M2 macrophages exhibit higher sensitivity to oxLDL-induced lipotoxicity than other monocyte/macrophage subtypes

Background In obesity, phenotypic switches occur in macrophage populations such that the predominantly M2-polarised anti-inflammatory state seen in lean individuals changes to a predominantly M1-polarised pro-inflammatory state in those who are obese. However, the mechanisms by which these phenotypic shifts occur have not yet been fully elucidated. Results The effects of oxLDL (1-40 μg/ml; 24 h) on several parameters relevant to the Unfolded Protein Response (UPR)-mediated lipotoxic effects of oxLDL (disruption of ER Ca2+ handling; activation of the UPR transcription factor XBP-1; upregulation of the UPR target genes BiP and CHOP; apoptosis; cell viability) were investigated in human primary monocyte-derived macrophages, and also in monocyte-macrophages derived from the THP-1 monocytic cell line. A consistent pattern was observed: M2-polarised macrophages were more sensitive to the lipotoxic effects of oxLDL than either non-polarised macrophages or non-differentiated monocytic cells. Specifically, M2-polarised macrophages were the only cell type to undergo significantly increased apoptosis (Primary cells: 1.23 ± 0.01 basal; THP-1-derived: 1.97 ± 0.12 basal; P < 0.05 in both cases) and decreased cell viability (Primary cells: 0.79 ± 0.04 basal; THP-1-derived: 0.67 ± 0.02 basal; P < 0.05 in both cases) when exposed to oxLDL levels similar to those seen in overweight individuals (ie. 1 μg/ml). Conclusions We propose that the enhanced susceptibility of M2-polarised macrophages to lipotoxicity seen in the present in vitro study could, over time, contribute to the phenotypic shift seen in obese individuals in vivo. This is because a higher degree of oxLDL-induced lipotoxic cell death within M2 macrophages could contribute to a decrease in numbers of M2 cells, and thus a relative increase in proportion of non-M2 cells, within macrophage populations. Given the pro-inflammatory characteristics of a predominantly M1-polarised state, the data presented here may constitute a useful contribution to our understanding of the origin of the pro-inflammatory nature of obesity, and of the pathogenesis of obesity-associated inflammatory disorders such as Type 2 Diabetes and atherosclerosis.


Background
Obesity and associated disorders such as Type-2 Diabetes (T2D) and atherosclerosis are associated with elevated levels of many lipids (eg. increased circulating oxidized low-density lipoprotein (oxLDL) [1,2]), and with chronic inflammation [3]. This can lead to intracellular lipid accumulation in non-adipocyte cells, which can in turn lead to cell death, a phenomenon known as "lipotoxicity" [2].
Due to their wide tissue distribution, monocyte/macrophages, which play vital roles in inflammation and the development of obesity, T2D and atherosclerosis [3], are involved in lipid accumulation within many tissues [4]. For example, the intracellular accumulation of oxLDL within macrophages is mediated by scavenger receptors (such as CD36, SR-A and possibly SR-BI, although the role of the latter is currently controversial [5]) recognizing altered molecular patterns present on oxLDL or other forms of modified lipoproteins such as acetylated LDL (as distinct from non-oxidised LDL), and facilitating its uptake [6].
Macrophage populations resident in adipose tissue upon high-fat feeding exhibit an M1-predominant state different from that of the M2-predominant population residing in adipose tissue under normal dietary conditions [12]. Moreover, similar phenotypic shifts have been reported for free cholesterol-loaded peritoneal macrophages in-vitro [13]. Mechanisms responsible for such obesity-linked shifts to predominantly M1 cells include increased infiltration of M1 cells from the circulation [1], and differentiation of mesenchymal stem cells into M1 cells (or trans-differentation of M2 cells into M1 cells) within adipose tissue [11]. Also, circulating peripheral blood monocytes can be primed for differentiation into functionally distinct macrophage subpopulations in certain circumstances, such as PPARγ-mediated priming of circulating monocytes for differentiation towards an antiinflammatory M2 macrophage phenotype [8,14]. However, additional mechanism(s) may also underpin the M2-to-M1 shifts seen within pre-existing macrophage populations during the development of obesity [4]; for example, a recent report has suggested that "lipidinduced toxicity is an important determinant of the obesity-linked proinflammatory switch in macrophage polarisation" [15].
The endoplasmic reticulum (ER) is a key organelle with regard to lipotoxicity in macrophages, as trafficking of free cholesterol to the ER membrane and incorporation of cholesterol into the normally cholesterol-poor ER membrane alters the physico-chemical properties of this membrane and leads to disruption of ER functions [16]. Disturbance of ER functions leads to the ER becoming overwhelmed with accumulated unfolded proteins, and under such conditions of ER stress, the cell responds by initiating an "unfolded protein response" (UPR [17]). The UPR involves either generalised inhibition of translation combined with specific upregulation of UPR genes that restore the ER's ability to function (eg. binding immunoglobulin protein (BiP; a.k.a. GRP78); [18]), or activation of the pro-apoptotic C/EBP homologous protein (CHOP) pathway [19] if these actions fail to resolve the original ER stress. Hence, UPRs lead either to restoration of normal cell physiology and cell survival, or to cell death [17]. Thus, excessive accumulation of cholesterol into macrophages can lead via UPR/CHOP-triggered apoptosis to lipotoxic macrophage cell death [20,21]; importantly, we and others have recently reported that oxLDL can trigger similar effects [22,23], due to liberation of free cholesterol from oxLDL particles that have been imported into macrophages, and trafficking of this free cholesterol to the ER membrane [16,20,23].
In the present study, we hypothesised that differential sensitivity of M2 macrophages to elevated levels of oxLDL in obesity, and thus to more frequent lipotoxic cell death of M2 cells, may contribute to the decrease in proportion of M2 cells seen within the macrophage population in obese individuals. Thus, using both primary cells and cells derived from the THP-1 monocytic cell lineage, we aimed to evaluate whether oxLDL induced the UPR and lipotoxic apoptosis to the same extent in control and M2-polarised monocyte-derived macrophages.

Results
i) Primary Cells IL-13/Rosiglitazone induced upregulation of markers of the M2 macrophage phenotype Two markers of the M2 phenotype, MR and IL-1Ra [8], underwent significant increases in mRNA expression after treatment of human primary macrophages (hMΦ) with IL-13 and rosiglitazone (see Methods; MR: 2.93 ± 1.03 basal; IL-1Ra: 3.76 ± 1.32 basal; P < 0.10 in both cases; Figure 1a), while MCP-1 (a marker of the M1 phenotype) did not undergo any significant increase (data not shown). Hence, cells that had undergone this treatment were confirmed as exhibiting the M2 macrophage (hM2MΦ) phenotype. IL-13/rosiglitazone treatment did not have any significant effect on cell viability (data not shown).
cholesterol into the ER membrane of oxLDL-treated THP1-Mon cells alters the physico-chemical properties of this membrane (as determined by Electron Paramagnetic Resonance spectrometry [23]), and so disrupts ER membrane protein function (as determined by coupledenzyme SERCA2b Ca 2+ ATPase assays [23]).

Expression of marker genes for the M2 macrophage phenotype
As shown in Figure 1b, specific upregulation of two M2 markers, MR and IL-Ra [8], was observed after 72 h treatment of dTHP1-MΦ with IL-13 and rosiglitazone (MR: 41.72 ± 1.92 basal; IL-1Ra: 25.76 ± 0.48 basal; P < 0.05 in both cases). Hence, cells that had undergone this treatment were confirmed as exhibiting an M2 macrophage-like (dTHP1-M2MΦ) phenotype. In contrast, non-differentiated THP-1 monocytic cells expressed both marker genes only at very low levels in either the presence or the absence of IL-13/rosiglitazone treatment (< 0.1 of that seen in dTHP1-MΦ; Figure 1b). Once again, IL-13/rosiglitazone treatment did not have any significant effect on cell viability (data not shown).

Discussion
The current study presents data indicating that M2polarised macrophages are more sensitive to the lipotoxic effects of oxLDL than other forms of monocyte/macrophage. The effects of oxLDL on disruption of ER Ca 2+ homeostasis, activation of the UPR transcription factor XBP-1, upregulation of the UPR target genes BiP and CHOP, and on apoptosis and cell viability, were determined in six cell types: hMΦ, hM2MΦ, THP1-Mon, THP1-M2Mon, dTHP1-MΦ, and dTHP1-M2MΦ. With regard to these parameters, a consistent pattern was observed: THP1-Mon were the least sensitive to oxLDL, while hM2MΦ and dTHP1-M2MΦ were the most sensitive, with hMΦ, dTHP1-MΦ and THP1-M2Mon being intermediate in their sensitivity. Importantly, the M2polarised macrophages (hM2MΦ and dTHP1-M2MΦ) were the only cell types to undergo significantly increased apoptosis and decreased cell viability when exposed to oxLDL levels similar to those seen in the circulation of overweight individuals (ie.~1 μg/ml -as seen in individuals with a body mass index of 25-30 kg/m 2 [31]).
The source of oxLDL-induced lipotoxicity in this instance is likely to be cholesterol, a component of oxLDL particles. It has been known for many years that intracellular accumulation of cholesterol results when macrophages take in more lipid than can be excreted; lysosomal cholesterol esterases are responsible for the generation of free cholesterol from lipoprotein-derived cholesteryl esters, and a proportion of the resulting free cholesterol is trafficked to intracellular membranes such as the ER membrane [32]. The ER membrane is a narrow, fluid structure containing < 10% cholesterol [33]; elevation of its cholesterol content results in a broader, less fluid structure which restricts the ability of ER integral membrane proteins to undergo conformational movements and so catalyse their respective enzymatic reactions [22,23]. This results in disrupted ER function [23], and the resulting ER stress has been linked via the UPR/CHOP pathway to the triggering of apoptosis   [19,20]. Due to the presence of an ERSE in its 3'UTR (see Results section), CHOP mRNA can be induced by prolonged and/or severe ER stresses to act as a transcription factor that controls several genes associated with apoptosis, including Bcl-2, GADD34, TRB3 and several caspases, including caspases 3 and 7 [19]. (NB. It should be noted that the autophagic pathway, which can lead to non-apoptotic cell death by engulfing, degrading and recycling of cell organelles and proteins [34], is also associated with ER stress [35]; because in the present study, lipotoxic cell death associated with UPRs was assessed solely by determination of CHOP expression and caspase 3/7 activity, the possibility cannot be ruled out that other forms of cell death may also play a role in obesity-related macrophage phenotypic shifts. Nevertheless, in line with several previous papers [19][20][21][22], the current study has focused on CHOP-mediated apoptosis as the predominant mode of lipotoxic cell death in macrophages.) As stated above, several mechanisms that may contribute to the obesity-linked shift to predominantly M1 cells been identified [1,11]. However, additional mechanism(s) may also underpin the M2-to-M1 shifts seen within preexisting macrophage populations during the development of obesity [4], and as described above, lipotoxicity has recently been proposed as a potential mechanism by which such shifts may occur [15]. In the light of our demonstration that M2 macrophages exhibit higher sensitivity to oxLDL-induced lipotoxicity than non-M2polarised macrophages, we propose that a higher degree of oxLDL-induced lipotoxic cell death within M2 macrophages may contribute to the decrease in numbers of M2 cells. Over time, a relative increase in proportion of non-M2 cells and hence a predominantly M1 macrophage population would ensue, leading to development of a more pro-inflammatory milieu in the tissues of obese individuals.
Having demonstrated heightened sensitivity to oxLDLinduced lipotoxicity in M2-polarised macrophages, we next considered the possible sources of this heightened sensitivity (ie. we addressed the question: "At what point in the ER stress/lipotoxicity pathway do M2 cells become more sensitive?"). There were no significant differences in oxLDL's impact on [Ca 2+ ] cyt in the different cell types (Figure 2), and dTHP1-M2MΦ did not react to a greater extent than dTHP1-MΦ in terms of XBP-1 activation (Figure 3). However, significant increases in BiP expression (cf. THP1-Mon) were only seen in dTHP1-M2MΦ after treatment with 1 μg/ml oxLDL, with only non-significant increases being seen in other cell types (Figure 4a). Moreover, as shown in Figure 4b, statistically significant increases in CHOP expression were seen only in dTHP1-M2MΦ after treatment with 40 μg/ml oxLDL (a similar pattern was seen in the case of 1 μg/ml oxLDL v. CHOP, albeit without attaining statistical significance). Finally, with regard to apoptosis and cell viability, the response of dTHP1-M2MΦ to oxLDL was significantly greater than the comparable responses of other THP-1 derived celltypes ( Figure 5). As stated above, with regard to primary cells, the same general pattern was observed: hM2MΦ exhibited significantly more apoptosis and less cell viability than hMΦ when exposed to oxLDL levels similar to those seen in sedentary individuals (i.e. 1 μg/ml [31]; Figures 1c  and 1d). Thus, it appears that a factor relevant to the UPR (perhaps involving differences in the ability of the UPR to restore normal cell physiology and so avoid the triggering of apoptosis) may be the source of M2-polarised cells' heightened lipotoxic sensitivity. Further elucidation of the source of this differential lipotoxicity may increase our understanding of the causal connections between obesity and inflammatory diseases such as T2D; however, definitive identification of this source is unfortunately beyond the scope of this preliminary study.
As stated above, macrophages are involved in lipid accumulation in a wide variety of tissues, including adipose tissue [12]. Intriguingly, quantitation of lipid-loading in different macrophage subtypes during the progression of obesity in ob/ob mice fed a high-fat diet revealed that, while macrophage lipid-loading in general increased, very few lipid-loaded M2 macrophages could be identified, with lipid-loading and adoption of a foam cell-like appearance being characteristic only of M1 macrophages [15]. One may speculate that this could be explained by an inability of M2 macrophages to tolerate large quantities of intracellular lipid, and thus increased M2 cell death in a high-lipid environment. Such an explanation could also account for the high levels of apoptosis and M1-predominant nature of the macrophage foam cell populations resident in the unstable regions surrounding the lipid cores of atherosclerotic plaques [8,21]. Thus, the relevance of differential macrophage lipotoxicity is likely to extend beyond adipose tissue macrophages. For example, the M2to-M1 shift seen in Kupffer cells during the progression of liver steatosis has recently led Mandal et al to suggest that "changes in the M1/M2 phenotypic balance can impact on diverse disease conditions" [36], a statement which lends increased importance to the need to understand the mechanism(s) by which macrophage polarisation can be affected by phenomena such as lipotoxicity.
Clearly, there are several limitations to the current study (eg. the exclusively in vitro nature of the experiments; the lack of direct characterisation of differences in cholesterol ER trafficking in each cell type). In particular, although the current study has specifically investigated the lipotoxic effects of oxLDL (and demonstrated that non-oxidised LDL is not associated with the same effects), it should also be recognised that in vivo, macrophages will be exposed to many other forms of lipid. Previous studies investigating different lipids have demonstrated the need to both activate the UPR/CHOP pathway and engage a scavenger receptor if apoptosis is to be triggered [20], suggesting that only lipoproteins that are ligands for scavenger receptors can trigger lipotoxic cell death. However, more recent studies have employed a lipidomic approach and found that a variety of lipid species underwent obesity-associated increases in levels within adipose tissue macrophages, and were associated with shifts towards the M1 phenotype; amongst these lipids were free cholesterol and saturated fatty acids [15]. Interestingly, the relatively rigid inflexible structures of both these lipids are compatible with disruption of ER membrane functions, and thus the triggering of the UPR. Meanwhile, lipids with more flexible structures that would not be expected to disrupt ER membrane properties (eg. polyunsaturated fatty acids, plasmalogens) were not associated with increases in the M1 subtype [15]. Thus, a capacity for triggering ER stress/UPR-linked apoptosis -and possibly a consequent heightened degree of damage to M2-polarised cells, which appear to be more sensitive to lipotoxicity -may underpin the detrimental effects of a variety of harmful lipids (including, but not limited to, oxLDL). Further studies are required to establish whether this is the case.

Conclusion
Given the elevated levels of circulating oxLDL and other lipids seen in obesity [2] (and hence the increased likelihood of the in vitro effects seen here occurring in the tissues of an obese individual in vivo), and also the importance of M1 macrophage-mediated inflammation in the pathogenesis of obesity-related diseases such as T2D and atherosclerosis [3], the current study's demonstration that M2-polarised macrophages are more sensitive to the UPR-mediated lipotoxic effects of oxLDL than other forms of monocyte/macrophage may constitute a useful contribution to our understanding of the origin of the pro-inflammatory nature of obesity, and of the pathogenesis of obesity-associated disorders such as T2D and atherosclerosis.

Materials
All reagents were purchased from Sigma-Aldrich (Poole, UK) unless stated otherwise. oxLDL, rosiglitazone, GW9662 and IL-13 were obtained from Autogen Bioclear (Calne, UK), GlaxoSmithKline (Uxbridge, UK) and R&D Systems (Abingdon, UK), respectively. Human primary monocyte-macrophages and human monocytic THP-1 cell lines were obtained from the Welsh Blood Transfusion Service (Llantrisant, UK) and European Collection of Cell Cultures (Salisbury, UK), respectively.

Generation of primary human monocyte-derived macrophages
Primary human monocyte-derived macrophages were obtained via the method of Iwashima et al [37]. Briefly, 10 ml of heparinised blood was diluted 1:1 in RPMI medium, layered over 10 ml of Histopaque-1077 Ficoll-Hypaque and centrifuged at 400 × G for 20 min. The mononuclear cell suspension was carefully removed from the Ficoll-Hypaque interface, and washed four times (500 × G; 10 min) in Phosphate-Buffered Saline (PBS), and mononuclear cells were then incubated (37°C; 5% CO 2 ) for 2 h, after which non-adherent cells (representing non-monocytes) were removed by discarding the media. Following adhesion to tissue culture plasticware for 7 days (with medium being changed every 48 h), monocytes were considered to have differentiated to human primary macrophages (hMΦ). M2-polarized macrophages (hM2MΦ) were then obtained via the method of Bouhlel et al [8], in which treatment for 72 h with 15 ng/ml IL-13 and 1 μM rosiglitazone induced M2 polarization.

Cell treatments
hMΦ, hM2MΦ, THP1-Mon, THP1-M2Mon, dTHP1-MΦ, and dTHP1-M2MΦ were treated for 24 h with 1 μg/ml oxLDL (which approximates to serum oxLDL levels seen in overweight individuals (ie. those with a body mass index of 25-30 kg/m 2 ) [31]) or 40 μg/ml oxLDL (a commonly-used supra-physiological dose [39]). In some cases, control samples were treated with nonoxidised LDL, or with thapsigargin (100 nM) as a positive control for disruption of ER homeostasis [16], respectively. Data were expressed for each sample in comparison to the readings obtained in basal samples (ie. cells of the corresponding cell-type that had not been treated with oxLDL).
For gene expression assays, expression of genes of interest (see below) was determined using SYBR ® Green or Taqman ® Gene Expression Assays and an Applied Biosystems 7500 Real-time PCR system (Applied Biosystems, Warrington, UK). In the case of MR, IL-1Ra and CHOP, semi-quantitative (relative to β-actin and/or Glyceraldehyde Phosphate Dehydrogenase (GAPDH)) SYBR ® Green assays were carried out using the follow- In the case of BiP, mRNA expression was assessed semiquantitatively (relative to GAPDH) via Taqman ® Gene Expression Assays (Applied Biosystems, Warrington, UK) for BiP (Gene Expression Assay Hs99999174_m1) and GAPDH (Gene Expression Assay Hs99999905_m1).

Statistical analysis
Data are expressed as mean ± standard error of the mean. Statistical significance was determined via ANOVA or 2-sample t-tests as appropriate; significance levels were set at P < 0.05 except where stated otherwise.