In this study we observed distinct inflammatory effects of HDL and LDL when co-administered with LPS, and describe marked peripheral metabolic changes associated with the acute phase response. We have shown that the coadministration of human HDL with LPS exacerbates brain markers of neuroinflammation in mice. By contrast, human LDL seemed to prevent LPS-induced inflammation in both the brain and liver. The behavioural tests showed that a mixed treatment LDL + LPS significantly reduced sickness behaviour compared to LPS treatment. The augmentation of LPS-induced brain inflammation in mice may be due to the functional capacity of HDL to bind LPS, and, through metabolic changes associated with the acute phase response, exert a pro-inflammatory effect on tissues distal to the liver. Indeed, LPS + HDL resulted in significantly increased levels of observable endotoxin in the brain while the levels observed in LPS + LDL animals were no different from LPS alone.

The APR is a well-orchestrated series of systemic physiological processes occurring at the onset of the innate immune response. Bacterial infections typically evoke a strong APR. LPS, a gram-negative bacterial endotoxin known to induce inflammation [42], can directly activate innate immune cells such as tissue macrophages, blood monocytes, and neutrophils [3]. This leads to the local induction of key pro-inflammatory cytokines, TNF and IL-1ß [44, 51], and the profound modulation of protein synthesis by hepatocytes in the liver [10]. These hepatic alterations translate to substantial changes in HDL protein composition, most notably with ApoA-I being displaced from HDL by the incorporation of SAA and LBP [52].

To determine whether the distinct behavioural effects of LPS in combination with HDL or LDL administration were due to underlying differences in brain or peripheral inflammation, we investigated the gene expression of key cytokines involved in the acute phase response, including IL-1ß, TNF, CXCL1, and CCL2. In the brain, LPS treatment led to a robust increase in the expression of all pro-inflammatory cytokines. The co-administration of HDL augmented this inflammatory response, significantly elevating expression of IL-1ß and TNF, though CCL2 expression was reduced (Fig. 2A). IL-1ß and TNF are two of the most sensitive cytokines associated with TLR4 signalling, so any increased inflammation due to HDL co-administration is likely to be discerned through increased transcription of these inflammatory mediators if not others [41]. Peripherally, premixing LPS with HDL also tended to exacerbate the inflammatory response, with a significant increase in TNF gene expression in the liver (Fig. 2). Importantly, the administration of human HDL or LDL alone did not elicit a pro-inflammatory immune response (Fig. 2).

Given the increase in central inflammation in the HDL cohorts, we hypothesised that HDL, but not LDL, may be shuttling bound LPS to the brain. To investigate this, we conducted an endotoxin assay on brain homogenates. Whilst mice that received an intraperitoneal injection of saline only or LPS only had equivalently low levels of endotoxin, mice that were injected with a premixed solution of LPS and HDL showed significantly increased levels of brain endotoxin (Fig. 6B). This may suggest that HDL, in an inflammatory environment, is able to transport LPS to the brain. It should be noted that we could not distinguish between the detection of endotoxin in brain vasculature from brain parenchyma, and thus whether HDL is able to transport LPS across the blood brain barrier and directly stimulate parenchymal macrophages, though this has been previously demonstrated [58].

In contrast to the effect of HDL on LPS-induced inflammation, LDL attenuated both central and peripheral pro-inflammatory gene expression (Fig. 2). As LDL may disperse its cargo away from the liver toward other cells, predominately through the interaction between ApoB-100 and LDL receptors [59], this observation may be due to the ability of LDL to scavenge LPS without overtly activating proximal innate immune cells. Oxidation of administered LDL is unlikely to contribute to the effects observed in this study. Our results are somewhat counterintuitive because the inflammatory properties of modified LDLs have been a principal focus of atherosclerosis research for decades where they contribute to inflammation in the arterial wall [47]. For example, modified LDLs, such as Oxidized LDL (oxLDL) and minimally modified LDL (mmLDL), are recognized by the scavenger receptor, and which mediates the uptake and formation of lipid-loaded foam cells that are typical of atherosclerotic lesions. It is known that LDL maybe modified in the arterial intima, and more slowly in plasma circulation. However, they have been detected in blood only at a very low concentration, which is strongly antioxidant enriched [13]. Fully oxLDL are reportedly quickly cleared from circulating blood mainly by hepatic Kupffer cells [32]. Indeed, the Kupffer cells removed > 90% or injected modified LDLs in 5 min. Thus it is unlikely that there is to be a contribution from modified LDLs in our experiments in the acute timeframe.

In addition, oxLDL reduces cytokine production after activation of plasmacytoid dendritic cells in vitro, suggesting that this may be one potential mechanism by which administered LDL may be anti-inflammatory in the presence of endotoxin [18]. More specific to this study, pre-treatment of LPS-stimulated macrophages with oxLDL reduced mRNA expression of the pro-inflammatory cytokine IL-6 and increased the expression of the anti-inflammatory cytokine IL-10 via interacting with macrophage scavenger receptors [48]. Thus, premixing LPS and LDL may partially reduce the extent of the acute phase response after LPS-induced inflammation. Further experiments are required to clarify the precise mechanisms by which HDL and LDL interact with LPS to modulate the immune response.

Consistent with the findings from qPCR, acute peripheral LPS administration also showed marked changes on peripheral metabolism that reflect altered energy requirements during the APR. One of the key changes in the plasma metabolome was the effect of LPS on increasing the level of plasma BCAAs (Fig. 3A and B). BCAA elevation may be downstream from LPS-mediated immune activation, as BCAAs are known to have pro-inflammatory effects via mTORC1 [62]. Other amino acids in plasma were elevated by LPS treatment, including glutamate, glutamine, lysine, and phenylalanine (Fig. 3 C-F). When stimulated, immune cells greatly increase their metabolic demand [55]. Consequently, in proliferating immune cells amino acids are preserved at the expense of glucose, as glycolysis becomes the key form of energy metabolism [2]. This switch to glycolysis is particularly pronounced in LPS-stimulated organisms via the activation of TLR4, known as the Warburg effect [49]. As a result, plasma glucose levels were significantly depleted (Fig. 4D). Like in plasma, liver glucose was also depleted in LPS-treated animals (Fig. 5F). Glutamate (Fig. 5A), glutamine (Fig. 5B), and acetate (Fig. 5E) were increased by LPS treatment relative to saline and may function as precursors for immune-related anabolic processes.

Peripheral metabolism was also differentially affected by the association of LPS with HDL or LDL. In line with the findings from behavioural and gene expression analysis, LDL attenuated the LPS-induced changes in glucose, citrate, and taurine (Fig. 4) while LPS + HDL tended to exacerbate those changes. In the liver, glutamine levels were significantly higher in LPS + HDL animals compared to LPS and LPS + HDL animals. Given the central role glutamine plays in the metabolic reprogramming of the innate immune response [5, 35, 36], this may reflect the increased liver inflammation in LPS + HDL treated animals as demonstrated by gene expression analysis of pro-inflammatory cytokines (Fig. 2).

In the brain, the effect of peripheral LPS administration on metabolism was less pronounced than that observed in either blood or liver suggesting that, while coadministration of lipoproteins resulted in increased endotoxin in the brain, the behavioural changes observed are largely a result of modulation of the peripheral immune response and metabolic pathways. The limited number of metabolic changes observed in the brain is somewhat expected as peripheral administration of LPS will produce a predominantly peripheral immune response although the use of whole brain lysates for metabolomics analysis may have also played a role; the heterogeneity of the tissue may dampen region-specific group differences in comparison to the more homogenous liver.

Interestingly, a significant reduction in brain NAD concentration were only observed when LPS was co-administered with (Fig. 6A-v) suggesting that this may be a direct result of increased brain endotoxin levels. NAD has shown to be reduced in both peripheral and brain inflammatory diseases [17, 63], and its levels are directly linked to neuronal function and survival [45]. Further research on the link between endotoxemia and neuroinflammation should investigate the link between NAD levels and the resolution of inflammatory activity.

Citrate was significantly elevated in the brain of LDL-treated animals, compared to saline (6Avi). This was irrespective of whether LDL was first premixed with LPS and suggests a unique role of LDL in being able to increase brain citrate levels. Citrate has previously been demonstrated to reduce brain inflammation and oxidative stress in a mouse model of LPS-induced systemic and central inflammation [1]. It is possible that here the LDL-mediated increase in brain citrate protected against LPS-induced brain inflammation, resulting in the reduced levels of brain IL-β and TNF observed, though further experiments are needed to clarify this.

While an abundance of research has demonstrated the involvement of specific HDL-associated proteins with the innate immune system, only recently have the advent of proteomic techniques in mass spectrometry demonstrated the functional capacity of HDL. This extends the general dogma of lipoproteins as vehicles of lipid transport to include complex roles in innate immunity and haemostasis [46]. Indeed, while the protein complement of LDL is dominated by apolipoprotein B (Figure S1E), a heterogeneous assortment of proteins have been described as constituents of HDL [52]. Under infectious and inflammatory conditions that elicit the APR, HDL undergoes significant remodelling and increased pro-inflammatory activity [57]. Recent research indicates that this is due in part to the association of SAA with HDL during infection [19]. In both humans and mice, SAA is one of the major acute phase proteins (APPs), in which synthesis can be upregulated > 1000-fold during the APR [10]. SAA is the primary apolipoprotein associated with HDL during the APR [8]. The ability of SAA to transiently but dramatically increase the levels of several cytokines, including IL-1ß, CCL2, and TNF through the stimulation of monocytes and macrophages has also been demonstrated [54].

Increased endotoxin levels in the brain likely explain the pro-inflammatory metabolite profiles and the increased levels of inflammatory cytokines observed in LPS + HDL treated animals. In contrast, the extent by which LDL modulated endotoxin transport and brain metabolism is unlikely to fully account for the significant amelioration of sickness behaviour and inflammatory cytokines observed. Thus, the anti-inflammatory properties of LDL observed in this study are most likely a result of modification of peripheral metabolic processes and the acute phase response as supported by the metabolic data presented. Lipoproteins – along with other components of the innate immune system – may act as a double-edged sword in endotoxemia and sepsis. Binding LPS may ultimately function to clear LPS from blood and tissues but may also induce an inflammatory response elsewhere. The association of LPS and HDL during infection may contribute to the pro-inflammatory activation of parenchymal macrophages of the CNS contributing to sepsis induced neuroinflammation.

While we confirmed the presence of ApoA-I, and absence of ApoB, on administered HDL particles (Fig. S1E), future work should aim to differentiate further between HDL subclasses and their immunomodulatory effects. Subsequent proteomic and lipidomic analysis of HDL and LDL lipoproteins isolated from mice in the aftermath of the acute phase response could help determine the key components of HDL that contribute to its pro-inflammatory effects. Moreover, although LPS displays greatest affinity for HDL amongst the lipoprotein subclasses [29], there is evidence to suggest that HDL subsequently redistributes endotoxin to LDL under the physiological conditions of the acute phase response. This process is complementary to the acute-phase remodelling of HDL and may contribute to its pro-inflammatory effects [30]. Therefore, the rate of exchange of LPS between lipoprotein subclasses when co-administered with HDL or LDL should also be investigated.

We recognise the limitations of this study. While we identified a protective role of LDL, when premixed with LPS, on brain and liver pro-inflammatory gene expression, liver endotoxin levels were not assessed. Subsequent experiments are required to determine the specific mechanisms behind the anti-inflammatory effects of LDL, for example, whether LDL shuttles LPS to the liver for subsequent secretion into the intestines [12]. This may be achieved by utilising a fluorescent labelled lipopolysaccharide conjugate, or isolating LPS + LDL particles from target organs. In addition, while the measurement of cytokine expression by mRNA levels was informative, this unilateral approach to assessing inflammation is a limitation of our study. Complementary protein and histological readouts of the inflammatory response would be a useful addition to future experiments.

It would also be useful to understand how different timepoints modify the relationship between lipoprotein administration and the host inflammatory response to LPS. Indeed, a further limitation of this study is the single dose regime of lipoprotein administration and single timepoint. While the single 6-hour timepoint was determined by our interest in the acute phase response and associated sickness behaviours, an extended timepoint and/or treatment regime may reveal additional roles for modified lipoproteins, for example oxLDL. Only male mice were used in these experiments, and sex differences should be a feature of follow-up studies. Lastly, it should also be noted that the translatability of the results to humans is not yet clear, due to the different lipoprotein profiles between mice and humans.

The results of this study present some possible therapeutic implications. Most notably, elevated levels of HDL, in the presence of LPS, may exacerbate neuroinflammation. Consequently, an ideal therapeutic range of HDL may be advisable, as opposed to simply a recommended minimum blood concentration. This is supported by recent clinical evidence, whereby both low and high HDL levels were associated with significantly increased odds of all-cause and cause-specific mortality [31, 60]. While HDL is also thought to play a protective immunomodulatory role in sepsis, this appears to be via an ApoA-I dependent mechanism [6, 33]. HDL and LDL may be useful targets in modulating the acute phase response, and greater insight into their immunomodulatory roles may help prevent neuroinflammation and subsequent neurodegeneration associated with endotoxemia.

In summary, our results indicate that HDL and LDL play distinct roles during the acute phase of LPS-induced inflammation. LDL but not HDL can ameliorate the LPS-induced sickness behaviour represented by an open field test. Furthermore, we showed HDL can shuttle LPS to the brain to promote neuroinflammation, but LDL has anti-neuroinflammatory properties in this LPS-induced mouse model. This study provides a platform to further explore how lipoproteins and other immunometabolites mediate brain-periphery communication during the acute phase of LPS-induced inflammation.