The role and consequences of the presence of Candida spp. in the lungs of mechanically ventilated patients are receiving growing attention as we gain insight into the interaction between fungi and bacteria in diseases in general [1] and in lung infection in particular [2]. In a 2009 issue of Intensive Care Medicine, Meersseman et al. provided us with convincing data on the absence of Candida pneumonia in the ICU despite a high rate of colonization (53 %) [3]. These authors could however only speculate on the significance of the presence of Candida spp. in their patients. We ended our accompanying editorial by stating that Candida lung colonization may have a significant role in bacterial pneumonia development [4]. This statement was prompted by several clinical and experimental clues.

Azoulay et al. were among the first to hypothesize that the presence of Candida spp. in the airways of mechanically ventilated ICU patients might not be anecdotal [5]. In their study, Candida colonization was associated with an increased risk for Pseudomonas aeruginosa ventilator-associated pneumonia (VAP). Another important hint towards the role of Candida spp. was provided by Nseir et al. [6], who showed that antifungal treatment decreased the risk for P. aeruginosa infection in colonized patients. An experimental validation to these clinical observations was later provided by the demonstration that when challenged with an inoculum of P. aeruginosa insufficient to induce bacterial pneumonia, rats colonized with C. albicans developed considerably more P. aeruginosa pneumonia that non-colonized rats [2]. The question that then arose was whether this effect was specific to P. aeruginosa? A preliminary experimental report suggested that colonization also favored the development of Staphylococcus aureus and Escherichia coli pneumonia [7], implying that the mechanisms by which Candida colonization promoted bacterial pneumonia were independent of bacterial species. Such an observation had not been made in the clinical field, probably because, given the prevalence of P. aeruginosa, studies were not powered enough to detect an effect on other bacteria.

In this issue of Intensive Care Medicine, Hamet et al. [8] provide further evidence for the significant role played by Candida colonization in bacterial VAP.

These authors studied all clinically suspected VAP episodes in their ICU during a 4-year period and reviewed the microbiological data for the presence of Candida spp. colonization, bacterial documentation of the VAP episode and resistance profile to antibiotics. A total of 323 episodes were analyzed. Important confirmatory but also unexpected interesting findings were obtained.

First of all, Hamet et al. confirm that Candida spp. colonization is frequent in mechanically ventilated patients, since more than half (56 %) of their patients harbored Candida in their airways, a figure very similar to the 53 % reported by Meersseman et al. [3], underlying the frequency of this phenomenon.

Second, they also confirm that C. albicans is the most frequently isolated yeast (56 %), in perfect agreement with Meersseman et al. (55 %)!

Unexpected, or at least unpublished so far, was the observation that Enterobacteriaceae were the most frequently isolated pathogens (25.1 %), quite ahead of P. aeruginosa (16.7 %) and way above S. aureus (13.3 %). These findings differ from the common knowledge that P. aeruginosa is the most frequent pathogen responsible for VAP and lend support to the growing importance in ICU infections of Enterobacteriaceae in general and E. coli in particular [9].

The authors also reported that 31.5 % of bacteria isolated in patients with Candida spp. colonization were multidrug resistant (MDR) versus 23.2 % in non-colonized patients (p = 0.13). In the univariate analysis, isolation of MDR bacteria was associated with antibiotic exposure. Since previous studies found an association between airway Candida colonization and previous antibiotic exposure [5, 10, 11], it is probable that the association between MDR bacteria and Candida colonization reported by Hamet et al. [8] is more likely due to shared risk factors than causal association.

Perhaps the most striking result of the Hamet et al. [8] study is the association found between Candida spp. airway colonization and increased mortality in mechanically ventilated immunocompetent critically-ill patients (44 % in colonized patients vs. 31 % in controls); this remained true after adjustment of other variables in a logarithmic regression model including SAPS II and age. This cannot be explained by actual fungal infection because no candidemia were observed during the whole period of this study. Even if no causal link can be raised, it is interesting to observe that previous clinical studies comparing the same set of patients (i.e., immunocompetent critically ill patients under mechanical ventilation with or without Candida colonization) showed the same trend. Azoulay et al. observed a mortality of 43 % in colonized patients and 36 % in controls (univariate analysis: p = 0.067) [5]. Delisle et al. found the same observation with a mortality of 34 vs. 21 % in colonized and control patients, respectively (p = 0.003) [10]. This was still the case in a logistic regression model that included age, comorbidities and admission for neurological disorders as co-variables.

So how can one explain this increased mortality in the context of airway Candida colonization? The first obvious answer is Candida colonization is simply a marker of severity that the retrospective design of the studies cited above cannot distinguish from cofounders despite ajustments performed in the analysis. Otherwise, two non-mutually exclusice hypotheses can be raised. The first is a modification in bacterial virulence and/or host immunity [1]. Back in the 1980s, Carlson showed that dual intraperitoneal injection of C. albicans and S. aureus at doses that separately caused no animal deaths resulted in 100 % mortality [12]. Interestingly, after this fungal-bacterial mixed inoculation, bacteria were highly predominant in peripheral organs. The latter was observed for gram-negative and gram-positive pathogenic bacteria, and even after a very small bacterial inoculum if the Candida inoculum was maintained [13]. These data suggest that the presence of C. albicans protects bacteria against normal clearance and enhances their virulence. In vitro, many studies have highlighted the importance of cross-kingdom communication, especially between fungus and bacteria [1]. If P. aeruginosa can influence the morphology of C. albicans with one of its quorum-sensing molecules (N-acyl homoserine lactones), the opposite is also true. Farnesol, which is C. albicans main quorum sensing molecule, is able to influence the motility and the expression of virulence factors of P. aeruginosa [14]. The second hypothesis is an indirect effect of Candida via the local and systemic inflammatory response to Candida colonization. Indeed, the presence of Candida spp. in airways of ventilated patients with suspected VAP (but without actual infection) is associated with an increased systemic inflammation as compared to control patients [11]. Inflammation has been associated with an increased risk of infection and mortality during ARDS [15], at least in part because of the enhanced growth, and possibly virulence, of pathogenic bacteria [15, 16]. Inflammation elicited by Candida colonization may also affect the immune defense, which in turn favors bacterial development. We developed a murine model of airway Candida colonization using single or iterative intratracheal instillations of clinical or laboratory strains of C. albicans [2, 17]. This model is consistently characterized by a Th1-Th17 immune response with very high pulmonary levels of interferon-gamma [17]. We showed that alveolar macrophages were unable to phagocyte gram-negative or gram-positive pathogenic bacteria after culture with this cytokine. Consequently, the incidence of bacterial pneumonia was higher in this context of fungal airway colonization as compared to control animals. Finally, an antifungal treatment was able to normalize the cytokine levels and prevent the development of bacterial pneumonia [17].

Taken together, Hamet et al.'s [8] study adds further evidence to the influence of Candida colonization in mechanically ventilated ICU patients. Clearly, Candida spp. should no longer be seen as a simple bystander and may constitute possible leads for new therapies to treat pneumonia [18, 19]. But in this case, the questions that remain unanswered are: should we give antifungal treatment to these colonized patients in order to decrease VAP and possibly mortality; when should we start it, and which drug should we use? But that’s another story.