ABSTRACT
Biological membranes form the functional, dynamic interface that hosts a major fraction of all cellular bioactivity. Proper membrane physiology requires maintenance of a narrow range of physicochemical properties, which must be buffered from external perturbations. While homeostatic adaptation of membrane fluidity to temperature variation is a ubiquitous design feature of ectothermic organisms, such responsive membrane adaptation to external inputs has not been directly observed in mammals. Here, we report that challenging mammalian membrane homeostasis by dietary lipids leads to robust lipidomic remodeling to preserve membrane physical properties. Specifically, exogenous polyunsaturated fatty acids (PUFAs) are rapidly and extensively incorporated into membrane lipids, inducing a reduction in membrane packing. These effects are rapidly compensated both in culture and in vivo by lipidome-wide remodeling, most notably upregulation of saturated lipids and cholesterol. These lipidomic changes result in recovery of membrane packing. This lipidomic and biophysical compensation is centrally mediated by the sterol regulatory machinery, whose pharmacological or genetic abrogation results in decreased cellular fitness when membrane homeostasis is challenged by dietary lipids. These results reveal an essential mammalian mechanism for membrane homeostasis wherein lipidome remodeling in response to dietary lipid inputs preserves functional membrane phenotypes.
INTRODUCTION
Lipidic membranes are the essential barriers between life and the abiotic world and also mediate most intracellular compartmentalization in eukaryotic cells. However, the role of membranes is not limited to passive barriers. Approximately one third of all proteins are membrane-embedded1, and many more are membrane-associated through post-translational modifications, lipid binding, and protein-protein interactions. Thus, a major fraction of cellular bioactivity occurs at membrane interfaces. Importantly, the physicochemical properties of the lipid matrix are key contributors to membrane physiology. A canonical example is membrane viscosity, which determines protein diffusivity, and thus protein-protein interaction frequency. Another is membrane permeability, which governs the diffusion of solutes into and out of the cytosol. Numerous other membrane physical parameters can determine protein behavior, including but not limited to fluidity, permeability, curvature, tension, packing, bilayer thickness, and lateral compartmentalization 2–12.
Because of their central role in protein function, effective maintenance of membrane properties is essential for survival in a complex and variable environment. In ectothermic (i.e. non-thermoregulating) organisms, a pervasive challenge to membrane homeostasis comes in the form of temperature variations. Low temperature reduces the motion of lipid acyl chains, causing membranes to laterally contract, stiffen, and become more viscous13. Organisms across the tree of life, from prokaryotes to ectothermic animals, respond to such perturbations by tuning membrane lipid composition, down-regulating tightly packing lipids (e.g. containing saturated acyl chains) and up-regulating more loosely packed ones containing unsaturations or methylations in their lipid tails13–15. This response was termed ‘homeoviscous adaptation’, as these lipid changes result in remarkable constancy in membrane fluidity in spite of variable growth conditions13,14,16. It is worth noting that while fluidity is maintained at a specific set point, it should not be assumed that this is either the control variable, or the physical property being sensed; rather membrane fluidity may correlate with other membrane/lipid control parameters. For example, membrane homeostasis in Bacillus subtilis is mediated by the DesK sensor that is believed to sense membrane thickness17 In the yeast endoplasmic reticulum (ER), recently discovered sensors are sensitive specifically to lipid packing5 and bilayer compressibility18.
Excepting a few specialized instances (e.g. hibernation), mammals and other warm-blooded animals are not subject to large-scale variations in body temperature; thus, there has been relatively little investigation of homeostatic membrane responsiveness in such organisms. However, it is a well-established but underappreciated fact that mammalian membrane homeostasis is extensively challenged by dietary inputs. Dietary lipids have major impacts on membrane compositions in vivo19,20, and these perturbations must presumably be buffered to maintain cellular functionality. Mammalian lipidomes are much more complex21–23 than either bacteria24 or yeast25,26, suggesting more potential control nodes required to balance the various conflicting demands of mammalian membrane physiology.
The possibility of a homeoviscous response in mammalian cells was suggested by two studies in the 1970s, which used a spontaneously arising mutant of Chinese Hamster Ovary (CHO) cells that is defective in cholesterol regulation27,28. These mutants accumulate cholesterol compared to wild-type CHO cells, but maintain normal membrane fluidity, possibly through modulation of their phospholipid profiles. However, the molecular etiology of the defects in these mutants remains unknown, and it was not reported where the cholesterol in these cells was accumulating (possibly storage organelles or lysosomes). Further, limitations of then-available technologies prevented direct demonstration of lipidomic responses to cholesterol modulation. Thus, the relevance of those insights to physiologically relevant perturbations of metabolically normal mammalian cells remains unclear. More recently, homeoviscous adaptation in mammals has been inferred from data-driven modeling approaches, which used the physical properties (melting temperature, intrinsic curvature) of pure lipids to extrapolate those of complex, biological membranes29,30. However, the inherent non-additivity31 and non-ideality32 of lipid mixtures suggests that extrapolation of physical parameters of complex membranes from pure lipids may not be a reliable approach. Finally, lipid composition and membrane properties have been implicated in the heat shock response, though usually with a specific focus on signaling in the proteostasis network33.
Here, we directly evaluate the hypothesis that mammalian membranes homeostatically adapt to dietary inputs by characterizing the lipidomic and biophysical responses to dietary fatty acids in several mammalian cell types and in vivo. We show that polyunsaturated fatty acids (PUFAs) are robustly incorporated into membrane phospholipids, introducing significant biophysical perturbations. This perturbation is counterbalanced by nearly concomitant lipidomic remodeling, most notable in the upregulation of saturated lipids and cholesterol. This remodeling normalizes membrane biophysical properties. These responses are centrally mediated by transcriptional sterol-regulatory machinery involving the Sterol Regulatory Element Binding Protein SREBP2. Finally, we show that the homeostatic membrane response is essential for cellular fitness, as abrogation of key response nodes leads to cytotoxic effects when membrane homeostasis is challenged by exogenous fatty acids.
RESULTS
Robust and specific incorporation of PUFAs into membrane lipids
Recent observations revealed that supplementation of cultured mammalian mast cells (rat basophilic leukemia cells (RBL)) with docosahexaenoic acid (DHA) leads to robust incorporation of this dietary polyunsaturated fatty acid into membrane lipids23. We observed similar effects in isolated human mesenchymal stem cells34, cultured Chinese hamster ovary (CHO) cells, and rat primary hippocampal neurons (Supp Fig S1), confirming that uptake and incorporation of exogenous DHA into membrane lipids is not cell-type specific. Supplementation designed to recapitulate DHA-enriched diets in mammals (see Materials and Methods) increased the fraction of DHA-containing glycerophospholipids (GPLs) by nearly 15-fold (from < 1 to ~15 mol%) (Fig 1A). Similarly, supplementation with the more common PUFA ω-6 arachidonic acid (AA) increased the fraction of AA-containing lipids by >3-fold (Fig 1B). Surprisingly, supplementation with monounsaturated (oleic; OA) acid or saturated (palmitic; PA) fatty acid produced only minimal lipidomic changes (Supp Fig S2). We speculate that this disparity in incorporation between PUFAs and more saturated FAs is associated with their availability in cell culture media: cells have access to sufficient levels of OA and PA such that supplementation has no effect, whereas PUFA levels are limited such that supplementation with the physiologically appropriate concentrations used here leads to robust uptake and incorporation.
To confirm that the membrane incorporation of supplemented PUFAs in cultured cells appropriately recapitulates in vivo conditions, we analyzed the membrane lipidomes of mouse hearts and livers after two weeks of free-feeding on chow containing either fish oil (FO; 22.5% w-3 PUFAs, detailed fatty acid composition in Supplement) or corn oil (CO; 1.3% w-3 PUFAs), as previously described35. Consistent with previous observations35–37 and the in vitro measurements here, dietary fish oil supplementation produced a robust incorporation of DHA into membrane lipids, with ~3-fold and ~30-fold more DHA-containing membrane lipids in the FO-fed versus CO-fed livers and hearts, respectively (Fig 1D-E). Interestingly, this effect was not observed for storage lipids (i.e. triglycerides, TAGs), as DHA did not incorporate into TAGs to a notable extent (not shown).
Lipidome remodeling associated with PUFA incorporation
Despite the copious incorporation of exogenous PUFAs into membrane lipids and the resulting increase in overall membrane unsaturation (Fig 1), cells in vitro did not show any obvious toxicity or differences in proliferation. This observation was somewhat surprising in light of the central role of lipid unsaturation in membrane physical properties38,39 and the critical role of those properties in regulating various cellular processes3,4,7 Thus, we hypothesized that mammalian cells may compensate for perturbations from exogenous FAs by remodeling their lipidomes. Indeed, while treatments with OA/PA had no effect on overall cell lipidomes (Supp Fig S2; consistent with their lack of incorporation), both AA and DHA supplementation reduced the abundance of other polyunsaturated (i.e. di-and tri-unsaturated) lipid species (Fig 2A-B). This effect could potentially be explained by replacement of these PUFA-containing lipids by AA/DHA-containing ones. Much more surprising was the highly significant increase in fully saturated lipids resulting from PUFA supplementation, due to significantly increased abundance of phospholipid-incorporated saturated fatty acids (Fig 2A-B). We note that for accurate estimates of lipidomic remodeling the supplemented FAs are removed from the analysis (e.g. ω-3-containing lipids from DHA-supplemented data), as the rather extreme over-abundance of those species upon supplementation suppresses the visualization of compensatory effects (original data are shown in Fig S3A). This analysis reveals approximately 2-fold and 4-fold more fully saturated lipids resulting from DHA and AA supplementation, respectively (Fig 2A-B). These changes were associated with significantly reduced overall unsaturation in phospholipids not containing the supplemented FAs (Fig 2A-B, insets). These effects were, at least in part, transcriptionally mediated, as evidenced by mRNA levels of the major fatty acid desaturase enzymes SCD1 and SCD2. As previously reported40, both were substantially down-regulated by DHA treatment (Fig S4), consistent with more saturated and fewer polyunsaturated lipids. Similar compensatory effects were observed for lipid length. Both AA (20-carbon) and DHA (22-carbon) are relatively long FAs, and thus their incorporation increased the overall length of membrane phospholipids (Fig S3B-C). However, there was also a notable increase in relatively short non-DHA/AA containing lipids (Fig S3D-E). The headgroup profile of membrane lipids was not notably affected by any of the treatments (Supp Fig S3F).
This lipidomic remodeling in response to PUFA feeding in vitro was not limited to a single cell type. The observations described so far were on a transformed leukocyte cell line (RBL); however, very similar DHA-induced lipid changes were observed in cultured CHO cells, primary rat hippocampal neurons, primary human MSCs, and MSCs that were differentiated in vitro into osteoblasts or adipocytes (Fig. 2C). Remarkably, not only the broad trends of the lipidomic remodeling (more saturated, less unsaturated lipids) were common between these disparate cell lineages and sources, but also the magnitude of the effects were quantitatively similar, despite drastic differences in overall lipid composition23,34,41,42. The similarity of the responses between freshly isolated cells, cultured primary cells, and long-term cultured cells lines derived from two different germ layers and three different organisms suggests that comprehensive lipidomic remodeling in response to exogenous PUFAs is a general phenomenon for mammalian cells.
Importantly, we confirmed that this response also occurs in vivo. The incorporation of dietary ω-3 PUFAs into mouse heart membrane lipids led to remarkably similar lipidomic remodeling as was observed in cultured cells. Namely, we observed ~3-fold increase in saturated lipids and 50% increased levels of monounsaturated lipids (Fig 2D). Concomitantly, there was a notable reduction in polyunsaturated lipids not containing ω-3 PUFAs. Altogether, the non-ω-3 PUFA lipids were significantly more saturated in heart membranes from FO-fed animals (Fig 2D, inset), fully consistent with compensatory, homeostatic lipidomic remodeling in vivo.
Lipidomic remodeling is nearly concomitant with exogenous PUFA incorporation
The remodeling associated with PUFA incorporation into membrane lipids suggested the induction of a homeostatic membrane response, wherein saturated lipids are upregulated to compensate for the fluidizing effect of PUFA-containing species. To support this inference, we analyzed the temporal profiles of PUFA incorporation and associated lipidomic changes in vitro. DHA was rapidly incorporated into membrane lipids, with significant increases in DHA-containing lipids observed within 1 h of supplementation and half-time of incorporation of ~4 h (Fig 3A-B). The recovery time course was significantly slower, as washout of DHA was followed by a return to baseline with a half-time of ~25 h (Fig 3C). This is the approximate doubling time of RBL cells in culture, suggesting that the “wash-out” effect likely results from dilution by new lipid synthesis rather than directed removal of DHA from membrane lipids. Remarkably, the associated lipidome remodeling proceeded with nearly identical temporal profiles to both DHA supplementation and wash-out (Fig 3D-I). The increase in saturated lipids (Fig 3E) and the decrease in di-unsaturated lipids (Fig 3H) were essentially concomitant with DHA incorporation and similar time courses were also observed for the wash-out (Fig 3F and 3I). These observations reveal unexpectedly rapid lipidomic changes to membrane lipid perturbations and suggest that such perturbations induce nearly simultaneous compensatory responses.
The nature of the broad compensatory lipidomic remodeling in response to PUFA supplementation in mammalian cells evokes classical observations of homeoviscous adaptation in ectothermic (e.g. nonthermoregulating) organisms13–16. There, perturbations of membrane physical properties produced by changes in ambient temperature are rapidly compensated by lipidomic changes apparently designed to renormalize membrane physical properties. Our findings suggest that a similar response to fluidizing stimuli occurs in mammalian cells, where it has been co-opted to cope with perturbations from exogenous lipids (e.g. from the diet).
Cholesterol upregulation by DHA via SREBP2
The above-described acyl chain remodeling in response to PUFA supplementation was incomplete, as the unsaturation index of PUFA-supplemented cells remained higher than those from untreated controls (Fig 1C and 1F). If this lipidome response were designed to normalize membrane properties, the acyl chain remodeling appears insufficient on its own, and there was no remodeling of GPL headgroups. In contrast to most prokaryotes where membrane homeostasis is mediated largely by glycerolipids14, membrane properties in more complex organisms are regulated to a very significant extent by sterols43,44. Thus, we hypothesized that the adaptive response to PUFA perturbation in mammalian cells is also mediated by cholesterol. Indeed, DHA supplementation significantly increased membrane cholesterol over the unsupplemented baseline in both murine heart tissue (Fig 4A) and cultured cells (Fig 4B). In cells, the increase in cholesterol abundance was again quite rapid, with effects observed within 1 h of DHA introduction and reaching a peak after ~4 h (Fig 4C), essentially concomitant with the acyl chain remodeling (Fig 3). There was also evidence of a small but notable overshoot, as the cholesterol increase was somewhat attenuated after ~4 hrs. These observations suggest that cholesterol upregulation is an early and potent response to membrane perturbation.
The machinery for cholesterol production in metazoans is regulated by proteolytic processing of transcription factors of the sterol regulatory element binding protein (SREBP) family45. Specifically, signals to upregulate cellular cholesterol levels are translated into proteolysis of a membrane-bound SREBP2 precursor to release a ‘mature’ cleaved fragment, which translocates to the nucleus to induce transcription of various target genes, including those for cholesterol synthesis and uptake45. Having observed a robust and rapid increase in cholesterol levels resulting from DHA supplementation, we evaluated whether SREBP2 processing was associated with this response. Indeed, DHA feeding increased the production of the ‘mature’ transcription factor form of SREBP2, with minimal effect on the precursor form (Fig 4D).
These findings suggested that processing of SREBP2 is involved in homeostatic membrane remodeling of mammalian lipid composition in response to perturbation by PUFA incorporation. To test this inference, we measured the response to PUFA supplementation in cells where SREBP processing was inhibited either genetically or pharmacologically. SRD-12B cells are clonal variants of CHO cells wherein a genetic defect in the Site 1 Protease (S1P; cleaves SREBPs to produce the transcriptionally active forms) prevents SREBP processing/activation46. In these cells, no notable upregulation of cholesterol (Fig 4E) or saturated lipids (Fig S5) was observed upon DHA supplementation. Similarly, chemical inhibition of SREBP processing by the pentacyclic triterpene betulin47 completely abrogated DHA-induced upregulation of cholesterol (Fig 4B, triangles) and saturated lipids (Fig S6). These results support the crucial role for SREBP2 in lipidomic remodeling induced by PUFAs.
Physical homeostasis following lipidomic perturbation
The above data reveal rapid and comprehensive lipidomic remodeling resulting from the incorporation of DHA into membrane lipids, wherein lipids that decrease membrane fluidity ? increase membrane packing (saturated lipids and cholesterol) are rapidly upregulated in response to introduction of PUFA-containing lipids that increase membrane fluidity. These changes are consistent with lipidome remodeling for the purpose of homeostatic maintenance of membrane physical properties. To directly evaluate this inference, we measured membrane packing in live cells using a fluorescence assay that relies on a solvatochromic dye (C-Laurdan) whose spectral characteristics are dependent on membrane properties48. Specifically, the emission spectrum of C-Laurdan is red-shifted in loosely-packed membranes (due to the enhanced polarity of the fluorophore nano-environment), and the relative extent of this spectral shift can be quantified by ratiometric spectroscopy or imaging49. The resulting dimensionless parameter called Generalized Polarization (GP) is a widely used proxy for membrane packing and fluidity50–54, with higher values reporting more tightly packed membranes.
Fig 5A shows a map of C-Laurdan GP in live cells generated by confocal spectral imaging55. The relatively fluid (low GP; blue/green pixels) internal membranes and relatively packed plasma membranes (high GP; yellow/red pixels) are characteristic of mammalian cells49,51 (see trace right). These clearly distinct regions enabled us to separately quantify the effects of DHA on plasma versus internal membranes (Fig 5B). The packing of internal membranes was significantly decreased shortly after introduction of exogenous DHA (purple circles; pooled data for 2 and 4 h time points are different from 0 with p < 0.01), suggesting reduced packing consistent with the known fluidizing effect of PUFA-containing lipids23,56. This response eventually reversed, with GP fully normalizing to baseline by 24 h. Thus, DHA incorporation combined with the associated remodeling produced no net change in overall membrane properties. This renormalization of membrane packing was markedly reduced when lipidomic remodeling was inhibited by targeting SREBP processing. Namely, betulin treatment suppressed the recovery of membrane packing following DHA treatment (Fig. 5B, orange symbols and line), leading to significantly reduced GP in DHA-treated RBL cells after 24 h (Fig. 5B-C). Similarly, while wild-type CHO cells did not show a significant effect of DHA on membrane fluidity after 24 h treatment, DHA supplementation of SRD-12B cells exhibited significantly reduced GP.
Disruption of membrane homeostasis reduces cell fitness
In unicellular organisms, the homeoviscous lipidomic response is necessary for maintaining membrane physical properties in a range compatible with the biochemical processes necessary for life13,15. Having observed similar lipidomic and biophysical responses in mammalian membranes, we hypothesized that this adaptation was necessary for cellular fitness under conditions of membrane stress. To test this hypothesis, we evaluated the cytological effects of DHA under chemical or pharmacological inhibition of key homeostatic response nodes. Neither DHA nor betulin alone showed significant cytostatic effects; however, DHA supplementation in the presence of betulin led to ~50% decrease in cell number after 3 days of culture (Fig 6A; full dose-response in Supp Fig S7). Similar effects were observed by inhibiting fatty acid synthase (via 10 nM GSK2194069; Fig 6B), which is presumably necessary for production of fatty acids towards the phospholipid remodeling described in Figures 2–3. These observations were confirmed in SRD-12B cells, whose genetic lesion renders them incapable of activating SREBP46 and mounting an adaptive lipidomic and biophysical response to DHA supplementation (Figs 4E and 5D). Consistent with the abrogated homeostatic response, DHA had a significant cytostatic effect on SRD-12B cells (Fig 6C, green). These results reveal that inhibition of lipidome remodeling in mammalian cells markedly reduced cellular fitness upon membrane perturbation with dietary fatty acids.
DISCUSSION
The incorporation of dietary fatty acids into mammalian membrane lipids has been widely observed both in vitro23,57,58 and in vivo19,20,36,37,59. The fact that exogenous fatty acids are so readily used for lipid synthesis is unsurprising in light of the fact that the enzyme required for de novo fatty acid production (fatty acid synthase) is minimally expressed in most adult human tissues60. Indeed, de novo lipogenesis is considered “negligible”61,62 in adult humans, suggesting that exogenous sources of fatty acids are the major raw material for maintenance and replenishment of membrane lipids. This reliance on exogenous inputs for production of components central to cellular architecture and function would seem to present a major complication for homeostasis.
In ectothermic organisms, homeostatic membrane control has been widely observed in the form of homeoviscous adaptation13–15. According to this hypothesis, the relative abundance of saturated to unsaturated membrane lipids is responsive to temperatures changes in order to maintain membrane fluidity at the level required for the many processes hosted and regulated by cell membranes. Although such adaptation has been proposed in prokaryotes14, single-celled eukaryotes63, and even cold-blooded animals64,65, studies on endothermic organisms have been limited28,29, and there has yet been no direct observations of HVA in mammals or isolated mammalian cells. Our data directly confirm the three major tenets of cell-autonomous homeoviscous adaptation: (1) lipidomic remodeling resulting from a perturbation of membrane physical properties (Fig 2–4); (2) recovery of baseline physical properties at a new lipid composition (Fig 5); and (3) necessity of this response for cellular fitness (Fig 6). These observations suggest that mammalian cells possess the capacity for homeostatic membrane adaptation analogous to HVA, and that this response can compensate for perturbations from dietary lipid inputs.
In cold-blooded animals, the homeostatic membrane response also involves modulation of cholesterol levels44,65, and we observe a similar response in mammalian cells, mediated at least partially through activation of SREBP2 (Fig 4). It is quite remarkable to note that disruption of SREBP activation either by betulin or in SRD-12B cells resulted in near-complete abrogation of the compensatory response. This abrogation was notable not only in the lack of cholesterol upregulation (Fig 4D and 6A), which may have been expected, but also in the failure to upregulate saturated lipids (Figs S5 and S6), a response not directly connected to SREBP2 target genes.
Our implication of SREBP2 as a critical node of the sense-and-respond module for membrane adaptation is consistent with the central role of SREBPs in membrane homeostasis45. SREBP transcription factors have been dubbed the “master regulators of lipid homeostasis”, because they direct not only cholesterol synthesis and uptake, but also proteins associated with membrane lipid metabolism, including those involved in FA synthesis, elongation, and desaturation66. The effect of PUFA supplementation on SREBP function has been extensively studied, and DHA is known to suppress the activation of SREBP1 and its target genes, both in cultured cells and in vivo67–69. We observed downregulation of two SREBP1 target genes (SCD1 and SCD2, Fig S3) suggesting a similar effect in our cells. In contrast, SREBP2 is not suppressed by PUFAs67,68, revealing that these two complementary regulators of membrane homeostasis have different functions, despite both being sensitive to membrane cholesterol. In our observations, SREBP2 is induced as a necessary part of the homeostatic response. Previous reports68 have not noted a significant effect of PUFAs on SREBP2, possibly because those experiments involved acute PUFA feeding of serum-starved cells, where high levels of activated SREBPs may have suppressed the DHA-mediated stimulation we observe here. We have confirmed this effect in our cells (not shown). It remains to be determined whether SREBP2 is simply an effector of membrane remodeling downstream of yet-unidentified sensing machinery, or whether this protein (and/or its regulatory machinery) may itself be capable of sensing perturbations in membrane physical properties. Direct demonstration of protein responsiveness to membrane packing has recently been described for two yeast ER proteins5,18, providing a conceptual and methodological toolbox for identifying other membrane sensors. Remarkably, despite the ubiquity and importance of membrane homeostasis, the machinery used for sensing membrane properties remains largely uncharacterized. It is an intriguing observation that Ire1, a core component of the unfolded protein response (UPR), exhibits a dual sensitivity to unfolded proteins and aberrant lipid compositions, demonstrating a tight connection between protein-folding and membrane properties18. This connection suggests that the machinery for ameliorating protein-folding stresses may also be involved in transducing and mitigating membrane stress70.
In summary, our observations strongly support the hypothesis that mammalian cells in culture and in vivo sense membrane physical properties and respond to perturbations by comprehensive remodeling of their lipidomes. In our measurements, the perturbations were induced by supplementation with polyunsaturated fatty acids (PUFAs); however, it is likely that other fats (e.g. cholesterol) or amphiphiles (bile acids, anesthetics) may induce similar responses. We implicate SREBP2 as a central response mediator and show that the compensatory response is important for cell fitness, confirming membrane adaptation as a central requirement for cellular homeostasis.
MATERIALS AND METHODS
Materials
Betulin and GSK2194069 were obtained from Sigma Aldrich. C-Laurdan was purchased from TPProbes (South Korea). Amplex Red kit to quantify cholesterol was purchased from Invitrogen. Antibodies used: actin (monoclonal clone AC-15, Abcam), SREBP2 (polyclonal, Abcam).
Cell culture
Rat basophilic leukemia (RBL) cells were maintained in medium containing 60% modified Eagle’s medium (MEM), 30% RPMI, 10% fetal calf serum, 100 units/mL penicillin, and 100 μg/mL streptomycin. Chinese hamster ovary (CHO) cells were maintained in DMEM:F12 (1:1) containing 5% fetal calf serum, 100 units/mL penicillin, and 100 μg/mL streptomycin. SRD12B cells were maintained in DMEM:F12 (1:1) containing 5% fetal calf serum, 50μM sodium mevalonate, 20μM oleic acid, 5μg/mL cholesterol, 100 units/mL penicillin, and 100 μg/mL streptomycin. All cells were grown at 37°C in humidified 5% CO2.
Fatty acid treatments
Fatty acid stock solutions were received as air-purged ampules (Sigma Aldrich) and loaded into BSA immediately upon opening. BSA loading was accompanied by stirring the fatty acid with BSA dissolved in water (2:1 mol/mol FA:BSA), sterile filtering, purging with nitrogen prior to aliquoting, and storing at -−°C. BSA loading, purging, and cold storage were all done to minimize FA oxidation. For all experiments with FA supplementation, cells were incubated with 20 μM FA for 3 days (cells were re-supplemented 24 h prior to analysis). Serum-containing, rather than serum-free, media was used to approximate physiological supplementation, rather than exclusive introduction of specific FAs. These conditions were chosen to approximate in vivo settings produced by DHA-enriched diets in mammals: plasma free FA concentrations range from 300 to 750 μM59,71 and up to 10 mol % of plasma fatty acids are ω-3 DHA in rats fed a high-fish-oil diet59. Further, diets rich in ω-3 PUFAs led to significant incorporation of these fats into cell membrane lipids19,37,59, similar to the levels we observed under our culture feeding conditions (Fig. 1). For these reasons, we believe our culture conditions reasonably approximate physiological dietary membrane perturbations.
Animals and diets
All experimental procedures using laboratory animals were approved by the University Laboratory Animal Care Committee of Texas A&M University. Pathogen-free female C57BL/6 mice (n=180; Frederick Research Facility, Frederick, MD) weighing 16–18 g, were randomly divided into two groups of 90. For 2 wk, mice had free access to one of the two semipurified diets, which were adequate in all nutrients35. Diets varied only in the oil composition, i.e., either corn oil (CO) or an (n-3) PUFA-enriched fish-corn oil (FO) mixture (4:1, w/w) at 5 g/100 g diet. The basic diet composition, expressed as g/100 g was: casein, 20; sucrose, 42; cornstarch, 22; cellulose, 6; AIN-76 mineral mix, 3.5; AIN-76 vitamin mix, 1, DL-methionine, 0.3; choline chloride, 0.2; Tenox 20A (containing 32% glycerol, 30% corn oil, 20% tert-butylhydroquinone, 15% propylene glycol, 3% citric acid) 0.1; and oil, 5. The fatty acid composition of the diets, as determined by gas chromatography, is shown in Supplemental Table 1. Mice were killed by CO2 asphyxiation after two weeks of feeding the CO/FO diet, and the livers and hearts were isolated and frozen in liquid nitrogen. 20-50 mg of tissue was mechanically homogenized in Dulbecco’s PBS (without Ca2+ and Mg2+). Samples were further diluted to 5 mg/mL in Dulbecco’s PBS (without Ca2+ and Mg2+), and lipidomics analysis was performed as below.
Drug treatments and cell number quantification
RBL cells were treated with or without 20 μM DHA in the presence or absence of betulin (inhibitor of SREBP processing) or GSK2194069 (FAS inhibitor). Cells were treated for 72 h and then the cell number determined via fluorescein diacetate (FDA) fluorescence. FDA is a cell viability probe which freely diffuses through cell membranes, but is trapped in cells following de-acytelation by cytoplasmic esterases in viable cells. The number of viable cells is then directly related to fluorescein fluorescence intensity. For the assay, cells were treated in 96-well plates, gently washed with PBS, and then incubated with fluorescein diacetate (5 μg/mL in PBS) for 2 minutes at 37°C. The plates were then washed again to remove excess FDA and fluorescence was measured at 488 nm excitation and 520 nm emission. For each experiment the number of cells per well were normalized to untreated, unsupplemented cells.
Lipidomics
Detailed lipidomic analysis was performed by Lipotype, GmbH, as previously described72. Briefly, for preparation of crude cell membranes, cells were washed with phosphate buffered saline (PBS), scraped in 10 mM Tris pH 7.4, and then homogenized with a 27-gauge needle. Nuclei were then pelleted by centrifugation at 300 xg for 5 min. The supernatant was pelleted by centrifugation at 100,000 xg for 1 h at 4°C. The membrane pellet was then washed and resuspended in 150 mM ammonium bicarbonate.
Lipids detected
Ceramide (Cer), Chol, SM, diacylglycerol (DAG), lactosyl ceramide (DiHexCer), glucosyl/galactosyl ceramide (HexCer), sterol ester (SE), and triacylglycerol (TAG), as well as phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), and phosphatidylinositol (PI), phosphatidylserine (PS), and their respective lysospecies (lysoPA, lysoPC, lysoPE, lysoPI, and lysoPS) and ether derivatives (PC O-, PE O-, LPC O-, and LPE O-). Lipid species were annotated according to their molecular composition as follows: [lipid class]-[sum of carbon atoms in the FAs]:[sum of double bonds in the FAs];[sum of hydroxyl groups in the long chain base and the FA moiety] (e.g., SM-32:2;1). When available, the individual FA composition according to the same rule is given in brackets (e.g., 18:1;0-24:2;0)
Lipid extraction
Samples were extracted and analyzed as previously described (39,47), which is a modification of a previously published method for shotgun lipidomics (48). Briefly, membrane samples were suspended in 150 μL of 150 mM ammonium bicarbonate in water, spiked with 20 μL of internal standard lipid mixture, and then extracted with 750 μL of a chloroform/methanol 10:1 (v/v) mixture for 2 h at 4_C with 1400 rpm shaking. After centrifugation (3 min, 3000 g) to facilitate phase partitioning, the lower, lipid-containing, organic phase was collected (first-step extract), and the remaining water phase was further extracted with 750 μL of a chloroform/methanol 2:1 (v/v) mixture under the same conditions. Again the lower, organic phase was collected (second-step extract). Extracts were dried in a speed vacuum concentrator, and 120 μL of a dried first-step extract underwent acetylation with 75 μL of an acetyl chloride/chloroform 1:2 (v/v) mixture for 1 h to derivatize Chol. After the reaction was completed, the mixture was dried. Then, 120 μL of a dried first-step extract and a derivatized extract were resuspended in an acquisition mixture with 8 mM ammonium acetate (400 mM ammonium acetate in methanol/chloroform/methanol/propan-2-ol, 1:7:14:28, v/v/v/v). Next, 120 μL of the second-step extract was resuspended in an acquisition mixture with 30 μL 33% methylamine in methanol, in 60 mL methanol/chloroform 1:5 (v/v). All liquid-handling steps were performed using a Hamilton (Reno, NV) STARlet robotic platform with the Anti Droplet Control feature, ensuring the accuracy and reproducibility of organic solvent pipetting.
Lipid standards
Synthetic lipid standards were purchased from Sigma-Aldrich (Chol D6), Larodan (Solna, Sweden) Fine Chemicals (DAG and TAG), and Avanti Polar Lipids (all others). The standard lipid mixtures were chloroform/ methanol 1:1 (v/v) solutions containing Cer 35:1;2, (D18:1;2, 17:0;0); Chol D6; DAG 34:0;0 (17:0;0, 17:0;0); DiHexCer 30:1;2 (D18:1;2.12:0;0); HexCer 30:1;2 (D18:1;2.12:0;0); LPA 17:0;0 (17:0;0); LPC 12:0;0 (12:0;0); LPE 17:1;0 (17:1;0); LPI 17:1;0 (17:1;0); LPS 17:1;0 (17:1;0); PA 34:0;0 (17:0;0, 17:0;0); PC 34:0;0 (17:0;0, 17:0;0); PE 34:0;0 (17:0;0, 17:0;0); PG 34:0;0 (17:0;0, 17:0;0); PI 32:0;0 (16:0;0, 16:0;0); PS 34:0;0 (17:0;0, 17:0;0); SE 20:0;0 (20:0;0); SM 30:1;2 (18:1;2, 12:0;0); and TAG 51:0;0 (17:0;0, 17:0;0, 17:0;0).
Lipid spectrum acquisition
Extracts in the acquisition mixtures were infused with a robotic nanoflow ion source (TriVersa NanoMate; Advion Biosciences, Ithaca, NY) into a mass spectrometer instrument (Q Exactive, Thermo Scientific). Cer, DiHexCer, HexCer, lysolipids, and SM were monitored by negative ion mode Fourier transform mass spectrometry (FT-MS). PA, PC, PE, PI, PS, and ether species were monitored by negative ion mode FT tandem MS (FT-MS/MS). Acetylated Chol was monitored by positive ion mode FTMS. SE, DAG, TAG, and species were monitored by positive ion mode FT-MS/MS.
Lipid identification and quantification
Automated processing of acquired mass spectra, and identification and quantification of detected molecular lipid species were performed with the use of LipidXplorer software (17). Only lipid identifications with a signal/noise ratio of >5, an absolute abundance of at least 1 pmol, and a signal intensity fivefold higher than in the corresponding blank samples were considered for further data analysis.
C-laurdan spectroscopy
Membrane packing (related to order and fluidity) was determined via C-laurdan spectroscopy as described73,74. Briefly, cells were washed with PBS and stained with 20 μg/mL C-laurdan for 10 minutes on ice. The emission spectrum from isolated GPMVs was gathered from 400-550 nm with excitation at 385 nm at 23°C. The GP was calculated according to the following equation:
C-Laurdan spectral imaging
C-Laurdan imaging was performed as previously described23,34,55,73,74 Briefly, cells were washed with PBS and stained with 10 μg/mL C-Laurdan for 10 min on ice, then imaged via confocal microscopy on a Nikon A1R with spectral imaging at 60x and excitation at 405 nm. The emission was collected in two bands: 433–463 nm and 473–503 nm. MATLAB (MathWorks, Natick, MA) was used to calculate the two-dimensional (2D) GP map, where GP for each pixel was calculated from a ratio of the two fluorescence channels, as previously described55. Briefly, each image was binned (2×2), background subtracted, and thresholded to keep only pixels with intensities greater than 3 standard deviations of the background value in both channels. The GP image was calculated for each pixel using the above equation. GP maps (pixels represented by GP value rather than intensity) were imported into ImageJ. Line scans were drawn across individual cells. PM GP values were calculated as averages of the peaks as values higher than 0.26. For internal membranes, average GP values were calculated from line scans around the nucleus (visible as a dark spot in C-Laurdan images). At least 10 cells were imaged and quantified per experiment with at least 4 individual experiments performed. Reported are the mean and standard deviations of the average GP values for the individual experiments.
Amplex Red assay
Amplex Red cholesterol assay was performed (according to manufacturer instructions; Invitrogen) to determine the abundance of cholesterol. Each reading was normalized to protein concentration (determined by bicinchoninic acid (BCA) assay) in the same samples. Technical triplicates were measured for each sample. Shown are values normalized to untreated cells. Reported are the average and standard deviations from at least 3 independent biological replicates.
Western Blot
Cells were washed with ice cold PBS, then scraped into Laemmli lysis buffer (50mM Tris-HCl, pH 8.0; 2% SDS; 5mM EDTA, pH8.0) supplemented with protease inhibitor cocktail. Protein concentration was determined using BCA (Pierce), and equal amounts of protein were mixed with reducing Laemmli sample buffer and loaded onto SDS-PAGE gels. Gels were transferred to PVDF membranes, which were blocked in 5% BSA. Membranes were incubated with primary antibodies overnight at 4°C, and detected with either AlexaFluor or HRP-tagged secondary antibodies. Membranes were imaged using a BioRad ChemiDoc imager. The intensities of the bands were quantified normalized to actin, and plotted as mean ± SD of n ≥ 3 experiments.
Real Time Quantitative PCR
Total RNA was isolated via Trizol (Sigma) following the manufacturer’s protocol. Reverse transcriptase PCR was performed using the High Capacity cDNA Reverse Transcription Kit from Applied Biosystems according to manufacturer’s protocol. To quantify mRNA expression, SYBR Fast MasterMix (2x) Universal Dye (#KK4602) from Kapa Biosystems was used in an Eppendorf Realplex2 Mastercycler. Each primer set for each sample was run in triplicate with 1 ng of cDNA per well. The primer sets used are:
Expression changes were calculated using the delta delta CT method. The data was standardized using a previously published protocol75 in which the average fold change was log transformed, mean centered, and confidence intervals determined to evaluate statistical significance.
Acknowledgements
All confocal microscopy was performed at the Center for Advanced Microscopy, Department of Integrative Biology and Pharmacology McGovern Medical School at the University of Texas Health Science Center at Houston. We acknowledge funding from the Cancer Prevention Research Institute of Texas (CPRIT; R1215), NIH/National Institute of General Medical Sciences (grant nos. GM114282, GM124072, GM120351), and the Volkswagen Foundation (grant no. 93091).