC3R1, insulin-like growth factor-1 (IGFR1) and nucleolin have been identified as potential cellular receptors for HRSV. CXC3R1 would probably not have been found if it were not for primary HAEC, since this receptor is absent (unless transfected) on immortalized cell lines (119, 120). The other two receptors have been confirmed and validated in primary HAEC (121, 156). Although closely related to HRSV, the proposed receptors for HMPV are not the same as for HRSV, but studies with primary respiratory epithelial cells from different locations in the RT have proposed integrins and heparin sulfate as receptors (126, 127, 157, 158).

Lastly, using primary HAEC_dif (tracheobronchial), it was confirmed that HPIV-3 uses α-2,6-linked sialic acid residues and HPIV-1 uses α-2,3-linked sialic acid residues as cellular receptors (133, 134). Primary HAEC have also been used to identify or verify the receptors for HAdV (138, 159).

All the above-mentioned examples clearly illustrate that HAEC are important for identifying the cellular entry receptor for respiratory viruses. Although initially the receptor is often found in cell lines, primary cell models are essential in confirming and validating this receptor and these results translate better to the in vivo situation.

Regarding viral tropism, in vivo models and ex vivo culture models provide valuable information. However, it has to be kept in mind that in in vivo studies the animal is often not the natural host of the virus, introducing a possible bias. Human post-mortem material from either healthy or infected individuals is also useful for binding studies, but this material is not always available. Cell lines are immortalized and often express unique gene expression patterns that are not representative of in vivo cell types. Primary airway models can overcome these hurdles and are a thus powerful tool to investigate viral tropism, since most cell types from the RT are represented in these models.

In the SARS-CoV-2 pandemic, differentiated AO (AO_dif) and primary HAEC rapidly helped elucidate that SARS-CoV-2 mainly infects ciliated cells, club cells and ATII, but not goblet cells (88–91, 109, 110). For other coronaviruses, similar models were used to determine their respective tropism, for instance HCoV-229E and MERS-CoV infect non-ciliated cells and HCoV-HKU1 predominantly infects ATII (105, 107, 160, 161). Multiple studies have been performed with influenza viruses in HAEC or AO. It was found that, for example, seasonal H1N1 was able to infect ciliated cells, goblet cells and alveolar cells, whereas pandemic H1N1 was also able to infect club cells (89, 93, 115). HRV is a common cause of respiratory infections in humans and both HRV-A and HRV-B can be cultured in conventional cell models, while this is not the case for HRV-C. However, by using 3D Matrigel cultures with primary HAEC_dif HRV-C can be propagated (159). For all three HRV subtypes the tropism could be determined in primary HAEC_dif (bronchial): HRV-A and HRV-B infect basal cells and HRV-C infects ciliated epithelial cells (101, 162). It was also confirmed that HRV can replicate in cells from both the URT (adenoids) and LRT (bronchus) (163). Clinical characteristics of HRV are also recapitulated in primary airway models. For instance, HRV-B infections are associated with less severe infections and studies in primary HAEC_dif (nasal and bronchial) reported that HRV-B replicates slower and leads to less cytotoxicity than HRV-A and HRV-C (164, 165). For both HRSV and HMPV it was confirmed in HAEC_dif and AO_dif mainly targets the ciliated epithelial cells (122, 123, 128, 166–168). One study showed that basal cells could also be infected by HRSV after epithelial injury, using commercially obtained primary HAEC_dif (bronchial) (124).

In conclusion, with primary airway cultures or SC-based cultures it is possible to culture viruses that are difficult to culture in immortalized cell lines and these models can provide reliable information on the viral tropism.

Studying pathogenesis is challenging in vitro, both in continuous and primary cell lines. A single or limited number of cell types are present in in vitro cultures, not resembling a full organism. As alternative, cytopathic effect (CPE) can be studied. CPE in primary HAEC_dif often resembles the natural situation more closely compared to continuous cells. HRSV offers a good example: in continuous cell lines it forms big syncytia (hence the name), but in primary HAEC HRSV forms little to no syncytia, which is in line with observations in vivo (both animal models and humans) (48, 122, 129, 168–170). A clinical observation in HRSV disease is neutrophil inflammation, which is involved in severe HRSV disease. This clinical phenomenon has been mimicked in HRSV-infected primary HAEC_dif (nasal) that were co-cultured with neutrophils (171). AO have significantly improved disease modelling opportunities. HRSV, HMPV and HPIV have been investigated in AO_dif: and for example for HRSV epithelial cell sloughing was shown, which was absent in HPIV infections. These observations both fit with clinical observations. Other hallmarks of HRSV and HMPV disease, such as infection of ciliated epithelial cells and mucus production were also recapitulated in AO and in primary HAEC (80, 87, 125, 155, 172). These hallmarks are impossible to mimic in 2D cell lines and thus highlight the usefulness of 3D AO to study respiratory virus pathogenesis.

When a virus enters a cell, this cell will directly mount an innate immune response, both in the infected cell and in surrounding cells. This usually includes the production of several cytokines and chemokines to attract more immune cells and activate the adaptive immune system. Studying the innate immune system in primary HAEC has as an advantage over continuous cell lines, because they potentially produce more clinically relevant cytokines that would also be produced by the natural target cells. For instance, for SARS-CoV it was described that ATII are less susceptible than ATI, which was linked to a robust innate immune response (161). This pronounced innate immune response was not found in primary HAEC_dif (bronchial, chemically differentiated) after MERS-CoV, SARS-CoV or HCoV-229E infection, indicating immune evasion (173). It appears that each corona-virus induces specific innate immune responses, with few cytokines and ISGs overlapping (90, 91, 110, 112–114). In response to influenza virus infection, similar cytokines, mainly IL-1β, IL-6 and IL-8, were produced in HAEC_dif from different parts of the RT (114, 116–118). For HRSV and HMPV infections it has also been shown that similar cytokines, predominantly type I and/or type III IFN, are produced in HAEC_dif (nasal, bronchial and small airway) (123, 126–132, 168).

Different subtypes of HPIV are associated with certain clinical observations: HPIV1 can stay undetected for days, whereas HPIV2 is associated with mild disease and HPIV3 with more severe respiratory disease. These clinical observations have been linked to cytokine profiles that were found in primary HAEC_dif (tracheobronchial). It was shown that for HPIV1 there was no early innate immune response, whereas for HPIV2 innate cytokines were produced at early time points and for HPIV3 cytokines increased overtime (61).

It was observed that HRV induced an innate immune response in 3D matrigel cultures with primary HAEC_dif that inhibited viral replication, but that this was only short-lived because a second peak of viral replication was measured. This correlated with the production of the cytokines TNF-α, IL-6, Il-8, and IP-10. Based on these results, the authors postulated that the second peak appeared due to the production of new virus, after the eclipse caused by the innate immune system (102). Similar replication kinetics and cytokines were seen in other studies, including type I and type III IFN (103, 104, 174, 175). In one study it was shown that upon HRV infection IL-17C was produced in the basolateral compartment and induced CXLC1, which is a neutrophil attractant and may contribute to exacerbations of lower airway disease (103).

To summarize, many studies underline the importance of using primary HAEC and AO when studying innate immune responses. Although these models seem to be a good proxy for clinically relevant cytokines in vivo, studies directly comparing the innate cytokine responses between continuous cell lines and primary cell models are lacking. Nevertheless, complex cell systems that mimic the in vivo target cells of viruses probably model the antiviral innate immune response better than cell lines.

Primary HAEC can mimic host factors related to lifestyle, such as smoking history and obesity. Smoking history might predispose to more severe SARS-CoV-2 disease, possibly to the upregulation of the ACE2 receptor, which was found in formalin-fixed paraffin-embedded human lung tissues and in brushings of the airway epithelium (176–178). This difference was not observed in primary HAEC_dif (bronchial), but the number of donors was limited (179). Obesity is related to more severe influenza symptoms, which has been shown in primary ATII, where cells from obese donors were more susceptible to a pandemic IAV than non-obese donors (52).

Primary HAEC are also useful for studying stability of viruses under environmental conditions. For example, primary HAEC_dif (bronchial) were used to study the impact of relative humidity on the stability of a pandemic influenza A virus in aerosols and droplets. Aerosols and droplets were created and supplemented with extracellular matrix material harvested from the apical surface of the HAEC and then the amount of virus was determined in MDKC cells (180). None of the experimental humidity conditions affected the stability of IAV H1N1 when aerosols and droplets were supplemented with extracellular matrix, but without supplementation, a humidity-dependent decay of virus infectivity was observed (180). This provided evidence that extracellular matrix can protect influenza viruses from decay and thereby promote spreading (181, 182).

Primary HAEC_dif have also been proposed as a screening model for evaluating and monitoring the infectivity, pathogenicity and antigenicity of virus variants, for instance during an outbreak. In the current SARS-CoV-2 pandemic many new, and supposedly more infectious, variants are arising. Information on these variants is important, so that interventions can be tailored. Using primary HAEC, it was found that SARS-CoV-2 variant (D614G) replicates better than the original wild-type strain in the nasal and proximal airways, which was not observed in VeroE6 cells. Increased replication in the URT might lead to more transmission (183). During influenza seasons of 2013-2014 and 2015-2016 the vaccine effectiveness against the circulating influenza strains was reduced. The reason for this was a reduction in sustained multi-cycle replication. This was found in primary HAEC_dif (nasal), but not in MDCK cells (182). These important data regarding novel virus variants would not have been available without primary airway models and are important for managing pandemics and developing interventions.

In this section, we illustrated that primary cell models may serve as an authentic model for in vitro respiratory viral pathogenesis studies, recapitulating viral infection in the host. They are useful for identifying viral receptors, determining viral tropism, mimicking disease and assessing innate immune responses in respiratory epithelial cells. Additionally, some viruses that are difficult to culture can be propagated in primary HAEC and often not in cell lines (184). Although continuous cell lines can give seminal insights into virus-host interactions, they also render selective pressure on viruses, leading to culture adaptations. One of the advantages of continuous cell lines is the possibility for genetic modification, which is thus far impossible in primary respiratory cell cultures. This hurdle can be overcome in the future by using AO derived from human SCs, which could be engineered to express desired host features.

Endothelial blood vessels are connected to the airway epithelial cells in vivo and therefore important to also include in an in vitro airway model. Co-cultures with endothelium have been performed, with primary HAEC placed on a chip and microvascular endothelial cells on the opposite side of the porous membrane, with a fluid flow underneath (185). This small airway-on-a-chip model recapitulates tissue-tissue interactions, physicochemical microenvironments, and vascular perfusion of the RT (185, 186). Unfortunately, these models are costly and often not compatible with a biosafety level (BSL)-II or BSL-III laboratory environment required for performing pathogenic virus infections. A way to mimic air flow has been addressed by applying mechanical forces to the airway-on-a-chip model, recreating a breathing movement (187). In another study, primary HAEC were cultured at the apical side of transwell filters and primary microvascular endothelial cells at the basolateral side to create an alveolar-capillary system. This system was used to study influenza virus and staphylococcus aureus (SA) co-infections (188). Thus, although progress is being made, the use of transwell filters and airway-on-a-chip models to model the epithelial-endothelial barrier to study respiratory virus pathogenesis is still limited.

Next to co-cultures with endothelial cells, co-culture models with bacteria and/or viruses to mimic co-infections or the microbiome are being developed. One option is an infection with a bacterium preceding viral infection. In HAEC_dif (alveolar), co-cultured with primary microvascular endothelial cells, it was found that methicillin-resistant SA (MRSA) dysregulated the host immune response and decreased the barrier function (188). Another option is a viral infection preceding an infection with a bacterium, for instance initial IAV infection resulted in increased replication of S. pneumonia in primary HAEC_dif (bronchial). In contrast, in primary HAEC it was then found that pre-exposure to HRV reduced the viral titers of subsequent influenza virus exposure, which was supported by clinical data (189). Co-culturing cells with bacteria poses a challenge, since these micro-organisms replicate fast and rapidly overgrow the respiratory epithelial cells, often leading to cell death. In some of these studies, bacteria were inactivated to overcome this problem, but it is questionable if this still mimics the real-life situation since several features of the bacteria are lost (190). However, it has been shown that inactivated bacteria can influence respiratory virus infections. For instance, UV-inactivated nontypable Heamophilus influenzae (NTHI) enhanced the susceptibility of human bronchial epithelial cells to HPIV, likely due to upregulation of ICAM-1, a cellular receptor for HPIV (191, 192). Another way around the toxicity problem could be only co-culturing bacteria for a short time. Obviously, this does not mimic the microbiome or a co-infection (193–196). Nonetheless, this still allowed investigation of bacterial adherence to the (infected) primary HAEC or transmigration of bacteria. Alternatively, co-cultures could be regularly washed. Using this method researchers were able to culture NTHI with commercially available primary HAEC for 10 days (197). Combining co-cultures of bacteria and AO have, to our knowledge, not been described yet, but lessons can be learned from co-cultures of bacteria and intestinal organoids or skin organoids (197, 198).

It is important to study the effect of bacteria and viruses on the innate immune state of the HAEC. Certain innate immune profiles influence the outcome of viral infections, for instance for HRSV. It was described, in humans, that neutrophilic inflammation at the time of viral challenge predisposed individuals to symptomatic HRSV infection (199). These data suggest that co-cultures with, for instance, neutrophils might be required to fully mimic the microenvironment of the RT. In conclusion, although there is only limited data available on this topic, more studies are being performed and we are confident it is feasible to create an in vitro airway model combined with the microbiome.

Another important feature of the airways is the (intra-epithelial) presence of immune cells that are often involved in the defense against respiratory virus infections. Co-cultures of neutrophils with primary HAEC_un (trachea) that were infected with HPIV were already performed in the 1990s (200, 201). These studies showed enhanced adhesion of neutrophils to HPIV-infected cells. This has later been shown in AO infected with HRSV, where it was even possible to visualize the preferential neutrophil movement to the infected AO. Co-cultures of primary HAEC (nasal or bronchial) with DCs or T cells have also been described, either with virus present prior to adding the cells, or after. In these studies, researchers investigated both the effect on the immune responses as well as viral replication (87, 202, 203). They showed that the presence of DCs or T cells enhanced antiviral and inflammatory responses and inhibited viral replication. A triple co-culture has also been described, in which primary HAEC_dif (bronchial) were co-cultured with monocyte-derived DCs (moDCs) and macrophages (204). Here, the moDCs were infected with HRSV and added apically to a layer of HAEC in a transwell system. Uninfected moDCs were simultaneously added to the basolateral membrane of the insert. This led to the transmission of HRSV to the HAEC, but not to the moDCs located basolaterally. However, when macrophages were added to the apical surface of this co-culture, the basolateral moDCs were infected too, indicating some type of trans-epithelial transport mechanism. This study shows the importance of having multiple factors in a culture to understand the infection process (205). All in all, combining primary HAEC_dif with endothelial cells, airflow, the microbiome and immune cells would be an ideal model to mimic our RT.