From the nose to the lungs: the intricate journey of airborne pathogens amid commensal bacteria.

The recent COVID-19 pandemic has dramatically brought the pitfalls of airborne pathogens to the attention of the scientific community. Not only viruses but also bacteria and fungi may exploit air transmission to colonize and infect potential hosts and be the cause of significant morbidity and mortality in susceptible populations. The efforts to decipher the mechanisms of pathogenicity of airborne microbes have brought to light the delicate equilibrium that governs the homeostasis of mucosal membranes. The microorganisms already thriving in the permissive environment of the respiratory tract represent a critical component of this equilibrium and a potent barrier to infection by means of direct competition with airborne pathogens or indirectly via modulation of the immune response. Moving down the respiratory tract, physicochemical and biological constraints promote site-specific expansion of microbes that engage in cross talk with the local immune system to maintain homeostasis and promote protection. In this review, we critically assess the site-specific microbial communities that an airborne pathogen encounters in its hypothetical travel along the respiratory tract and discuss the changes in the composition and function of the microbiome in airborne diseases by taking fungal and SARS-CoV-2 infections as examples. Finally, we discuss how technological and bioinformatics advancements may turn microbiome analysis into a valuable tool in the hands of clinicians to predict the risk of disease onset, the clinical course, and the response to treatment of individual patients in the direction of personalized medicine implementation.

INTRODUCTION

Airborne pathogens include a vast array of microbes that enter the respiratory tract via aerosol, dust, or droplets and may be the cause of significant morbidity and mortality, especially in immunocompromised individuals. Bacteria, fungi, and viruses, including the SARS-CoV-2 responsible for the recent COVID-19 pandemic (1), may exploit air transmission in their life cycle to find susceptible hosts. But the ability to be transported in air particles and find their way to get in contact with potential hosts is just the first step in the etiopathogenesis of airborne diseases. Indeed, microbes must survive multiple environmental barriers, such as temperature (2) and relative humidity (3). In addition, microbes must overcome the hostile environment of the host mucous membranes of the respiratory tract, from the nostrils and upper respiratory tract down to the lungs, each with its arsenal of defensive strategies that encompass the local immune system, the epithelial barrier lining as well as structured communities of commensal microorganisms, or microbiota (4). It is becoming increasingly clear that the respiratory tract harbors a complex and variegated microbiota, shaped by the physicochemical characteristics of the relative site, such as pH, relative humidity, temperature, as well as partial pressure of oxygen and carbon dioxide, and distinct bacterial taxa take the main stage from the nose to the lungs (5). Although exposure to ambient air and contiguity with the skin made it reasonable to hypothesize the presence of microbiota in the nose and upper respiratory tract, the presence of commensal microbes in the lower respiratory tract came as a surprise. Initially considered a sterile site in health, the lungs contain defined microbial taxa that are regulated by ecological rules of immigration from the upper respiratory tract and elimination, making the presence of microorganisms, although at low density compared with the gastrointestinal tract, an undisputed assumption in respiratory biology (5). But the mucous membranes of the respiratory tract are more than just protection from respiratory diseases, as revealed in the recent COVID-19 pandemic. The nose in particular represents not only the first part of the respiratory tract but also the site of entrance to the central nervous system via the olfactory mucosa (6). This dual role, indicated by the distinct embryological origin of the nose, from the ectoderm as opposed to the endodermal origin of the lining from the pharynx to the lungs, makes the nose a gateway to respiratory and neurological diseases from airborne pathogens. Neurological symptoms, from headaches, ageusia, and anosmia to the more severe encephalopathies, were common findings in COVID-19 (7). The nose, therefore, must be equipped with effective protective barriers against airborne pathogens, but, at the same time, be tolerant to the innocuous external stimuli to which it is constantly exposed (8). It is in the nose that the fine concept of mucosal tolerance finds one of its apexes, and the definition of its molecular and cellular underpinnings is important not only to better appreciate the physiological activity of the nose but also to design therapeutic strategies for mucosal intervention. Indeed, COVID-19 has also brought to attention the limits of current vaccination strategies and the need to exploit intranasal vaccination to boost the local immune defenses against infection (9). The development of intranasal vaccination cannot overlook the mucosal tolerance, but rather co-opt its mechanisms to refine the discriminating ability of targeting invading pathogens while sparing innocuous or beneficial encounters (10).

In this review, we will embody potential pathogens and envision their pathway as they move along the respiratory tract. We will give particular emphasis to the composition and function of the microbiota and the mechanisms of protection against colonization and infection. We will take advantage of our recent studies on fungal pneumonia as well as published literature on COVID-19 to contextualize these concepts into experimental evidence. Finally, we will describe the potential of the studies on the microbiota not only as a way to get further insights into the pathogenesis of airborne diseases but also as a tool for personalized medicine to predict the diagnosis, the prognosis, as well as the response to treatment, again taking advantage from studies from our group and others. This will lead to final considerations on the applicability of bioinformatic approaches such as machine learning and deep learning into clinical practice to make personalized medicine in respiratory infection a tangible reality.

ENTERING THE NOSE: THE FIRST GATE TO AIRBORNE DISEASES

Once transmitted through air, potential pathogens enter the nasal passages, which represent not just inert communication pathways or passive filters from the outside to the central nervous system or lower airways, but rather vibrant environments of host and microbial components. The former includes both an epithelial lining, either mucus-secreting as respiratory epithelium or endowed with sensory function as part of the olfactory system, and local immune cells as well as more defined mucosa-associated lymphoid tissues (11). The latter is primarily composed of Firmicutes, Proteobacteria, Actinobacteria, and Bacteroidota at phylum level, and Staphylococcus, Corynebacterium, Streptococcus, and Cutibacterium at the genus level, in line with the nares being mainly colonized by lipophilic bacteria (5). An interesting feature of the nasal microbiome is the coexistence of opportunistic pathogens that may transition from a commensal to a pathogenic behavior and be a significant cause of morbidity and mortality (12). For instance, Staphylococcus aureus is present in the nasal microbiota of ∼30% of the human population and is controlled by commensal microbes via multiple mechanisms that include competition for space and nutrients as well as production of antimicrobial molecules or modulation of immune response (13, 14). The failure of these protective mechanisms may cause S. aureus overgrowth with diverse clinical pictures, including, among other, potentially severe respiratory infections. Other examples are represented by Streptococcus pneumonia, Haemophilus influenzae, and Moraxella catarrhalis, commonly present in the nose of healthy children (15), further strengthening the idea that a healthy nasal microbiota is not only required to protect from environmental microbes but also to keep commensal opportunistic pathogens under control. Various clinical conditions, including allergic rhinitis, chronic rhinosinusitis, asthma, and respiratory tract infections from both viral and nonviral pathogens, have been associated with changes in the composition of nasal microbiome (8, 16, 17). Although causative relationships are difficult to draw, it is becoming increasingly clear that associations can be unveiled by combining the specific changes of the microbiota with the onset and/or the clinical course of the disease. For instance, the composition of the nasopharyngeal microbiome during the first year of life was associated with severity and lower respiratory infection by viral and bacterial pathogens as well as with allergic sensitization and risk of asthma in the subsequent years (18). Moving from this same concept, we have recently assessed whether changes in the microbiome could be associated with the risk of fungal pneumonia in hematological patients (19, 20). It is known that immunocompromised patients, such as those undergoing solid organ or hematopoietic stem cell transplantation, are at particular risk of developing invasive fungal infections, with significant morbidity and mortality (21, 22). However, the risk of fungal infection and the related treatment should be carefully titrated because antifungal therapy comes with side effects that should be avoided if unnecessary. The search for criteria to efficiently define the risk of fungal infections has led to the identification of risk factors and more elaborated dynamic algorithms that take into account the type of disease and treatment to assign a relative score during time (22). It is becoming increasingly clear that clinical parameters, such as neutrophil count, antibiotic, or steroid use, should be accompanied by biological and genetic aspects, at the basis of the individual variability. One such aspect is represented by the airway microbiota, and the commensal microbes that first come in touch with airborne pathogens are represented by the nasal microbiota. We, therefore, designed a prospective, multicenter study termed SNIF (Survey of Nasal Infection) in which patients diagnosed with hematological malignancies were enrolled and their nasal and oropharyngeal microbiome were analyzed every month for a 6-mo period. Upon stratifying the samples as high or low risk of infection based on clinical criteria, we characterized their nasal microbiota and found that high-risk samples were characterized by reduced diversity, loss of beneficial bacteria, and the expansion of potential pathogenic microbes (19). In particular, we found a reduced abundance of Corynebacterium and Dolosigranulum in the high-risk group. It is known that Corynebacterium is able to promote the shift of S. aureus to a commensal state (23), whereas Dolosigranulum promotes a variety of protective antimicrobial and anti-inflammatory activities in the nasal barrier (24). Interestingly, similar microbiome changes have been associated with SARS-CoV-2 infection. In particular, antibodies or cytokines in the nose and blood associated with nasal microbiota and patients with severe and critical disease showed reduced microbial diversity and lower abundance of Corynebacterium and Dolosigranulum compared with healthy donors and patients with moderate disease (25). This would indicate that the presence of certain taxa, or the function associated with their activity, provides a general mechanism of mucosal protection against a variety of pathological agents and could be targeted for therapeutic intervention.

All in all, this experimental evidence suggests that a proper configuration of the commensal microbes in the nose may not just be markers for identification of the risk of infection but represent at least one of the mechanisms employed by the microbiota to protect from colonization.

MOVING THROUGH THE PHARYNX: WHERE THE DIGESTIVE AND RESPIRATORY TRACTS CONNECT

Advancing the nasal cavity, a potential pathogen encounters the pharynx in its nasal, oral, and laryngeal sections. The oropharynx is particularly interesting because it is located in the back of mouth. Thus, as the nose represents a joint between the neurological and respiratory systems, the oropharynx plays a similar role between the digestive and respiratory tract. As such, it is exposed to external clues from both respiratory and digestive entrances and transmits along the two tracts accordingly. The importance is multifold. First of all, as part of the upper respiratory tract, the pharynx represents a major source of commensal microbes to the lungs, thus, representing a proxy for lung microbiota (5). Second, airborne pathogens may also be transmitted through the mouth. For instance, the oral cavity is an important site for SARS-CoV-2 infection and saliva represents a potential route of SARS-CoV-2 transmission (26).

From the compositional point of view, the major phyla of the oropharynx are represented by Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, whereas common genera of the pharynx are Streptococcus, Veillonella, Prevotella, Neisseria, and Rothia (5, 27, 28). It has been suggested that the oropharyngeal microbiome may play a protective role against respiratory tract infections (29). In the recent COVID-19 pandemic, the oropharyngeal microbiome was found to differ according to the severity of the disease and early microbiome signatures were able to predict severity and mortality (30–32). Discordant results have been instead reported on the association between changes in the oropharyngeal microbiome and wheeze in children (33, 34). As discussed by the authors, this may be linked to the diversity of the microbiome in the pediatric population that may hinder the identification of potential associations (34), thus making the selection of the target population and the number of samples analyzed critical criteria for association studies.

In the previously mentioned SNIF study, we also tried to determine whether changes in the oropharyngeal microbiota could predict the risk of fungal infection (20). First of all, we found that, as expected (28, 35), the oropharynx had a different microbiota compared with the nasal cavity, with a higher richness. Upon stratification for the risk of fungal infection, it was found that similar to the nose, the high-risk group was characterized by a lower diversity, loss of beneficial bacteria, such as Clostridiales, Bacteroidetes, and common oral taxa (Veillonella, Neisseria, Leptotrichia, and Gemella), and expansion of potential pathogens such as Staphylococcus and Enterococcus. Interestingly, a recent study has shown that patients admitted to a hospital with moderate or severe COVID-19 showed dysbiotic oropharyngeal microbial configuration (30), and machine learning models indicated that Neisseria and Haemophilus species abundances were the most important factors to predict hospital COVID-19 mortality (30), in line with a previous study (32). The authors further speculated that these changes in Neisseria or Haemophilus spp. may play a role in SARS-CoV-2 infections, such as by regulating the innate immune responses (30), thus representing not only predictive markers but also key players involved in the mechanisms of protection.

One such mechanism may involve the production of metabolites. In the nose and oropharynx of high-risk patients, we found reduced levels of the essential amino acid tryptophan, a building block of proteins, but also the substrate for bioactive metabolites production by both the host and commensal microbes (36). Interestingly, the reduced levels of tryptophan translated in reduced levels of host-derived kynurenine, implicated in immune tolerance (37), and microbial indole-3-aldehyde (3-IAld), active in mucosal barrier homeostasis (38), in the nose and the oropharynx, respectively, suggesting the existence of distinct, site-specific, immunoregulatory mechanisms. In line with this idea, the nose had higher levels of kynurenine compared with the pharynx (C.C, personal communication). This is consistent with the higher levels of IDO1, the rate-limiting enzyme in the kynurenine pathway (39), in the nose as revealed by single-cell RNA sequencing (40–42). This would suggest that IDO1-dependent tolerance is an important mechanism in the nose and should be taken into consideration when evaluating host-pathogen cross talk in the nose as well as when designing intranasal vaccination strategies. For instance, a recent study has shown that conventional dendritic cells in the nasal-associated lymphoid tissue were able to suppress T-cell responses in steady state to guarantee tolerance to inhaled antigens (10). This tolerogenic mechanism was suppressed upon infection by recruitment of monocyte-derived dendritic cells (10), thus providing a mechanistic explanation for tolerance breaking in host-pathogen interaction. This same mechanism was co-opted for a rational design of an intranasal vaccination strategy to improve T-cell responses upon intranasal immunization (10), thus coupling the knowledge of mucosal immunity with the development of a therapeutic intervention. Investigation of the suppressive activity of conventional dendritic cells on T-cell responses revealed the existence of contact-independent mechanisms via the production of prostaglandin E2 and reactive oxygen species (10). However, a previous study has shown that nonplasmacytoid dendritic cells expressed high levels of IDO in the nose-draining lymph node compared with other peripheral lymph nodes and that IDO blockade was able to break tolerance in a model of intranasal ovalbumin administration. The mechanism at the basis of IDO activity was likely related to an increase in the number of regulatory T cells able to counteract effector T cells (43), thus potentially implicating kynurenines in the nasal mucosal tolerance, in agreement with our findings. It is, therefore, likely that multiple mechanisms operate at the nasal mucosal surface to promote tolerance and their deciphering is crucial to understand host-pathogen interaction and to develop effective intranasal vaccination strategies.

THE LUNG AS FINAL DESTINATION

Moving down the respiratory tree, potential pathogens settle in the lungs where clinical manifestations of pulmonary infection may develop. Our current understanding of lower airway microbiome has moved from the idea of a sterile environment to that of a dynamic microbial population. The latter is governed by ecological mechanisms of immigration from the upper respiratory tract and elimination by means of cough, mucociliary clearance, and immune-mediated responses. Exploration of the lung microbiota is usually performed by analysis of sputum or bronchoalveolar lavage fluid, each with advantages and drawbacks. For instance, sputum is more likely a mix between upper and lower airways, and this contamination can be bypassed by bronchoalveolar lavage fluid collection, which is, however, an invasive procedure (44). Alternative sampling approaches, as reviewed in Ref. 44, include endotracheal aspirates in intubated patients, protected specimen brushing, bronchial biopsies, exhaled breath condensates, and tissues from explanted lungs. In healthy conditions, it is believed that bacteria in the lungs are at low biomass largely reflecting the upper respiratory tract, whereas changes in regional growth conditions, such as nutrients, oxygen, and inflammation, may drive the expansion of bacteria and the onset of pathological conditions (5, 44, 45). For instance, lung dysbiosis has been described in chronic obstructive pulmonary disease (46), lung cancer (47), interstitial lung disease (48, 49), cystic fibrosis (50, 51), asthma (52), acute respiratory distress syndrome, and ventilator-associated pneumonia (53), as well as infectious diseases, such as tuberculosis (54), and may be targeted for therapeutic intervention (55). In the context of fungal pneumonia, the lung microbiome has been investigated by bronchoalveolar lavage fluid collection in patients with proven or probable invasive pulmonary aspergillosis (56). As a result, it was found that patients diagnosed with invasive fungal infection had a lower diversity with expansion of Staphylococcus, Escherichia, Paraclostridium, and Finegoldia genera and reduced abundance of Prevotella and Veillonella genera (56). It is interesting to note that similar changes occurred in the oropharynx (20), indicating that less invasive approaches, such as the use of pharyngeal swabs, may be used to monitor the changes occurring in the respiratory tract and inform on the risk of colonization and infection in the lower airways as well as to predict the outcome of infection (56).

However, generalization of this concept should be taken with caution. A recent study on critically ill, mechanically ventilated patients with COVID-19 revealed distinct compositional structures in the lower and upper respiratory tract and only the former was predictive of death and prolonged mechanical ventilation time (57, 58). This is in line with the lung microbiome being exposed to regional growth in disease conditions and deviating from upper respiratory tract microbiome.

All in all, these data indicate that airborne pathogens travel along distinct regions of the respiratory tract, each contributing with specific physical and biological properties to the protection against colonization and infection (Fig. 1). All of these peculiarities, while posing the basis for deciphering the etiopathogenesis of airborne diseases, may also be exploited by bioinformatics approaches to elaborate predictive algorithms of disease diagnosis and prognosis.

Figure 1.

The picture depicts the mucosal membrane at different sections of the respiratory tract. The distinctive microbial composition, metabolic and immune responses in health (left) and fungal pneumonia (right) are shown. Details are described in the main text and summarized results as presented in Refs. 19, 20, 56. Images were taken from Servier Medical Art (https://smart.servier.com) and modified by the authors under the following terms: Creative Commons Attribution 3.0 Unported License.

FROM BIOLOGY TO PERSONALIZED MEDICINE: OMICS-INSPIRED ARTIFICIAL INTELLIGENCE IN PREDICTIVE DIAGNOSTICS AND THERAPEUTICS

Machine and deep learning are making major advances in solving problems related to the study of microbial communities, ranging from the classification and prediction of microbial taxa to the prediction of disease and biomarker discovery (59, 60). Indeed, the classical culture-based methodology of studying microorganisms in a particular environment has given way to metagenomics as an alternative way to explore and describe hidden microbial communities through culture-independent models by direct DNA isolation (61). However, microbiome data show special characteristics that challenge traditional methods of analysis (62). Machine learning (ML), a branch of artificial intelligence (AI) that implements models directly designed on the available data, allows to analyze through predictive statistics the adaptability, behaviors, and niche associations of microbes in a simple and accurate manner (63). Table 1 contains a list of the most widely used ML tools in the analysis of microbiota and their corresponding applications. ML methods have been successfully used in the development of microbiota-derived indicators of host phenotypes of infant age, sex, breastfeeding, antibiotic use, country of origin, and type of birth (74); in predicting positive and negative categories in bacterial vaginosis (75); in the association between pediatric irritable bowel syndrome and abdominal pain with intestinal microbes and fecal metabolites (76) exploiting the Least Absolute Shrinkage and Selection Operator algorithms, Random Forrest and Support Vector Machine models. With specific regard to the respiratory microbiome, the complex intertwining between the host and microbes associated with the heterogeneous nature of the microbiota itself also demands for investigational methods in which technological advances synergize with potent bioinformatics analytic models to translate this complexity into precision medicine. Recent work has provided compelling evidence of the successful use of ML to show that certain microbiome profiles of the lung following lung transplantation can provide prognostic information (77).

Table 1.

Selected machine learning (ML) tools in the analysis of microbiota

ML Tool	Application	Ref.
Meta-Signer	Feature ranking through ensemble learning and metagenome signature identifier	(64)
DeepMicro	Deep representation learning for infection/disease prediction using the microbiome data	(65)
mAML	Automated human disease classification through reproducible models	(66)
PaPrBaG	Detection of novel pathogens from NGS data	(67)
MicrobiomeAnalystR	Comprehensive functional, statistical, and meta analysis of microbiome data	(68)
mothur	Handling of multiple microbial datasets for community analysis	(69)
QIIME2	End-to-end analysis of microbiome data	(70)
BiomMiner	Exploratory microbiome analysis through auto-tuning of optimal parameters for visualization of clinical datasets	(71)
Scikit-learn	Predictive analysis	(72)
MIPMLP	Microbiome preprocessing for accurate data analysis	(73)

Ultimately, given the increased ability provided by AI methodologies to select biomarkers based on taxonomy-informed characteristics, to establish associations between the microbiome and host phenotypes, to reliably anticipate different disease states, and so on, the clinical utilization of the ML technology constitutes a paradigm shift that will change how medicine is practiced.

CONCLUSIONS

Airborne diseases have recently received a major boost by the outbreak of COVID-19 pandemic, and researchers have concentrated their efforts to improve our understanding of airborne pathogen transmission, colonization, and infection. Such efforts have required an increased awareness of the mechanisms of protection at airway mucosal barriers and the recognition of the crucial role played by the microbes that already live in our respiratory tract. In a life-long alliance, these commensal microbes work at our side to maintain mucosal integrity, but, as with all relationships, things can go awry. But the microbes can still come to our aid. Indeed, the changes in the composition and function of the microbiome may be inserted into automatic learning algorithms to improve our ability to predict the risk of disease onset and/or progression as well as the response to therapy, thus representing an innovative element in patient management. And since the microbiome is unique to the individual that carries it, the basis for personalized medicine is all set, just waiting for more steps forward in the use of omics-inspired artificial intelligence to become part of clinical practice.

GRANTS

This work was supported by the FunMeta Project (ERC-2011-AdG 293714), MicroTher (ERC-2018-PoC-813099), and HDM-FUN (European Union’s Horizon 2020 research and innovation program, number 847507) (to L.R.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.C., E.N., and L.R. drafted manuscript; C.C., E.N., and L.R. edited and revised manuscript; C.C., E.N., and L.R. approved final version of manuscript.