Background
The human body harbours trillions of bacterial cells on and within its surfaces and mucous membranes [
1]. These microorganisms (microbiota) are largely commensals and mutualists that can confer health advantages to the host. The microbiome is essential for numerous features of host physiology, including metabolism (by degrading otherwise non-digestible molecules), resistance to infection (e.g., via colonization resistance), and immune maturation and homeostasis [
2,
3]. Normally, the microbiota of healthy individuals is stable over time, although it is sensitive to changes in lifestyle, diet, and illnesses [
1,
4]. Perturbations of these microbial ecosystems can be associated with several diseases, including inflammatory bowel disease, and
Clostridium difficile infection, as well as conditions associated with critical illness (i.e., sepsis, acute respiratory distress syndrome, and multiple organ dysfunction syndrome) [
1,
5‐
9].
The microbiota of patients in the intensive care unit (ICU) fluctuates considerably due to acute disease states associated with critical illness, and common interventions such as mechanical ventilation, antimicrobials, gastric acid suppression, and enteral nutrition [
6,
10]. Studies using culture-dependent and culture-independent methods have demonstrated that microbial diversity in the gastrointestinal (GI) and respiratory tracts of critically ill patients decreases following ICU admission, and that critically ill patients experience pronounced disturbances of their microbial communities which become more severe over time [
8,
11‐
17].
The consequences of microbial dysbiosis on illness severity and mortality have been relatively unexplored, particularly in the lower respiratory tract. A better understanding of microbial disturbances in the ICU setting and their impact on clinical outcomes is needed, given the emerging evidence suggesting that therapeutics targeting the microbiota in critical illness may be promising to prevent or treat complications [
18‐
21].
The objectives of this prospective observational study were to investigate the microbial composition at distinct anatomical sites of the respiratory and GI tracts during critical illness, and to evaluate whether the microbial diversity in the first week of the ICU stay of mechanically ventilated patients is associated with illness severity and mortality.
Discussion
In this prospective observational study, we profiled the composition of microbial communities in a cohort of mechanically ventilated critically ill patients with a high incidence of pneumonia within their first week in the ICU, and how these communities relate to illness severity and clinical outcomes. Our results demonstrate that, in contrast to findings in healthy individuals [
39], the bacterial community structures are considerably less defined by body site in the upper GI tract and lower respiratory tract. Conditions associated with admission to the ICU and interventions in this setting could compromise normal host barriers and lead to compositional overlap between the airway and stomach, reflecting a loss of microbial separation across anatomical sites in critical illness. This work builds upon early studies of the microbiome in ICU patients [
7,
11,
14,
43], by including a control group with lower respiratory tract specimens to compare our findings with those of healthy individuals, and by including additional gastric samples in the ICU cohort to expand the multi-anatomical site analysis. While the proximal gastrointestinal tract is established as an important reservoir of ICU-associated pathogens [
44], the role of the stomach as a source of tracheal colonization is controversial, and is not always considered to be a substantial contributor to the pathogenesis of VAP [
45]. The prominent gastric colonization in ICU patients may be primarily due to use of prophylactic acid suppression to prevent stress-related gastric bleeding [
46]. The overlap in microbial compositions of the respiratory tract and stomach has also been demonstrated in non-critically ill pediatric patients receiving proton-pump inhibitors [
47], suggesting that these agents may contribute to the observed loss of biogeographical distinction.
The limited compositional similarities between healthy and ICU cohorts could be explained by the loss of commensal microorganisms during critical illness. We have demonstrated that ‘normal’ host-associated taxa, including the Lachnospiraceae family,
Faecalibacterium, and
Blautia genera in the GI tract and
Veillonella, Prevotella, and
Neisseria genera in the lower respiratory tract [
1,
40,
41] were decreased in relative abundance in the ICU cohort. The decrease of these commensal taxa in the GI tract is consistent with other work [
11,
14]. Several of the depleted OTUs are known to confer host advantages, such as anti-inflammatory and nutritional benefits via the production of short-chain fatty acid (SCFA) [
3,
48,
49]. The level of fecal SCFA is drastically decreased upon admission to the ICU [
9,
50]. Moreover, perturbation of the indigenous microbiota could lead to harmful repercussions and allow colonization by opportunistic secondary potential pathogens such as
C. difficile,
Candida albicans or facultative anaerobic gammaproteobacteria such as
Pseudomonas [
2,
51]
. Interestingly, the OTUs that were increased in the ICU cohort in comparison to the healthy cohort were common pathogens associated with ICU-acquired infections from the
Enterococcus, Pseudomonas, and
Staphylococcus genera [
52]. The limited number of OTUs that were significantly increased in the ICU cohort versus healthy cohort reflects the heterogeneity of the critically ill population due to various comorbidities and clinical interventions. We also demonstrate a lack of a shared microbial community structure in the ICU cohort toward a microbiota dominated by only few taxa. These results suggest that the emergence of pathogens is patient-specific, while the decrease in relative abundance of commensal taxa is observed more uniformly within the ICU population.
In this study, we demonstrated an inverse association between ETA microbial diversity and both illness severity and hospital mortality. We observed a significant decrease in respiratory microbial diversity in patients who died versus survived their hospital stay. Several studies have established that microbial composition tends to collapse in ICU patients toward the dominance of only a few taxa [
12‐
14]. It has been suggested that the microbial collapse is driven by aggressive antimicrobial administration during critical illness. A large prospective study including 14,414 critically ill patients has demonstrated that 71% of the patients were receiving antibiotics [
53]. Nevertheless, we did not see a correlation between concomitant antimicrobial exposure and ETAs microbial diversity in our cohort. This could be due to the fact that the samples included in this study were collected early during critical illness and an effect could be observed with later time points. Zakharkina et al. recently demonstrated no association between respiratory tract microbial diversity and antimicrobials in mechanically ventilated patients [
8]. Moreover, it has been demonstrated that lower respiratory tract samples from mechanically ventilated patients with pneumonia tend to have lower α-diversity in comparison with patient without suspected pneumonia [
15]. The loss of diversity could be due to the predominance of potential pathogens and could explain why this decrease in diversity is observed only with metrics accounting for taxa relative abundances (e.g., Shannon and Simpson diversity). However, survival groups exhibited no detectable difference of bacterial biomass nor rates of pneumonia. These results highlight the potential utilization of α-diversity metrics as an index of severity of illness and could, in addition to other clinical markers, improve patient’s stratification based on their survival prognosis. To the best of our knowledge, respiratory α-diversity has not been identified as a potential predictor of survival outcome in ICU patients. Microbial diversity of the GI tract has been associated with outcome in patients undergoing allogeneic hematopoietic stem cell transplantation, severe inflammatory response syndrome, and in high-risk patients admitted to the ICU [
16,
54,
55]. By contrast, in one study of 34 ICU patients, no association between survival and microbial diversity of the GI tract was observed [
17].
Our analysis was performed on samples collected during the first week of ICU stay (median collection time was 3 days following ICU admission for ETA and GA and 6 days for stool samples). This suggests that early time points could potentially be used to study the association between dysbiosis and an outcome occurring further downstream during the hospital stay (median 28.5 days IQR 14–50). Although there was no significant difference in survival curves between low and high ETA microbial diversity groups, of 29 patients included in the survival analysis, 35% died in hospital whereas 80% of the deceased patients were in the low diversity group. This suggests how ETA microbial diversity may be a complementary prognostic variable. As an indicator of illness severity akin to organ dysfunction, loss of microbial diversity can be conceptualized as a marker of poor outcome that potentially could be modifiable.
The limited number of samples and patients influences the power of these analyses. Larger studies will be required to increase our confidence in the association between microbial diversity and mortality. The comparison of critically ill patients and healthy individuals was constrained by lack of BAL specimens from the ICU cohort; therefore, more easily obtained ETA samples were used. Moreover, in clinical practice, BAL would not typically be performed early enough or commonly enough in patients without classical immunocompromised states for inclusion in our study. Although the ICU patients received several clinical interventions that could potentially affect the microbial structure of host-associated communities (e.g., antimicrobials, acid suppressants), these are intrinsic to critical care management. Adjustment for confounders such as sex, age, specific medications, and comorbidities was not suitable with our limited sample size, although no significant difference in α-diversity measurement was detected between the admission diagnoses. Moreover, despite the great heterogeneity observed in ICU patients, a proinflammatory state is a unifying feature of critical illness, regardless of the reason for admission (i.e., non-infectious conditions such as trauma and pancreatitis, as well as various infectious problems). Changes in colonization resistance and on the host’s biologic processes mediated by the microbiota may be similar across critically ill subgroups. Our study was confounded by the collection of samples over the first 7 days in the ICU. However, there was no correlation of ETAs α-diversity with time of sample collection and outcome (deceased or discharged alive).
Strengths of this study include the prospective data collection, protocolized specimen procurement and complete follow-up. We compared specimens from critically ill patients in 2 centers with a control group of healthy individuals. Our study was nested in a randomized trial testing the probiotic L. rhamnosus GG versus placebo; to avoid confounding results due to probiotic administration, only samples collected in the first week were included. However, we also performed a sensitivity analysis in which we removed OTU sequences assigned to L. rhamnosus GG and our results were not different, increasing confidence in the findings.