The role of high-density lipoproteins in sepsis

Abstract

High density lipoproteins (HDLs) are best known for their role in atherosclerotic cardiovascular diseases. However, efforts to reduce cardiovascular risk by increasing the concentration of cholesterol in HDL have failed, raising the question of whether HDL may have other, more central aspects to its role in health and disease. In epidemiological studies, low levels of HDL cholesterol are strongly associated with risk of infectious diseases and death from sepsis and, during sepsis, a larger decline in HDL cholesterol predicts a worse outcome. Many preclinical studies have examined strategies to augment HDL genetically or pharmacologically and have shown that this leads to protection from sepsis in animal models. Data in humans are more limited, but suggest a clinically relevant role of HDL in mediating the response to pathogen-associated lipids and preventing excessive inflammation. Collectively, these data raise the intriguing possibility that a clinically important biological function of HDL is as a component of the innate immune system and suggest that pharmacological manipulation of HDL may be a strategy to reduce the organ dysfunction and death that results from uncontrolled inflammation during sepsis. This review article discusses recent advances in our understanding of the role of HDL in sepsis.

HDL particles are best known for their inverse association with the risk of cardiovascular disease. This relationship, which has been appreciated since at least the 1950s, was bolstered by a large body of preclinical work demonstrating that HDL particles have antiatherogenic properties in cells and animal models. Decades later, these observations led to the “HDL hypothesis,” which predicts that raising levels of HDL in humans would lead to a reduction in the occurrence of major adverse cardiovascular events. This hypothesis has been robustly tested in clinical trials using a number of molecules that increase the concentration of HDL cholesterol (HDL-C), including niacin (1, 2), inhibitors of cholesteryl ester transfer protein (CETP) (3, 4, 5, 6), and infusion of recombinant HDL (7), which have, in general, failed to demonstrate a cardiovascular benefit of increasing HDL-C, particularly in the statin era.

Possible explanations for these failures include that perhaps HDL is not cardioprotective after all, that interventions tested to increase HDL-C failed to enhance HDL function, or that the focus on measurement of HDL based on the concentration of cholesterol in these particles has been inappropriate. Regardless, these data have had led to a re-evaluation of the role of HDL in normal physiology, health, and disease. Assuming that HDL does play an important biological function, the intense focus on HDL in relation to cardiovascular health over the past 70 years may have failed to recognize the true primordial function of HDL.

HDL particles have a number of properties that are relevant to sepsis, including their ability to bind and inactivate pathogen-associated lipids such as lipopolysaccharide (LPS) and lipoteichoic acid (8, 9), as well as their anti-inflammatory, antithrombotic, and vasoprotective actions and their ability to enhance inflammatory responses in macrophages (10, 11, 12, 13). In this article, we review current knowledge regarding the role of HDL in infectious diseases and sepsis. Other recent review articles have described the multiple pleiotropic functions of HDL that may be relevant to sepsis (14, 15), and the focus of this review is on evaluating the strength of evidence that supports a biologically and clinically significant role of HDL in sepsis in humans.

Evolutionary perspectives on HDL

HDLs are evolutionary ancient, appearing as far back in evolutionary time as the arthropods, which appeared 521 million years ago (16), and in which they play a role transporting apolar lipids through the physiological fluid, hemolymph (17). One of the major physiological roles of HDL in these species is as a component of the innate immune system by enhancing the activity of phenoloxidases and key enzymes in the immune system of invertebrates (17). Other lipoproteins, such as VLDL and LDL appeared later in evolutionary history, being first identified in vertebrates (18). These observations imply that HDL may be the original lipoprotein molecule and that it evolved primarily as a component of the immune system (19).

Epidemiological studies of HDL and infectious diseases

Levels of HDL-C are inversely associated with the risk of infection. In a large study in population study including 97,166 individuals from the Copenhagen General Population Study and 9,387 from the Copenhagen City Heart Study, a low HDL-C at baseline (<0.8 mmol/L) was associated with a 1.75- to 2-fold increased risk of infection after multifactorial adjustment (20). Interesting, a high HDL-C was also associated with an increased risk for infection, albeit to a lesser extent. Low levels of apoA-I were also strongly associated with increased risk of infection.

During sepsis, a low level of HDL-C at presentation is a strong predictor of organ dysfunction and death. In a clinical cohort of 200 patients with early stages of sepsis, an HDL-C level of <0.65 mmol/L was associated with a significantly increased risk of 28-day mortality (21). Intriguingly, HDL-C levels were a stronger predictor of mortality than biomarkers such as lactate which are commonly used to assess the severity of sepsis. Many studies have replicated the strong association between lower levels of HDL-C and increased risk of morbidity and mortality from sepsis in a variety of populations. In a meta-analysis of 24 studies including 2,542 participants, survivors of sepsis had significantly higher HDL-C levels on day 1 of intensive care unit admission than nonsurvivors (22).

HDLs are highly heterogeneous populations of particles. Whether specific subpopulations of HDL are most strongly associated with protection from sepsis is not well understood. For instance, small HDL (also called HDL₃), which have a distinct proteome (23) and lipidome (24), are known to have specific anti-inflammatory and antioxidant properties (25), which may be relevant to their role in protection from sepsis. One study of patients with septic shock found that levels of small HDL particles, assessed using gel electrophoresis, were reduced to a greater extent than were large HDL particles, compared to non-septic patients (26), suggesting that the drop in HDL during sepsis may preferentially impact small HDL. Consistent with this, in a large population-based study of >30,000 individuals, low levels of small and medium HDL particles, assessed using nuclear magnetic resonance spectroscopy, were associated with an increased risk of infectious-related death, whereas low levels of large and extralarge HDL particles were not (27). Low levels of small HDL were also associated with risk of sepsis and sepsis death in the UK Biobank (28).

The small intestine is responsible for the biogenesis of ∼20% of the plasma HDL pool (29), and interestingly, a population of small, enterically derived HDL, called HDL₃, has been shown to carry LPS-binding protein and are able to bind and inactivate LPS, limiting its ability to activate toll-like receptor 4 in Kupffer cells (30). It is intriguing to speculate whether this small HDL species produced by the small intestine also plays a role in protection from sepsis.

Mendelian randomization

While the epidemiological data surrounding low levels of HDL-C and increased risk of infectious disease and death from sepsis are robust, this does not imply a causal relationship between HDL-C and sepsis pathogenesis. Low HDL-C is also robustly associated with an increased risk of atherosclerotic cardiovascular disease, but human genetic and clinical trial evidence has not substantiated a causal relationship between HDL and atherosclerotic cardiovascular disease. Thus, a critical question is whether low HDL-C has a causal relationship to increased risk of infection.

One method to infer such a causal relationship is through Mendelian randomization, which is a type of instrumental variable analysis that considers whether a genetically predicted trait (such as low HDL-C) is associated with an outcome of interest (increased risk of infectious disease) (31). Because the genetic variants are randomly assigned at meiosis and are relatively fixed throughout life, this type of analysis can overcome two of the critical pitfalls of observational epidemiology, namely confounding and reverse causation.

The first study to use a Mendelian randomization approach to infer causality between HDL-C and infection used the UK Biobank population (32). In the UK Biobank of nearly 500,000 participants, low levels of HDL-C were strongly associated with an increased risk of hospitalization for infectious disease, consistent with previous epidemiological studies. Using a polygenic score composed of 223 single nucleotide variants, higher levels of genetically predicted HDL-C were associated with a decreased risk of hospitalization for infectious disease. Overall there was a 16% lower risk of hospitalization for infectious disease per 1 mmol/l increase in genetically predicted HDL-C (32). There was also a lower likelihood of out-patient antibiotic prescriptions in individuals with a high HDL-C polygenic risk score, indicating a lower risk of mild to moderate infections in individuals with greater genetically determined HDL-C.

A subsequent study examined the association of HDL subclasses based on nuclear magnetic resonance in the UK Biobank and FinnGen cohorts and found that higher concentration of small HDL particles was strongly associated with protection from sepsis (28). In Mendelian randomization analysis using genetic variants associated with small HDL particle count, there was no predicted causal effect of small HDL particle count on the risk of sepsis or 28-day mortality, but there was a statistically significant causal prediction of lower risk of admission to an intensive care unit for sepsis with greater levels of small HDL (28). The authors then performed an elegant bidirectional Mendelian randomization analysis between genetically predicted interleukin-6 (IL-6) and C-reactive protein levels and HDL subclasses, and found that genetically proxied increased IL-6 signaling was associated with reduced small and medium HDL particle count, whereas genetic variants associated with C-reactive protein had no effect on HDL particle count, and genetically proxied HDL particle count did not have a predicted causal effect on IL-6. These findings support a model in which the observed association of low HDL with sepsis may be downstream of IL-6, which itself has causal effects on both infection and small HDL particle level.

A limitation of these analyses is that they are generally performed in general population cohorts in which the number of individuals with severe sepsis is relatively small. The sepsis phenotype is imperfectly defined by diagnostic codes used in such datasets, which likely leads to heterogeneity and lack of precision in the phenotype. An additional challenge is to control for changes in other lipoprotein classes, such as LDL, which are also altered in sepsis, as well as reciprocal changes in triglyceride (TG) levels which can occur due to changes in HDL metabolism.

Preclinical studies of HDL in sepsis

A number of studies using animal-models of sepsis have provided support for a physiologically relevant role of HDL in sepsis. One of the first such studies reported that mice transgenic for human APOA1 showed improved survival following LPS administration (33). Subsequent studies reported that mice-deficient in ApoA1 due to genetic deletion of the ApoA1 gene had reduced survival following cecal ligation and puncture compared to nonlittermate control mice (10). Mechanistically this was associated with a reduced ability of serum from ApoA1-deficient mice to prevent LPS-induced activation of toll-like receptor 4 (10). In the same study, mice transgenic for human ApoA1 had numerically but not statistically significant improved survival after cecal ligation and puncture (10). Scavenger receptor B1 (SRB1) is the major hepatic HDL receptor (34) and may play a role in the hepatic uptake of HDL-associated pathogen lipids such as LPS (35). Mice-deficient for SRB1, despite having significantly higher levels of HDL-C, show evidence of an increased inflammatory response to LPS administration or intraperitoneal injection of Escherichia coli (36). Deficiency for SRB1 was also associated with reduced appearance of labeled LPS in the liver (36). A separate study found that mice with targeted inactivation of SRB1 have increased mortality from cecal ligation and puncture compared to littermate controls, whereas mice transgenic for SRB1 were protected from cecal ligation and puncture-induced mortality (37). The increased mortality from cecal ligation and puncture observed in mice with targeted inactivation of SRB1 was recapitulated in Scarb1^I179N mice with hepatic-specific deficiency of SRB1 expression (38), identifying the liver as the key site at which SRB1 protects from sepsis-related mortality. These studies suggest that the ability of HDL to bind LPS and return it to the liver is a key component of the role of HDL in immunity.

SRB1 is a multiligand receptor, and in addition to HDL, can also bind lipid-soluble vitamins and silica, and plays a role in cholesterol efflux and reverse cholesterol transport (RCT) (39). Of particular relevance to sepsis, SRB1 may play a role in cellular uptake of the hepatitis C virus (24) and phagocytosis of apoptotic cells (40). Macrophages that lack SR-B1 also display enhanced cytokine production in response to LPS (37). As a result, SR-B1 may play a role in sepsis through a variety of mechanisms in addition to its role in LPS clearance.

Phospholipid transfer protein (PLTP) is a protein involved in HDL metabolism by facilitating the transfer of phospholipids as well as modulating the size of HDL particles (41). Mice-deficient for PLTP have reduced HDL-C levels and reduced survival in a mouse model of endotoxemia (42). Conversely, transgenic overexpression of PLTP increases survival from endotoxemia (42).

In addition to these studies of transgenic mice, a number of pharmacological interventions have been studied in mouse models of sepsis. Administration of an ApoAI mimetic peptide, 4F, to rats increased levels of HDL-C and improved survival after cecal ligation and puncture (43). A 22-amino acid ApoAI mimetic peptide, ETC-642, also increased HDL-C and improved survival in cecal ligation and puncture model of sepsis (44). Recombinant HDL infusions have also been shown to attenuate the response to LPS in rabbits (45, 46). A constituted HDL preparation, CSL-111, which consists of ApoA1 complexed with phosphatidyl choline, has been shown to improve survival from cecal ligation and puncture (47).

The cholesteryl ester transfer protein mediates bidirectional transfer of cholesteryl esters and TGs between HDL and TG-rich lipoproteins. A gain-of-function genetic variant in CETP is associated with a greater reduction in HDL-C during sepsis in humans (48). This genetic variant, rs1800777, is robustly associated with increased 28-day mortality from sepsis in multiple cohorts (49) as well as with the risk of sepsis-related acute kidney injury (50). This finding raised the hypothesis that inhibition of CETP may be protective from sepsis. This concept was tested using the CETP inhibitor, anacetrapib, which was shown to significantly decrease mortality in the cecal ligation and puncture model of experimental sepsis (49). Although cecal ligation and puncture is considered a gold standard for sepsis research, it has a number of limitations, including variability induced by the length and degree of ligation (51). In humans, pneumonia is the most common cause of sepsis (52). Recently, treatment with anacetrapib was shown to reduce sepsis mortality in a Streptococcus pneumoniae model of pneumosepsis in mice (53). This finding provides further validation of the potential role of CETP inhibition as a treatment for sepsis.

The most well-known physiological function of HDL is in RCT, which consists of the trafficking of excess cholesterol from peripheral tissues to the liver, where it can excreted into the bile (54). RCT can occur though the direct pathway, in which cholesterol in HDL is removed via SR-B1, or the indirect pathway in which cholesterol in HDL is first shuttled to LDL and very low density lipoprotein via CETP, and then taken up via the LDL receptor (55) (Fig. 1). HDL can also transport cholesterol to the intestine in a process known as transintestinal cholesterol efflux (56). The first step in the RCT pathway is the efflux of cellular cholesterol by ABC transports (principally ABCA1 and ABCG1) to an apolipoprotein or lipoprotein acceptor. ABCA1 appears to play a role in the efflux of LPS from macrophages, and treatment of macrophages with an agonist of the liver X receptor increases LPS efflux in the presence, but not in the absence of ABCA1 (57). LPS-induced tolerance occurs to a greater extent in the absence of ABCA1, and deficiency of ABCA1 prolongs the LPS-induced tolerant state (57). Deficiency of ABCA1, ABCG1, or both, in macrophages enhances proinflammatory response to LPS (58, 59, 60). In addition to the role of RCT in promoting clearance of LPS (61), inflammation itself impairs RCT (62, 63), which may compromise the ability of HDL to clear pathogen-associated lipids during an infectious challenge. RCT and inflammation therefore appear to have a bidirectional relationship.

Fig. 1.

HDL promotes LPS clearance in sepsis. Invading gram negative bacteria release lipopolysaccharide (LPS) which interacts with toll-like receptors including toll-like receptor 4 (TLR4) in immune cells such as macrophages. This leads to the overwhelming immune response and organ dysfunction that is characteristic of sepsis. LPS can be bound by HDL and returned to the liver via the scavenger receptor B1 (SRB1). SRB1 plays a number of other roles in immune response in addition to LPS clearance. Macrophages express ABC transporters such as ABCA1 and ABCG1 that can efflux LPS and play a role in LPS-induced tolerance. The cholesteryl ester transfer protein (CETP) transports cholesterol from HDL to LDL or VLDL in exchange for triglycerides. Gain-of-function genetic variants in CETP are associated with reduced HDL cholesterol levels during sepsis and increased mortality. In mouse models of sepsis, inhibition of CETP leads to preserved HDL cholesterol levels during sepsis and improved survival.

Overall, these studies provide consistent and robust data that enhancing HDL levels via either genetic manipulation or pharmacological interventions is protective against sepsis in mice. One notable finding is that flux through the HDL system, rather than the static level of HDL-C, may be the most important determinant of the effect on response to sepsis as demonstrated by the finding that mice-deficient for SRB1 have increased susceptibility to sepsis, despite elevated HDL-C levels (36, 38). Nonetheless, despite these consistent preclinical results, it is important to recognize that findings from animal models of sepsis have a poor track record of clinical translation (64), and testing interventions that modulate HDL in humans will be essentially to definitively assess the clinical relevance of these findings.

Clinical studies of HDL in sepsis

Few studies have directly tested the hypothesis that HDL protects from sepsis in humans. The first cardiovascular outcome trial of a CETP inhibitor, the ILLUMINATE trial of torcetrapib, unexpectedly reported an increase in death from infection in the torcetrapib versus placebo group (9 versus 0) (3). However, this trend was not observed in other trials of CETP inhibitors (4, 5, 6), and CETP inhibition does not appear to impair LPS clearance (65), suggesting that the increased death from infection in ILLUMINATE may relate to known off-target effects of torcetrapib (66).

In a small study of 8 healthy volunteers, administration of 40 mg/kg of recombinant HDL prior to infusion of E. coli endotoxin 4 ng/kg significantly blunted the increase in TNF, IL-6 and IL-8, relative to placebo (67). Endotoxin-associated symptoms, including chills, myalgia and nausea were also reduced.

An open-label study of patients admitted to an intensive care unit with an intra-abdominal infection or urosepsis randomized patients to standard of care (n = 5), or one of 4 doses of CER-001, a preparation of synthetic HDL (68). Treatment with CER-001 was associated with reductions in plasma LPS as well as inflammatory cytokines (68). There were no statistically significant differences in 30-day survival. However, length of intensive care unit stay was shorter in patients who received CER-001 than placebo.

Intravenous administration of lipid emulsion therapy has been hypothesized to stabilize reductions in HDL and other lipoproteins during sepsis. In a phase 2 trial of patients with sepsis and low levels of total or HDL-C, administration of lipid emulsion did not result in a change in total or HDL-C, but was associated with an increase TG concentration (69).

Summary and future directions

A large body of work has established a robust association of low levels of HDL with infectious diseases and septic death. Mechanistic and preclinical studies add further support of an important immunomodulatory role of HDL, have identified a number of putative mechanisms by which HDL may impact sepsis survival, and have established that in animal models, genetic or pharmacological enhancement of HDL levels is protective from sepsis. Some, but not all, Mendelian randomization studies support a causal role of HDL in sepsis. Human data are more limited but also suggest a clinically relevant role of HDL in LPS clearance and limiting excessive inflammation.

Nonetheless, many questions remain. Which of the many functions of HDL are most important for their role in protection from sepsis? Are specific subclasses or specific lipid or protein constituents of HDL most critical for protection from sepsis? Are the effects of HDL most pronounced for sepsis due to Gram negative organisms (which release LPS), or are these effects equally relevant regardless of the microbiological cause? Are the effects of HDL in sepsis most important in macrophages, or in other cell types or tissues? And, what mechanism of augmenting HDL (small molecules, recombinant HDL infusion, nutritional supplementation) are most suitable for studies in humans with sepsis?

The findings reviewed here raise many intriguing possibilities about future clinical translation and whether HDL therapeutics, after failing in cardiovascular disease, may be successful in reducing morbidity and mortality from sepsis. However, these expectations should be tempered by past experience. Will HDL find its elusive primary purpose as a component of the immune system, or will sepsis join the long list of other disease states for which HDL serves as a marker but ultimately plays a noncausal role?

Conflict of interests

The authors declare that they have no conflicts of interest with the contents of this article.