There has been a surge in the worldwide use of over-the-counter probiotics in recent years in tandem with an increased understanding of the vital role of the gut microbiota on human health1. The commercial development of these formulations has progressed into multibillion-dollar endeavors, and the potential benefits of probiotics have been studied in a wide range of disorders. In the intensive care unit (ICU), studies using probiotics have specifically focused on their administration for the prevention of ventilator-associated pneumonia, pancreatitis and sepsis, as well as the treatment of neonatal late-onset sepsis (LOS)2,3,4. However, differences in the size and methodological rigor of the individual studies conducted, as well as large variations in the type, duration and dose of administered probiotic strains, continue to limit the strength of any conclusions regarding the effects of these formulations1.

Although that probiotics are generally considered safe, concerns have been raised that the administration of these live organisms could contribute to antibiotic resistance through horizontal transfer of antibiotic-resistance genes in the gastrointestinal environment5. In addition, administration of probiotic formulations has been associated with an increased risk of bacteremia in preterm neonates and ICU patients6. However, it has been challenging to determine a causal link between administration of probiotics with bacteremia, as these species—predominantly Lactobacillus spp. and Bifidobacterium spp.—are also common constituents of the human residential gut flora, which could therefore have been the underlying source of bacteremia, rather than probiotic administration itself. In this issue, two contributions shed further light on the contrasting roles of lactobacilli. While one study shows that Lactobacillus rhamnosus GG strains can indeed directly cause bacteremia and adaptively evolve in critically ill patients7, a second shows that the development of late-onset sepsis in a mouse model through intestinal overgrowth of Klebsiella pneumoniae can be prevented by administration of certain other Lactobacillus strains8.

Yelin and colleagues7 carried out a large epidemiological study using surveillance data that was collected from over 22,000 ICU patients at Boston Children’s Hospital between 2009 and 2014. A total of 522 ICU patients received the probiotic L. rhamnosus GG as part of their treatment (LGG), most commonly because they had been taking it prior to ICU admission. Of note, there were no ICU-specific guidelines for probiotic administration. The authors found that patients receiving this probiotic formulation had significantly higher risk of developing L. rhamnosus bacteremia. The authors then carried out whole-genome sequencing of the bacteria to determine strain-level similarity, and found that the blood isolates of L. rhamnosus were phylogenetically inseparable from probiotic isolates administered at that time, confirming that these probiotic strains were indeed the source of bacteremia. Additionally, the authors found some mutations in the blood isolates in which they differed from the probiotic bacteria, suggesting that the probiotic strains could have mutated within the host environment. Strikingly, in one instance they identified a de novo mutation conferring resistance to the antibiotic rifampin in an isolate from a patient who received concurrent treatment with LGG and with the rifampicin derivate rifaximin during the 3 months before developing bacteremia, indicating that this probiotic strain may be susceptible to environmental evolutionary pressure.

Lactobacillus bacteremia occurs infrequently in critical illness, and its clinical significance is poorly defined9. However, although Lactobacillus bacteremia was diagnosed in only 6 of 522 ICU patients (1.1%) receiving LGG-containing probiotics, only 2 of 21,652 ICU patients (0.009%) not receiving the LGG probiotic developed Lactobacillus bacteremia, indicating that the risk of bacteremia increased significantly upon probiotic administration. In addition, several key questions remain to be answered, most importantly which underlying mechanisms and risks drive progression to bacteremia, as none of these 6 critically ill patients had factors that would predispose them to infection, such as severe immunosuppression or signs of bowel disintegrity. Nonetheless, these findings fuel the need to develop a deeper mechanistic understanding of how live micro-organisms drive health-promoting effects in order to develop a new generation of probiotics that can be employed with less potential for unwanted consequences.

In another study in this issue, Singer and colleagues8 investigate LOS, a neonatal infection in premature infants that is thought to be caused by commensal bacteria and fungi, using a model system in mice. They colonized the gut of neonatal mice with bioluminescent and fluorescent strains of Klebsiella pneumonia so that they could be traced. First, the authors showed that gut overgrowth with K. pneumoniae preceded the spread of the pathogen into the systemic circulation, subsequently leading to extra-intestinal infection and LOS. Then, using a combination of germ-free animals without a microbiome and mouse models in which dams were given vancomycin and gentamycin to alter the microbiome inherited by their pups, they found that pups in which Lactobacillus murinus was largely eradicated had a higher susceptibility to K. pneumoniae overgrowth and subsequent sepsis, whereas enrichment with L. murinus conferred relative protection. These findings were confirmed by the observation that probiotic administration of L. murinus, and to a lesser extent Lactobacillus johnsonii—but not LGG—provided resistance to K. pneumoniae colonization. Of interest, a single probiotic strain of Escherichia coli was similarly effective in preventing dysbiosis, showing that the efficacy of probiotics in preventing infection was not limited to lactobacilli. A later series of experiments suggested that low microbiota diversity and elevated intestinal oxygen levels in the immature neonatal gut were major drivers of K. pneumoniae overgrowth and risk of LOS, whereas older pups with more mature microbiome communities—containing anaerobic bacteria—displayed resistance to the disease.

These two contributions7,8 shed light on the potential therapeutic uses and risks of probiotics. One important takeaway is that host–microbiota interactions remain extremely complex (Fig. 1) and are likely to be specific to both context (e.g., prior antibiotic exposure, intestinal maturation and integrity, age) and disease. Therefore, a one-size-fits-all approach to treating patients with probiotics is not only potentially limited in efficacy, but also likely to be harmful. We foresee that these findings, along with recent discoveries that the commensals Blautia producta and Clostridium scindens are capable of providing colonization resistance against vancomycin-resistant Enterococcus10 and Clostridium difficile11, respectively, will provide further grounds for recommending that untargeted administration of probiotics should be supplanted by tailored microbiome therapy. This also extends to identifying microbiota-derived secreted factors, cell components and metabolites that could help overcome the challenges of introducing live organisms into the host environment12. It is clear that human studies and randomized clinical trials, with an emphasis on therapeutic reproducibility and patient safety, should be highly anticipated as means to translate these exciting preclinical steps into new therapeutic applications.

Fig. 1: Understanding the contrasting roles of probiotics in critical illness.
figure 1

(a) Yelin and colleagues7 show in humans that L. rhamnosus GG strains can directly cause bacteremia and adaptively evolve in critically ill patients becoming antibiotic resistant. (b) Singer and colleagues8 show that the development of late-onset sepsis due to oxygen-dependent overgrowth of K. pneumoniae can be prevented by certain Lactobacillus strains. They found that aerobic conditions in the immature neonatal gut were major drivers of K. pneumoniae overgrowth, whereas older pups with more mature microbiome communities—containing anaerobic bacteria—had resistance to the disease. However, administration of mature microbiota to neonatal pups led to limited intestinal engraftment of anaerobic communities. Abx, antibiotics; E. coli, Escherichia coli; K. pneumoniae, Klebsiella pneumoniae; L. murinus, Lactobacillus murinus; LGG, Lactobacillus rhamnosus GG; LOS, late-onset sepsis.

Full size image