Impact of the type of bacteria on high-density lipoprotein cholesterol level during sepsis

Abstract

High-density lipoproteins (HDL) are nanoparticles that enable the reverse transport of cholesterol and have multiple endothelioprotective properties that may play a role in sepsis, where HDL-cholesterol (HDL-C) concentrations typically plummet, which is strongly linked to poor outcomes. While HDL has the capacity to capture and eliminate lipopolysaccharides (LPS), which may explain this decrease in HDL-C concentration, this effect is less clear for Gram-positive bacteria via binding to lipoteichoic acid (LTA). The aim of this study is to determine whether the type of bacteria influences HDL-C concentrations in septic patients. This was a prospective, observational, single-center study of septic patients hospitalized in the intensive care unit (ICU). Patients were stratified into three groups according to the type of bacterial sepsis: infections caused by one or more Gram-negative bacteria, infections caused by one or more Gram-positive bacteria, or mixed infections involving both Gram-negative and Gram-positive bacteria. A total of 202 patients were prospectively and consecutively included in this study. HDL-C concentrations were below reference values. Additionally, there was no association between the type of bacteria and clinical outcomes. Interestingly, baseline HDL-C concentrations were similar between groups. HDL-C concentrations are low in septic conditions, and there is no relationship between HDL-C levels and the type of bacteria, which may indirectly support the scavenger action of HDL particles on all types of bacteria, including Gram-positive. Before conducting a phase 3 trial comparing recombinant HDL injection versus placebo during sepsis to improve HDL levels, further mechanistic studies are needed.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-18473-1.

Introduction

HDL are nanoparticles that provide reverse cholesterol transport (RCT) from peripheral tissues to the liver¹. The concentration of HDL-C has been shown to be inversely correlated with the occurrence of cardiovascular events, giving HDL a major cardioprotective effect². In addition to this major function, HDL has pleiotropic, endothelioprotective, antioxidant, anti-inflammatory and anti-apoptotic effects³. It has been also demonstrated that HDL particles have the capacity to capture bacterial lipopolysaccharides (LPS) from Gram-negative bacteria, with ultimate biliary elimination⁴. Although less described, HDL also has the capacity to bind lipoteichoic acid (LTA) from Gram-positive bacteria to remove them⁵.

During sepsis, there are both a drastic reduction in HDL-C concentration and a major dysfunction of HDL particles, which lose their endothelioprotective properties. This loss of function and dramatic decrease of HDL-C level during sepsis are strongly associated with poor outcome³.

There are several possible, albeit partially proven, explanations for the reduction in HDL-C concentration during sepsis: 1-/ a reduction in hepatic production of HDL and/or Apolipoproteine A1 (major apolipoproteins in HDL particles) leads to a decrease in HDL-C concentration, while up-regulation of the hepatic HDL receptor (SRB-1) increases HDL clearance. 2-/ Sepsis-induced microvascular dysfunction leads to capillary leakage, allowing HDL to accumulate in the target organs. 3-/ the scavenger effect of HDL, resulting in endotoxin-bacteria binding and hepatic elimination, may explain this decrease in plasma HDL-C concentrations.

While LPS and HDL interactions appear to be well documented, binding to LTA in Gram-positive bacteria is less clear. Thus, the aim of this study was to compare HDL-C concentrations in septic patients as a function of bacterial type.

Methods

HIGHSEPS cohort is a prospective, observational monocentric study conducted in the surgical ICU of Bichat-Claude Bernard University Hospital, Paris, France. The methods of this cohort have been previously published in a study involving 205 patients⁶. This study was approved by the French Society of Anesthesiology and Critical Care Medicine Research Ethics Board (HIGHSEPS study, IRB number 00010254). Written informed consent was obtained from participating patients. All methods were carried out in accordance with relevant guidelines and regulations.

All patients recruited from May 2016 to April 2020 admitted for septic shock or sepsis according to the criteria of the Surviving Sepsis Campaign were included⁷. All patients with preexisting liver disease, such as cirrhosis, fatty liver disease or liver cancer, and immunocompromised patients (acquired immune deficiency syndrome or solid organ transplantation) were excluded from the study. Furthermore, for this analysis on the relationship between Gram and HDL-C changes during sepsis, patients with negative microbiological samples were excluded. For the included patients, three groups were defined based on bacterial Gram stain: Gram-negative group, comprising patients with infections caused by one or more Gram-negative bacteria; Gram-positive group, comprising patients with infections caused by one or more Gram-positive bacteria; and mixed group, comprising patients with infections caused by both Gram-negative and Gram-positive bacteria.

Patient demographics, diagnosis, Simplified Acute Physiology Score II (SAPSII) and Sepsis-related Organ Failure Assessment (SOFA) severity scores, organ supportive therapies, including renal replacement therapy and vasopressor use, and clinical data were prospectively collected. Data regarding the site of infection and microbiology were compiled. Data on ICU and in-hospital mortality at 28 days (Day 28), duration of mechanical ventilation, number of days alive without mechanical ventilation at Day 28, length of ICU and hospital stay were collected.

At admission (day 1), at 48 h (day 3), and on the day of ICU discharge (discharge day), plasma concentrations of total cholesterol, HDL-C, LDL-C, and triglycerides were measured plasma concentrations of total cholesterol, HDL-C, low-density lipoprotein (LDL-C), and triglycerides were measured. These analyses were performed in the Biochemistry Laboratory of Bichat Claude-Bernard Hospital. Total cholesterol (TC), HDL-C, LDL-C and triglyceride concentrations were determined by routine enzymatic assays (CHOL, HDL-C and TRIG methods, Dimension VISTA^® System, Siemens Healthineers™). The reference values for these assays were HDL-C > 1.40 mmol/l, TC4.40 < N < 5.20 mmol/l and triglycerides 0.50 < N < 1.7 mmol/l. According to the recommendations of the French National Authority for Health 2017 and the European Society of Cardiology 2016, LDL-C concentration targets have been established based on vascular risk factors.

When a patient was included in this study, the physician in charge of the patient asked the patient himself or his family whether a complete lipid panel had been performed within the 2 years prior to the patient’s ICU admission. This assessment had to have been performed outside of any infectious episode. If available, these results were considered as the prehospitalization lipid test.

Statistical analysis

Continuous variables are expressed as medians with interquartile ranges (IQRs) and were compared with the Mann-Whitney U test or the Kruskal–Wallis test as appropriate. Categorical variables were expressed as counts and percentages and were compared with Fisher’s exact test or the chi-square test. Two sensitivity analyses were conducted to assess the robustness of our results.

First, we performed a subgroup analysis including only the most severely ill patients, defined as those with a SOFA score greater than or equal to the population median. Second, we conducted a ponderal analysis with an inverse probability of treatment weighting (IPTW) to control enables adjustment for the main confounding. First, inverse probability weights were estimated using the WeightIt package⁸, based on a multinomial logistic regression modeling the probability of Gram staining type as a function of age and SOFA score at admission. Covariate balance was assessed using cobalt package⁹. Second, the analysis of the lipid panel and estimation of p-values were conducted using weighted models from the survey package¹⁰, incorporating the previously estimated weights. In this analysis mean and standard error of the mean are plotted, and the Wald test was used to compare the groups.

Comparisons over time for biological variables were performed using a mixed model.

All statistical analyses were performed using R software version 4.4.1 (R Core Team, 2014, R Foundation for Statistical Computing, Vienna, Austria, https://www.r-project.org/). Figures were produced using the ‘ggplot2 package’, and statistics were produced using the ‘stat package’. Missing values were not imputed. A p value < 0.05 was considered statistically significant.

Results

226 patients were consecutively and prospectively included from May 2016 to April 2020. Among these patients, twenty-one had non-documented sepsis and three had fungal sepsis. After exclusion of these twenty-four patients, 202 patients were finally included in the analysis. Of the 202 septic patients, 50 patients has a Gram-positive sepsis, 73 patients has a Gram-negative sepsis, and 79 had mixed Gram sepsis with both Gram-negative and Gram-positive bacteria (Table 1).

Table 1.

Type of sepsis and microorganisms isolated according to the different groups.

Characteristic	Overall, N = 202 (100%)1	Gram-positive, N = 50 (25%)1	Gram-negative, N = 73 (36%)1	Mixed Gram, N = 79 (39%)1	p-value2
Peritonitis	71 (35)	12 (24)	17 (23)	42 (53)	< 0.001
Pyelonephritis	39 (19)	3 (6.0)	31 (42)	5 (6.3)	< 0.001
SSTI	35 (17)	16 (32)	5 (6.8)	14 (18)	0.001
Pneumonia	33 (16)	10 (20)	11 (15)	12 (15)	0.721
Other infections	20 (9.9)	7 (14)	9 (12)	4 (5.1)	0.143
Bacteriemia	64 (32)	16 (32)	32 (44)	16 (20)	0.008
Gram negative infections
Enterobacterales	131 (65)	0 (0)	69 (95)	62 (78)	< 0.001
E. coli	86 (43)	0 (0)	43 (59)	43 (54)	< 0.001
K. pneumoniae	28 (14)	0 (0)	14 (19)	14 (18)	0.005
Enterobacter spp	27 (13)	0 (0)	11 (15)	16 (20)	0.004
Other enterobacterales	31 (15)	0 (0)	13 (18)	18 (23)	0.002
Anaerobic gram-negative bacteria	14 (6.9)	0 (0)	2 (2.7)	12 (15)	< 0.001
Pseudomonas aeruginosa	12 (5.9)	0 (0)	2 (2.7)	10 (13)	0.005
Other Gram negative bacteria	9 (4.5)	0 (0)	3 (4.1)	6 (7.6)	0.125
Gram positive infections
Streptococci	51 (25)	26 (52)	0 (0)	25 (32)	< 0.001
Enterococci	43 (21)	2 (4.0)	0 (0)	41 (52)	< 0.001
Staphylococci	32 (16)	24 (48)	0 (0)	8 (10)	< 0.001
Anaerobic gram-positive bacteria	15 (7.4)	4 (8.0)	0 (0)	11 (14)	0.001
Other Gram positive bacteria	8 (4.0)	5 (10)	0 (0)	3 (3.8)	0.015
Oropharyngeal flora	10 (5.0)	0 (0)	0 (0)	10 (13)	< 0.001
Fungus	17 (8.4)	3 (6.0)	4 (5.5)	10 (13)	0.259

Continuous variables are expressed as median and interquartile range (IQR) and were compared using the Mann-Whitney U test. Categorical variables are expressed as n (%) and were compared with Fisher’s exact test.

Other infections is a mix of vascular infection, osteomyelitis, mediastinitis, meningitis, endocarditis, angiocholitis.

Other Enterobacterales includes the following bacteria : Campylobacter jejuni, Citrobacter spp, Hafnia alvei, Morganella morganii, Proteus spp, Providencia rettgeri, Serratia ureilytica.

Other Gram-negative aerobic bacteria includes the following bacteria : Acinetobacter baumanii, Haemophilus spp, Moraxella catarrhalis, Mycoplasma pneumoniae, Neisseria subflava, Pasteurella spp.

Other Gram-positive aerobic bacteria includes the following bacteria : Corynebacteirum spp, Rothia spp.

Gram-ànegative anaerobic bacteria includes the following bacteria : Bacteroides fragilis, Bacteroides thetaiotaomicron, Bacteroides uniformis, Prevotella spp.

Gram positive anaerobic bacteria includes the following bacteria : Anaerococcus spp, Actinomyces spp, Clostridium spp, Lactobacillus spp, Bifidobacterium spp.

SSTI: Skin and soft tissue infections.

¹n (%).

²Pearson’s Chi-squared test; Fisher’s exact test.

No difference was noted in severity on admission or mortality according to the Gram staining of the bacteria. The only significant difference between the groups was the age of the population, though this had minimal clinical relevance (Table 2).

Table 2.

Patient’s general characteristics and outcome.

	Overall, N = 202 (100%)1	Gram-positive, N = 50 (25%)1	Gram-negative, N = 73 (36%)1	mixed Gram, N = 79 (39%)1	p-value2
Age (years)	63 [51, 72]	59 [42, 64]	66 [54, 74]	64 [53, 73]	< 0.001
Male	109 (54)	29 (58)	38 (52)	42 (53)	0.796
SAPS II on admission	56 [40, 68]	50 [38, 65]	58 [40, 68]	57 [42, 69]	0.421
SOFA on admission	7.0 [4.0, 9.0]	6.0 [2.0, 8.8]	8.0 [5.0, 10.0]	7.0 [5.0, 8.0]	0.075
Lactate (mmol/l)	2.30 [1.45, 3.60]	2.00 [1.45, 3.45]	2.30 [1.40, 4.10]	2.40 [1.60, 3.40]	0.689
TC (mmol/l)	2.24 [1.71, 2.74]	2.24 [1.85, 2.72]	2.45 [1.86, 3.05]	2.08 [1.58, 2.61]	0.048
Triglyceride (mmol/l)	1.62 [1.07, 2.29]	1.53 [1.01, 2.09]	2.12 [1.45, 2.67]	1.36 [0.91, 1.92]	< 0.001
HDL-c (mmol/l)	0.41 [0.28, 0.67]	0.48 [0.29, 0.67]	0.32 [0.26, 0.66]	0.51 [0.32, 0.70]	0.089
LHDL-c (mmol/l)	0.99 [0.56, 1.50]	1.08 [0.68, 1.49]	1.02 [0.60, 1.72]	0.83 [0.49, 1.34]	0.071
Length of MV (days)	2 [0, 8]	2 [0, 9]	2 [0, 6]	3 [1, 8]	0.054
ICU length of stay (stay)	7 [3, 15]	6 [3, 19]	6 [3, 11]	8 [5, 16]	0.257
ICU mortality	36 (18)	8 (16)	13 (18)	15 (19)	0.911
28-day mortality	36 (18)	8 (16)	13 (18)	15 (19)	0.911
90-day mortality	48 (24)	10 (20)	17 (24)	21 (27)	0.693
One year mortality	57 (29)	11 (23)	18 (26)	28 (36)	0.215

Continuous variables are expressed as median and interquartile range (IQR) and were compared using the Mann-Whitney U test. Categorical variables are expressed as n (%) and were compared with Fisher’s exact test. HDL-c, High-density lipoproteins; ICU: intensive care unit; LDL-c, Low-density lipoproteins; MV: mechanical ventilation; TC, total cholesterol.

In the overall population, TC concentration was 2.24 mmol/l [1.74–2.73], TG concentration was 1.62 mmol/l [1.07–2.29], HDL-C 0.42 mmol/l [0.28–0.67] and LDL-C 1.00 mmol/l [0.56–1.49]. Except for TG, all values were below the reference values.

To assess the effect of bacterial type on lipid and lipoprotein levels, bacteria were first compared according to their group within their Gram stain. No difference were noted between Gram negative bacteria (Supplemental figure S1) and between Gram positive bacteria (Supplemental figure S2).

Then, Gram-positive bacteria were compared with Gram-negative bacteria, as well as with cases involving both types (mixed Gram) (Fig. 1). In this analysis, the main significant finding was a lower triglyceride concentration at admission in the presence of Gram-positive bacteria. This difference remained significant in a sensitivity analysis including only the largest groups within each Gram stain category—Enterobacterales for Gram-negative and Streptococci/Enterococci for Gram-positive bacteria (Supplemental Figure S3). No differences were observed in HDL concentrations.

Fig. 1.

Comparison between total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol and triglyceride concentrations on admission, according to Gram staining.

Furthermore, to assess the robustness of our results, we conducted two sensitivity analyses. First, we performed a subgroup analysis including only the most severely ill patients, defined as those with a SOFA score greater than or equal to the population median⁷. This analysis confirmed our findings and further emphasized the difference in triglyceride levels between the groups (Fig. 2).

Fig. 2.

Comparison between total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol and triglyceride concentrations on admission, in a sub-group analysis that only included patients with a SOFA score greater than or egal to 7.

Second, we performed an IPTW analysis. In this analysis, the differences in age and SOFA score between groups disappeared (Supplemental Figure S4). Regarding the patients’ lipid profiles, the results were consistent with the standard analysis: no major differences were observed in total cholesterol, HDL, or LDL levels. The previously noted difference in triglyceride levels between Gram-positive and Gram-negative bacteria no longer reached statistical significance (p = 0.051; Fig. 3).

Fig. 3.

Comparison between total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol and triglyceride concentrations on admission, in a sensitive analysis using inverse probability of treatment weighting.

Moreover, we evaluated whether differences could be observed according to the type of infection (Fig. 4). In these analyses, the only significant differences were noted between pyelonephritis and peritonitis/other infections regarding total cholesterol and triglyceride concentrations. These findings are consistent with the distribution of bacterial pathogens, as pyelonephritis cases were predominantly caused by Gram-negative bacteria (80%), whereas peritonitis cases involved mixed infections in approximately 60% of cases.

Fig. 4.

Comparison between total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol and triglyceride concentrations on admission according to the type of infection.

Finally, we evaluated the evolution of the lipid profile during the first week (Fig. 5). Once again, the results were quite similar across groups, with no significant differences for total cholesterol and HDL cholesterol, both of which remained very low on days 1 and 3 regardless of gram staining. In contrast, triglyceride levels showed a significant increase, peaking at day 3 in patients with gram-negative infections, while remaining stable in those with gram-positive or mixed infections.

Fig. 5.

Comparison over time between total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol and triglyceride concentrations according to Gram staining. D1, Day 1 (admission); D3, Day 3; DD, discharge day.

Discussion

The main conclusion from these results is that basal HDL-C concentration upon ICU admission for sepsis is the same regardless of the type of bacteria. Interestingly, our study shows that patients with Gram-negative sepsis have a statistically higher TG concentration than others.

The first finding is notable as it offers insights into the interaction between lipoproteins and bacteria. The causes of low HDL-C concentrations during sepsis are poorly documented. Major capillary leakage during sepsis and initial hepatic dysfunction, leading to reduced HDL synthesis, are two primary hypotheses. A third possible explanation is bacterial capture by HDL, resulting in an HDL-bacteria complex that is eliminated via the biliary route. Clinical and preclinical studies have successfully demonstrated this hypothesis, particularly through imaging and LPS assays^11,12. However, there is still uncertainty regarding HDL’s capacity to capture and eliminate Gram-positive bacteria. The absence of any difference in HDL-C levels between Gram-positive and Gram-negative bacteria in our study further supports the potential scavenger effect of HDL, regardless of the type of bacteria (i.e., in relation to LTA and LPS). Nevertheless, this remains a hypothesis that requires confirmation through preclinical studies.

Beyond HDL-C, it is interesting to note in our study that TC and LDL-C also decrease over time, with a secondary increase when the patient is discharged from the ICU in connection with the resolution of sepsis. These findings are not surprising and have already been described in other studies^13,14 highlighting lipids and lipoproteins dysregulation during inflammation. Indeed, liver dysfunction during sepsis probably plays a significant role in the synthesis of all lipoproteins (HDL, LDL, VLDL…), which may explain the decrease in LDL-C¹⁵. Furthermore, as with HDL, the scavenger effect of LDL has also been described⁵ and a decrease in the LDL pool will also have an effect on this LPS and LTA clearance function. Finally, as described for HDL, major capillary dysfunction during sepsis also explains a significant part of the decrease in both LDL nanoparticles and cholesterol.

The dramatic decreased concentration of HDL-C during sepsis and the major dysfunction of HDL particles have motivated numerous teams to inject functional HDL into animal models of sepsis with encouraging results on survival and morbidity^11,16,17. More recently, Stasi et al. injected recombinant HDL particles (CER-001) into patients with Gram-negative sepsis during a phase 2 trial with very encouraging results¹². A phase 3 trial in shock septic patients is currently being designed. The results of our study provide further justification for testing these functional HDL particles in all septic patients, regardless of the type of bacteria, including Gram-positive bacteria.

The significant higher TG concentrations in the Gram-negative group is somewhat perplexing, as there is no clear pathophysiological rationale for it. Inflammation has been shown to increase TG levels, with positive correlations to several inflammatory markers, such as cytokines. Although we did not measure these parameters in our study, it is interesting to note that there was no difference between groups in terms of severity (SOFA and SAPSII scores) or lactatemia. Measuring specific pro-inflammatory cytokines could have helped to explain these notable differences in our study.

Our study has several limitations. First, it is based on an extrapolation of data and lacks mechanistic insights. For example, LTA and LPS assays would have been useful. Second, our results only account for basal HDL-C levels and do not consider HDL particle dysfunction, which seems to play an important role in patient outcomes. If the injection of functional HDL is to be tested, the basal functionality of HDL particles, in addition to HDL-C concentration, is crucial information to gather.

More broadly, our results contribute to a better understanding of why HDL-C concentrations decrease dramatically during sepsis and warrant further mechanistic preclinical studies.

Conclusion

We have shown that HDL-C concentrations upon ICU admission for sepsis were low but had no clear link to the type of bacteria. Before conducting a phase 3 trial (HDL particles versus placebo), a better understanding of the interaction between HDL and bacteria is necessary.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: CDT, ST; Data curation: CDT, TR, AB; Investigation, JS, NZ, AS, BL-J, EA; ABMethodology: CDT; Software: CDT; Supervision: PM; Writing—original draft, ST; Writing—review and editing: CDT, ST, PM, OMAll authors have read and agreed to the published version of the manuscript.

Data availability

Data are available upon request by contacting Christian de Tymowski [christian.detymowski@aphp.fr](mailto: christian.detymowski@aphp.fr).

Declarations

Competing interests

The authors declare no competing interests.

Ethics

This study was approved by the French Society of Anesthesiology and Critical Care Medicine Research Ethics Board (HIGHSEPS study, IRB number 00010254). Written informed consent was obtained from participating patients.