Experimental Models of Hospital-Acquired Infections After Traumatic Brain Injury: Challenges and Opportunities

TBI-induced systemic immunosuppression

TBI initiates a systemic stress response associated with activation of the HPA axis and the sympathetic branch of the autonomic nervous system, resulting in the release of glucocorticoids and catecholamines, respectively. ^{73 –76} Catecholamine release into the circulation remained elevated for 14 days post-injury and was found to be proportional to the severity of injury in TBI patients. ⁹² While this response may acutely compensate for TBI effects, a prolonged stress response has deleterious effects and can lead to multi-organ dysfunction. ^43,73,93

In addition to the increased release of corticoids and catecholamines, TBI induces a systemic immune response that can progress to systemic immune response syndrome, releasing immune mediators of inflammation (e.g., cytokines, chemokines) into the circulation. ^{58,94 -96} Systemic inflammation can persist for months in patients suffering from mild to severe TBI. ^94,95,97,98 In prospective cohort studies, TBI patients had elevated serum interleukin (IL)-1β, IL-6, IL-8, IL-10, and tumor necrosis factor (TNF)-α levels over the first year post-injury when compared with age-matched healthy controls. ⁹⁷ A subacute systemic cytokine load score identified individuals at risk for unfavorable outcomes following TBI, and was predictive of global outcomes at 6 and 12 months. ⁹⁸ In particular, a higher IL-6/IL-10 ratio was associated with increased risk of poor outcomes at 6 months. Moreover, the acute phase cytokine load score predicted cognitive performance in moderate-to-severe TBI patients, ⁹⁹ such that IL-1β, TNF-α, soluble IL6 receptor (sIL-6R), regulated upon activation, normal T cell expressed and secreted (RANTES), and macrophage inflammatory protein (MIP)-1β were negatively associated with cognition. Thus, there is a significant correlation between circulating cytokines and chemokines and worsened patient outcomes. Once in circulation, these immune mediators exert deleterious effects on other peripheral organ systems. ^58,100

Animal studies indicate that TBI results in acute and chronic changes in peripheral immune function, prominently skewed towards widespread immunosuppression in the innate and adaptive arms of systemic immunity acutely after TBI, and then transforming to a hyper-inflammatory and dysfunctional state during chronic recovery periods. ^{80 -82} In the acute phase, systemic immune changes include suppressed pro-inflammatory cytokine production and impairments in critical respiratory burst and phagocytosis functions in myeloid cells (neutrophils and monocytes), as well as marked thymic atrophy and circulating T lymphopenia. ⁸⁰ In contrast, during the chronic phase, excessive inflammation ensues with increased ROS and pro-inflammatory cytokine production in myeloid cells, and disrupted adaptive immune function as demonstrated by chronic T lymphopenia, CD4:CD8 inversion, fewer naïve T cells, and greater memory T cell conversion. ⁸⁰ Such changes in systemic immunity parallel robust activation of neuroinflammatory pathways in the injured brain, which have been shown to drive secondary neurodegeneration at distant sites (e.g., hippocampus and thalamus), leading to cognitive impairments and long-term neurological dysfunction in TBI mice. ⁵⁹

TBI-induced immunosuppression is hypothesized to result from three major mechanisms: 1) HPA axis activation leading to excess glucocorticoid release from the adrenal cortex; 2) excess sympathetic activation followed by catecholamine release from the adrenal medulla and sympathetic terminals; and 3) increased vagal nerve activity. ^46,78,81,101 These mechanisms have been implicated in driving injury-induced central nervous system (CNS) immunosuppression in ischemic stroke, spinal cord injury, and TBI. ^78,79

Clinical studies investigating cortisol and the HPA axis demonstrate high circulating cortisol levels following acute brain injury. ¹⁰² In an experimental mild TBI study in mice, decreased numbers of circulating T cells was inversely correlated with a transient and robust increase in plasma cortisol levels. ¹⁰³ Cortisol reduces pro-inflammatory cytokine production and boosts levels of anti-inflammatory cytokines, such as IL-10 and transforming growth factor (TGF)-β. ¹⁰⁴ TBI-induced transient increase in cortisol impaired circulation of lymphocytes and sequestered T cells in lymphoid tissue. ¹⁰³ High cortisol levels have been shown to suppress phagocytosis, cytokine production (e.g., TNF-α, IL-1β, IL-6), and downregulate toll-like receptor 4 (TLR4) signaling in macrophages throughout the body, resulting in an anti-inflammatory state, ^105,106 which may contribute to increased susceptibility to infection.

Acutely following TBI, there are increased circulating catecholamines in patients. ¹⁰⁷ Sympathetic storm or increased norepinephrine and epinephrine have also been shown to induce immunosuppression in experimental models of spinal cord injury, ¹⁰¹ and both spinal cord injury and stroke-induced immunosuppression is sufficient to cause spontaneous pneumonia, typically attributed to the translocation of gut-derived bacteria. ^101,108 In contrast, spontaneous pneumonia has not been demonstrated following TBI in experimental models. ¹⁰⁹ Pre-clinical studies on the role of epinephrine in TBI have shown that there is a profound effect of sympathetic activation on the activation of T cells. ¹¹⁰ However, the effects of the sympathetic nervous system on monocytes and macrophage function in the context of TBI have not been elucidated.

The vagus nerve, a key bidirectional mechanism for neuro-immune communication, is activated following acute brain injury to regulate peripheral inflammatory responses. Vagal parasympathetic innervation of the celiac ganglion activates the splenic nerve to control the release of norepinephrine and epinephrine in the spleen. ^77,111 This activates T cells in the spleen via β2-adrenergic receptors, resulting in acetylcholine release. ¹¹² Acetylcholine binds α7 nicotinic acetylcholine receptors on macrophages, which inhibits TLR signaling resulting in decreased TNF-α production. ¹¹² In studies of experimental stroke with secondary P. aeruginosa lung infection, there is a strong association between immunosuppression and the activation of α7 nicotinic acetylcholine receptors. ⁵⁵ Interestingly, antagonism of this receptor abrogated stroke-induced immunosuppression, increasing phagocytosis and protecting the lung from Pseudomonas infection. ⁵⁵ However, whether this occurs in TBI as the overriding mechanism of immunosuppression, and whether potential targeting of this mechanism can modulate the risk of infection after acute CNS injury, is unknown.

HMGB1 and inflammasome activation

High mobility group box protein-1 (HMGB1) is a DAMP that has been implicated as pivotal to the brain-lung axis after injury. HMGB1 is released from necrotic neurons, ⁵⁶ enters the bloodstream, and travels to the lungs, where it acts as an agonist for TLR4 and RAGE receptors and activates NF-κB signaling. ¹¹³ In trauma patients without brain injury, circulating HMGB1 has been shown to mediate ALI, and higher levels of HMGB1 are associated with poorer outcomes in ALI patients. ¹¹⁴ TBI patients also exhibit enhanced HMGB1 levels that correlate with poorer outcomes. ^115,116

In mouse experiments, increased expression of inflammasome components (e.g., AIM2 and ASC) have been reported in the lung at 4 and 24 h following moderate-severe TBI, and have been associated with lung pathology. ⁴⁷ Extracellular vesicles containing HMGB1 and inflammasome components are released from damaged endothelial cells in the brain, enter circulation and traffic to the lung where they induce local cell death and pathology. ⁴⁷ The NLRP3 inflammasome also plays a prominent role in the pathophysiology of ALI and TBI. ^62,63 Following acute brain injury, the NLRP3 and AIM2 inflammasomes have been implicated in cell death and alveolar collapse in the lung. ⁴⁷ In TBI patients, lung cell pyroptosis has been reported, ¹¹⁷ and animal TBI studies have demonstrated that NLRP3/AIM2 inflammasome activation in the lung results in increased IL-1β and pyroptosis. ⁴⁷ Thus, post-TBI lung injury due to circulating HMGB1-induced inflammasome activation may be perpetuated by increased TLR4/RAGE signaling and increased NLRP3 and AIM2 inflammasome ligands in the lung. ¹¹³

Animal models of experimental TBI

Various animal models have been developed to better understand the pathophysiology of TBI as well as to explore potential therapeutics. The choice of animal used must take into account theoretical and ethical considerations. This includes focusing on certain aspects of TBI; namely, the nature, severity, and location of injury. It is important to note that each model chiefly focuses on a certain aspect of TBI, yet none of these models can perfectly mimic all aspects of TBI in humans.

Rodents are most commonly used in TBI research due to their small size, modest cost, and standardized outcome measurements. Among the many different models that have been developed, the three most prominent in the field include the fluid percussion injury (FPI), controlled cortical impact (CCI) injury, and the weight-drop (WD) model. ¹¹⁸ Other less common, and more recently developed, models include blast-induced TBI, ¹¹⁹ the Closed-Head Impact Model of Engineered Rotational Acceleration (CHIMERA), ^120,121 and the awake closed-head injury (ACHI) model. ¹²² Readers are referred to several excellent reviews that describe these models in depth. ^118,123,124

Despite the frequent use of these animal models to explore injury mechanisms and potential therapeutics, it has been difficult to translate pre-clinical findings to the clinic to date. Certain clinical and pathological features have been successfully modeled in animals, such as the neuroinflammatory response; yet other aspects, such as the severity of the injury, are rarely recapitulated to the same extent. For example, inconsistent with the human condition, rodents with a “moderate or severe TBI” are ambulant and resume normal behavior shortly after the injury. ¹²⁵ Functional deficits seen may range from sensorimotor to cognitive and social behavior changes, depending upon several variables including the age at the time of injury, injury location, and severity. Of note, most studies use healthy young adult male animals, which does not necessarily reflect the human population who sustain TBI (i.e., presence of pre-existing comorbidities, alcohol/drug abuse, obesity, young or old age, different sexes, and post-TBI complications such as infection). Therefore, further studies are needed to address such translational gaps.

Animal models of lung infection

Various animals, from rodents to non-human primates, have been used to study the pathophysiology of lung infections. ¹²⁶ Mouse is the most common species, particularly for models of pneumonia, due to practical considerations. ¹²⁷ There are several routes to administer microorganisms into the lungs to model a respiratory infection. ¹²⁸ Aerosol infection involves the inhalation of aerosolized microorganisms into whole–body or nose-only chambers: the aerosol is generated by one or more nebulizers and sprayed into an enclosed chamber (“whole–body” model), or directly onto the restrained animal's snout (“nose-only” model). This method usually results in the symmetrical and uniform deposition of microorganisms in both lungs, ¹²⁹ and is relatively simple and noninvasive. However, it is not suitable for all pathogens, as some pathogens survive poorly in aerosols, and the extent to which the bacteria reach the lower respiratory tract is variable due to multiple factors (e.g., the performance of the device and size of the droplet or particle).

An alternative method is intranasal inoculation, which better mimics natural infection. Droplets of the inoculum are administered onto the animal's nares, while the animal is held in a vertical position, and then aspirated by the animal upon inhalation. Anesthesia is often required to prevent the inoculum from being swallowed or sneezed out. ¹³⁰ While this technique is simple, it can also result in a highly variable, non-uniform and asymmetric deposition of microorganisms in the lungs, dependent upon the technician's skill. A more reproducible method is instead intratracheal inoculation via endotracheal or oropharyngeal injection. ¹³¹ This involves directly administering the pathogen into the trachea, with or without a tracheotomy, thereby bypassing the upper airways. While this technique requires no surgery and causes the animal minimal pain and discomfort, it is technically more difficult to perform. It has the advantage of being highly consistent and allows for greater control over the dose of the pathogen entering the lungs. ¹³²

Following pulmonary infection, animals may demonstrate behavioral changes such as lethargy and lack of response to environmental stimulus, as well as physical changes such as piloerection, weight loss, fever, and nasal discharge. Severe infection often results in signs of respiratory distress such as gasping or wheezing. Some animals may succumb to the infection, with mortality rate depending on many factors such as the infectious agent, isolate and dose, and host defense. ¹³³

Experimental studies investigating infection in TBI models

To date, only limited studies have considered the experimental combination of a TBI and infection. Infection is most commonly induced by administering isolated or synthetic endotoxins such as lipopolysaccharide (LPS), heat-killed bacteria, or viable bacteria. Although these mechanisms all serve to mimic infection in vivo, the use of live bacteria most closely emulates human disease in animal models, is thereby most clinically-relevant, and as such is the focus of this review. Relevant animal studies investigating infection in TBI models are summarized in Table 2. ^{134 –140} It should be noted that animal models of TBI and infection has so far been established in the three most common models of TBI: FPI, CCI and WD, but not in blast injury, CHIMERA, or ACHI models.

Most experimental studies modeling hospital-acquired infections after TBI have examined pulmonary injury and respiratory infections. For example, Vermeij and colleagues ¹³⁵ used the Marmarou WD model to induce mild TBI in male rats, and induced pneumonia via intratracheal inoculation of heat-killed S. aureus (1 × 10⁹ CFU) at 24 h after TBI. Rats with TBI alone had increased plasma levels of phosphorylated neurofilament heavy chain (pNF-H), a marker of neuronal damage; alongside indices of lung damage such as increased relative lung weight (reflecting pulmonary edema), vascular permeability, and increased levels of interleukin (IL)-1 and IL-6 cytokines in the bronchoalveolar lavage fluid. TBI rats that sustained S. aureus infection showed a marked reduction in TNF-α release, indicative of systemic immune suppression. Additionally, extensive histological pulmonary inflammation, pulmonary edema, and lung damage was observed at 24 h post-infection. However, TBI prior to lung infection was not found to affect the severity of subsequent pneumonia, perhaps due to the mild nature of the injury model. ¹³⁵

A recent study by Pittet and colleagues ¹³⁶ sought to investigate sex differences in the response to pneumonia post-TBI, revealing a beneficial effect of estrogen. In this study, C57BL/6 male and female mice underwent CCI-induced severe TBI, followed by intratracheal instillation of P. aeruginosa (5 × 10⁷ CFU K-strain) 24 h post-TBI. Increased mortality and decreased lung bacterial clearance (i.e., increased bacterial burden) were observed in TBI male mice challenged with secondary bacterial pneumonia, compared with similarly infected sham-operated (i.e., craniotomy only) mice. Male mice were significantly more vulnerable than females to mortality resulting from bacterial pneumonia after TBI (25% vs. 100% survival), while oophorectomized female mice had a similar mortality rate to male mice. The lung bacterial load was also reduced in female mice, alongside increased production of alveolar macrophages and higher levels of TNF-α secretion compared with male mice. Administration of estradiol to male and oophorectomized female mice were found to rescue these deficits, i.e., reduced mortality and increased lung bacterial clearance. ¹³⁶ Together, these findings demonstrate greater mortality and decreased lung bacterial clearance in infected mice post-TBI compared with similarly infected sham-operated mice. These outcomes were found to be sex-dependent, with estrogen playing a salutary role.

Paradoxically, contrasting results were reported by Vaickus and colleagues, ¹³⁷ using P. aeruginosa at the same concentration (5 × 10⁷ CFU), but from a different source (Boston 41501 strain; ATCC), when administered at 48 h post-TBI. Mild TBI was induced using a closed-head WD model, with tail trauma as a control group, in female ICR132 (CD-1) mice. The authors reported that mild TBI prior to infection improved survival at least in part due to release of the neuropeptide substance P, which acts as a chemokine and increased neutrophil recruitment within the lungs, thereby facilitating increased bacterial clearance. ¹³⁷ Similar findings were reported by Ruan and colleagues, ¹³⁸ when they induced a mild TBI using the WD model in FVB/N mice of both sexes, followed 30 min later by an intratracheal injection of P. aeruginosa following a tracheotomy to model a lung infection concurrent to a TBI. In this study, AN engineered strain of P. aeruginosa was used (Xen5 at 2 × 10⁴ CFU) which exhibits bioluminescence, allowing for in vivo imaging of the infection. ¹³⁸ Despite this difference, they reported similar findings to those of Vaickus and colleagues in terms of the mild TBI reducing bacterial burden and pulmonary microvascular permeability; fewer immune cells, and lower levels of the pro-inflammatory cytokines IL-1β, IL-6, and TNF-α in the bronchoalveolar lavage fluid and lung tissue of mice that sustained the dual insult compared with infected mice without a TBI. Additionally, alveolar macrophages isolated from mild TBI mice demonstrated enhanced bactericidal capacity.

Altogether, these results suggest that a mild TBI attenuate, rather than exacerbate, inflammation and ALI following a subsequent pulmonary infection. ¹³⁸ Such discrepancies in findings may be attributed to differences in experimental model and study design, ¹³⁶ including baseline differences in the immunological profiles of rodent strains. ¹⁴¹ Together, this highlights the need for further research in animal models of TBI and infection to gain a more comprehensive understanding of interacting pathophysiological mechanisms in this context.

Inducing pneumonia caused by a different bacterium, Streptococcus pneumoniae, Doran and colleagues ¹³⁹ first investigated the consequences of TBI combined with infection across a more prolonged time course. Twelve-week-old male mice with a moderate CCI injury were administered intranasal S. pneumoniae (1500 CFU) at 3- or 60-days post-injury, then euthanized at 3- or 7-days post-infection. Mice that received the dual insult suffered higher mortality rates (60% vs. 5% when infected at 60-days post-injury, for TBI + S. pneumoniae vs. Sham + S. pneumoniae), poorer motor function recovery, and enhanced neuroinflammation. Further, infiltrating lung monocytes were found to be acutely immunosuppressed, as they were not able to produce IL-1ß, TNF-α or ROS. Notably, delayed infection resulted in the opposite outcome, with lung monocytes becoming hyper-reactive and producing increased levels of pro-inflammatory cytokines at chronic time points. Further, there was increased expression of NLRP3, AIM2, cleaved caspase-1 and ASC, indicative of chronic inflammasome activation in the lungs in the delayed infection model. In both experimental paradigms, mice with both TBI and S. pneumoniae were found to have increased bacterial burden in the lung, worsened lung pathology (infiltrating neutrophils, and alveolitis), and increased circulating HMGB1 levels. This study highlights the importance of the timing of a post-injury infection and identified differential responses dependent upon the pre-existing immune state. ¹³⁹

To the best of our knowledge, only one study to date has considered the consequences of viral pathogens on TBI outcomes. Mastorakos and colleagues ¹⁴⁰ challenged adult (8-12 weeks old) male and female C57Bl/6J mice with mild TBI, induced by compression of the meningeal space by thinning the skull bone and applying a downward pressure. ¹⁴² They combined this with one of several pathogens/mimics, including: 1) the lymphocytic choriomeningitis virus Armstrong strain, a non-cytopathic arenavirus; 2) vesicular stomatitis virus Indiana strain; 3) polyinosinic:polycytidylic acid, a synthetic analogue of double-stranded RNA that activates antiviral pathogen-associated molecular pattern (PAMP) receptors; 4) the pathogenic yeast Candida albicans; or 5) LPS derived from Escherichia coli. They found that these immune challenges, regardless of their etiology, interfered with meningeal vascular repair and prevented neurological recovery after mild brain injury. Viral infection specifically resulted in altered angiogenic programming and reparative macrophage distribution; alongside a marked and prolonged increase in Type I interferon (IFN) expression, which then acts on Type I IFN receptors on myelomonocytic cells, preventing meningeal repair. Failed revascularization then led to neuronal loss and sustained BBB breakdown. Therefore, this study revealed the deleterious effects of single or recurrent infections of various infectious agents in impeding neurological recovery post-TBI. ¹⁴⁰

Although we have focused here on reviewing models of respiratory infections post-TBI, it is pertinent to note that the gut-brain axis has also been implicated in TBI, which may influence how other types of infection interact with a brain injury. ¹⁴³ Many studies have now shown that TBI alters the gut microbiota, ¹⁴⁴ with microbial dysbiosis associated with gastrointestinal dysfunction and intestinal inflammation, ¹⁴⁵ as well as exacerbated neuroinflammation. ¹⁴⁶

Two key experimental studies to date have examined intestinal infections after TBI. Firstly, Ma and colleagues ¹³⁴ examined the effects of the enteric pathogen Citrobacter rodentium (5 × 10¹⁰ CFU administered via oral gavage at 28 days post-injury) on gut inflammation after moderate TBI, in young adult (8-10 weeks old) male mice. Infection with C rodentium after TBI worsened brain injury and neuroinflammation, in addition to changing colon morphology and mucosal barrier function at 12 days post-infection (but to a similar extent in sham and TBI mice). ¹³⁴ Another study by the same group sought to understand the effects of colonic inflammation induced by dextran sodium sulfate (administered via drinking water, from 28 days post-TBI for 7 days), on neurobehavior in young adult (9 weeks old) male mice after a moderately-severe CCI injury. ¹⁴⁷ Persistent neurological deficits (e.g., impaired cognition and increased anxiety-like behavior), exacerbated TBI-associated hippocampal neurodegeneration, and altered autonomic balance, were observed in chronic TBI mice with acute colonic inflammation, ¹⁴⁷ which may have been due to how injured mice handled the prolonged systemic immune response following delayed administration with dextran sodium sulfate. Together, these studies demonstrate that a post-TBI gastrointestinal inflammatory challenge adversely impacts outcome in TBI mice.

The experimental host animal

There are obvious practical benefits to the use of rodents for these types of studies, alongside the relative ease of measuring bacterial load, immunological responses, and morbidity and mortality in rats and mice. Indeed, while large animal models of both TBI and pneumonia exist (e.g., pigs, sheep, and rabbits), ^133,150,151 most published studies to date have used rodents in the common experimental TBI models, FPI, and CCI. The choice of injury severity in the model is important, given that clinically, infections are most often associated with severe TBI. Ideally, models should also incorporate common secondary insults such as hypoxemia and hypertension, which are known risk factors both for hospital-acquired infections as well as poorer outcomes more broadly post-injury. ^19,152,153 Such secondary physiological insults can be more readily incorporated and monitored in large animal models, to mirror the complex ICU experience of severe TBI patients. ¹⁵⁴

It is also worth noting several limitations that stem from host species-specific responses to infections. The rodent lung and respiratory tract differ considerably from humans in terms of their anatomy, distribution of cells, and connective tissue, ^155,156 which can influence the distribution of bacteria infecting the lungs. ¹⁵⁷ Species-specific physiological and immune responses should also be considered. The causes of mortality in the context of pneumonia also differ between humans and laboratory animals—while humans may succumb to cardiac dysfunction, stroke, or pulmonary embolism associated with pneumonia, these events are typically rare in mice. ¹⁵⁸ As alluded to earlier, even within the same species, strain-specific differences in susceptibility to bacterial infections have been reported. ¹⁵⁸ For example, C57BL/6 mice show marked resistance to lung infection with 10⁴ CFU K. pneumoniae (strain ATCC 43816 serotype 2), while 129/Sv mice are highly susceptible to the same strain/dose. ¹⁴¹ There are also important differences in how bacteria interact with the host species (e.g., how virulent a particular strain is in the rodent vs. human).

Another consideration for the study design and choice of host animal is whether to include sex as a biological variable. Experimental TBI studies have historically utilized mostly male animals. However, this bias has shifted in recent years with increased awareness of sex-specific differences in injury responses, including neuroinflammation. ^159,160 By contrast, in the microbiology literature, the majority of studies inducing pneumonia to assess antimicrobial efficacy have been conducted in female mice. ¹⁵⁰ Sex is likely to be an important variable in dual-hit insult scenarios as host-pathogen interactions may be sex-dependent, mediated at least in part via the effects of steroidal sex hormones. ¹⁶¹ Limited evidence has alluded to differential vulnerability to bacterial infections in male and female patients after TBI, ⁴⁶ while a recent experimental study of pulmonary P. aeruginosa infection in CCI mice, as described above, reported that males had higher mortality, impaired bacterial clearance, and an altered immune response compared with females. ¹³⁶ Further, the potential effect of age on outcomes after TBI and infection also represents a knowledge gap in the field.

The infectious agent

For optimal clinical relevance, experimental models should focus on microorganisms that drive the infections most frequently reported in TBI patients. For example, P. aeruginosa, K. pneumoniae, S. aureus, and A. baumannii are among the most common, treatment-resistant opportunistic pathogens in nosocomial infections such as ventilator-associated pneumonia. ^46,150 Therefore, the specific strain of interest must be decided upon, including whether to use a clinical isolate or laboratory-derived strain. The choice of bacterial strain can be a key determinant of the manifestation of a pneumonia phenotype in the model. For example, different bacterial strains may have different degrees of pathogenicity depending upon the host species, through variable expression of different virulence factors depending on the microenvironment. ¹³³ Clinically-isolated strains can be adapted for increased virulence in rodents; yet, such manipulations may result in the infection eliciting an altered immune response of reduced relevance to the clinical scenario. ^162,163 Inoculations are typically performed with bacteria in logarithmic phase due to their increased ability to survive and expand during this growth phase, although such a detail is rarely reported. ¹⁵⁰ A dose–response curve is likely required to determine the optimal concentration of live bacteria to be inoculated to elicit the desired infection-associated symptoms and pathology in each new experimental paradigm.

Administration of bacteria

As noted earlier, a bacterial infection can be induced via multiple routes into the animal host. For lung infection models, the bacterial inoculum may be delivered intranasally or intratracheally as liquid or powder aerosol, with the choice of delivery method being an important factor for the dose of bacteria that reaches the lower respiratory tract and lungs, and the likelihood of off-target effects. ^150,164 A single bolus inoculation is most common, ¹⁵⁰ involving forced aspiration of a high concentration of bacteria in suspension, although this carries the caveat that this mode of administration typically induces acute infection in rodent models that differs from what is seen clinically. ^158,165 Intratracheal inoculation, involving direct delivery of the bacterial suspension into the trachea, results in a reproducible dosing strategy. However, both intratracheal and intranasal inoculation routes typically require a skilled operator and anesthesia to perform, which represents an additional stressor to the experimental animals.

Some studies of ventilator-associated pneumonia incorporate mechanical ventilation of the animal in addition to delivery of a bacterial inoculation, as the ventilation process itself is considered to be an important driver of lung inflammation. While not yet examined in combination with experimental TBI, the addition of mechanical ventilation compared with animals with bacterial inoculation alone results in exacerbated lung infections and higher mortality in mice. ^57,166 In future, other comorbidities and/or host factors that can influence the response to infection and/or TBI should also be incorporated into such models. For example, prior exposure to illness, the role of genetics and epigenetics in outcomes after injury and infection, and the influence of medications such as antibiotics warrant investigation. ¹⁵⁸

Another decision is whether to induce neutropenia in the experimental animals prior to inoculation. Often used in studies to evaluate the efficacy of antibiotics, neutropenia induced by drugs such as cyclophosphamide or vinblastine allows for more rapid bacterial growth and evaluation of the killing effect of the antibiotic itself, and minimizes spontaneous resolution of the infection. ¹⁵⁰ In addition, some opportunistic pathogens (e.g., P. aeruginosa, K. pneumoniae, and A. baumannii) may not be able to establish reproducible lung infections in immunocompetent animals.

To date, most animal studies of pneumonia (and all studies of TBI plus a bacterial infection) have used just a single microorganism. ¹⁶⁷ However, future studies might consider the incorporation of polymicrobial infection to better reflect the human scenario. Micro-aspiration to sample patients' oropharyngeal secretions have found that numerous species of bacteria (both colonized microorganisms and causative pathogens) are typically present, at least for community-acquired pneumonia, which creates a uniquely complex immune response. ^158,168 For example, in an adult population of community-acquired pneumonia cases in Norway over a 3-year period, multiple pathogens were detected in 41% of cases, with S. pneumoniae being most commonly identified in co-infected cases. ¹⁶⁹

In summary, there are many experimental variables to consider when establishing a model of hospital-acquired infection in rodents. Bacterial inocula should be prepared, and infections performed in a rigorous, standardized manner; methodology should be reported in depth for optimal transparency and reproducibility. Pilot studies are usually required to determine the most appropriate inoculum for each strain of bacteria and host strain/species. In addition, parameters of the experimental TBI model must be determined, including which model and injury severity. Finally, the timing of infection relative to TBI (i.e., pre-injury, concurrent, or at x time point post-injury) is another critical determinant of the model phenotype, as demonstrated by Doran and colleagues, when differences in timing of post-injury infection dramatically altered the lung immune state (i.e., immune suppression versus hyper-reactive) and mortality outcomes following TBI. ¹³⁹

Outcome measures

Once the two-hit insult model is established, a strong interdisciplinary team to evaluate outcomes is likely required, ranging from clinicians to neuroscientists, microbiologists, and immunologists, to ensure relevant expertise in considering the impact of an infection post-TBI on the whole host animal. For TBI and pneumonia combined models, outcome measures typically include neuroinflammation, brain histology, and neurobehavioral outcomes relative to the TBI insult, accompanied by lung histology and lung inflammation read-outs, and mortality. ^170,171 Some practical and logistical considerations can influence the choice of outcome measures. For example, due to the use of live infectious material in these models, most of the live animal work must be performed in an appropriate biocontainment environment (e.g., Class 2 Biosafety Cabinet or facility), which may limit the extent of neurobehavioral testing that can be performed.

Broadly, an experimental TBI typically causes a neuroinflammatory response involving the activation of resident glial cells, alongside neuronal and axonal damage. ¹⁷² In general, pneumonia models in rodents cause a phenotype of vascular congestion, alveolitis and alveolar collapse, together with significant neutrophil infiltration into the lung tissue, coinciding with a peak in sickness symptoms up to 24 h post-infection and resolution within days. ¹³³ Lung pathology assessments should be performed in a blinded fashion by a trained individual, preferably a pathologist, with expertise in infectious disease and lung pathology in rodents. In addition, assessment of bronchoalveolar lavage fluid and sera for immune cells and cytokines can provide important evidence linking the injury consequences with the infection. Assessment of the bacterial burden in the blood and lungs of infected mice, to gauge disease progression or severity, is usually achieved by postmortem homogenization of lung tissue and plating of serial dilutions onto nutrient-rich agar.

Finally, while immunological and pathological markers of TBI-induced lung injury and the consequences of respiratory infection provide mechanistic insight, there is also a need to test pulmonary function in mice in vivo in the two-hit insult models. This is possible using several approaches to assess pulmonary function in mice, ^173,174 but there is a trade-off between accuracy, invasiveness, and convenience between approaches. Nevertheless, investigating physiological and functional responses in the lung following post-traumatic respiratory infection (e.g., lung resistance and dynamic compliance) will be needed in the future to mimic the clinical scenario more accurately.

As should be the case for all experimental TBI studies, those that also incorporate an infection model are recommended to provide detailed reporting of all data and methodology that could influence the results (e.g., by following the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines). ¹⁷⁵ Key variables associated with establishment of the infection model, such as a clear description of consistent culture conditions and methods, are important yet often omitted details essential for reproducibility and transparency.