The leaky gut and microbiome in critical illness: emerging insights into microbial “translocation”

Abstract

Purpose of review

Microbial translocation has long been viewed as a nonspecific consequence of gut barrier failure in critical illness, with downstream effects attributed primarily to the host immune response. This review examines emerging evidence that the gut microbiome plays a more active and specific role in this process than previously appreciated.

Recent findings

Dysbiosis during critical illness directly contributes to barrier dysfunction through depletion of metabolites that sustain epithelial integrity. Culture-independent approaches have revealed that gut-derived organisms are a major reservoir for secondary infection, with translocation governed by microbial virulence, community dynamics, and immune cell-mediated transport rather than barrier permeability alone. In parallel, organism-specific microbial components - including structurally diverse forms of lipopolysaccharide and bacterial DNA detected across multiple blood fractions - enter the circulation and differentially modulate host immune responses. Recent studies link circulating microbial DNA composition to distinct inflammatory phenotypes in sepsis and acute respiratory distress syndrome, suggesting these signals contribute to clinical heterogeneity.

Summary

These findings support a revised framework in which translocation reflects the composition of the dysbiotic gut, not barrier integrity alone. Integrating microbial data with host phenotyping may enable more precise risk stratification and microbiome-informed therapeutic strategies in critical illness.

INTRODUCTION

Microbial translocation – the passage of microorganisms and microbial components from the gut lumen into the systemic circulation – has been implicated as a driver of organ dysfunction and secondary infection for decades [1–3]. Traditionally, this phenomenon has been viewed through a host-focused lens, in which inflammation disrupts gut barrier integrity and permits entry of nonspecific gut-derived factors with downstream effects determined by host immune activation. Growing insights into the gut microbiome and the application of culture-independent techniques now support a more integrated understanding. In addition to being affected by critical illness, the microbiome contributes to the maintenance of barrier integrity and shapes the composition of microbial signals that may access the circulation [4–6]. These observations suggest that translocation is not solely a consequence of barrier failure, but shaped by the specific organisms present in the dysbiotic gut and their interactions with host physiology. In this review, we highlight recent advances that have contributed to this shift in perspective. We first examine how dysbiosis contributes to barrier dysfunction, emphasizing that derangements in the microbiome may directly impair barrier integrity. We then review evidence that the gut serves as a reservoir for secondary infection by specific pathogens, and examine how microbial components-including lipopolysaccharide (LPS) and bacterial DNA identified by culture-independent approaches-exert organism-specific effects on host immune responses. Together, these findings support an updated view of translocation that incorporates microbial composition, barrier dynamics, and mechanisms of microbial signal trafficking, with implications for immune responses and clinical outcomes in the intensive care unit (ICU).

GUT MICROBIAL DYSBIOSIS DRIVES BARRIER DYSFUNCTION

The gut barrier is a multilayered ecosystem consisting of the microbiota, metabolites (dietary and those ingested and produced by the microbiota), mucus, epithelial cells joined by tight junctions, antimicrobial peptides, secretory immunoglobulin A (IgA) (sIgA), gut-associated lymphoid tissue (GALT), and coordinated immune signaling networks [6–8]. Under physiologic conditions, the gut microbiota and metabolites they produce are central to maintaining gut barrier integrity and immune homeostasis. Example gut microbiota-associated metabolites include short-chain fatty acids (SCFAs), tryptophan-derived indoles, and bile acids. SCFAs (e.g., acetate, propionate, and butyrate) are produced through microbial fermentation of dietary fiber, and play a role in protecting the gut lining from injury by reinforcing tight junctions, limiting autophagy, and modulating mucosal immunity through effects on T-cell differentiation, antibody production, and cytokine signaling [6,8–10]. Tryptophan-derived indoles maintain epithelial regeneration through aryl hydrocarbon receptor (AhR)-mediated signaling, complementing the role of SCFAs by enhancing epithelial integrity and shaping host immune responses. Bile acids contribute to gut barrier regulation by exerting context-dependent effects on permeability and inflammation after conversion from primary to secondary forms by gut microbes (Fig. 1) [11,12]. In the ICU, this ecosystem transforms into a dysfunctional driver of systemic inflammation and organ failure [6,8,13,14]. Inflammation and exposures including antibiotics, alterations in enteral nutrition, opioids, and other therapies alter composition, favoring expansion of pathobionts such as Enterobacteriaceae and Enterococcus over the obligate anaerobic commensals responsible for secondary bile acid conversion and SCFA production [15,16,17▪,18]. The loss of butyrate-producing species further increases luminal oxygen by impairing colonocyte metabolism, creating a self-reinforcing cycle of pathobiont expansion and epithelial injury [18–20].

FIGURE 1.

Gut barrier dynamics in a healthy state vs. critical illness. LPS, lipopolysaccharide.

Recent clinical studies reinforce the link between dysbiosis and barrier dysfunction and suggest that microbiota-targeted approaches may mitigate this process. Kimura et al. used a novel endoscopic impedance technique to measure intestinal permeability in real time in patients with type 2 diabetes and found that lower ileal impedance-indicating greater permeability-correlated with alterations in microbial composition and reduced SCFA metabolism [21▪]. A recent meta-analysis demonstrated probiotic, synbiotic, and prebiotic interventions reduced circulating markers of endotoxemia (LPS binding protein i.e. LBP) and barrier disruption (zonulin) across clinical trials [22▪], and a randomized trial in critically ill children with sepsis found that probiotic therapy reduced inflammatory markers with trends toward improved barrier biomarkers [23▪]. While direct evidence linking dysbiosis to barrier dysfunction in adult ICU populations remains limited, these findings collectively support the concept that barrier integrity is not solely a host-mediated process but is shaped by microbial composition-and potentially amenable to microbiome-directed intervention.

TOWARDS A BETTER UNDERSTANDING OF “TRANSLOCATION”

Beyond the role of the microbiome in maintaining gut barrier defenses, the advent of culture-independent sequencing has enabled more detailed characterization of microbial material in compartments traditionally considered sterile, including blood. These observations challenge conventional assumptions about sterility and suggest that microbial translocation may be more prevalent and mechanistically diverse than previously appreciated. Broadly, this work highlights two related insights: first, that the gut can serve as a reservoir for primary and secondary infection even in the absence of overt barrier injury; and second, that organism-specific microbial components (e.g., LPS and bacterial DNA) entering the circulation can directly influence immune activation and contribute to clinical disease phenotypes.

Live organisms and the gut as a reservoir of infection

Across diverse patient populations, organisms residing within the gastrointestinal (GI) tract have been repeatedly identified as sources of infection at extraintestinal sites. In patients undergoing hematopoietic stem cell transplant (HCT), mucosal barrier injury and subsequent bacteremia are well recognized complications, supported by studies demonstrating that strains genetically identical to bloodstream pathogens can be detected in patients’ stool weeks prior to infection [24,25]. Siranosian et al. further demonstrated that these infections are predominantly endogenous: in a cohort of 149 HCT recipients, pathogen acquisition from hospital contacts was rare, while opportunistic pathogens within the gut remained relatively stable over the course of hospitalization [26]. Similar observations have been reported in surgical populations. Long et al. found that 86% of postoperative wound infections were caused by bacteria present in the patient’s preoperative microbiome, with 59% resistant to perioperative prophylaxis [27▪]. Even organisms typically considered commensal may contribute under specific conditions. In a landmark study, a strain derived from a probiotic Lactobacillus formulation administered during hospitalization was identified as the cause of bacteremia in pediatric patients [28]. Together, these findings support the concept that the gut microbiota serves as a persistent and patient-specific reservoir for subsequent infection.

The mechanisms driving gut-derived infection remain incompletely defined. A recent study found no association between indirect markers of gut barrier injury (intestinal fatty acid–binding protein, trefoil factor-3, and citrulline) and ICU-acquired enterococcal bacteremia [29▪]. While these biomarkers have known limitations in sensitivity [30,31], such findings add to a growing body of evidence arguing against a purely linear relationship between barrier dysfunction and translocation. Two potential-and likely overlappingmechanisms may explain this. First, microbial virulence is not a fixed attribute, but is shaped by host and environmental context. Alverdy et al. demonstrated that stress-associated signals during critical illness – including opioids, ischemia, and phosphate depletion – can induce a shift in commensal organisms toward hypervirulent phenotypes capable of causing lethal gut-derived sepsis following otherwise sterile surgical injury [32]. Complementing this, Stoma et al. showed that in HCT recipients, the risk of Gram-negative bloodstream infection was better predicted by compositional flux within the gut microbiota-dynamic changes in community structure over time-than by single timepoint measures of diversity or domination [33]. Together, these findings suggest that progression to bacteremia depends not simply on the presence of a pathobiont, but also on ecological reorganization and activation of virulent phenotypes that enable specific organisms to expand and access the circulation. Second, certain organisms may bypass mucosal defenses altogether through immune cell–mediated transport. For example, dissemination of Salmonella enterica from the gut occurs through both autonomous migration and dendritic cell–mediated transport to draining mesenteric lymph nodes [34], while methicillin-resistant Staphylococcus aureus (MRSA), internalized by gut neutrophils, can be transported to distal surgical sites to cause localized infection without ever resulting in overt bacteremia [35]. This mode of dissemination appears to be facilitated by organism-specific virulence traits. For example, in in vitro models of human infection, S. aureus – but not other Staphylococcus species – was found to survive intracellularly within macrophages through pH resistance, enabling migration across sites [36]. Together, these observations suggest that gut-derived infection reflects not only barrier integrity, but dynamic interactions between microbial behavior, community structure, and host physiology. Identifying and defining these relationships will be critical for developing risk stratification strategies and targeted interventions to reduce secondary infection in critically ill patients.

Characterizing the blood microbial landscape: gut-derived lipopolysaccharide and bacterial DNA

Complementing studies of viable organisms, culture-independent approaches have enabled characterization of microbial signals present in the circulation. The techniques and terminology used to study these signals vary, and individual approaches often capture distinct-though overlapping-components of the circulating microbial landscape. Here, we focus on two well studied classes of signals: LPS, a glycolipid component of Gram-negative bacteria, and bacterial DNA signatures identified through sequencing-based methods, which may reflect both viable organisms and nonviable microbial material.

Immunogenicity of lipopolysaccharide is species-specific

LPS has long been considered the prototypical translocating gut microbial inflammatory signal. Foundational studies of LPS derived from E. coli and other Proteobacteria have shown that binding to toll-like receptor 4 (TLR4) activates nuclear factor κB and interferon regulatory transcription factors, promoting inflammasome activation and contributing to inflammatory responses and organ dysfunction in sepsis [37,38]. However, it is now clear that the immunogenicity of LPS-largely determined by its Lipid A structure-is not uniform across bacterial species.

Lipid A consists of a glucosamine disaccharide acylated with fatty acid chains, and structural variation in this domain dictates TLR4 activation. The hexa-acylated form of Lipid A, characteristic of Proteobacteria, elicits a prototypical pro-inflammatory response and is the form captured by clinical assays of “endotoxin.” In contrast, other Gram-negative species, including many commensal gut microbes, produce structurally distinct forms of LPS that act as weak agonists or even antagonists of TLR4 [39▪, 40,41]. In functional profiling studies of healthy stool, the majority of LPS has been attributed to Bacteroidales, whose LPS can antagonize the immunostimulatory effects of Proteobacteria-derived LPS-a mechanism thought to contribute to immune homeostasis [42]. Structural studies further support this heterogeneity: cryo-EM analyses of lipid A–TLR4 interactions demonstrate distinct binding modes with potential therapeutic implications [43▪▪].

These distinctions extend to our interpretation of endotoxemia and widely used markers like LPS binding protein (LBP) in clinical settings. In a study of postmenopausal women, elevated plasma LBP and anti-LPS IgA levels were associated with increased abundance of gut Proteobacteria and microbial expression of lipid A biosynthesis genes (lpxA, lpxB) [44▪]. In critical illness, LBP functions as an acute phase protein and prognostic marker in both infectious and noninfectious states [45], and is frequently used as an indirect marker of gut barrier “leakiness” [46]. However, rather than solely reflecting permeability, these findings suggest that LBP may also signal a shift in gut microbial composition toward Proteobacteria and other organisms whose LPS is more immunostimulatory.

BACTERIAL DNA IN BLOOD

Unlike LPS and other microbial components shared across bacterial species, bacterial DNA captures the genetic signature, or fingerprint, of specific organisms. The principal sequencing methods for detecting circulating microbial DNA are summarized in Table 1 (reviewed in detail by Wensel et al.[47]). Application of these approaches to blood and other typically sterile (“low biomass”) sites requires careful consideration, as these factors fundamentally shape how translocation is defined and interpreted. Unlike DNA in stool, which largely reflects live organisms, bacterial DNA in blood may derive from viable cells (freely circulating or within or bound to host cells, including leukocytes) and nonviable microbial fragments. Fragments can exist unbound in plasma (plasma cell-free DNA) or be shuttled as cargo by host cells – including erythrocytes, leukocytes, and neutrophil extracellular traps (NETs) – or within bacterial extracellular vesicles (BEVs) [48,49]. As a result, the microbial DNA detected is influenced by collection and processing strategies, including selection of blood fractions [plasma cell-free vs. whole blood (also termed whole cell)], enzymatic treatments, and isolation approaches targeting vesicle-associated material. In addition, sequencing of low-biomass samples inherently includes contaminant DNA, necessitating rigorous controls and bio-informatic filtering to distinguish biological signal from background noise [50▪].

Table 1.

Open-ended sequencing methods for detection of bacterial DNA

Platform features	16S rRNA PCR with Sanger sequencing	16S rRNA amplicon sequencing (short read)	Full length 16S rRNA gene Sequencing (Long Read)	Cell-Free mNGS sequencing (shotgun metagenomic)	Whole-cell mNGS sequencing (shotgun metagenomic)
Target	Near-full-length 16S rRNA gene (~1400 bp) from cultured isolates	Variable regions (V1-V9, ~300 bp) of 16S rRNA gene	Full-length 16S rRNA gene (~1500 bp) or 16S-ITS-23S operon (~4500 bp)	Fragmented DNA circulating freely in plasma/body fluids	All DNA in sample including cell-free, intact organisms (free and bound), and cell-associated bacterial fragments
Organisms detected	Bacteria, archaea	Bacteria, archaea	Bacteria, archaea; fungi only if ITS region is included in operon sequencing.	Bacteria, fungi, DNA viruses, parasites	Bacteria, fungi, DNA viruses, parasites
Typical taxonomic resolution	Species level (limited for closely related species)	Genus level (species for some taxa depending on region and database)	Genus/Species level	Species/strain level	Species/strain level
Common sequencing platforms	Capillary electrophoresis	Illumina (MiSeq, NextSeq), Ion Torrent	PacBio SMRT, Oxford Nanopore (MinION)	Illumina, Nanopore, Ion Torrent	Illumina, Nanopore, Ion Torrent
Primer-specific amplification bias	Yes	Yes (primer selection, amplification bias)	Yes, but reduced by targeting conserved flanking regions	Less than 16S, risk for amplification bias in library prep remains	Less than 16S, risk for amplification bias in library prep remains
Functional information	No	No (requires inference)	No (requires inference)	Yes (resistance genes, virulence factors)	Yes (resistance genes, virulence factors)
Host DNA interferencea	Minimal	Minimal (bacteria-specific primers)	Minimal (bacteria-specific primers)	High	Moderate/High
Sensitivity for pathogens	High for cultured isolates; limited by culture requirements.	Moderate	Moderate	High for viruses, fungi, and bacteria that are being shedding DNA or actively killed/degraded; useful for monitoring treatment response	High overall, including for viable and cell-associated organisms; may miss low abundance taxa due to host background
Polymicrobial detection	No	Yes	Yes	Yes	Yes
Best use case	Species-level identification from cultured isolates; reference-standard for taxonomic confirmation	Community-level profiling; work-horse in microbiome studies especially large-scale or cost-constrained studies	Community profiling; improved resolution over short-read for species-level or closely related taxa	Hypothesis-free detection of organisms; enhanced sensitivity for organisms not captured by amplicon approaches; accessible from banked plasma; treatment monitoring	Hypothesis-free detection of organisms; captures viable and cell-associated organisms including those with low DNA shedding
Applied in clinical practice?	Yes – standard for identification of cultured isolates in clinical microbiology laboratoriesb	Limited – primarily research; not routinely used in clinical diagnostics	Limited – emerging in clinical research settings	Yes – commercial platforms available (e.g., Karius Test) for bloodstream and deep-seated infections	Limited – primarily research and select reference laboratories

bp, base pairs; cfDNA, cell free deoxyribonucleic acid; DNA, deoxyribonucleic acid; NGS, next generation sequencing; RNA, ribonucleic acid.

All tests provide qualitative information (relative abundance) but not absolute quantification.

^aHost DNA interference varies substantially by sample type, processing method, and use of host depletion techniques.

^bMultiplex PCR platforms (e.g., BioFire FilmArray) are widely used in clinical practice for pathogen detection but are not included in this table as they use targeted amplification of predefined pathogen panels rather than open-ended approaches.

Despite these challenges, bacterial DNA derived from commensal microbes has been consistently detected in the circulation of healthy individuals [51]. In a large-scale study of nearly 10 000 adults, over 100 microbial species were identified in blood, the majority of which were commensals associated with mucosal surfaces rather than environmental contaminants. Shared species were present in fewer than 5% of individuals, arguing against a stable “blood microbiome” and instead supporting a model of transient microbial trafficking [52]. Paired analyses of gut and blood samples further suggest that this process is selective rather than indiscriminate. Zhai et al. demonstrated that circulating microbial DNA was enriched for Gram-negative and aerobic or facultatively anaerobic organisms, with only limited overlap (~7%) with stool microbial profiles, indicating that translocation likely reflects specific microbial and host factors rather than stochastic leakage [53▪]. Complementary analyses of blood fractions demonstrate that the majority of microbial DNA is cell-associated, localizing predominantly to erythrocytes and the buffy coat, with only a small fraction (~0.03%) present as plasma cell-free DNA [49].

In critically ill patients, a large body of literature has focused on the application of metagenomic next-generation sequencing (mNGS) approaches for pathogen detection in blood and other body fluids [54]. Meta-analyses demonstrate higher circulating mcfDNA levels in nonsurvivors of sepsis compared with survivors [55,56]. A single measurement at ICU admission predicted 28-day mortality with an area under the curve (AUC) of 0.79, outperforming APACHE II scores (AUC 0.68) [57], with integrative models incorporating mcfDNA concentration, fragmentation patterns, and microbial composition further improving predictive performance, achieving AUCs of 0.992 for sepsis diagnosis and 0.802 for mortality [58].

Beyond their diagnostic utility, circulating bacterial DNA and its trafficking pathways may provide a mechanistic link between translocation and heterogeneous immune responses in diverse critical illness syndromes. Immune cells expressing nucleic acid–sensing receptors, including TLR9, can bind extracellular microbial DNA, promoting chemokine production, cellular trafficking, and enhanced effector function. Neyton et al. used paired host transcriptomic and plasma cell-free metagenomic sequencing in 189 critically ill patients with sepsis and acute respiratory distress syndrome (ARDS) and found that the hyperinflammatory phenotype was characterized by increased circulating Enterobacteriaceae DNA and elevated innate immune gene expression, while the hypoinflammatory phenotype was enriched in adaptive immunity pathways-providing direct evidence that circulating microbial composition differs across established critical care endotypes [59▪▪]. Separately, Lam et al. demonstrated that the composition of RBC-bound microbial DNA correlated with circulating IL-6 levels independent of illness severity [60▪▪], suggesting that microbial signals across multiple blood fractions may contribute to shaping host inflammatory responses.

Finally, BEVs represent an additional microbiota-mediated mechanism by which microbial DNA and associated cargo are disseminated systemically. These nanoscale particles, generated through membrane blebbing or bacterial lysis, encapsulate DNA, lipids, proteins, and metabolites, enabling protected transport to distant tissues. In conditions characterized by gut barrier disruption-including inflammatory bowel disease, HIV, and chemotherapy-induced mucositis-BEV-associated LPS has been identified as a dominant driver of systemic TLR4 activation and correlates with markers of barrier dysfunction such as zonulin [61]. In Gram-positive organisms, vesicle formation is less well characterized but BEVs containing lipoteichoic acid and peptidoglycan have been shown to engage host immune pathways via TLR2 [62,63▪].

Taken together, these studies suggest a conceptual shift from traditional notions of “single microbe/single disease” to one that additionally accounts for circulating bacterial DNA as dynamic and biologically active signals that have broader roles in immune responses and disease trajectory.

CONCLUSION

The concept of microbial translocation in critical illness has evolved from a nonspecific consequence of barrier failure to a dynamic, microbiome-dependent process that shapes host response. Disruption of the gut microbiota-driven by inflammation, antibiotics, and altered nutrient delivery-not only weakens barrier integrity but also determines the composition of microbial signals entering the circulation. In this framework, the gut functions as both a source and a filter, influencing not only whether translocation occurs, but which organisms and microbial products predominate. Importantly, translocation is not a single entity. Viable organisms may access extraintestinal sites through barrier disruption or host-mediated transport, contributing to primary and secondary infections. In parallel, microbial components-including LPS, bacterial DNA, and vesicle-associated cargo-enter the circulation through distinct pathways and act as immunologically active signals. These signals are not interchangeable: their effects depend on microbial origin, structural features, and mode of delivery, providing a mechanistic basis for the heterogeneity of immune responses observed in critically ill patients.

This perspective supports a revised model in which the gut microbiome shapes systemic physiology through its influence on both barrier function and the composition of circulating microbial signals. For the practicing intensivist, these signals may offer insight into patient-specific host–microbe interactions not captured by conventional diagnostics. At the same time, the limited success of broadly applied microbiome-targeted therapies in the ICU likely reflects a failure to account for this specificity. The field now faces a critical next step: moving from descriptive associations toward mechanistically informed, patient-specific strategies. Defining which microbial signals are present, where they originate, and how they influence immune trajectories will be essential to translating these insights into improved outcomes. Achieving this will require deliberate investment in sample biobanking – particularly in studies of immune endotypes and critical illness phenotyping – to enable retrospective characterization of host-microbe interactions in well defined clinical cohorts. With this updated view, the question is no longer whether the gut matters in critical illness, but which microbial signals, in which patients, and at which time point can be leveraged to predict trajectory and guide intervention.

KEY POINTS.

• Dysbiosis during critical illness directly impairs gut barrier integrity through loss of barrier-sustaining metabolites and expansion of pathobionts, creating a self-reinforcing cycle of epithelial injury that may be amenable to microbiome-directed intervention.

• Secondary infections in critically ill patients frequently originate from endogenous gut organisms whose translocation depends on microbial virulence properties, community dynamics, and immune cell-mediated transport rather than barrier permeability alone.

• The immunogenicity of translocating lipopolysaccharide (LPS) is species-specific with Proteobacteria-derived LPS driving pro-inflammatory toll-like receptor 4 activation while LPS from certain commensals (Bacteroidales) exerts antagonistic effects-implying that shifts in gut microbial composition qualitatively alter the inflammatory character of endotoxemia.

• Circulating bacterial DNA detected across multiple blood fractions-including plasma cell-free DNA and RBC-bound DNA-differs in composition between critical illness phenotypes and correlates with inflammatory markers independent of illness severity, suggesting that microbial signals shape host immune trajectories.

• Advancing the field will require standardized approaches to characterizing circulating microbial signals, deliberate biobanking in critically ill cohorts, and integration of microbial data with host immune phenotyping.