Abstract
Despite the need for effective treatments against chronic respiratory infections (often caused by pathogenic biofilms), only a few new antimicrobials have been introduced to the market in recent decades. Although different factors impede the successful advancement of antimicrobial candidates from the bench to the clinic, a major driver is the use of poorly predictive model systems in preclinical research. To bridge this translational gap, significant efforts have been made to develop physiologically relevant models capable of recapitulating the key aspects of the airway microenvironment that are known to influence infection dynamics and antimicrobial activity in vivo. In this review, we provide an overview of state-of-the-art cell culture platforms and ex vivo models that have been used to model chronic (biofilm-associated) airway infections, including air–liquid interfaces, three-dimensional cultures obtained with rotating-wall vessel bioreactors, lung-on-a-chips and ex vivo pig lungs. Our focus is on highlighting the advantages of these infection models over standard (abiotic) biofilm methods by describing studies that have benefited from these platforms to investigate chronic bacterial infections and explore novel antibiofilm strategies. Furthermore, we discuss the challenges that still need to be overcome to ensure the widespread application of in vivo-like infection models in antimicrobial drug development, suggesting possible directions for future research. Bearing in mind that no single model is able to faithfully capture the full complexity of the (infected) airways, we emphasise the importance of informed model selection in order to generate clinically relevant experimental data.
Shareable abstract
Using physiologically relevant models of chronic respiratory infections in preclinical research has the potential to improve the predictive value of in vitro data and accelerate the development of novel (and more effective) antimicrobial therapies https://bit.ly/3VIxNzs
Introduction
Chronic respiratory diseases (CRDs) are among the leading causes of death and disability worldwide, collectively accounting for up to 4 million deaths each year [1, 2]. Bacterial infections are a major contributor to lung function decline and reduction in life expectancy in people suffering from CRDs, such as COPD, cystic fibrosis (CF), non-CF bronchiectasis and primary ciliary dyskinesia (PCD) [3–5]. The establishment of difficult-to-treat chronic infections is a common occurrence in these individuals due to pre-existing lung damage and a compromised immune system [6]. Chronic colonisation by various respiratory pathogens is usually supported by the presence of excess mucus in the airways and/or impaired mucociliary clearance, which in turn depend on various disease-specific factors [7]. In particular, the ability of different pathogens to form biofilms under these conditions is recognised as the main driver of their survival and persistence in the lungs of people with CRDs. Microscopic examination of clinical samples has shown that biofilms of most respiratory pathogens consist of small clusters of bacteria embedded in a self-produced extracellular matrix that adhere to the airway epithelium (e.g., Haemophilus (H.) influenzae) and/or float in the mucus (e.g., Pseudomonas (P.) aeruginosa and Moraxella sp.) [8–11]. Regardless of their association with the respiratory mucosa, these bacterial aggregates constitute protected niches within the host that are highly tolerant to both antibiotic therapy and the immune system. Crucial factors in biofilm protection include the extracellular matrix, the altered chemical microenvironment and the presence of metabolically inactive bacterial subpopulations, which collectively promote antibiotic treatment failure and prevent the eradication of the infection (reviewed in [12, 13]). As a consequence of their chronic nature, biofilm-associated infections are typically accompanied by an excessive and sustained immune response, thus contributing significantly to progressive lung injury and exacerbation of underlying respiratory diseases [14]. Since eradication of established biofilms is often not possible, current therapeutic regimens mainly rely on the early aggressive treatment of lung infections to prevent biofilm formation and on the administration of high doses of antibiotics in the later stages to delay disease progression [15]. In light of the limited efficacy of these therapies, the development of novel treatment strategies against chronic lung infections is essential to ensure the effective and long-term management of CRDs.
Despite the urgent need for therapeutic options capable of controlling biofilm infections, only a few new antimicrobials have reached the market in recent decades. High attrition rates in the late stages of drug development underlie this lack of innovation in the antibacterial pipeline [16]. It has been estimated that up to 97% of antimicrobial drugs resulting from preclinical studies fail in clinical trials due to lack of efficacy or safety issues [17]. While different factors hamper the successful advancement of drug candidates from the bench to the clinic, a key reason is the use of poorly predictive model systems in drug discovery and preclinical testing [18, 19]. Most of the currently available in vitro biofilm models provide a simplistic representation of biofilm-related infections, commonly relying on the use of abiotic substrates (e.g., Calgary device) and standard (nutrient-rich) culture media for biofilm formation (reviewed in [20, 21]). Although useful for high-throughput testing of novel antibiofilm agents, these models neglect several aspects of the in vivo environment that are known to influence the infectious disease process and antimicrobial efficacy (reviewed in [22–24]). In this regard, clinical trials in people with CF (pwCF) have highlighted the lack of association between biofilm susceptibility data obtained with these models and in vivo outcome of the antibiotic treatment [25, 26]. With the aim of producing data of increased clinical relevance, significant progress has been made in the development of laboratory media and matrices able to recapitulate the local microenvironment found at the infection site, such as artificial sputum media [27]. While facilitating the formation of bacterial aggregates with structural and functional features similar to in vivo biofilms [28], these media do not recapitulate the host cellular component, which is essential to assess biofilm-induced immune responses and potential adverse effects of new treatments. The latter aspects have traditionally been studied using a wide range of mammalian and nonmammalian animal models (reviewed in [29]). However, existing animal models are rarely able to replicate the multiple aspects of the pathophysiology of human respiratory diseases due to interspecies differences in anatomy, immunity and tropism of microbial pathogens, thus proving inadequate to accurately predict the response to treatment [30, 31]. In this scenario, increasing efforts have been made to generate translationally relevant models capable of integrating key features of the microenvironment of (diseased) lungs and the biofilm phenotype of most pathogens observed in chronic infections. In this review, we provide an overview of advanced cell culture platforms and ex vivo models that have been implemented in chronic infection studies to date. We mainly focus on discussing recent advances in modelling biofilm infection using three-dimensional (3D) cell culture models, while refraining from including two-dimensional (2D) cell cultures due to their limited ability to capture essential features of the human airway epithelium [32]. In particular, our aim is to underline the advantages of these infection models over traditional in vitro biofilm methods as well as to indicate the limitations that still need to be overcome to improve their accuracy in mimicking chronic respiratory infections.
Mimicking chronic respiratory infections using in vitro models relevant for human disease
A significant number of in vivo-like models of increasing complexity have been developed in the last decade with the aim of increasing the predictive value of preclinical data and accelerate the identification of effective therapies for respiratory diseases [33]. Herein, our main focus is on describing state-of-the-art cell culture systems that have been optimised or specially designed for biofilm infection studies (table 1 and figure 1). While some of these platforms have already been used extensively to model biofilm infections (i.e., air–liquid interfaces (ALIs) and 3D cell cultures generated with rotating-wall vessel (RWV) bioreactors), others have started to be exploited for this purpose in recent years (i.e., AirGel and lung-on-a-chip). Although differing in their degree of refinement and throughput, these models recapitulate various structural and physiological characteristics of the native lung tissue. In addition, they are able to mimic the biofilm phenotype of the most common respiratory pathogens as well as crucial aspects of the in vivo host response to biofilm-associated infections (e.g., inflammatory response) [38, 60, 64]. As described in more detail in the next sections, the ever-growing use of these models has offered the opportunity to gain valuable insights into the mechanisms of biofilm development in the airways of people with CRDs and to assess the effectiveness of novel antibiofilm strategies in a more relevant setting compared to standard biofilm methods.
Overview of in vitro cell culture models of increasing complexity used for biofilm infection studies
Schematic representation of in vitro and ex vivo models used for biofilm infection studies. Microscopy images are reported for each model to show biofilm formation by Pseudomonas (P.) aeruginosa (used as representative example) in association with the airway epithelium or ex vivo lung tissue. a) Scanning electron microscopy (SEM) image of air–liquid interfaces of CFBE41o- cells with human tracheobronchial mucus (black arrow) and pre-formed biofilms of P. aeruginosa PAO1 (white arrows) after 1 h of co-incubation (scale bar: 3 μm) (reproduced from [53] under the terms of the CC BY licence). b) SEM images of P. aeruginosa PAO1 biofilms (white dotted frame) formed in association with three-dimensional A549 cells after 6 h of incubation (magnification: ×500; scale bar: 50 μm) (reproduced with modifications from [60] under the terms of the CC BY licence). c) Confocal laser scanning microscopy (CLSM) image of biofilm-like aggregates of green fluorescent protein labelled P. aeruginosa (ATCC 10145GFP) (green) on the epithelium of the cystic fibrosis (CF) lung chip at 24 h post-infection (scale bar: 10 μm) (reproduced with permission from [64]). d) CLSM image of P. aeruginosa PAO1 clusters (orange) formed in association with primary bronchial epithelial cells (pink) and cell-secreted mucus (green) at 13 h post-infection (scale bar: 200 μm) (reproduced from [65] under the terms of the CC BY licence). e) Visual observation of a pig lung explant after 4 days of incubation with a CF isolate of P. aeruginosa (reproduced with permission from [67]). SEM image of P. aeruginosa PA14 biofilms in the ex vivo pig lung model at 48 h post-infection. Image was processed with GIMP1.2 for false colouring (red: lung tissue; green: bacteria and biofilm matrix) (scale bar: 2 μm) (reproduced from [68] under the terms of the CC BY licence). Figure created with Biorender.com. PNEC: pulmonary neuroendocrine cell; RWV: rotating-wall vessel; SCFM: synthetic CF medium.
ALI cultures
Often considered the gold standard in respiratory research, ALIs represent a relatively simple culture model that consists in growing airway epithelial cells on the semi-permeable membrane of Transwell inserts. The distinctive feature of this system is that only the basal side of the cells is kept in contact with the culture medium, while their apical surface is exposed to atmospheric air [69]. In addition to simulating the extracellular environment of the respiratory tract, air exposure promotes differentiation of cells on the Transwell membranes, leading to the formation of polarised monolayers with morphological and functional features of the in vivo airway epithelium (e.g., barrier function, mucus production and ciliation) [69, 70]. Furthermore, the presence of apical and basal chambers in ALI cultures provides a convenient platform to establish co-cultures of airway epithelium with endothelial cells and/or immune cells, and thus more closely mimic physiological and immunological processes of the respiratory mucosa [71, 72]. Nevertheless, it is worth mentioning that the ability to reproduce specific characteristics of human airway observed in health and disease largely depends on the cell source selected to establish ALI cultures [73, 74]. For instance, primary airway epithelial cells obtained from clinical samples usually ensure a better representation of the native (diseased) tissue compared to immortalised cell lines, which often exhibit inadequate differentiation into certain cell types in ALI cultures [75–77].
A variety of human airway epithelial cells (primary cells and cell lines) of diverse origin have been used to generate ALIs for biofilm infection studies, particularly in the context of CF (table 1). ALIs of the CFBE41o- cell line (i.e., CF bronchial epithelial cells derived from a patient homozygous for the delta F508 mutation, the predominant mutation of the CF transmembrane conductance regulator gene in the CF population) represent one of the most widely used models to investigate chronic colonisation of the CF airway. Indeed, these cells have demonstrated the ability to reflect distinctive traits of the CF lung microenvironment (e.g., high iron levels) and to sustain biofilm formation by bacterial pathogens prevalent in the CF population, such as P. aeruginosa and Staphylococcus (S.) aureus [40, 57]. Although cytotoxicity has usually resulted from short-term exposure of ALIs to these pathogens, bacterial aggregates formed in association with airway epithelial cells have been reported to exhibit high levels of antibiotic-tolerance and gene-expression patterns compatible with mature biofilms [40, 56, 57]. Remarkably, this biofilm infection model has been recently found to outperform a mouse pneumonia model in replicating the gene expression profile of P. aeruginosa observed in the sputum of pwCF [78]. Nevertheless, despite resembling the in vivo phenotype of CF pathogens, this model has been shown to poorly replicate the viscosity of CF airways due to the limited secretion of mucus by polarised CFBE41o- cells [76, 79]. Since the presence of mucus is a key factor in the onset and treatment of chronic lung infections, this issue has been increasingly addressed by integrating primary cells from pwCF or exogenous mucus into the system [48, 52, 53]. In this regard, the addition of synthetic CF medium (SCFM) to ALIs of CFBE41o- cells improved the accuracy of the model in capturing the in vivo transcriptome of P. aeruginosa [48]. Analogously to CF ALI cultures, a refined infection model of the airways of PCD patients has been recently developed by using primary human nasal epithelial cells instead of cell lines, such as Calu-3 (i.e., human lung adenocarcinoma cells) and 16HBE14o- (i.e., immortalised healthy bronchial epithelial cells), which are known for inadequate ciliation [75]. Due to their ability to retain the ciliary dysfunction of the parental tissue in culture, ALIs of PCD patient primary cells allowed to better simulate the infection process of H. influenzae in PCD patients. In particular, infection of the PCD epithelium with H. influenzae resulted in a greater development of cell-associated biofilms compared to the non-PCD control [38].
Studies with ALI cultures have provided relevant insights into the pathogenic mechanisms underlying chronic respiratory infections, often leading to the identification of new potential targets for antibiofilm therapy. In this regard, ALI models of CRDs (especially CF) have proven to be useful tools to elucidate the impact of host cell factors on biofilm development. For instance, Moreau-Marquis et al. [40] were able to establish a causal relation between increased iron release from CF airway epithelial cells and enhanced P. aeruginosa biofilm formation by using the above-mentioned CF ALI cultures. The same model contributed to the discovery of CF modifier genes able to influence P. aeruginosa biofilm formation through the modulation of biofilm-promoting metabolites in the airways (e.g., L-arginine) [46]. Furthermore, the possibility of modelling polymicrobial infections in ALI cultures has revealed the influence of respiratory viruses on a host's susceptibility to chronic colonisation by different bacterial pathogens, such as Moraxella catarrhalis, P. aeruginosa and S. aureus [35, 45, 57]. Importantly, these mechanistic studies provided evidence for the role of host cells in mediating virus–bacteria interactions during co-infection. For example, excessive release of the host iron-binding protein transferrin in response to respiratory syncytial virus infection has been implicated in the transition of P. aeruginosa to a biofilm mode of growth [45, 80]. Analogously, rhinovirus-induced production of H2O2 by airway epithelial cells has been reported to trigger the dispersal of P. aeruginosa biofilms pre-formed on ALI cultures [49].
ALIs are undoubtedly the most widely used cell culture system for evaluating the effectiveness of novel antibiofilm agents. On the one hand, ALI infection models have been particularly useful for assessing the impact of in vivo environmental factors (e.g., host cells, mucus, pH, salt concentrations) on the activity, stability and bioavailability of a wide range of antimicrobials, thereby enabling a more realistic prediction of their therapeutic potential [40, 43, 44, 53]. On the other hand, the use of this model system has aided the development of antibiofilm strategies capable of overcoming the inhibitory effects of the host microenvironment, such as nanostructured delivery systems and combination therapies [39, 41, 42, 50–55, 58]. Nevertheless, it should be pointed out that most studies using ALIs for antimicrobial susceptibility testing have been conducted by maintaining the airway epithelium under submerged conditions during exposure to the tested antibiofilm agents, which is not particularly relevant for the evaluation of drugs intended for inhalation therapy. To address this issue, Horstmann et al. [55] recently developed an optimised model of P. aeruginosa infection, which involves the deposition of pre-formed biofilms on the surface of CFBE41o- cells that are cultured as ALIs and exposed to air for the entire duration of drug efficacy testing. This advanced infection model allowed to evaluate (over a period of 1–3 days) the efficacy of antibiofilm therapies involving the repeated administration of aerosolised drugs [53, 54]. Analogously, a proof-of-concept infection model has been established by bioprinting pre-formed biofilms of Escherichia coli (used as model microorganism) on ALIs of Calu-3 cells [34].
3D cell cultures generated with RWV bioreactors
Originally developed by NASA engineers, the RWV bioreactor is an advanced suspension culture technology based on the growth of cells on the surface of porous extracellular matrix-coated microcarrier beads under physiologically relevant low fluid-shear conditions. The main design feature of this system is a horizontally rotating vessel completely filled with culture medium, in which the constant rotation of the fluid prevents the sedimentation of cells [81]. These dynamic culture conditions enable cells to grow in three dimensions, self-assemble and form complex 3D structures that exhibit structural and functional features of in vivo human tissues. RWV-derived 3D cell culture models have been successfully developed for a broad range of anatomical sites (e.g., lung, small intestine, colon, liver and bladder), consistently outperforming conventional 2D cultures in mimicking parental tissues (reviewed in [82, 83]). In addition, the possibility of obtaining 3D cells in suspension makes this platform compatible with small-scale testing (e.g., 96-well plate format), which is convenient for high-throughput assessment of the efficacy and toxicity of drug candidates.
Optimised 3D cell culture models of human lung epithelium have currently been generated using both alveolar and bronchial epithelial cell lines (A549, CFBE41o-, CFBE-wt and 16HBE14o-). Culturing these cells in the RWV bioreactor has been reported to increase the expression of tissue-like features (e.g., cell polarity, tight junction formation and mucin secretion) and downregulate cancer-specific markers compared to their growth as monolayers [79, 84]. Notably, RWV-derived 3D cultures of bronchial epithelial cell lines have been shown to exhibit differentiation patterns comparable to standard ALI cultures, suggesting a possible application of this model system as a viable alternative to ALIs [79]. When exposed to P. aeruginosa, 3D A549 cell cultures were found to more closely replicate the in vivo host response to infection than corresponding monolayers, exhibiting lower susceptibility to bacterial invasion and increased production of pro-inflammatory mediators [83]. In addition, P. aeruginosa formed antibiotic-tolerant biofilm-like clusters on the surface of 3D A549 cells, with negligible effects on cell viability and epithelial barrier function over a prolonged period of incubation [60]. Given the key role of the host immune response to biofilms in the pathogenesis of chronic lung infections [85], this model has been optimised for use with NF-κB reporter A549 cells in order to enable the high-throughput evaluation of inflammatory processes under physiologically relevant conditions [62, 86]. Although not yet exploited for infection studies, co-cultures of 3D A549 cells and immune cells (i.e., monocyte-derived macrophages) have also been established to further increase the physiological relevance of the model [87].
The use of the 3D lung epithelial cell model for antimicrobial susceptibility testing of biofilms has provided relevant information on the influence of host cells and their metabolites on the efficacy of both traditional and nontraditional antibiotics. For instance, aminoglycoside antibiotics have been shown to exert enhanced activity against P. aeruginosa biofilms formed in association with 3D A549 cells as compared to standard abiotic biofilms [60]. On the other hand, the presence of 3D cells caused a reduction in the antibiofilm efficacy of both colistin and the antimicrobial peptide DJK-5 [60]. Similarly, loss of activity in the presence of 3D cells was observed for a combination treatment that successfully inhibited the formation of Burkholderia cenocepacia biofilms on a plastic surface [59]. Interestingly, this observation correlated with the lack of protective effect of such a combination in different animal models, thus emphasising the higher predictive value of cell culture-based infection models compared to conventional in vitro biofilm methods. In addition to ensuring a more realistic evaluation of the efficacy of different antimicrobials, the 3D cell culture model has also proved to be crucial for the identification of host factors capable of potentiating the antibiofilm activity of aminoglycoside antibiotics against P. aeruginosa, potentially contributing to the development of novel approaches to treat chronic lung infections caused by this pathogen [88].
Given the relevance of interspecies interactions for the in vivo infection process, the 3D cell culture model has been further refined to mimic the polymicrobial colonisation of the lungs that commonly occurs in people with CRDs. Dual-species biofilms of P. aeruginosa and the emerging CF pathogen Mycobacterium (Myc.) abscessus have successfully been established on 3-D A549 cells, allowing to evaluate the influence of their interaction on antibiotic activity [63]. Of note, the results obtained using this infection model aligned with clinical data, suggesting that antibiotic therapy directed at P. aeruginosa might provide a competitive advantage in lung colonisation to other opportunistic bacteria (such as Myc. abscessus) [63, 89]. In addition to giving insights into the impact of mixed infections on the outcome of antibiotic therapy, the exposure of 3D cell cultures to polymicrobial communities has allowed to elucidate the immunomodulatory properties of the lung microbiota. In particular, microbiome members belonging to the genus Rothia have recently been shown to inhibit the inflammatory response of 3D lung epithelial cells to P. aeruginosa [62]. A negative correlation between the abundance of Rothia species and pro-inflammatory markers was also observed in a cohort of patients with bronchiectasis, corroborating the translational potential of in vitro observations.
Lung-on-a-chip
Organ-on-a-chip devices are certainly one of the most advanced cell culture systems developed to date, commonly consisting of continuously perfused microchannels covered by living human cells. In addition to recapitulating organ-level structures (e.g., tissue–tissue interfaces), these microfluidic devices mimic relevant biochemical and mechanical aspects of the in vivo environment that are known to influence organ function, such as tissue-stretching, vascular perfusion and chemical gradients (reviewed in [90]). As for the respiratory system, lung-on-a-chip devices have been designed to simulate in vivo breathing mechanics, airflow shear and vascular flow rates (reviewed in [91]). Several chip-based devices have been developed to model different regions of the respiratory tract, especially alveoli and conducting airways [92–95]. Notwithstanding variations in design, most lung-on-a-chip devices are composed of two parallel channels separated by a porous membrane, which is lined with lung epithelial cells (on one side) and vascular endothelial cells (on the other side). While a continuous flow of culture medium is usually supplied to the endothelium-lined channel of the chip to mimic vascular perfusion, the epithelium-lined channel is exposed to air with the aim of recreating the ALI barrier of human lungs [92, 93]. Importantly, the presence of separate chambers subjected to independent flow conditions (i.e., airflow and liquid flow) allows to emulate the in vivo exposure of the airways to pathogens and drug treatments, and to evaluate host defence responses mediated by circulating immune cells [92].
In recent years, continuous adaptations of lung-on-a-chip devices have led to the development of model systems capable of replicating crucial physiopathological characteristics of different CRDs, including asthma, CF and COPD [64, 93, 96, 97]. Despite their ability to mimic clinically relevant aspects of the disease (e.g., inflammation) as well as in vivo-like responses to immunomodulatory therapies, chip-based models of COPD and asthmatic airways have not yet been exploited to model chronic lung infections [96, 97]. To the best of our knowledge, the CF lung chip recently developed by Plebani et al. [64] is the first microfluidic device to simulate bacterial colonisation of the airways. Such a model has been established by culturing primary bronchial epithelial cells from pwCF on chip at the interface with pulmonary microvascular endothelial cells. It accurately recapitulated several hallmarks of the CF airway, including abundant mucus accumulation and baseline secretion of pro-inflammatory cytokines. When infected with P. aeruginosa, CF chips exhibited higher susceptibility to colonisation than healthy chips, as demonstrated by the formation of large bacterial clusters in the mucus layer adjacent to the airway epithelium. The CF lung chip was also found to initiate in vivo-like responses to P. aeruginosa infection, including enhanced inflammation and increased recruitment of neutrophils to the airway compartment, which are highly relevant for the lung pathology of pwCF [98].
AirGels
A novel tissue-engineered model of the airway mucosa has recently been developed by Rossy et al. [65] with the aim of exploring the biophysical mechanisms of P. aeruginosa biofilm formation in a more realistic context than abiotic surfaces. Known under the name of AirGel, this model has been generated by culturing normal and CF bronchial epithelial cells on the luminal cavity of a tubular extracellular matrix scaffold under ALI conditions. A microfluidic chip has been integrated into the system to enable luminal access to the scaffold and noninvasive infection of epithelial cells (as represented in figure 1). In addition to reflecting the architecture and size of human small bronchi, AirGels have been shown to closely mimic the airway epithelium in terms of cellular composition, mucus secretion and mucociliary clearance function. Infection of AirGels with P. aeruginosa resulted in the formation of mucus-associated biofilm-like structures within hours of colonisation. In this regard, the compatibility of AirGels with high-resolution live microscopy allowed the authors to monitor infection dynamics at the single-cell level and to better define the contribution of airway mucus to the biofilm formation process. In particular, real-time visualisation of AirGel colonisation revealed that P. aeruginosa forms biofilms at the airway mucosal surface via a mechanism of active mucus remodelling involving type IV pili. Nevertheless, it should be noted that these mechanistic studies have mainly been conducted with AirGels containing bronchial epithelial cells of healthy donors. Therefore, although beyond the scope of the research, differences in mucus secretion and bacterial colonisation between normal and CF AirGels remain to be explored. Similarly to the aforementioned lung-on-a-chip device, the AirGel platform has not yet been used to evaluate antimicrobial compounds, but it is anticipated that the overall low throughput of both model systems could represent a limitation in view of their possible application in preclinical drug screening.
Mimicking chronic respiratory infections using ex vivo lung tissue models
Compared to the above-described cell culture platforms, a major advantage of ex vivo models is their ability to reflect the complex architecture of the lung both in terms of topography and richness of cell types (i.e., epithelial cells, endothelial cells, fibroblasts and tissue-resident immune cells) [33]. Among the existing ex vivo models for respiratory research (reviewed in [99, 100]), lung tissue explants and precision-cut tissue slices are undoubtedly the most common in infection studies (reviewed in [101, 102]). Despite being widely used to investigate human infectious diseases, these models have scarcely been optimised to simulate chronic lung infections. In this regard, an optimised ex vivo model of CF lung infection has recently been developed by using sections of porcine alveolar and bronchial tissue immersed in SCFM [67] (figure 1). Given the close similarity of porcine airway anatomy and immunity to human lungs [103], ex vivo pig lung (EVPL) tissue represents a valuable alternative to human explants, which commonly exhibit limited availability and considerable interpatient variability [100]. Studies aimed at validating and/or exploiting this infection model for the assessment of antibiofilm agents are summarised in table 2.
Overview of biofilm infection studies performed using the ex vivo pig lung model
When infected with P. aeruginosa and S. aureus, the EVPL model supported the formation of biofilm-like structures with in vivo-like morphology, with negligible effects on overall tissue integrity for an extended period of time (up to 7 days) [105, 110]. Based on transcriptome analysis, such a model closely replicated the gene expression profile of P. aeruginosa found during human CF infections, especially with regard to genes involved in antimicrobial resistance, regulation of virulence and metabolism [105, 106]. In this regard, Harrington et al. [106] demonstrated the ability of EVPL-associated biofilms to better capture aspects of the in vivo infection as compared to other common biofilm models (such as the SCFM alone), including downregulation of phenazine biosynthesis pathways and overexpression of genes involved in lipopolysaccharide modifications. In addition to morphological and functional similarities with in vivo biofilms, bacterial aggregates formed in association with the EVPL tissue exhibited high levels of tolerance to therapeutically relevant concentrations of antibiotics [107, 110]. When compared to standard laboratory media (i.e., Mueller–Hinton broth) and/or conventional in vitro biofilm models (i.e., Calgary device), a drastic reduction in the susceptibility of P. aeruginosa to different classes of antibiotics was observed in the EVPL model [107, 109]. Indeed, concentrations of antibiotics able to eradicate abiotic biofilms (even in the presence of artificial sputum) exerted a minor effect against EVPL-associated biofilms. Hence, analogously to previous research [59, 60], these comparative studies underline once again the limited predictive power of current antimicrobial susceptibility testing methods.
Strengths, limitations and challenges of in vivo-like biofilm infection models
Significant progress has been made in the development of in vitro models capable of integrating the biofilm phenotype of common respiratory pathogens and several features of the in vivo microenvironment, thus providing useful platforms to study chronic bacterial infections and/or explore new antimicrobial strategies against biofilms. However, several challenges remain that limit a more widespread application of these in vivo-like models in lung research. First of all, more attention should be given to model validation as a crucial (but still overlooked) step in the development of new model systems. Indeed, it is often unclear to what extent the above-described models are able to faithfully replicate relevant aspects of the in vivo physiology of both host and pathogen(s). In most studies, the characterisation of biofilms obtained in these models is largely based on the assessment of their morphological similarity with in vivo bacterial aggregates, while a functional evaluation of extracellular matrix production and/or antibiotic tolerance is frequently missing. Only a few models have been benchmarked with clinical samples to verify their likeness to the in vivo infection. Most of the work done in this regard has mainly relied on transcriptome analysis to compare the gene expression profile of P. aeruginosa in the model with that observed in CF sputum [48, 78, 106]. Interestingly, this approach allowed to demonstrate the accuracy of different infection models (i.e., SCFM, ALIs, EVPL and organoids) in reproducing key P. aeruginosa traits found during chronic lung infections, such as mechanisms of antibiotic resistance and metabolic adaptation [48, 106, 112]. Nevertheless, it can be anticipated that the combination of multi-omics approaches and advanced imaging techniques would provide a more complete picture of the degree of fidelity of each model to patient samples. When these infection models were compared, differences emerged in their ability to recapitulate the in vivo transcriptome of P. aeruginosa. For example, while better mimicking P. aeruginosa nucleotide metabolism than a CF ALI model, the SCFM less effectively captured fatty acid and phospholipid metabolism (and vice versa) [78]. This emphasises the importance of fully understanding which biological functions are (or are not) replicated in each model system to make a rational model selection. Indeed, since there is no model capable of reflecting the whole complexity of the infected airways, it is essential to choose the most appropriate model depending on the specific research question to be answered. As an integral part of the validation process, it is also critical to evaluate whether the data obtained using in vivo-like models could be meaningful for the clinical setting [19]. In this regard, it could be useful to compare the activity of known antibiotics in a given model with available clinical data from patients with chronic lung infections. In addition to frequently used reference strains, a representative set of clinical isolates should be employed for this purpose in order to cover the genetic diversity and resistance patterns of the pathogen(s) of interest [113, 114]. Although preliminary, this approach could help gain insights into the predictive power of newly developed models and/or the need for further improvement.
A variety of host factors have been incorporated into in vitro biofilm models to increase physiological relevance, but there are still important aspects of the in vivo microenvironment that have not yet been considered and/or fully explored. Except for the lung-on-a-chip [64], the infection models described above do not include cellular components of the immune system. Therefore, the evaluation of the host immune response to biofilms is mostly limited to the inflammatory processes that occur at the level of the airway epithelium. However, it is worth mentioning that some cell culture platforms (ALI cultures and RWV-derived 3D cells) have already been optimised to include immune cells (such as neutrophils and macrophages), but further research is needed to model biofilm-related infections in these systems [71, 87]. In addition to differences in reflecting host immunity, these cell culture models differ considerably in their ability to reproduce the various cell types of the respiratory epithelium (as depicted in figure 1). Such a difference in cellular composition has important implications for the possibility of mimicking functions of the respiratory mucosa that are relevant to infection, such as mucus secretion and mucociliary clearance. In this regard, ALI cultures of primary airway epithelial cells recapitulate the highest fraction of cell types, including both well-known and recently discovered epithelial cells (e.g., pulmonary ionocytes) [115]. Along with the host cellular component, another aspect to consider when developing in vivo-like models of lung infection is the polymicrobial nature of the in vivo airways. Although it is well known that the presence of complex microbial communities in the airways has the potential to influence the pathogenesis of chronic lung infections, there is still a limited number of studies that have modelled polymicrobial interactions under in vivo-like conditions [45, 62, 63]. Therefore, while the focus of current research has been on the development of single-species biofilm models of major pathogens, more attention should be paid to optimising the experimental conditions to enable multi-species infection of cell cultures and/or ex vivo lung tissue explants.
It is generally accepted that models with a higher level of complexity have a greater ability to predict the in vivo activity of antimicrobial candidates. However, it should be taken into account that increasing model complexity through the inclusion of a large number of (host and microbial) factors typically leads to a reduction in feasibility and throughput, which are desirable characteristics of model systems intended for preclinical studies. It is therefore essential to establish a compromise between the fidelity of a given model with respect to the in vivo environment and ease of implementation from a practical and economical point of view. In order to reach such a delicate balance, it could be useful to define a minimal number of biological parameters that are relevant to the infectious disease of interest and/or the specific scientific question to be addressed. As for the airways of people with CRDs, there is still little consensus on which components of the host microenvironment should be included in the model to adequately represent the physicochemical conditions that both pathogens and antimicrobials encounter in vivo. In this regard, the microenvironmental heterogeneity observed within and between patients with CRDs (especially CF) further complicates the design of relevant (yet relatively simple) model systems. As a result, a wide range of in vivo-like infection models are currently available that incorporate distinct microenvironmental factors and most likely result in variable susceptibility profiles of the tested pathogens to the antimicrobial treatment. In this scenario, along with a deeper knowledge of the in vivo microenvironment, it is necessary to understand which feature of the infected host have a central role in determining the (in)effectiveness of antimicrobial therapies. A systematic evaluation of the influence of different microenvironmental factors (alone and/or combined) on pathogen behaviour and antimicrobial activity could lead to the identification of the key contributors to the net antimicrobial susceptibility of respiratory pathogens in vivo [23]. Modelling and statistical approaches based on clinical data could also help establish which factors have greater impact on infection dynamics and/or treatment outcomes [116].
Despite providing physiologically relevant conditions for assessing novel antimicrobials, most of the model systems described in this review (especially microfluidic devices) have a more limited throughput compared to traditional (abiotic) biofilm models. Increasing efforts have been made to overcome this obstacle and facilitate the application of these platforms in drug efficacy testing. The most significant progress in this regard has been achieved through the development of miniaturised (96-well format) ALI cultures of both healthy and diseased lung tissue [117, 118]. In addition to miniaturisation, the integration of robotic devices for liquid handling has proven to be a successful strategy to fully automate several steps in the preparation and long-term maintenance of ALIs (e.g., cell seeding and media exchange), thereby increasing their potential as high-throughput screening platforms [117]. Advances in the development of computational tools for high-throughput data acquisition (e.g., automated imaging) and analysis (e.g., machine learning) are also critical to enable a broader use of these cell cultures systems for high-throughput assays [119]. In this regard, high-throughput microfluidic platforms (known under the name of PREDICT96-ALI and AKITA) have recently been developed that integrate ALI cultures of airway epithelial cells (established in 96- and 384-well plates), electrical sensors for real-time measurement of barrier function and/or live cell imaging systems [120, 121]. While these developments are promising, it is important to note that the majority of these (miniaturised) cell culture models have not yet applied to infection studies and/or antimicrobial activity testing. Therefore, further experimentation is needed to corroborate their ability to support the formation of in vivo-like biofilms and simulate physiopathological processes relevant to chronic lung infections.
While this review mainly examined biofilms formed at the interface with the respiratory mucosa, it is important to mention that biofilm formation also occurs on medical devices used for mechanical ventilation of critically ill patients (e.g., endotracheal and tracheostomy tubes) [122]. Biofilms associated with these devices significantly contribute to the onset of lower respiratory tract infections, such as ventilator-associated pneumonia (VAP), by acting as a reservoir for pathogenic microorganisms [123, 124]. The complexity of clinical conditions in mechanically ventilated patients represents a significant challenge for the development of robust in vitro models of airway infections, thereby leading to a reliance on animal models (especially mammals) to investigate their pathogenesis and potential treatment [125, 126]. Although not primarily intended for infection studies, there are some recent examples of in vitro models that have accurately replicated the effects of mechanical ventilation in ALI cultures and lung-on-a-chip devices [92, 126, 127]. Through the application of mechanical forces to cultured cells, these models were able to reproduce relevant aspects of ventilator-induced lung injury and pathology associated with acute respiratory distress syndrome [127, 128]. It would be of interest to further refine these platforms to incorporate aspects of infection, such as simulating airway device contamination and/or the development of VAP [123]. This type of in vitro model could potentially advance our understanding of the impact of artificial ventilation on respiratory infections and support the development of effective antimicrobial strategies for infection prevention and control.
Conclusions and perspectives
It is now clear that the in vivo microenvironment exerts a strong influence on antimicrobial activity and that mimicking the biofilm phenotype alone is not sufficient to predict the in vivo susceptibility of pathogens to antimicrobial therapy. In this regard, it is very likely that the extensive use of standard (abiotic) biofilm models in preclinical research has led to overlooking antimicrobial compounds that are actually effective in vivo (and vice versa). Reproducing relevant features of the host microenvironment observed during infection has the potential to significantly increase model predictivity and provide realistic insight into the activity of novel antibiofilm strategies. The implementation of in vivo-like model systems in antimicrobial discovery could also offer early toxicity read-outs, thus accelerating the identification of promising (i.e., effective and nontoxic) candidates. Although in vivo-like infection models offer numerous advantages over traditional in vitro biofilm methods, their selection and application to the study of chronic lung infections requires a thorough understanding of both their strengths and drawbacks. Indeed, while reflecting several traits of the infectious disease process in individuals with CRDs, these models still have limitations in terms of throughput and accessibility. Their successful implementation in preclinical studies will depend on a better understanding of which factors of the in vivo microenvironment need to be modelled to ensure biological relevance and feasibility of the experimental design. Advances towards model miniaturisation and automation will also contribute to increase throughput and reproducibility of these model systems. Although often neglected, rigorous validation is imperative at each stage of model development (and optimisation) in order to verify the degree of fidelity and actual predictive value of in vivo-like biofilm infection models. Finally, it is worth pointing out that no single model will be able to faithfully capture the full complexity of the human airways. In our opinion, it may not be necessary to strive towards achieving this goal to obtain physiologically relevant experimental outcomes. Instead, using (sets of) models capable of capturing specific features of the in vivo microenvironment (and hence of pathogen behaviour) that are meaningful for the specific experimental question is anticipated to strongly improve the clinical relevance of the generated findings. In this regard, a comprehensive knowledge of the aspects recapitulated (or not) by the available infection models is crucial to make a rational selection and/or undertake further optimisation.
Questions for future research
Which aspects of the in vivo infection process are reproduced in biofilm models? In order to understand the accuracy of a given model in simulating physiopathological processes relevant to chronic lung infections, more attention should be given to model validation and benchmarking with clinical samples.
Which factors of the in vivo microenvironment really matter for the treatment of biofilm infections? In order to select the microenvironmental factors to include in a given model, it is important to fully understand their impact on infection dynamics and antimicrobial activity in vivo.
Are the data obtained with in vivo-like biofilm infection models predictive of clinical outcomes? A systematic comparison of in vitro antimicrobial activity with clinical data has the potential to provide valuable insight into the predictive power of a given model.
What is the right model? In order to obtain clinically meaningful data, it is crucial to use model system(s) capable of recapitulating features of the in vivo microenvironment that are relevant to the specific scientific problem to be addressed.
How can model predictivity be further improved? It is anticipated that model predictivity can be enhanced by identifying and incorporating microenvironmental factors that influence biofilm behaviour and susceptibility to treatment (question for future research above) in model systems that are predictive for the specific research question to be answered.
Footnotes
Provenance: Submitted article, peer reviewed.
Conflict of interest: The authors declare no competing interests.
Support statement: L. Grassi was supported by the Research Foundation Flanders (FWO) (Junior Postdoctoral Fellowship, 12X6322N) and the Ghent University Special Research Fund (BOF.DPO.2019.023.01). Funding information for this article has been deposited with the Crossref Funder Registry.
- Received March 20, 2024.
- Accepted June 18, 2024.
- Copyright ©The authors 2024
This version is distributed under the terms of the Creative Commons Attribution Non-Commercial Licence 4.0. For commercial reproduction rights and permissions contact permissions{at}ersnet.org
References
Jump To
- Article
- Abstract
- Shareable abstract
- Introduction
- Mimicking chronic respiratory infections using in vitro models relevant for human disease
- Mimicking chronic respiratory infections using ex vivo lung tissue models
- Strengths, limitations and challenges of in vivo-like biofilm infection models
- Conclusions and perspectives
- Footnotes
- References
- Figures & Data
- Info & Metrics