Bloodstream infections: mechanisms of pathogenesis and opportunities for intervention

Abstract

Bloodstream infections (BSIs) are common in hospitals, often life-threatening and increasing in prevalence. Microorganisms in the blood are usually rapidly cleared by the immune system and filtering organs but, in some cases, they can cause an acute infection and trigger sepsis, a systemic response to infection that leads to circulatory collapse, multiorgan dysfunction and death. Most BSIs are caused by bacteria, although fungi also contribute to a substantial portion of cases. Escherichia coli, Staphylococcus aureus, coagulase-negative Staphylococcus, Klebsiella pneumoniae and Candida albicans are leading causes of BSIs, although their prevalence depends on patient demographics and geographical region. Each species is equipped with unique factors that aid in the colonization of initial sites and dissemination and survival in the blood, and these factors represent potential opportunities for interventions. As many pathogens become increasingly resistant to antimicrobials, new approaches to diagnose and treat BSIs at all stages of infection are urgently needed. In this Review, we explore the prevalence of major BSI pathogens, prominent mechanisms of BSI pathogenesis, opportunities for prevention and diagnosis, and treatment options.

Introduction

Bloodstream infections (BSIs) are a life-threatening condition with high prevalence worldwide^1–5, with an estimated 250,000 cases of BSI reported in hospitals yearly, and the rates are increasing. About half of hospital-associated BSIs occur in intensive care units, with other wards, including internal medicine, surgery and paediatrics, also commonly reporting these infections^3,6. The annual incidence of BSIs is estimated at 150 cases per 100,000 population, with a 17% crude mortality defined as death within 30 days of a positive blood culture⁵. Due to the substantial links between BSIs and other infections, such as pneumonia and urinary tract infections, increased antimicrobial resistance (AMR), and the direct associations with sepsis, BSIs are a public health concern. The leading causes of BSIs are the bacterial pathogens Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, coagulase-negative Staphylococcus (CoNS) species and the fungal pathogen Candida albicans. BSIs can be caused by multiple species, originate from various primary sites, lead to secondary infections and initiate different degrees of host responses (Fig. 1).

Fig. 1 |. Primary sites that can seed bloodstream infections and opportunities for prevention and treatment.

Shown are the various primary sites across the body that can seed bloodstream infections. Potential opportunities for the prevention and treatment of bloodstream infection originating at each site are listed.

Most BSIs are caused by bacteria^3,5. Bacteraemia is defined as the presence of viable bacteria in the blood and can range from a transient, asymptomatic infection to a life-threatening BSI⁷. Thus, bacteraemia encompasses the presence of bacteria in the blood, regardless of whether or not this event leads to host responses, damage or disease⁸. For example, routine dental cleanings can introduce members of the oral microbiome into the blood through mechanical disruption of vessel barriers⁹. Bacteraemia can also be initiated through minor skin wounds, like a cut or scrape. Bacteraemia caused by such events can be rapidly cleared from the bloodstream either through microbial inability to adapt to this niche or through host responses that effectively eliminate the invading pathogen. In asymptomatic bacteraemia, this response results in minimal damage to the host⁸.

In some cases, bacteraemia can lead to a BSI, defined clinically as a positive blood culture from a patient with signs and symptoms of systemic infection¹⁰. These infections encompass a range of clinical presentations that require unique and rapid interventions. In terms of pathogenesis, BSIs may involve bacterial growth within vessels and the heart, colonization of secondary sites, and perpetuation of disease by other affected tissues¹¹. Members of the Staphylococcus genus can exhibit high bacterial burden during endovascular infection through foci formation that adhere to vessel walls¹². By contrast, Gram-negative species are associated with lower bloodstream abundance. Despite varying circulatory abundance, S. aureus, K. pneumoniae and related pathogens can have harmful consequences, including abscesses, endophthalmitis and meningitis^13–17.

BSIs can trigger a rapid and intense host immune response and lead to sepsis, which is defined as immune dysregulation in response to an infection. Sepsis is a leading cause of global morbidity and mortality^18,19. The current diagnosis of sepsis is predicated on bedside assessments of clinical instability (high respiratory rate, low blood pressure and altered mentation), coupled with evidence of end-organ dysfunction. Septic shock is defined as sepsis coupled with circulatory shock, where abnormalities in metabolic and cellular functions increase mortality risk. Although sepsis occurs in response to infection, detection of the infectious agent is not required for the diagnosis. Given the high mortality currently associated with BSIs, the continued increase in AMR among causative pathogens, and the predicted impact of resistant infections on global mortality in the coming decades, new preventative, diagnostic and therapeutic approaches for such infections are needed^20,21.

In this Review, we discuss the major pathogens, pathogenesis, risk factors, prevention and treatment of BSIs. The aim of this article is to highlight fundamental concepts about BSIs that can guide future interventions to prevent and treat them. We review the incidence, epidemiology and mortality rates of major pathogens that cause BSIs and highlight examples of pathogenesis mechanisms that enable progression from initial site colonization to dissemination, survival and replication in the bloodstream. To identify opportunities for improvement in the management of BSIs (Fig. 1), we explore risk factors for infection and current methods of prevention, detection and treatment. As these topics are broad, multifactorial and covered individually elsewhere^2,11,22, our goal is to provide a framework for understanding BSIs that can be used to synthesize clinical and basic research. We do not intend to provide an extensive review of pathogenesis or clinical care but, instead, a balanced overview at the interface of these disciplines. We conclude with opportunities for translational research to improve, prevent and manage deadly BSIs.

Major pathogens in BSIs

Pathogens that cause BSIs have been ranked by prevalence in multiple studies, encompassing broad patient cohorts and ranging across many centres worldwide^2–5 (Table 1). Notably, rates of BSI prevalence for individual species vary based on the source of infection and whether the infection was acquired in the community or in the health-care setting. Most BSIs are monomicrobial, with polymicrobial infections accounting for 5–13% of clinical cases^1,3,4. Although the major species that cause BSIs are consistent, frequencies of infection vary across populations with risk factors and also geographically. These frequencies are based on culture results, and improved molecular detection is likely to highlight previously underdiagnosed pathogens. In the early 2000s, S. aureus was the leading cause of BSIs with E. coli closely following as the second most common pathogen². Specifically, in studies prior to 2004, S. aureus accounted for ~20–22% of BSIs and E. coli for 5–20%^2,3. This trend subtly shifted after 2004, as E. coli became the leading BSI pathogen^1,2 — E. coli then accounted for 20–27% of BSIs and S. aureus for ~19–25%^2,5. Combined, E. coli and S. aureus cause half of clinical BSIs, making these species major areas of interest for future interventions. It is unknown whether BSIs with certain species are more likely to initiate sepsis. Identifying such trends would be difficult given that the infectious agent in sepsis is often not found.

Table 1 |.

Pathogens that are consistently associated with high prevalence across cohorts in bloodstream infections

Pathogen	Frequency (%)a	Crude mortality (%)a	Common initial sites	Associated clinical syndromes	Preferred antimicrobial therapyb
Escherichia coli	5.6–26.9	12.1–22.4	Urinary tract, gut	Pyelonephritis, cholangitis, intra-abdominal abscess, CLABSI	β-Lactams (carbapenems if extended-spectrum β-lactamases present), fluoroquinolones, trimethoprim-sulfamethoxazole
Staphylococcus aureus	15.4–20.7	22.8–31.0	Skin, nares, indwelling device	CLABSI, endocarditis, vascular graft infection, skin and soft tissue infection	First-generation cephalosporins, antistaphylococcal penicillins, vancomycin, linezolid, daptomycin, ceftaroline
Coagulase-negative Staphylococcus	9.2–31.3	19.7–20.7	Skin, indwelling device	CLABSI, prosthetic valve endocarditis, vascular graft infection	First-generation cephalosporins, antistaphylococcal penicillins, vancomycin
Klebsiella species	4.8–13.2	17.6–34.7	Lungs, gut, urinary tract	Pyelonephritis, cholangitis, intra-abdominal abscess, CLABSI, pneumonia	β-Lactams (carbapenems if extended-spectrum β-lactamases present), fluoroquinolones, trimethoprim-sulfamethoxazole
Candida species	2.6–9.0	32.0–39.2	Skin, indwelling device, gut	CLABSI, intra-abdominal abscess, endocarditis	Echinocandins (azoles for step-down therapy)
Pseudomonas aeruginosa	3.4–8.9	24.7–48.9	Lung, indwelling device	Pneumonia, CLABSI	Variable (piperacillin-tazobactam, cefepime, carbapenems, ceftazidime-avibactam, ceftolozane-tazobactam, cefiderocol, plazomicin, fluoroquinolones)
Enterococcus species	2.8–9.4	23.6–49.5	Gut, urinary tract	Cholangitis, intra-abdominal abscess, CLABSI, endocarditis	Ampicillin (linezolid or daptomycin if ampicillin/vancomycin resistant)
Acinetobacter baumannii	1.3–12.5	34.0–52.1	Skin, indwelling device, lungs	Pneumonia, CLABSI	Variable (ampicillin-sulbactam, piperacillin-tazobactam, cefepime, carbapenems, cefiderocol, plazomicin, eravacycline, tetracyclines)
Enterobacter species	2.6–6.1	19.8–30.2	Gut	Pyelonephritis, cholangitis, intra-abdominal abscess, CLABSI	β-Lactams (cefepime or carbapenems if Enterobacter cloacae), fluoroquinolones, trimethoprim-sulfamethoxazole
Citrobacter species	~1.7%	Not available	Gut	Cholangitis, intra-abdominal abscess	β-Lactams (cefepime or carbapenems if Citrobacter koseri has been identified), fluoroquinolones, trimethoprim-sulfamethoxazole

CLABSI, central line-associated bloodstream infection.

^aData represent ranges for selected studies, when available. The selected studies included cohorts with nosocomial infections only or infections from multiple sources, including community, health care-associated and nosocomial origin.

^bPreferred therapy may vary by region based on epidemiology and practice guidelines.

The leading causes of Gram-positive bacteraemia are S. aureus, a coagulase-positive facultative anaerobic coccus, followed by CoNS. S. aureus can infect the heart valves, which can lead to infective endocarditis, and spread to prosthetic materials in the vasculature and joints²³. S. aureus is flexible in its niche and colonizes many different body sites, particularly the skin or nasopharynx of nearly 20% of asymptomatic individuals^2,3,5,17. Although S. aureus can live commensally in the nares, it is a common source of skin and soft tissue infections, such as purulent cellulitis, soft tissue or intramuscular abscesses, and necrotizing fasciitis, and these infections can be the source of S. aureus BSI. However, the primary infection source is often difficult to identify and, in these cases, invasion from a central line or skin is typically presumed to be the cause²⁴. CoNS are a diverse group of skin commensal species that are usually less pathogenic than S. aureus²⁵. However, Staphylococcus lugdunensis is a CoNS that possesses similar virulence features as S. aureus, including the ability to establish infective endocarditis⁵. CoNS such as Staphylococcus epidermidis are common skin commensals but also cause prosthetic valve endocarditis and catheter-related BSIs. CoNS BSIs are particularly prevalent in hospitalized patients.

Other Gram-positive species are also notable causes of bacteraemia. Streptococcus agalactiae (Group B streptococcus (GBS)) poses a specific BSI risk in newborns due to colonization in pregnant women but invasive disease is increasing in non-pregnant women with increased risk with age and underlying disease²⁶. Moreover, Enterococcus, in particular Enterococcus faecalis and vancomycin-resistant Enterococcus faecium, are common causes of BSI². Clostridium species are associated with some of the greatest mortality risks in BSI even though their relative BSI disease burden is low. The mortality risk following Clostridium BSI is 41.7% after 30 days⁵, and remains high even a year after hospital admission. This underscores the complex nature of BSIs often having lasting effects after the onset of infection.

Gram-negative species are of particular concern in bacteraemia due to the increasing prevalence of antimicrobial-resistant species. BSIs are the second most common condition associated with death owing to AMR, following respiratory infections²⁰. The most prominent species causing Gram-negative BSIs is E. coli, which is also linked to the largest 30-day mortality rate⁵. E. coli, a highly diverse facultative anaerobic rod species, is metabolically flexible and can colonize the gut, urinary tract and skin and is commonly found in the environment. Extra-intestinal pathogenic E. coli (ExPEC) are a distinct but heterogeneous pathotype with a pangenome of sequenced bacteraemia isolates containing ~70,000 unique genes^27,28. The primary phylogenetic groups for ExPEC strains are B2 and D, with the most prominent sequence type being ST131 (refs. 29,30). Primary bloodstream isolates are more often multidrug resistant than commensal isolates^27,31,32.

Although E. coli and S. aureus are the two most common BSI species, K. pneumoniae is consistently the third². Klebsiella species are a diverse group of bacteria, equipped with a substantial accessory genome that makes each strain largely distinct³³. Klebsiella is particularly problematic as a frequent cause of nosocomial infections³⁴. The respiratory tract, gastrointestinal tract and urinary tract are common reservoirs for Klebsiella species, and these species must use many different fitness strategies to survive in such diverse niches.

Other Gram-negative rods, including Citrobacter, Enterobacter and Acinetobacter species, are also prominently detected in bacteraemia. Although less is known about their mechanisms of pathogenesis, candidate genes linked to fitness during BSIs have been identified for some of these species using transposon sequencing in a mouse model of bacteraemia^35–39. E. coli, Klebsiella, Citrobacter and Enterobacter are members of the Enterobacterales order, and treatment of infections is complicated by resistance against β-lactam antibiotics^20,40,41. The frequency of multidrug-resistant isolates among Enterobacterales increased globally from 6.2% in 1997–2000 to 15.8% in 2013–2016 (ref. 2). Acinetobacter baumannii and Pseudomonas aeruginosa are also common causes of Gram-negative BSIs with increasing AMR, prevalence in health-care settings and high mortality rates^2,42,43.

Fungi are often isolated in BSIs, with increased prevalence in recent decades^3,4,44. Candida species, the principal fungal bloodstream pathogens, are associated with high infection rates in immunocompromised individuals^45,46. Since 2015, Candida have been estimated to cause 10% of BSIs and are associated with some of the highest mortality of BSI pathogens^5,47. Candida are diverse and flexible, colonizing many niches⁴⁸. Prominent species in fungal BSI include C. albicans, Candida auris, Candida glabrata, Candida parapsilosis and Candida tropicalis^49–51. The Centers for Disease Control and Prevention have monitored drug-resistant Candida species due to increasing resistance⁴¹. C. auris has emerged as a particular concern due to resistance across all classes of available antifungals. C. auris can be misidentified with standard laboratory diagnostics, increasing the time to appropriate treatment.

Although many microorganisms can cause bacteraemia, only a few species cause most clinical cases of BSI. Prominent BSI pathogens, like S. aureus, E. coli and Candida species are also associated with the highest attributable mortality. Their frequency and mortality suggest incredibly effective virulence mechanisms. Other pathogens, like Klebsiella and C. auris, are associated with increasing AMR and cause opportunistic infections in hospitalized patients. Through understanding how these pathogens cause BSIs, we can improve diagnostic, preventative and therapeutic approaches.

The bloodstream is naturally free of microorganisms and there is no defined microbiota for the blood or filtering organs. Accordingly, any bacteria that enter the blood must arrive by entry from a primary reservoir¹¹, crossing host barriers to disseminate. Once in the blood, bacteria must adapt to a new microenvironment through metabolic flexibility and resist host clearance mechanisms to establish a BSI. These three phases of pathogenesis can be broadly applied to pathogenic microorganisms, although dissemination mechanisms vary by primary site and bacterial adaptation during earlier phases could influence fitness in the bloodstream.

Entry into the bloodstream can occur without a known prior infection. This is defined as direct or primary bacteraemia and can cause both community and nosocomial infections. The pathogen can originate from the environment or sites of asymptomatic colonization in the patient. Microorganisms can be introduced to the blood through a wound, where environmental strains may readily cross the skin barrier (Fig. 2) or be directly introduced into the bloodstream. In community settings, intravenous drug users can acquire bacteraemia through contaminated needles. In nosocomial settings, direct bacteraemia can be introduced through non-sterile equipment utilized during procedural intervention or from interventions in non-sterile sites like the gastrointestinal or urinary tract. Central venous catheters can be a source of primary bacteraemia if they become contaminated by the patient or environment (Fig. 2). Indeed, in hospitals, there are many reservoirs in which pathogenic bacteria can survive such as sink drains and bedrails, which have been linked to outbreaks⁵².

Fig. 2 |. Pathogens use diverse strategies to infect initial sites and disseminate into the bloodstream.

Bacteria have evolved unique mechanisms to colonize and establish infection across multiple primary niches, and those mechanisms are dependent on species and primary site. a, In the lung, Klebsiella pneumoniae relies on a polysaccharide capsule to resist host defences and iron acquisition by siderophores for replication. Dissemination into the perivascular space and the bloodstream are dependent on siderophore interactions with the host factor hypoxia-inducible factor 1α (HIF1α). For Pseudomonas aeruginosa, lung fitness and dissemination rely on type III and type II secretion systems that inject effector proteins (exotoxins) into host cells to evade innate immunity and permeabilize barriers to the bloodstream. b, In the bladder, Escherichia coli and other urinary tract pathogens use complex adhesion mechanisms (for example, fimbriae) to resist urine flow and invade the bladder epithelium. Leveraging toxin secretion and adhesion mechanisms, E. coli can degrade host barriers and ascend upward into the kidneys, where it can then disseminate into the bloodstream. c, Pathogenic members of the endogenous gut microbiome, like enterococcus, Citrobacter freundii, E. coli and K. pneumoniae can outcompete commensal members (dysbiosis; for example, following antibiotic treatment), which promotes translocation to the bloodstream. d, Colonization of the skin and nares by Staphylococcus aureus is aided by a complex polysaccharide capsule that masks bacterial surface antigens. S. aureus also uses matrix-binding proteins to anchor at these sites. When skin barriers are disrupted through wounds, S. aureus can leverage these immune evasion and matrix-binding strategies to disseminate into the blood. e, Indwelling devices and medical equipment may be contaminated by fungal Candida species, which use highly effective adherence mechanisms to form biofilms. Filamentous Candida albicans are resistant to clearance by neutrophils, perpetuating infection.

Bacteria can also enter the blood after asymptomatic colonization. For S. aureus, colonization of the skin and nasopharynx is a significant risk factor for the development of a BSI⁵³. In certain patients, and often following antibiotic treatment, pathogenic members of the gut microbiome can outcompete commensal members (dysbiosis), which promotes translocation to the blood^54,55 (Fig. 2). Similarly, disruption of mucosal barriers, as occurs during mucositis, may promote the translocation of commensal species into circulation. Enterococcus, ExPEC, Citrobacter freundii or K. pneumoniae are examples of BSI pathogens that may originate from the gut microbiome (Fig. 2).

In cases of secondary bacteraemia, pathogens cause a clinical infection at an initial site prior to dissemination into the blood. Multiple infected primary sites can lead to BSIs; specifically colonized mucosal surfaces have been implicated in shedding pathogens. Pneumonia and urinary tract infections are common sources of secondary bacteraemia (Fig. 2), caused either through intermittent or persistent shedding. The ability for a pathogen to damage host tissue at a primary site of infection can increase the chances of bacteraemia by breaking down tissue-specific epithelial and endothelial barriers to bloodstream entry.

Once in the circulation, immune clearance mechanisms in response to bacteraemia are not uniform. At least two separate immune responses to systemic infection have been identified across multiple bacterial species⁵⁶. In a rapid response within the liver, immune cells capture and degrade circulating pathogens. In a slower response within the spleen, the adaptive immune system is trained to produce long-term memory against circulating pathogens. Both the spleen and liver are important in bacteraemia clearance, and contributions from each likely depend on the infectious species. It is likely that additional immune clearance tracks occur during bacteraemia clearance. Further research is required to determine the relative contributions of immune clearance mechanisms to the initiation of sepsis. It is also possible that bacteria circulating through tissues like the lung or kidneys may induce additional immune clearance mechanisms.

A diverse set of bacterial species colonize and infect primary sites but only a small subset can disseminate to the bloodstream and survive within that niche. Each step of BSI pathogenesis presents an opportunity for future interventions to control these deadly infections. However, individual pathogens are uniquely equipped to cause infections within the host, and these mechanisms vary across the initial site and even within strains of the same species (Fig. 2 and Box 1). Complete models of pathogenesis have not been developed for most BSI species. However, attributes that are critical for colonization or initial site infection, dissemination into the bloodstream, and survival in the blood and blood-filtering organs have been described for a few prominent pathogens (Fig. 2).

Box 1 |. Pathogenesis of intracellular pathogens during BSIs

Listeria monocytogenes and Salmonella enterica subsp. enterica serovar Typhi are facultative intercellular bacterial pathogens that can cause bloodstream infections (BSIs). Their pathogenic mechanisms vary in terms of life cycle, host cell interactions, and immune evasion and their intracellular lifestyle affects the choice of antibiotic therapy. These two bacterial species use distinct strategies to invade host cells, evade the immune system and cause systemic disease. Their complex interactions with the host underscore the challenges in treating BSIs caused by these versatile pathogens.

L. monocytogenes

L. monocytogenes is a Gram-positive, rod-shaped bacterium that causes listeriosis, an infection that primarily affects pregnant women, newborns, older adults and immunocompromised individuals. After ingestion of contaminated food, L. monocytogenes survives the acidic gastric environment and reaches the small intestine, where it adheres to and invades the intestinal epithelium. The surface protein internalin binds to the host receptor E-cadherin¹⁸⁵. Once inside, L. monocytogenes uses listeriolysin O (LLO) and two phospholipases (PlcA and PlcB) to escape the phagosome and fuse with lysosomes, avoiding degradation^186–188. In the cytosol, L. monocytogenes hijacks host actin polymerization machinery to propel through the cytoplasm and into adjacent cells without exposing itself to the extracellular environment. This forward propulsion is mediated by the bacterial surface protein ActA, which leads to the formation of actin ‘comet tails’¹⁸⁹. When L. monocytogenes breaches the gut endothelium, it can initiate bacteraemia^187,190. In the blood, the bacteria may disseminate to sites like the liver and spleen, where they are phagocytized by macrophages^191,192. Inside the macrophage, L. monocytogenes can survive and replicate, exacerbating infection, and leading to systemic listeriosis.

S. Typhi

S. Typhi is a Gram-negative, rod-shaped facultative anaerobic bacterium. Certain serotypes of Salmonella cause typhoid fever or, more commonly, self-limiting gastroenteritis. Typhoidal strains disseminate from the intestines to the blood, which leads to more severe infection¹⁹³. Following ingestion, S. Typhi survives the acidic stomach conditions and adheres to and invades non-phagocytic intestinal epithelial cells using a type III secretion system encoded by the SPI-1 pathogenicity island¹⁹⁴. Bacterial effector proteins are injected into host cells, manipulating the cytoskeleton to promote bacterial uptake¹⁹⁵. Within intestinal epithelial cells, S. Typhi induces inflammation, leading to diarrhoea, which can facilitate bacterial spread. A second type III secretion system (encoded by SPI-2) is also used to survive and replicate within macrophages¹⁹⁶. The bacteria manipulate host vesicular trafficking to avoid transport to lysosomes and induce the formation of a modified phagosome called the Salmonella-containing vacuole¹⁹⁴, inside which bacteria replicate, shielded from the immune system. As with L. monocytogenes, macrophages carrying S. Typhi can circulate through the bloodstream, which leads to systemic infection and colonization of various tissues^197,198. Persistent bacteraemia can occur as bacteria reemerge from these protected intracellular niches and re-enter the blood¹⁹⁹. The systemic spread of both L. monocytogenes and S. Typhi is combated by innate immune defences. Neutrophils and other immune cells can clear the bacteria from the bloodstream, and the adaptive immune response targets and eliminates the free bacteria and infected cells.

Colonization and infection of initial sites

Bloodstream pathogens use multiple mechanisms to colonize and infect diverse primary sites but common themes are deployment of adherence factors, iron acquisition systems and polysaccharide capsules (Fig. 2). For ExPEC, there are substantial associations between bacteraemia isolates and strains expressing adherence factors, including P fimbriae (pap) and other fimbriae such as S (sfa) and F1C (foc), and an afimbrial adhesin (afa)^57–59. Similarly, P. aeruginosa relies on adherence mechanisms to approach and attach to the epithelium, particularly flagella and type 4 pili⁶⁰. Candida species are equipped with many, and sometimes species-specific, adherence mechanisms that enable the robust attachment to abiotic and biological surfaces to promote biofilm formation during primary infection⁶¹. Thus, Candida species can readily form biofilms on indwelling devices, which are commonly used in intensive care units and in immunocompromised patients. Indeed, a noted risk factor for mortality is the delayed removal of central venous catheters during candidaemia^62–64. A newly described specific adhesin, Scf1, which is produced by C. auris, contributes to surface binding, colonization and virulence⁶¹. Combined, adhesins promote sustained colonization of the skin, abiotic surfaces and mucous membranes^61,65,66.

Capsule and iron acquisition have critical roles in initial colonization and infection. S. aureus relies on its polysaccharide capsule to mask surface antigens and prevent phagocytosis. Bacterial binding to fibrinogen also creates a protective matrix that prevents the binding of the complement component C3b and subsequent clearance by immune mechanisms⁶⁷. K. pneumoniae leverages diverse strategies to invade primary sites like the lung, gut and urinary tract^11,68–71. K. pneumoniae lung infection leads to strong inflammatory responses, and the bacteria rely on multiple capsule types to survive at this primary site^72–76. K. pneumoniae also leverages multiple, non-redundant siderophores to scavenge iron across many environments, including the lung^77,78. Siderophore-null mutants of K. pneumoniae are unable to invade and replicate in the perivascular space⁷⁹.

Invasion and dissemination into the bloodstream

To disseminate into the bloodstream, bacteria must breach epithelial, endothelial and tissue-specific barriers like basal lamina, mucosa or the perivascular space (Fig. 2). As barriers from the primary sites to the bloodstream vary widely, bacteria must be equipped with many virulence strategies to ensure translocation. Common themes for bacterial dissemination include the use of toxins, siderophores and leveraging inflammation (Fig. 2).

ExPEC strains commonly produce toxins, including haemolysin (hlyA), cytotoxin necrotizing factor (cnf) and cytolethal distending toxin (cdt), to disrupt epithelial barriers⁸⁰. P. aeruginosa also relies on toxin secretion through multiple mechanisms. The type III secretion system injects ExoU, ExoS, ExoT and ExoY cytotoxins to kill host cells and inhibit wound healing^81–88. P. aeruginosa can also delay repair of the epithelium using the type II secretion system toxin ExoA, which has ADP-ribosylating activity that affects protein synthesis^89,90. P. aeruginosa also uses additional virulence factors to disrupt tight junctions, like the protease LasB^89,91. The quorum-sensing autoinducer M-(3-oxododecanoyl)-I-homoserine lactone disrupts epithelial barrier integrity by modifying the expression levels and phosphorylation status of E-cadherin, β-catenin, occludin and ZO-1 (ref. 92). Moreover, P. aeruginosa lipopolysaccharide (LPS) increases epithelium permeability^93,94.

Genes that encode components for siderophore synthesis and uptake in ExPEC are also critical for infection such as iroN, fyuA, ireA and iutA^95–97, which encode importers. The Sit transporter (sit) that imports iron and manganese is encoded on a pathogenicity island and has also been highlighted as a contributing factor to ExPEC dissemination in large-scale whole-genome sequencing projects^31,98. For K. pneumoniae, siderophores promote replication in the perivascular space. Through iron depletion in the host environment, these siderophores also lead to stabilization of the host transcription factor hypoxia-inducible factor 1α (HIF1α) in lung epithelial cells, which in turn promotes dissemination^79,99, although the underlying mechanisms and how this leads to the progression to bacteraemia remain to be determined. Overall, dissemination is difficult to measure as it is affected by bacterial fitness during the other phases of bacteraemia, and novel research approaches are needed to identify targets for intervention in this critical step of pathogenesis.

Survival in the bloodstream and blood-filtering organs

Once in the bloodstream, microorganisms may circulate but are rapidly filtered by organs like the spleen and liver. Numerous strategies are used by pathogens to evade or modify interactions with leukocytes and the complement system (Fig. 3). To survive in this environment, bacteria must fend off antimicrobial peptides, serum complement, osmotic stress, oxidative bursts and other factors. Common themes of virulence strategies include the use of a polysaccharide capsule, metabolic flexibility and inactivation of immune cells.

Fig. 3 |. Mechanisms of pathogen survival in the bloodstream.

Microorganisms use multiple strategies to survive in the bloodstream and blood-filtering organs. a, Staphylococcus aureus forms foci in blood vessels that function as reservoirs in which bacterial replication occurs and which promote re-seeding of the blood. S. aureus upregulates endothelial expression of the von Willebrand factor and can anchor to endothelial cells using a von Willebrand factor-binding protein. Bacterial effector proteins, including staphylocoagulase, then cause additional aggregation, which leads to the formation of clots within vessels. b, Gram-negative species like Escherichia coli, Klebsiella pneumoniae, Serratia marcescens and Citrobacter freundii rely on metabolic flexibility to survive in this microenvironment. Central carbon metabolism is required to survive in the bloodstream, and separate species have differential requirements for the metabolism of various amino acids to survive in this niche. c, All pathogens that cause bloodstream infections must also evade killing by immune mechanisms in the blood, such as complement deposition and the formation of a membrane attack complex, or targeting by antimicrobial peptides. Many microorganisms use an extensive polysaccharide capsule to evade membrane attack complex formation on the surface, cleave complement with proteases or modify the charge of the lipopolysaccharide (LPS) to repel charged antimicrobial peptides. d, Polysaccharide capsule can also influence the interactions with different immune subsets. Unencapsulated K. pneumoniae and capsule types that cannot prevent phagocytosis are readily cleared by tissue-resident Kupffer cells in the liver. K2 capsule-type K. pneumoniae can replicate within Kupffer cells, which leads to host cell death and the formation of abscesses, in which the bacteria can further replicate and seed the bloodstream.

To survive in the bloodstream, pathogens must first evade killing by complement and antimicrobial peptides. E. coli carries several mechanisms to resist antimicrobial peptides, including the sap (sensitivity to antimicrobial peptides) gene products. E. coli and other Gram-negative species can modify the charge on moieties of LPS to repel positively charged antimicrobial peptides¹⁰⁰. A protease encoded by prc, involved in regulating peptidoglycan synthesis, has also been shown to proteolytically inactivate complement, thus conferring resistance to serum killing^101,102. As in the initial site of colonization and infection, many BSI pathogens, including E. coli, K. pneumoniae and S. aureus, rely on their polysaccharide capsule for survival in the bloodstream. For K. pneumoniae, many capsular serotypes have been identified that broadly protect against complement-mediated killing, although specific host interactions may vary by serotype. Modifications of K. pneumoniae capsule properties, such as chain length, are linked to a striking hypermucoid phenotype that is regulated in response to host interactions^103,104. This hypermucoviscosity is particularly prominent in hypervirulent strains associated with sequelae of BSI, including pyogenic liver abscess, meningitis and endopthalmitis^{14,77,78,105,106}.

S. aureus uses vessel wall adhesion, matrix formation and coagulation initiation to form foci that can function as reservoirs in which bacterial replication and additional shedding into the blood can occur^12,107,108. S. aureus uses alpha-toxin and staphylocoagulase to initiate platelet aggregation and crosslink fibrin to promote bacterial–platelet aggregation^109,110. S. aureus can then use a von Willebrand factor-binding protein to anchor bacterial foci to vessel walls¹¹¹. Upon entry into the blood, S. aureus induces recruitment of circulating neutrophils and production of inflammatory cytokines¹¹². Of particular concern, foci on the heart valves and lining of the chambers can cause infective endocarditis, increasing risk of mortality¹¹³. Another common agent of infective endocarditis are Enterococcus species, specifically E. faecalis and E. faecium. Enterococci possess virulence mechanisms like the collagen-binding protein Ace and the biofilm-associated pili ecb that promote adhesion of the bacteria to blood vessels and heart valves¹¹⁴.

For many microorganisms, evasion of the host immune response is key to survival in the bloodstream. P. aeruginosa uses its type III secretion system to escape neutrophil-mediated killing. Specifically, ExoS inhibits actin polymerization in neutrophils whereas ExoU impairs internalization of bacteria¹⁰². Bacteria can also block the formation of the pore-forming membrane attack complex (C5b–C9 complex) on their surfaces by evading complement deposition. For example, LasB and the alkaline protease AprA degrade specific complement effectors C1 and C3 (ref. 89), and protease inhibitors can block complement activation^115–119. Exopolysaccharide production, including alginate, further interferes with complement cascade by masking bacterial surface determinants, thus limiting opsonization and phagocytosis^120,121. P. aeruginosa can lengthen the O-specific antigen to prevent membrane attack complex insertion¹²², or the bacteria recruit negative complement regulators or complement regulator-acquiring surface proteins to their surface, which inhibits complement deposition¹²³.

During infection, Candida species can grow in yeast or filamentous morphologies and as planktonic or biofilm forms, leading to distinct immune responses. For example, fungal β-glucans are widely detected by Toll-like receptors and Dectin 1, yet C. albicans in yeast morphology induce a T helper 17 response while the filamentous morphology predominantly induces a T helper 1 response^124–126. While C. albicans can induce neutrophil cell death via the highly inflammatory neutrophil extracellular trap (NET) release, NETosis is dependent on fungal morphology^127,128. C. albicans yeast elicit extensive NETosis while filamentous biofilms actively impair the release of extracellular traps by inhibiting reactive oxygen species, although this interaction is strain dependent^128,129. Other Candida species can inhibit NET release in general. For example, C. auris can broadly inhibit NETosis and thus resists killing by neutrophils in a more effective manner than C. albicans¹³⁰.

Additional factors that enhance bacterial fitness in the bloodstream have been identified and have revealed previously unidentified host–pathogen interactions during systemic infection. For example, capsule type influences infection beyond complement resistance. The K. pneumoniae K2 capsule type aids in the evasion of clearance by liver-resident Kupffer cells¹³¹. Other K. pneumoniae bacteraemia factors are tissue specific. Spleen-specific fitness factors are diverse but are not always required in the liver¹³². Some K. pneumoniae bacteraemia factors are specific to the bloodstream. For example, GmhB, a partially redundant LPS biosynthetic enzyme, is required for spleen and liver fitness but is dispensable for infection of the lung¹³³.

Risk factors and prevention

Several risk factors in susceptible hosts predispose to BSIs. BSIs are disproportionately increased at the extremes of age, with increased rates observed in both neonates and patients >65 years of age^134,135. In elderly patients, this is likely to be due to immune senescence, coupled with age-related changes in skin integrity, airway protection and bladder function that can increase the risk of primary site infections and BSIs¹³⁵. BSIs are more common in males than in females overall and across pathogens, and the difference widens with age¹³⁶. The notable exception is that BSIs caused by E. coli are more common in females, which is likely to be due to prior urinary tract infections. Certain comorbid conditions seem to increase the risk of BSIs with Gram-negative bacteria, notably immunocompromising conditions (such as haematopoetic stem cell transplantation, end-stage renal disease, HIV/AIDS, critical illness and cirrhosis). Community-onset Gram-negative bacteraemia is frequently observed in patients with anatomic abnormalities (such as structural lung disease, obstructive uropathy and/or obstructive hepatobiliary pathology) that impede the ability of the host to clear colonizing microorganisms^137,138. Patients requiring intensive care are particularly vulnerable to BSIs and the risk increases with longer hospital stays, use of immunosuppressants and use of invasive equipment, including catheters or ventilators^4,5,24,34. Additionally, conditions that affect the integrity of mucosal surfaces that are colonized with potential pathogens (for example, inflammatory bowel disease, chemotherapy-induced gastrointestinal mucositis) increase the risk for BSIs. BSIs caused by Staphylococcus species are disproportionately observed in patients with integumentary disruptions such as indwelling vascular catheters or injection drug use^139,140.

In conjunction with intrinsic patient risk factors, colonization with pathogens increases the risk for subsequent BSI. Bacteraemia caused by E. coli or K. pneumoniae peaks in the summer^141,142, correlating with increased K. pneumoniae colonization rates¹⁴³; the mechanism of this seasonality is unclear. The link between colonization and subsequent infection provides an opportunity for intervention. For example, early-onset neonatal sepsis could be prevented when colonization of pregnant women with S. agalactiae (GBS) is identified. GBS colonization increases the risk of BSI >25-fold, with 1–2% incidence of early-onset GBS infection in newborns if their mother was colonized; with colonization rates of 10–30%, it is standard-of-care to offer antibiotic prophylaxis to the mother at the time of delivery^144,145, an approach that requires a substantial allocation of resources. However, it has lowered overall rates of neonatal sepsis to nearly that of mothers who are not colonized (<0.05%), with rare serious antibiotic side effects in the mother. Colonization with methicillin-resistant S. aureus (MRSA) is also associated with BSI, with a negative nasal surveillance swab having a 97% negative predictive value for BSI caused by MRSA within 7 days¹⁴⁶. Decolonization approaches for MRSA include nasal administration of the antibiotic mupirocin and bathing with the antiseptic chlorhexidine. The universal application of both approaches in patients in intensive care units led to a statistically significant reduction in BSIs caused by MRSA and in BSIs overall, and is currently a recommended approach as is targeted decolonization in other populations of patients at risk¹⁴⁷. Preoperative use of antibiotics that target skin commensals represents another standard-of-care practice that can reduce the incidence of surgical site infections and potentially BSIs¹⁴⁸. Characterization of the colonizing isolates can also be used for tailored interventions to prevent BSI. Culture-directed antimicrobial treatment of bacteria asymptomatically colonizing the genitourinary tract can reduce post-procedural BSI^149,150. Similarly, tailored prophylactic antibiotics based on a rectal culture can reduce the risk of BSI after transrectal prostate biopsies from colonizing pathogens such as E. coli and K. pneumoniae¹⁵⁰. Alternatively, transperineal biopsies avoid the gastrointestinal tract and have low rates of infection¹⁵¹.

The greatest reservoir for BSI pathogens, and therefore the greatest opportunity for prevention, is the gastrointestinal tract. With the notable exception of S. aureus, many of the leading causes of BSI colonize the gastrointestinal tract (Table 1). Disruption of the gut community, often by antibiotics, can lead to dominance (>30%) of both Gram-positive and Gram-negative pathogens and increase the risk of subsequent BSIs by fivefold to ninefold¹⁵². For K. pneumoniae, this increased risk based on dominance is independent of other patient risk factors¹⁵³. The genetics of the colonizing strain may drive infection risk through dominance by nutritionally outcompeting other microbiome members or directly killing competing species¹⁵⁴. Combined with our knowledge of patient risk factors, the characterization of the colonizing isolates could guide targeted interventions and decrease BSI risks. Selective decolonization using non-systemic antibiotics can be effective but, as for prostate biopsies, using antibiotics to prevent antibiotic-resistant infections is unlikely to be sustainable as resistance will rapidly evolve. Instead, novel, non-antibiotic decolonization approaches are being actively pursued. Many are based on the proof-of-principle that faecal transplants and microbiota products, gaining regulatory approval around the world, can reduce colonization with pathogens¹⁵⁵.

Treatment

Contemporary management of BSIs involves the use of appropriately dosed antimicrobial therapy with in vitro activity against the infecting pathogen coupled with ‘source control’ procedures to eliminate the initial site of infection and secondary BSIs. Source control procedures vary based on clinical syndrome but may include removal of infected intravascular catheters, percutaneous or operative drainage of suppurative foci of infection (intra-abdominal abscesses, soft tissue structure abscesses, empyema), procedural decompression of the genitourinary and hepatobiliary tracts, and operative debridement of infected endovascular reservoirs in infective endocarditis and vascular graft infections.

Accurate diagnosis of the causative agent of a BSI is critical to the treatment of these infections, and multiple approaches are used to identify the pathogens (Table 2). Blood cultures are the gold standard but have important limitations that must be overcome to improve treatment (Box 2). The clinical manifestations of BSIs are often dependent on the causative pathogen and infected host. In general, patients present with constitutional symptoms such as fevers, chills, sweats, lassitude or anorexia. The presentation may be fulminant in patients infected with Enterobacterales, S. aureus or haemolytic streptococci, whereas patients infected with enterococci and other Gram-negative bacteria can present more indolently. Of these potential symptoms, shaking chills (rigours) in a febrile patient has the highest probability of predicting a positive blood culture¹⁵⁶. Certain simultaneous infections make BSIs more likely. Meningitis and catheter-related infections have the highest likelihood of detectable bacteraemia (>50%), followed by acute pyelonephritis, pyogenic liver abscess and severe community-acquired pneumonia (20–50%), whereas lower tract urinary tract infections have a lower likelihood of leading to bacteraemia (<10%)¹⁵⁷. Thus, the diagnosis requires a combination of physical examination, thorough history and laboratory testing.

Table 2 |.

Diagnostic approaches for bloodstream infections

Diagnostic approach	Detection	Identification	AST	Analytical target	Time to result	Advantages	Disadvantages
Automated blood-culture incubation systems	Complete	NA	NA	CO2 production; changes in gas pressure	Several hours to 5 days	Standard of care; high throughput, low cost, short hands-on time	Long time to detection, occasional false positives, sensitive to collection errors (contamination, underfilling), lower performance with prior antibiotic treatment, fastidious pathogens may not grow
Gram stain from positive blood cultures	NA	Limited	NA	Bacterial cell wall and cell morphology	<1 h	Standard of care; rapid, low cost	Limited identification, prone to errors in staining process and interpretation, requires high expertise
MALDI-TOF mass spectrometry	NA	Complete	Limited	Pathogen proteins	Minutes per sample	Becoming standard of care; definitive identification for nearly all common pathogens, some targeted AST detection, inexpensive reagents	Expensive instrument, ambiguous identification of some closely related organisms; requires a culture isolate or purified suspension
Nucleic acid amplification from positive blood-culture bottles	NA	Complete	Limited	Pathogen gene sequences	1–3 h	Becoming standard of care; can provide definitive identification, early detection of certain resistance genes	Limited breadth of targets, expensive, false positives from blood-culture bottle manufacturing process
Biochemical testing	NA	Complete	NA	Varies, combinations of reactivity used	Hours to overnight	Extensively validated, can distinguish closely related species, low cost	Long time to result, may not provide a definitive identification, requires expertise for interpretation, can be highly manual
Gene or genome sequencing	NA	Complete	Limited	Pathogen DNA sequences	Overnight to days	Gold standard for identification (for example, 16S sequencing for bacteria)	Requires an isolate, can be slow, requires expertise for interpretation
Phenotypic susceptibility testing	NA	NA	Complete	Growth inhibition in known concentrations of antibiotics	Overnight	Standard of care for antimicrobial susceptibility; low cost, standardized protocols and interpretative guidelines	Long time to result, requires an isolate, requires expertise for interpretation and quality control
Direct PCR from blood	Complete	Complete	NA	Pathogen DNA sequences, amplicon detection approaches vary	Hours	Rapid, can be highly sensitive and specific	High cost, currently limited inclusivity of pathogens, possibility of false positives
Cell-free plasma DNA sequencing	Complete	Complete	Limited	Pathogen DNA sequences	Overnight to days	Direct detection, good sensitivity even if collected after the patient received antibiotics, does not require growth in culture	High cost, usually sent to a reference laboratory (delayed time to result), specificity and additional yield over standard of care can be low, still need to determine best use cases
Concentration and purification from blood-culture bottles	NA	Complete	Complete	Prepares viable bacteria for rapid testing	<1 h	Eliminates second culture step on agar plates, compatible with standard identification, AST methods (disk diffusion or MIC)	Incremental decrease in time to results, may get different results or failures versus standard identification, AST procedures
Manual rapid phenotypic susceptibility testing	NA	NA	Limited	Growth inhibition by disk diffusion after direct plating from positive bottle	Hours	Rapid, low cost	Limited to certain species, may need repeat reading
Automated rapid phenotypic susceptibility testing	NA	NA	Complete	Varies, measurements of cell growth, metabolism	Hours	Eliminates second culture step prior to AST, faster than current AST methods	May not have comprehensive results for certain pathogens, high cost, limited throughput
Inflammatory markers	Complete	NA	NA	Procalcitonin, C-reactive protein	Hours	Rapid direct testing, independent of pathogen detection, can guide antibiotic use	False positives and false negatives, does not identify the pathogen or guide antibiotic choice
Immune response patterns	Complete	NA	NA	Combinations of cytokines	Hours	Rapid direct testing, can distinguish bacterial versus viral infection, can guide antibiotic use	Does not identify the pathogen or guide antibiotic choice

AST, antimicrobial susceptibility testing; MALDI-TOF, matrix-assisted laser desorption ionization–time of flight; MIC, minimal inhibitory concentration; NA, not applicable.

Box 2 |. Laboratory diagnosis of bloodstream infections.

Blood cultures are the gold standard for the diagnosis of the causative agents of bloodstream infections (BSIs) but they have notable limitations such as low sensitivity and specificity and a long time to diagnosis (Table 2). Although BSIs are life-threatening, the concentration of bacteria in blood can be low, even <1 CFU/ml. To optimize sensitivity and specificity, multiple blood-culture sets (≥2) should be collected. In addition, the pathogen may only circulate sporadically, and cultures may be negative, as exemplified by the diagnosis of Candida species²⁰⁰. Blood cultures are collected from venipuncture through the skin or from indwelling catheters, both of which are sites of bacterial colonization. Contamination with colonizing bacteria during collection can cause false positives, compromising test specificity. If the same organism is isolated in multiple sets, this is likely the cause of bacteraemia as opposed to a skin contaminant. The blood is incubated on automated instruments that continuously monitor for bacterial growth. Once bacteria are detected, characterization typically progresses from Gram stain and morphological description to identification at the species level by matrix-assisted laser desorption ionization–time of flight mass spectrometry or PCR, and finally phenotypic susceptibility to a panel of antibiotics that are appropriate for treatment.

The slow time to blood-culture results has prompted the development of many strategies to reduce the time to obtain actionable information (Table 2). These include molecular and phenotypic assays for pathogen identification, and antibiotic susceptibility testing directly from positive blood-culture bottles. However, the largely effective use of empirical antibiotic therapy has limited the impact on patient outcomes of these interventions that do not shorten or eliminate the culture step²⁰¹. There are recent advances in direct detection from blood samples, including PCR with novel detection of amplicons and metagenomic sequencing of cell-free DNA in plasma (Table 2); in this early stage of development, these assays have tradeoffs in inclusivity of pathogens detected, and in specificity and turnaround time, respectively²⁰².

As blood cultures are only as rapid as the doubling time of the pathogen (Box 2 and Table 2), they are too slow to guide initial choice of antibiotics. Instead, empirical, broad-spectrum antibiotics are frequently prescribed in patients with suggestive clinical syndromes, which are then transitioned to a narrow-spectrum antibiotic following pathogen identification and antimicrobial susceptibility testing¹⁵⁸. Both the preferred duration and route of definitive antimicrobial therapy for BSIs are largely based on historical precedent. Antimicrobial treatment courses can be shortened for BSIs when coupled with effective source control procedures^159–161. Highly bioavailable oral rather than intravenous antibiotics are appropriate options for definitive therapy in selected patients^162–164. In general, a single antimicrobial agent is recommended to treat BSIs; routine use of combination antimicrobial therapy for definitive treatment of BSIs may reduce time to clearance of bacteraemia but does not confer patient-oriented benefits and can precipitate drug toxicity^165–168. Combination therapy may be needed to treat extensively drug-resistant pathogens without effective source control procedures^169,170.

The choice of antibiotic is based on an integrated appraisal of the patient, the microorganism and clinical phenotype. Although in vitro susceptibility testing may predict efficacy of several different antibiotics against an offending pathogen in controlled laboratory settings, comparative in vivo trials suggest that the specific choice of antibiotic may influence morbidity and mortality for specific pathogens^171,172 (Table 2). For patients with BSIs caused by difficult-to-treat pathogens and persistent reservoirs of infection not amenable to source control procedures, the choice of antimicrobial agent may affect the risk of relapse and emergence of subsequent resistance¹⁷³. Additionally, the site of infection is of critical importance to antimicrobial selection as certain agents may fail to achieve adequate concentrations in infected tissue compartments (notably, bone, the central nervous system, the genitourinary tract and epithelial lining fluid within alveoli)^174–176.

Equally important to the choice of antibiotic is dosing of the antibiotic. Patients with BSIs, particularly those with sepsis, are frequently affected by physiological alterations, including augmented renal clearance, hypoalbuminaemia and altered volumes of distribution, all of which may impact the pharmacokinetic exposures to antimicrobial therapy¹⁷⁷. Therapeutic drug monitoring is routinely used with selected antimicrobial agents (aminoglycosides, glycopeptides in particular) to inform dosing strategies. Future studies are required to establish the role of therapeutic drug monitoring in more commonly used antibiotics such as β-lactam antibiotics.

Despite the revolutionary advancements in patient outcomes during the antibiotic era, existing antimicrobial therapy remains unsuccessful in many BSI cases for a multitude of reasons, not the least of which is the emergence of AMR. Novel strategies for the treatment of BSIs are required to surmount these challenges. Promising avenues for future research and therapy include antimicrobial agents with novel mechanisms of action, obviating existing mechanisms of bacterial drug resistance¹⁷⁸. Patients with high-bioburden endovascular reservoirs of infection (such as patients with prosthetic valve endocarditis, cardiac electronic implantable device infections and vascular graft infections) may fail treatment due to the presence of microbial biofilm, in which bacteria are encased by a matrix of extracellular polymeric substance that can protect the bacteria from conventional antimicrobial agents¹⁷⁹. Use of adjunctive therapies to penetrate biofilms represents a promising avenue to increase cure rates in selected patients, although the efficacy of this practice remains uncertain¹⁸⁰. For patients with recurrent catheter-related BSIs, catheter ‘locking’ solutions comprising biocidal agents may be used preventatively. Bacteriophages, which are viruses that infect bacteria, may represent attractive therapeutic options for treatment-refractory BSIs, though operationalizing them for contemporary clinical use remains challenging^181,182. Last, adjunctive therapies targeting bacterial virulence factors (such as toxin neutralization, quorum-sensing inhibition, pathogen immune evasion or anti-adherence strategies) remain an ongoing area of investigation that requires preclinical and clinical study^22,183,184.

Conclusions and outlook

BSIs are a serious threat to public health, leading to death due to sepsis, cardiovascular collapse and organ failure. The immunocompromised, elderly, neonates and those receiving treatment in intensive care units are at a particularly high risk of BSIs and death. A small group of pathogens cause a large burden of BSIs, likely due to frequent exposure and their pathogenic mechanisms to evade clearance mechanisms in the blood. BSIs arise from pathogens in the environment and endogenous sites of colonization or infection and may lead to endovascular foci and spread to any organ, including the liver and spleen, which normally filter the blood. Diagnosis and treatment are intrinsically linked. Empirical treatment with antibiotics compensates for the slow process of culture, and subsequent laboratory identification and antibiotic susceptibility testing enable definitive treatment. Given an ageing population, the continued emergence of AMR and the increasing number of therapies such as organ transplantation with concomitant immunosuppression, the numbers of patients at risk will continue to rise, highlighting the need for continued improvement in medical care for BSIs.

There are key gaps in our knowledge of BSI pathogenesis that are impeding progress. The mechanisms by which pathogens disseminate from primary sites into the bloodstream and the relative contribution of microbial and host factors are unclear for the top causes of bacteraemia. Similarly, the critical reservoirs that perpetuate bacteraemia through re-seeding the blood are unclear. The kidney, liver and spleen may be critical reservoirs that are unmeasured by current diagnostic approaches. Basic and translational research characterizing microbial fitness at primary sites, routes of dissemination and mechanisms of bloodstream survival will provide targets for novel therapies and diagnostics approaches. These insights will then need to be applied to diagnostic testing, drug development and clinical trials to bring the next generation of tools to bear on BSIs.

To counteract increasing rates of BSIs and substantially improve patient outcomes, disruptive technologies will be required. The ideal test for BSIs would have the inclusivity of blood cultures, better sensitivity and specificity, and a turnaround time that can guide the first dose of antibiotics. To achieve this, diagnostic tests that are culture independent are needed. Direct-from-sample detection using next-generation sequencing is a promising disruptive technology but will need to be rapid, easy to perform, decentralized to hospital laboratories, and affordable to be covered by insurance and also be used in middle-income and low-income settings. Novel treatment strategies need to be developed, proactively selecting leads to which evolution of resistance is highly unlikely. This may require moving away from microbial products, for which resistance has inevitably co-evolved, and exploring new biological spaces computationally. Another disruptive approach is a pivot to measuring and modulating host responses to improve speed of diagnosis and counteract the damage from sepsis, respectively. Finally, prevention through decolonization could be effective for the management of BSIs in select cases. There are promising leads based on the high efficacy of faecal transplants that may lead to safe and inexpensive approaches. However, implementation will require a financial structure that rewards preventative care and, to be sustainable, the benefits must be measurable and apparent to the patient, the health system, and the payors. In conclusion, there are many exciting advances in basic research, diagnostic testing and drug development under way that can be brought to bear on reducing the burden of BSIs.