Very-low and low-density lipoproteins induce neutral lipid accumulation and impair migration in monocyte subsets

Blood monocytes are heterogeneous effector cells of the innate immune system. In circulation these cells are constantly in contact with lipid-rich lipoproteins, yet this interaction is poorly characterised. Our aim was to examine the functional effect of hyperlipidaemia on blood monocytes. In the Ldlr−/− mouse monocytes rapidly accumulate cytoplasmic neutral lipid vesicles during hyperlipidaemia. Functional analysis in vivo revealed impaired monocyte chemotaxis towards peritonitis following high fat diet due to retention of monocytes in the greater omentum. In vitro assays using human monocytes confirmed neutral lipid vesicle accumulation after exposure to LDL or VLDL. Neutral lipid accumulation did not inhibit phagocytosis, endothelial adhesion, intravascular crawling and transmigration. However, lipid loading led to a migratory defect towards C5a and disruption of cytoskeletal rearrangement, including an inhibition of RHOA signaling. These data demonstrate distinct effects of hyperlipidaemia on the chemotaxis and cytoskeletal regulation of monocyte subpopulations. These data emphasise the functional consequences of blood monocyte lipid accumulation and reveal important implications for treating inflammation, infection and atherosclerosis in the context of dyslipidaemia.

strongly suggest a possible inhibitory effect of dyslipidemia on monocytes and macrophages in infection and pro-atherosclerotic effects in cardiovascular disease.
Previous work on lipoprotein biology and the mononuclear phagocyte system has mainly focused the role of LDL on macrophages and dendritic cells (DC) in atherosclerosis progression 17 . However, circulating lipophages have been described in-vivo since the 1960s, which are most likely neutral lipid positive blood monocytes after high fat feeding [18][19][20][21][22][23][24] . More recently, it has been shown that these 'foamy' monocytes can enter early atherosclerotic plaques 25 and hypertriglyceridemia can mediate non-classical margination and macrophage tissue accumulation 26 . Together these data raise the possibility that dyslipidaemia can affect monocyte phenotype and possibly functionality. Therefore given the increasing disparity between monocytes and tissue macrophages 1 and the distinct kinematic phenotypes of monocyte subpopulations in homeostatic and inflammatory conditions during atherosclerosis 25 , further investigation was warranted into the effects of dyslipidaemia on monocyte migration during inflammation independent of cardiovascular disease.
We demonstrate here that monocytes accumulate cytoplasmic neutral lipid droplets in response to LDL and VLDL, which subsequently alters their cytoskeletal dynamics both in vitro and in the hypercholesterolemic Ldlr −/− mouse. Strikingly, the extravascular chemotaxis of monocytes is impaired by lipid accumulation, in part mediated by RHOA inactivation. These findings underscore the functional role of blood monocytes and suggest that dyslipidaemia associated with monocyte neutral lipid accumulation, may result in monocyte immunosuppression.

Materials and Methods
See extended methods in supplemental information. All animal procedures were carried out according to the Institutional guidelines for the care and use of experimental animals and the ARRIVE guidelines. Animal studies were approved by the UK Home Office. Blood from healthy human donors was collected under institutional guidelines with informed consent approved by NRES Committee London.
Fluorescence microscopy. Cells were stained for neutral lipid using LipidTox-Green (Life Technologies).
To image the actin cytoskeleton cells were stained with phalloidin-AlexaFluor 488 (Life Technologies). Mounted cells were imaged using either a Zeiss AxioObserver widefield or a Leica SP5 confocal microscope with a 63× /1.4 objective. Images were analysed using Imaris software (Bitplane) or ImageJ, as indicated.
Monocyte adhesion and migration. Purified monocyte subsets were analysed for migration using transwells (Corning) and 2D-chemotaxis chambers (IBIDI). Adhesion and transendothelial migration (TEM) was assessed in static co-cultures using transwells and chamber slides (IBIDI). Intravital imaging of monocyte intravascular migration was assessed in high fat fed Cx3cr1 gfp/+ mice as previously described 27 .
Monocyte phagocytosis and cell death. Phagocytosis was examined using fluorescence latex bead (Life Technologies) and cell death examined using Annexin V (eBioscience).
Cytoskeletal signaling. Purified monocytes were treated with LDL and VLDL (100 μ g/ml) for 2 hours and lysates collected for RHOA activation using luminescence-based G-LISA RhoA activation kit according to manufacturer instructions (Cytoskeleton, Inc.). CDC42 activity was assessed using colorimetric-based G-LISA kit according to manufacturer instructions (Cytoskeleton, Inc.) and PAK activity was assessed using anti-phospho-PAK1/2/3 (Novus Biologicals) via Western blot.
Statistics. Experimental data is presented as mean + /− standard error of the mean (SEM). Populations were compared using a two-tailed Mann-Whitney U test to avoid assumptions of parametric distribution. P < 0.05 was considered significantly different.

Results
In vivo phenotyping of monocyte subsets in response to hyperlipidaemia. We aimed to characterise the interactions of monocyte subsets (GR1 hi and GR1 low ) with blood lipoproteins in the hyperlipidaemic Ldlr −/− mouse with raised plasma VLDL and LDL 28 . When analysed by flow cytometry, monocyte side-scatter (SSC) was significantly elevated in both subsets after 8 weeks HFD when compared to chow, indicating increased cell granularity (Fig. 1a,b), predominantly on GR1 low monocytes (Fig. 1c). Monocyte CD11b levels decreased on both subpopulations on HFD (Supplemental Fig. 1a), while MHC-II (I-A) and GR1 expression did not change  Fig. 1b,c) indicating no monocyte activation as measured by these surface protein markers. However, previous work has shown an increase in CD11c during hyperlipidaemia 22,25 . In our hands, only 4% of monocytes expressed CD11c, mainly on GR1 low , which was not elevated after high fat diet as percentage of cells or MFI (Fig. 1d,e). However, there was a modest but significant increase in percentage of monocytes expressing CD11c in SSC hi monocytes from both chow and HFD Ldlr −/− mice (Fig. 1f). More recently others have reported CD36 expression to indicate CD11c+ 'foamy monocyte' during HFD in ApoE −/− mice 25 , therefore we also examined CD36 expression. While 80-85% of all monocytes expressed CD36, there was no change in expression after HFD or between subsets in the Ldlr −/− mouse (Supplemental Fig. 1d). Finally, GR1 hi neutrophils showed no change in granularity or CD11b expression (Supplemental Fig. 1e,f).
To investigate the composition of this increased monocyte granularity, we sorted (gating strategy Supplemental Fig. 2a) blood monocytes from the Ldlr −/− mouse, stained for neutral lipid using the lipophilic fluorescent dye LipidTOX and visualized using confocal microscopy. HFD causes a marked accumulation of cytoplasmic neutral lipid vesicles in both monocyte subsets (Fig. 1g). This was true for approximately 70% of monocytes from animals on HFD and no such vesicles were observed on a chow diet (data not shown). Further analysis demonstrated that HFD causes a significant increase in the number of neutral lipid vesicles per cell in both monocyte subsets (Fig. 1h). There were no significant differences in lipid accumulation between monocyte subsets when analysed for SSC or mean LipidTOX vesicle intensity (Fig. 1b,i).

HFD impairs monocyte accumulation in the inflamed peritoneum due to retention in the omentum.
To investigate the functional impact of blood monocyte lipid loading on their response to inflammation, we performed thioglycollate peritonitis in the Ldlr −/− mouse and FACS analysed cells from the peritoneal lavage at 72 hours post-injection; a time point well characterized to represent pronounced monocyte/macrophage infiltration 29 . (FACS gating strategy for peritoneal monocyte/macrophages is shown in Supplemental  Fig. 2b). Thioglycollate (Thio) increased total CD11b+ CD115+ monocyte/macrophages in the peritoneum and this was significantly decreased by HFD (Fig. 2a). As expected 29 , very few (20-fold less than monocytes) CD11b+ CD115 neg GR1 high granulocytes were found in the peritoneum at 72 hours (Supplemental Fig. IIc). To further characterize this phenotype, we assessed gene expression by qPCR of cells from the peritoneum following thioglycollate peritonitis. As previously observed by others 30 , these cells exhibited a 'de-activated' phenotype characterized by decreased expression of inflammatory genes Il1β and Cxcl10 after HFD (Fig. 2b).
We hypothesized that neutral lipid loading may be causing a chemotactic defect in blood monocytes, based on previous work in macrophage foam-cells 31,32 . To determine any blood monocyte migratory defect, we labelled blood phagocytes using intravenous (IV) injection of fluorescent latex beads, as previously reported 19 . As expected, bead injection labelled both subsets, with some preference for GR1 low monocytes (Supplemental. Fig. 2d). Thioglycollate peritonitis was induced 4 hours after bead injection, and peritoneal lavage analysed at 72 hours. While bead fluorescence intensity was equal in chow and HFD mice (Supplemental Fig. 2e), indicating that phagocytosis was not impaired, there was a significant decrease in the number of bead-positive CD115+ monocytes in the peritoneum (Supplemental Fig. 2f,g). We therefore concluded that blood monocytes exposed to a HFD have an intrinsic migratory defect. In keeping with this hypothesis, peritoneal cells from HFD mice had decreased expression of the cytoskeletal small GTPase RhoA, which is required for efficient leukocyte chemotaxis ( Fig. 2d) 32,33 . Interestingly, we also noted decreased expression of RhoA in published microarray data from monocytes of patients with familial hypercholesterolemia, supporting an effect of hypercholesterolemia on cytoskeletal regulation 34 (NCBI GEO GSE6054, data not shown).
At this stage, it was unclear whether HFD monocytes were failing to extravasate into the sub-endothelial space, or whether they were being retained in tissue before reaching the peritoneal cavity. Previous work has demonstrated that the greater omentum is the major site of leukocyte extravasation during rodent peritonitis 35 , therefore we harvested and analysed the greater omentum by FACS at 72 hours post-thioglycollate peritonitis to assess the number of leukocytes (Fig. 2e). On both chow and HFD, peritonitis decreased the number of CD45+ leukocytes in the omentum (Fig. 2f), as macrophages migrate from reserves in the omental 'milky spots' into the peritoneal cavity. However, HFD causes an approximate 10-fold increase in the number of CD115+ F4/80 low monocytes in the omentum during peritonitis (Fig. 2g). These data provide strong evidence that the decreased accumulation of monocytes/macrophages in the peritoneum on high fat diet is due to blood monocyte retention in the omentum following extravasation.
Human monocytes accumulate cytoplasmic neutral lipid in response to LDL or VLDL. Following observations in the Ldlr −/− mouse, we attempted to recapitulate the effects of hypercholesterolemia on human monocytes in vitro. It is well established that VLDL is the blood lipoprotein fraction most elevated by HFD and most abundant in the Ldlr −/− mouse 28,36 . As such we specifically investigated the effects of VLDL, which is often overlooked in in vitro studies, as well as further characterizing the effects of LDL. When exposed to 100 μ g/mL VLDL for 150 minutes, both monocyte subsets accumulated neutral lipid vesicles, although vesicles were present in approximately 70% of CD16 neg cells and 45% of CD16 pos (Fig. 3a). When quantified as number of lipid vesicles per cell over time, peak accumulation was at 30 minutes, with CD16 neg monocytes containing approximately 15 lipid vesicles per cell and CD16 pos approximately half this number (Fig. 3b). A similar phenotype was seen with LDL treatment, although peak lipid accumulation was delayed to 60 minutes (Fig. 3c,d) and showed a modest increased number of cells contained neutral lipid vesicles overall (compared to VLDL). Representative images are shown in Fig. 3e, and are visually similar to blood monocytes from the Ldlr −/− mouse on HFD (Fig. 1g).
Previous work examining macrophages from ABCA/G1 knockout mice, but not acLDL treated cells, demonstrated an increased membrane cholesterol content that was suggested to be responsible for a migratory defect 31 . Therefore we analysed at membrane cholesterol content using filipin. CD16 neg monocytes showed a modest but significant 10-20% increase in filipin staining after LDL or VLDL treatment, which was not seen in CD16 pos cells (Fig. 3f). However, given that neutral lipid accumulation increased 60-70% in both subsets after lipid treatments, we concluded that VLDL and LDL treatment predominantly increases neutral lipid content in all monocytes. Despite this striking neutral lipid accumulation, we observed no effect of VLDL or LDL on CD11c, HLA-DR, CD14 or CD16 surface expression, potential markers of monocyte activation (Supplemental Fig. 3a-h). Moreover, there was no change in cell death or phagocytosis (Supplemental Fig. 3i-l).
Lipid accumulation impairs monocyte extravascular migration. Given the reduced accumulation of mouse monocytes following peritonitis on HFD, we next investigated whether neutral lipid loading would impair monocyte chemotaxis in vitro. Using transwell chambers, we found that pre-treatment with 100 μ g/mL VLDL inhibited monocyte chemotaxis towards C5a, mainly in CD16 pos monocytes (Fig. 4a,b). A similar effect was seen with LDL treatment (Supplemental Fig. 4a). To examine any defect in migration in greater detail, we utilized a more realistic C5a chemotactic model using real-time microscopy with 2D migration chambers. VLDL treated CD16 neg and CD16 pos monocytes showed decreased track displacement and confinement ratio, accompanied by an almost complete failure to polarize towards the chemoattractant (Fig. 4c-e) decreased migratory speed towards C5a (Fig. 4f), all indicative of chemotactic inhibition after VLDL treatment, which did not appear to be subset specific. Interestingly, this migratory inhibition also existed during monocyte monocytes. * and ** represents P < 0.05 and P < 0.01 respectively analysed by Mann-Whitney U test. See also Sup. Fig. 2. chemokinesis, when the chemoattractant was uniformly distributed and the cells were not required to polarize (Supplemental Fig. 4b-e). Importantly, to examine if the defect in C5a induced migration was simply an effect of VLDL on the C5a receptor (CD88), we examined CD88 expression levels before and after lipid treatments. CD88 expression was not significantly different between monocyte subsets or after VLDL treatment (Fig. 4g), indicating that the defect in migration was not a reduction in C5a receptor signaling.
We next examined monocyte-endothelial interactions after lipid treatments. Monocyte adhesion to a HUVEC endothelial layer was unchanged by VLDL treatment of the monocytes (Fig. 4h) or the endothelium (Fig. 4i). However, LDL treatment did increase CD16 neg endothelial adhesion, as reported by others 37 (Supplemental Fig. 4f,g). There was no effect of VLDL on CD16 pos migration across a HUVEC monolayer, although VLDL did increase CD16 neg monocyte trans-endothelial migration (TEM) (Fig. 4j). A similar effect was seen with LDL treatment during TEM (Supplemental Fig. 4h).
No defect in monocyte TEM suggests that lipid loading is inhibiting only extravascular migration. To confirm that intravascular monocyte migration is not affected, we performed live intravital imaging of the vasculature of the ear in Cx3cr1 gfp/+ mice fed HFD or chow for 6 weeks. Without further genetic manipulation we were able to significantly increase both total cholesterol and LDL levels after 6 weeks HFD (Fig. 5a), as has been reported by others 38 . This increased neutral lipid loading in blood monocytes (Fig. 5b), similar to Ldlr −/− mouse on HFD. Corroborating in vitro endothelial migration findings, HFD did not cause any clear migratory defect in GR1 low monocytes crawling on the vascular endothelium (Fig. 5c-h). While monocytes exhibited a modest increase in speed on HFD (Fig. 5f), this was not accompanied by changes in track displacement, straightness, length or duration (Fig. 5c-e,g,h). VLDL modulates cytoskeletal-signaling in monocytes. Given previous work implicating cholesterol in cytoskeletal disruption in macrophages 31,32 and the decreased RhoA expression which we observe in the inflamed peritoneum after HFD (Fig. 2f), we hypothesized that neutral lipid loading may be altering cytoskeletal rearrangement in monocytes, thus impairing migration.
Initially, we used fluorescent phalloidin to visualize actin after monocytes had been VLDL treated and allowed to adhere to tissue-culture plastic. When analyzed by confocal microscopy clear effects of VLDL treatment on monocytes were observed, and were distinct between monocyte subsets. While CD16 neg cells had a decreased cell spreading after lipid loading resulting in a lower cell area (Fig. 6a,c), CD16 pos cells displayed a decreased circularity due to pronounced filopodia-like protrusions (Fig. 6b,c). Interestingly, in some CD16 pos polarised monocytes after VLDL treatment, vesicles accumulated within the uropod away from the leading edge (Fig. 6d). RHOA activation via GTP binding is required for leukocyte chemotaxis 39 and has previously been implicated in monocyte uropod retraction 33,39 therefore we assessed RHOA activation in response to VLDL and the RHOA activator CN03. VLDL alone had little effect on RHOA-GTP levels, yet when cells were treated with CN03, VLDL significantly impaired RHOA activation in both monocyte subsets (Fig. 6e). We also investigated the activation of the cytoskeletal GTPase CDC42, which is involved in filopodia formation in leukocytes, and the CDC42 and RAC1 effector proteins PAK1 and PAK2, but saw no effect of VLDL or LDL treatment on these two candidates (Supplemental Fig. 4i,j).

Discussion
We show here that monocytes can be extensively loaded with neutral lipids in vivo during high fat diet and in vitro following LDL and VLDL treatments. It is unclear whether this is due to endocytosis of the entire lipoprotein, or due to hydrolysis of fatty acids at the cell surface, as hypothesized in smooth muscle cells by Ira Goldberg and * and ** represents P < 0.05 and P < 0.01 respectively analysed by Mann-Whitney U test. colleagues 40 . In peripheral tissues, SR-BI may act synergistically with CD36, heparan sulphate proteoglycans and cell surface lipases to mediate the hydrolysis and uptake of FFAs from triglyceride-rich lipoproteins (TGRL) such as VLDL 41 . While in macrophages the LDL receptor (LDLR) has been implicated in VLDL uptake and trafficking 42 , we believe monocyte lipid loading is independent of LDLR, as lipid uptake occurs in the hypercholesterolemic LDLR knockout mouse. LDLR independent uptake of TGRL would be an interesting study in monocytes.
Monocyte 'foam cells' or lipophages have been previously demonstrated in hyperlipidaemic rats 21 , Ldlr −/− peritoneal macrophages 30,43 , ApoE −/− monocytes during atherosclerosis 22,25 , familial hypercholesterolemia patients 18 and in postprandial monocytes 23 . In some of these studies it is reported that this can lead to monocyte activation exemplified by CD11c expression 22,23 . Surprisingly, under our experimental settings we did not observe an obvious monocyte activatory phenotype after neutral lipid accumulation. The disparities are most likely due to the differences in methods of monocyte isolation, treatment dose and time, nuances in lipid species or effects specific to one murine atherosclerosis model, which we did not investigate. Moreover, only a very small fraction of mouse monocytes actually expressed CD11c, which is similar to previous reports (reviewed in 44 ), but not by others 22,25 . More work is needed to investigate CD11c expression on monocytes in various murine strains under different experimental conditions and the conditions in which lipoproteins may activate monocyte subsets.
Regardless of activation phenotype, a feature of the advanced atherosclerotic plaque is the inhibition of cellular efflux or migration from the plaque 45,46 . This led us to explore the effect of lipids on monocyte migration and the role of cytoskeletal genes in monocytes loaded with neutral lipids. We show that in-vivo during hyperlipidaemia monocytes become trapped in the omentum and are unable to effectively migratory into the peritoneum during peritonitis. We hypothesise this phenotype is responsible for increased bacterial and fungal pathogenicity reported during hyperlipidaemia [13][14][15][16] . Moreover, we can recapitulate this dysfunctional migration in-vitro using transwell and 2D chemotaxis and chemokinesis assays. Lipid loading significantly inhibited C5a induced migration and polarization. This migratory dysfunction after lipid loading is thought to be extravascular, based on 3 key findings: 1) monocytes are retained in the omentum following peritonitis on HFD 2) luminal endothelial monocyte crawling is not reduced during hypercholesterolemia in-vivo and 3) monocyte endothelial adhesion or transendothelial migration is not reduced after lipid treatments in vitro. Under our conditions, LDL but not VLDL increased monocyte adhesion to endothelial cells. This is in contrast to a previous report showing increased adhesion after VLDL treatment 47 . The differences are most likely that we used a static adhesion assay, which does not take into account the complex relationship between shear stress and adhesion molecule dynamics 48 and/or may be dependent on lipoprotein concentration or composition, which we did not investigate.
RhoA transcripts were down regulated during HFD peritonitis; therefore we investigated whether lipid loading inhibited monocyte cytoskeletal signaling. Cytoskeletal regulation involves many proteins including the small GTPases RAC, CDC42 and RHO and is a complex and dynamic network that responds to a variety of agonists 49 . A series of phosphorylation events regulate cellular morphology and migration through formation of lamellipodia, filopodia and the extension and retraction of the uropod 49 . In our hands, lamellipodia and filopodia formation is different in each subset after neutral lipid loading with a significant decrease in circulatory index of non-classical monocytes. Interestingly, we also noted an accumulation of vesicles in the uropod in some of the polarised CD16 pos non-classical monocytes, which may be involved with inhibition of uropod retraction and impairment of forward cell migration 33 . Given that we did not see subset specific differences in the inhibition of 2D migration towards C5a, we therefore proceeded to investigate cytoskeletal signaling. VLDL treated monocyte subsets exhibit no change in CDC42 activity or phosphorylation of the RAC/CDC42 downstream effector PAK, supporting our hypothesis that this migratory phenotype is not mediated by an increase in membrane cholesterol as described in ABCA1/ABCG1 −/− macrophages 31 . However we did note a significant decrease in gene expression of RHOA after peritonitis during high fat diet and therefore, we investigated RHOA activity after lipid loading.
Our studies on the RHOA pathway were performed using a direct RHOA activator (CN03), which is based on the catalytic domain of bacterial cytotoxic necrotizing factor toxins 50 . While there was a significant inhibition of RHOA activity in monocytes after VLDL treatment, this was not subset specific. The inhibition of RHOA activity by neutral lipids may indicate a direct binding to RHO GTPase by a neutral lipid moiety. Overall we believe that there are multiple cytoskeletal signals involved in monocyte lipid accumulation, some of which are subset specific and mediate changes in morphology, while others potentially affect RHOA downstream signaling in all monocytes. More work is needed to dissect these multiple cues and the mechanisms that dictate subset specific phenotypes.
During atherosclerosis others have already shown that dyslipidaemia is able to perturb macrophage migration through modulation of cytoskeletal rearrangement by free cholesterol in the cell membrane 31,32,51 , perhaps contributing to increased macrophage plaque retention 19 . For example, a paper by the Tall group showed that the defect in migration was mediated by RAC1 and involved the cholesterol transporter ABCA/G1 31 . However our observations are independent from these on a number levels: 1) Neutral lipid and not unesterified cholesterol accumulate in vesicles following lipoprotein exposure 2) we observe no change in membrane cholesterol levels in LDL/VLDL treated CD16 pos monocytes, which have a migratory phenotype. This confirms Pagler et al. 31 , describing no effect of LDL treatment on membrane cholesterol accumulation in wild-type macrophages. 3) We show migratory dysfunction is only dependent on RHOA and not PAK1 (downstream of RAC1) or CDC42. These different observations between monocytes and macrophages highlight the discrete functional independence of mononuclear phagocytes during dyslipidaemia. Furthermore, we hypothesise that independent mechanisms govern the effects of unesterfied cholesterol efflux versus neutral lipid droplet accumulation in these cells.
In summary we demonstrate the effects of hyperlipidaemia on the migration and cytoskeletal regulation of monocyte subpopulations. This work extends the function of 'foamy monocytes' or lipophages beyond their recently demonstrated role in delivering lipid during early atherogenesis 25 . Although our work may highlight a mechanism for monocyte retention in the atherosclerotic plaque, perhaps more importantly there may be broader implications for dyslipidaemia-induced monocyte immunosuppression. Our study emphasises the role of dyslipidaemia in infection and innate immunity and the need to further study the effects of lipid accumulation in both blood and tissue compartments.