Live-Cell Imaging and Lipidomics of Low-Density Lipoprotein Containing Intrinsically Fluorescent Cholesteryl Esters

Abstract

Low-density lipoprotein (LDL) delivers cholesterol to mammalian cells in the form of cholesteryl esters (CEs), but due to the lack of suitable tools, our understanding of the intracellular transport and hydrolysis of CEs is limited. We present a novel approach for studying LDL-derived CEs in cells using fatty acyl chain conjugates of the intrinsically fluorescent cholestatrienol (CTL). We demonstrate that CTL esters reconstituted into LDL particles are hydrolyzed in late endosomes and lysosomes (LE/LYSs) by acid lipase, while an LDL-derived CTL ether analogue cannot leave LE/LYSs. Using live-cell imaging, lipidomics, and multimodal Bayesian modeling, we discover a sequential biphasic transport of LDL-derived CTL to LE/LYSs with a half-time of 3.0 h. Hydrolyzed CTL derived from LDL-associated CTL esters is rapidly re-esterified with a similar half-time and stored in lipid droplets, demonstrating efficient sterol transport to the endoplasmic reticulum. Lack of functional Niemann-Pick C1 or C2 protein caused lysosomal accumulation of LDL-derived sterol. Using lipidomics and kinetic modeling, we also track LDL-derived CEs and triacylglycerols in cells and determine the uptake kinetics for each lipid species individually. Our novel approach allows for precise measurement of postendocytic trafficking and metabolism of LDL-derived cholesterol and other lipids in living cells.

Introduction

Mammalian cells have a regulated homeostasis of cholesterol, with de novo synthesis and cellular uptake controlled by transcriptional processes located primarily in the endoplasmic reticulum (ER). The abundance of cholesterol in the ER is low under normal conditions, and dedicated mechanisms exist in the ER to maintain cholesterol homeostasis. Excess cholesterol is esterified by sterol O-acyltransferases (i.e., SOAT1, SOAT2), also named acyl-CoA acyl transferase (ACAT), and stored as cholesteryl esters (CEs) in lipid droplets , A drop below about 5 mol % signals cellular cholesterol depletion to the cholesterol-sensing SCAP protein and triggers transcription of cholesterol target genes via the sterol response element-binding protein (SREBP). These genes code for enzymes of the biosynthetic pathway as well as the receptor for low-density lipoprotein (LDL-R), which mediates uptake of cholesteryl ester-rich LDL particles by clathrin-mediated endocytosis (CME). Internalized LDL dissociates from its receptor in the endosomes due to acidification of these organelles via the vacuolar proton pump. , The LDL particles released from their receptor in early endosomes are transported to late endosomes and lysosomes (LE/LYSs), where their degradation starts after a lag phase of about 1–3 h post internalization. , Biochemical studies using radioactive tracers showed that lysosomal acid lipase (LAL, i.e., lysosomal integral protein A, LIPA) mediates the hydrolysis of the majority of the LDL-associated CEs in fibroblasts. −

The trafficking itineraries and metabolic fates of LDL-derived cholesterol, once CEs are hydrolyzed, are less well-defined. , By reconstituting LDL particles with radioactive CEs, Neufeld and coworkers provided evidence for a direct trafficking route of the liberated cholesterol from LE/LYSs to the plasma membrane (PM). They suggested that this route comprises about 70% of the LDL-derived cholesterol, while the remaining 30% traffics to the ER. Similar conclusions were drawn by Liscum and coworkers, who employed small molecule inhibitors of lysosomal cholesterol export, such as U18666A and imipramine, to show that a portion of LDL-derived cholesterol directly traffics to the ER. Others have argued that the entire pool of liberated LDL-derived cholesterol first passes through the PM before reaching the ER. , Dysregulated uptake of cholesterol from modified LDL particles takes place during the development of atherosclerosis, which leads to massive reesterification of excess cholesterol by ACAT1 and foam cell formation in macrophages. − In neurodegenerative diseases, such as Wolman and Niemann-Pick type C (NPC) disease, impaired hydrolysis and/or transport of the LDL-derived cholesterol out of the lysosomal compartment leads to lysosome dysfunction. ,− LDL-derived cholesterol also plays an important role in the development of brain cancers, as glioblastoma cells allegedly shut down their endogenous production of cholesterol and rely primarily on imported LDL cholesteryl esters for growth and survival. Thus, understanding the mechanisms underlying the cellular processing of LDL-derived cholesterol is instrumental for developing novel therapies against these devastating diseases.

Our current understanding of intracellular transport and metabolism of LDL-derived cholesterol relies mainly on reconstituting radioactively labeled CEs into LDL and following their transport to subcellular organelles by membrane fractionation and scintillation counting. While being powerful, this approach suffers from impurities arising from organelle purification, limited time resolution, and the fact that no directional information is available. Also, subcellular organelles of many cells are pooled before lipid analysis, preventing any insight into the cell-to-cell heterogeneity of lipoprotein transport. Classical fluorescence imaging-based approaches use a combination of targeted gene knockouts and fluorescence staining with the sterol-binding polyene filipin or cholesterol-sensing proteins, like AloD4, perifringolysin, or the GRAMD1 protein domains, to obtain insight into the spatial distribution of various cholesterol pools. − These tools were employed to identify a variety of proteins whose genetic ablation causes cholesterol accumulation in LE/LYSs and/or impaired delivery of LDL-derived cholesterol to ACAT in the ER, including members of the oxysterol-binding protein family, − cytosolic or ER-bound sterol transfer proteins containing the START domain − as well as ER-resident transporters containing a sterol-binding GRAM domain. − However, given the complexity of sterol trafficking pathways, cholesterol buildup in a given compartment or lack thereof does not necessarily translate into an understanding of the sterol transport dynamics or trafficking routes between individual organelles. For example, increased filipin staining in a subcellular compartment after loading cells with LDL could be the result of not only cholesterol accumulation from the LDL source but also the buildup of endogenous cholesterol. Since filipin is not specific for cholesterol, accumulation of other sterols, such as oxysterols, cannot be distinguished by filipin staining either. Furthermore, the kinetics of sterol trafficking between organelles cannot be determined, and the actual trafficking route of LDL-derived cholesterol is thereby overlooked by this standard approach.

Here, we introduce a novel method to directly follow the intracellular trafficking and metabolic processing of LDL-derived cholesterol using minimally modified fluorescent analogs of CEs. We chemically link cholestatrienol (CTL), differing from cholesterol only in having two additional double bonds in the ring system (Figure A), to the fatty acid oleic acid either as a hydrolyzable ester (CTL-ester) or a nonhydrolyzable ether (CTL-ether). By combining live-cell imaging with lipidomics and Bayesian kinetic modeling, we establish a novel toolbox for direct visualization and quantitative analysis of trafficking and metabolic conversion of lipoprotein-derived lipids in living cells. Our primary focus is on human fibroblasts, since groundbreaking work including detailed pulse-chase studies has been carried out in these cells. ,,, Also, LDL trafficking has been studied in skin fibroblasts from patients suffering from lysosomal storage disorders, ,,, which are readily available and have been used in this study. To demonstrate the broad applicability of our novel approach, we also present results obtained in macrophages, a cell type with tremendous importance in immunometabolism and cardiovascular disease. Our new methods should find wide application in understanding cholesterol homeostatic pathways and disease pathogenesis.

1.

Structure and properties of intrinsically fluorescent CTL 18:1 ester and ether. A, Cholestatrienol (CTL) oleate ester and ether are shown with differences from the corresponding cholesterol ester and ether being two additional double bonds in the steroid ring system (indicated in red for the ester and green for the ether). B, Fluorescence emission spectra for 0.2 mg/mL LDL reconstituted with either CTL-oleate ester (red line and left axis) or CTL-oleate ether (green line and right axis), excited at 328 nm. The spectrum of the same amount of unlabeled LDL is shown for comparison (blue). C, Thin layer chromatography (TLC) of CTL-ester and ether after treatment with cholesteryl esterase. TLC plates were developed with copper sulfate, and the red circle indicates hydrolyzed CTL obtained from the CTL-ester. The same position is empty for the CTL-ether (green circle).

Experimental Section

Reagents

The human lipoprotein-depleted serum (LPDS, S5519), LDL (SAE0053), cholesteryl oleate (303-43-5), Sigmacote (SL2), potato starch (9005-25-8), U18666A (U3633), and Lalistat-2 (SML2053) were purchased from Sigma Chemicals (St. Louis, MO). Fetal bovine serum (FBS) and DMEM were obtained from GIBCO BRL (Life Technologies, Paisley, Scotland). Rhodamine-labeled dextran (D1819), ER-tracker red (E34250), Nile Red (N1142), Alexa Fluor 488 Protein Kit (A10235), and NBD-C6-Ceramide (N1154) were from Invitrogen/Molecular Probes (Inc., USA). The Amplex Red kit (A12216) for quantification of cholesterol and BODIPY 493/503 (D3922) were purchased from Thermo Fisher (Denmark office, Odense, DK). Avasimibe (APOSBISN0114) was purchased from BioNordika (Herslev, Denmark). CTL, CTL-oleate ester, and ether were synthesized as described in the Supporting Information. Niemann-Pick C2 protein (NPC2) was purified from bovine milk and fluorescence-labeled as described.

Cell Culture and Labeling

Human skin fibroblasts from a healthy donor (Coriell Institute, GM08680), from a patient lacking functional NPC1 (Coriell Institute, GM03123) or NPC2 (Coriell Institute, GM18455) as well as J774A.1 murine macrophages (ATCC, TIB-67) were cultured in T25 culture flasks at 37 °C in an atmosphere of 5% CO₂. Fibroblasts were cultured in complete DMEM medium supplemented with 1% glutamax, 1% penicillin–streptomycin (P/S), and either 10% FBS for control fibroblasts or 25% FBS in the case of NPC disease fibroblasts. J774A.1 cells were cultured in DMEM medium supplemented with 1% P/S and 10% heat-inactivated FBS. Cells were labeled with fluorescent LDL and organelle markers as described in the Supporting Information.

Live-Cell Imaging and Image Analysis

All fluorescence microscopy experiments and accompanying image analysis were carried out as described in the Supporting Information.

Lipid Extraction and Analysis

Cells were prepared as before and incubated in LPDS media for 24 h prior to incubation with the LDL-CTL-oleate ester for 2, 6, and 24 h. Cells were washed with PBS and harvested by scraping in cold PBS. After centrifugation at 500g at 4 °C for 3 min, the supernatant was removed, and the sample was snap-frozen in liquid nitrogen. Frozen cell pellets were incubated for 5 min at 90 °C with preheated 155 mM ammonium formate buffer and intermittent vortexing and placed on ice. The samples were sonicated for 20 s at 40% power using a Branson tip sonicator. An aliquot of each cell lysate was used to determine the protein concentration by using the BCA assay (Thermo Fisher Scientific). Details of the Lipid-MS analysis are provided in the Supporting Information.

Kinetic Modeling and Bayesian Model Inference

Kinetic modeling of the data for transport and hydrolysis of LDL-associated CTL-esters was based on a sequential 2-pool model (see Supporting Information for model equations), solved analytically and fitted to the experimental data using nonlinear regression as implemented in the SciPy library in Python. Alternatively, the data were fitted to the same model in a Bayesian setting with normal distributions for the mean and a half-normal for the standard deviation as informed priors using the Python library PyMC3. Further details of this analysis are provided in the Supporting Information.

Results and Discussion

Binding and Endocytosis of LDL Containing Fluorescent Cholesteryl Esters and Ethers

Our CTL 18:1-ester and -ether probes contain identical conjugated double bonds in the steroid ring system and differ only in the linkage of the C18:1 hydrocarbon chain to the sterol backbone (Figure A). Accordingly, upon reconstitution into LDL (Figure S1A), their fluorescence spectra are very similar (Figure B). The reconstituted particles contain more than 90% sterols, mostly CEs and some cholesterol, with the remaining lipids being composed of glycerophospholipids, di- and triacylglycerols, as well as sphingolipids (Figure S1B). This composition is somewhat different from LDL particles isolated from human plasma, whichon averagecontain about 48% CEs (corresponding to ∼1536 CE molecules per particle). , Also, in our particles, CTL-esters reconstituted into LDL must be tightly packed in the core of the lipoproteins, as adding Triton X-100 caused a slight increase in CTL fluorescence. This is likely the result of dequenching upon solubilization of the particles (Figure S1C). The main glycerophospholipid in LDL was phosphatidylcholine (PC), followed by various species of phosphatidylethanolamine (PE, Table S1).

Our reconstituted LDL additionally contained sphingomyelin (SM) and small amounts of ceramides (Table S2), similar as native LDL. The fluorescent ester, but not the CTL-ether, can be hydrolyzed by a neutral lipase, as inferred from thin layer chromatography (TLC) plates (Figure C). The overall low abundance of nonsterol lipids in the reconstituted LDL particles suggests that they resemble phospholipid-depleted LDL, as described previously. Incubating human fibroblasts with LDL containing CTL-ester overnight resulted in strong fluorescence in the PM and intracellular vesicles (Figure S2A). When excess unlabeled LDL was added during this incubation, the cell-associated CTL fluorescence was greatly reduced (Figure S2A-D). Similar results were found for LDL containing CTL-ether (Figure S3A, B), indicating that the nature of the sterol esters does not affect endocytosis of the LDL particles. This is in line with earlier findings, showing that the apoprotein portion in LDL controls the cellular uptake of the lipoprotein. Remaining cell-associated fluorescence of LDL containing either CTL-esters or -ethers is likely due to the binding of the particles to other receptors than the LDL-R, as has been observed already in the seminal studies by Goldstein and Brown (1974 and 1975). , Phospholipid-depleted LDL has been shown to have increased affinity to proteoglycans compared to native LDL, and such a pathway could contribute to internalization of our LDL particles.

Hydrolysis of CEs derived from LDL has been shown to take place in LE/LYSs primarily by acid lipase, and to test whether this also applies to our novel fluorescent sterol esters, we incubated fibroblasts with LDL containing CTL-ester in the presence of the acid lipase inhibitor Lalistat (Figure S2E and F). We found that, in the presence of Lalistat, lysosomal staining of CTL-ester is strongly increased, while that of the PM is very low compared with cells incubated in the absence of Lalistat. Lysosomal targeting was confirmed by costaining cells with rhodamine-dextran (Rh-dextran; Figure S2E and F, inset). Since the fluorescence of CTL-ester is indistinguishable from that of free CTL, we can only indirectly infer that the PM staining observed for nontreated cells stems from free CTL, as such staining was much lower in the presence of Lalistat (Figure S2E′ and F′). Strong endolysosomal fluorescence of the sterol was also found in fibroblasts incubated with LDL containing CTL-ether (Figure S3). We did not detect any free CTL by Lipid-MS when loading cells with LDL containing CTL-ether, confirming that the CTL 18:1-ether cannot be hydrolyzed in cells. The fluorescence of CTL in the PM of cells after 24 h incubation with LDL containing CTL-ester in the absence of the lipase inhibitor is homogeneous, in stark contrast to the patchy PM staining observed in cells incubated with LDL containing CTL-ether.

While our current setup does not allow us to distinguish these populations unequivocally, we surmise that a homogeneous PM staining stems from membrane-intercalated CTL, while LDL particles bound to their receptor on the cell surface result in a patchy signal of CTL esters in the LDL core. Supporting that notion, we have compared the membrane fluorescence of fibroblasts incubated with either Alexa488-LDL or LDL containing CTL ester to that of cells stained with CTL delivered from albumin. We confirmed that the PM signal of Alexa488-LDL and of CTL derived from LDL is patchy after 2 h of incubation, while that of CTL is homogeneous in the PM after overnight labeling from either LDL-associated esters or from albumin (Figure S4).

Kinetics of Lysosomal Transport of LDL and of Metabolism of Its CTL-Esters

Using quantitative live-cell fluorescence microscopy, we measured the kinetics of transport of LDL to LE/LYSs containing Rh-dextran. We found that the sterol fluorescence from LDL containing CTL-esters increases in endolysosomes in a slightly sigmoidal fashion over time (Figure A and B). Such sigmoidal kinetics are well described by a sequential transport model (Figure B, inset, and Experimental Section). The rate constants for the first and second phases (i.e., k ₁ and k ₂) were first determined by standard nonlinear regression of this sequential model to the lysosomal accumulation of CTL/CTL-esters. We found the two rate constants to be equal, and they corresponded to a half-time of 72.3 min for each step (see Table S3). To ensure that the measured CTL fluorescence in endolysosomes indeed comes from LDL-derived CTL-esters, we labeled the LDL particles additionally with an Alexa488 fluorophore on the apoB portion after reconstituting the CTL-esters and incubated cells with such double-labeled LDL particles for 24 h (see Experimental Section). Unfortunately, the second labeling step strongly reduced the observable fluorescence of the sterol in cells, which meant that this experiment is limited to the long incubation period of 24 h to enable sufficient cellular uptake. Still, we can show that the remaining fluorescence derived from CTL-esters coincides with the Alexa488 fluorescence in intracellular vesicles located in the perinuclear area (Figure C). These results support the idea that the majority of the lysosomal fluorescence of the CTL-esters at early time points is associated with LDL particles. Using Lipid-MS, we were able to measure the kinetics of intracellular accumulation of hydrolyzed CTL derived from LDL-associated CTL-oleate esters as well as of re-esterification of the liberated CTL (Figure A and B). Formation of free CTL, indicating hydrolysis of the CTL-ester, was slightly sigmoidal, resembling that measured for lysosomal transport of these LDL particles (Figure A, compared with Figure B).

2.

Kinetics of lysosomal accumulation of LDL-derived CTL (ester). Fibroblasts were incubated in LPDS medium supplemented with 0.5 mg/mL Rh-dextran overnight. On the following day, the cells were incubated in LPDS medium containing 0.1 mg/mL LDL with CTL-oleate ester for the indicated times. A, Representative deconvolved images are shown for each time point with CTL ester fluorescence pseudocolored in green and Rh-dextran in red. Co-localization appears to be orange. B, Quantification of CTL ester fluorescence overlapping with the Rh-dextran signal in the endolysosomes and its kinetic modeling using a sequential transport model (see Experimental Section; blue dots, data mean ± S.D., orange line, model fit) is shown. C, Cells were incubated in LPDS medium with 0.2 mg/mL Alexa488-LDL containing CTL-oleate ester for 24 h and imaged. Co-localization of LDL’s apoprotein labeled with Alexa488 (green) with the fluorescent sterol, CTL (ester), pseudocolored in red, is indicated with arrows. Scale bar: 10 μm.

3.

Lipidomic analysis of cells loaded with reconstituted LDL particles. Fibroblasts were incubated in LPDS medium containing 0.1 mg/mL LDL with CTL-oleate ester for the indicated times. A, Hydrolysis of CTL-oleate ester was detected as a sigmoidal increase in the amount of free CTL over time (mean ± SD of the data in blue). The data were fitted to a sequential kinetic model, either with both rate constants, k ₁ and k ₂, allowed to vary (orange line) or with k ₁ being fixed at k ₁ = 0.0096 min^–1 (green line). This is the value determined from the CTL (ester) transport data (compare Figure B and Table ). B, Re-esterification of CTL quantified for the indicated ester species (colored symbols at 24 h). Only for the formation of CTL-stearate (18:0), a time course could be measured and is shown as the mean ± SD (blue symbols), together with a fit to a single exponential function (orange line, see also Table ). C–H, Hierarchical Bayesian modeling of the time courses for fluorescence increase (compare Figure B) and of the hydrolysis of LDL-associated CTL-ester in LE/LYSs. The mean values of the data are shown as red dots, and the simulated time courses generated with parameters sampled from the posterior distribution are shown as blue lines (n = 100) for the fluorescence imaging data (C) and the Lipid-MS data (D), respectively. From the posterior distribution, the values for the amplitudes (E and F) and shared rate constants k ₁ (G) and k ₂ (H) are shown with the 95% highest density interval (HDI) indicated.

To analyze these data further, we employed the same sequential transport model as used for quantifying the lysosomal fluorescence of CTL/CTL-esters. When fitting this sequential kinetic model to the hydrolysis kinetics, we found that the rate constant for the first step could be set to the value determined for the first step of the lysosomal accumulation of LDL, while the second step of the hydrolysis kinetics was slower with a half-time of 198.0 min (Figure A data, blue symbols; green line is fit to this model). These values were obtained using a nonlinear regression routine, which only provides point estimates for the parameters and asymptotic confidence intervals. To better account for the variability and sparsity of the measured data, we used the kinetic two-step model in a Bayesian framework, which not only provides information about model uncertainty but also allows for analyzing multiple data sets with shared parameters (see below and Experimental Section). Since the data, we collect, are inherently noisy, we assume that the intensities, we measure, are part of a normal distribution with a mean equal to the average intensity of the determined kinetics and an unknown standard deviation, σ. Bayesian modeling of the sequential model provides not only the mean but also the entire posterior distribution of the estimated kinetic parameters. This posterior distribution is the conditional probability of the parameter values given the observed data. It has as many dimensions as there are estimated parameters and allows for simulating time courses by sampling from it by Markov Chain Monte Carlo (MCMC) simulations (Figures C–H and S5). Using the Bayesian approach for model parameter inference on the fluorescence kinetics data of LDL transport to LE/LYSs, identical mean values as those for the nonlinear regression were found, and samples drawn from the posterior distribution are coinciding well and in a narrow range with the experimental data (Figure S5A-E and Table S3). By applying the same model separately to the time course of hydrolysis of LDL-derived CTL-ester, we found a very similar distribution for the first rate constant but obtained a slower second kinetic rate constant with a broad distribution of values (Figure S5F–I and Table S3). The exact value of the second rate constant, k ₂, as inferred by Bayesian analysis is rather imprecise for the CTL-ester hydrolysis data, as the posterior is relatively wide with the 95% highest (posterior) density interval (95% HDI) varying from k ₂ = 0.2 to 0.62 s^–1 (Figure S5I). Accordingly, the simulated time courses sampled from the posterior distribution by MCMC simulations spread much more around the measurement data than those obtained by Bayesian analysis of the CTL-ester uptake kinetics (Figure S5B and F). The larger uncertainty for the second rate constant in the hydrolysis experiment is likely the result of the sparsity of the measurement data.

This limitation can be overcome using a hierarchical model in which the rate constants are shared between several data sets. We used this approach to determine the rate constants of the kinetic two-step model using the fluorescence kinetics of endolysosomal accumulation of LDL-derived CTL (Figure B) as well as the hydrolysis kinetics of LDL-derived CTL esters (Figure A, symbols) as input. With such a hierarchical Bayesian model, we could increase the confidence level in the parameter estimates and confine the time courses sampled from the posterior closer to the measured data (Figure C–H). This Bayesian analysis confirms the parameter values obtained by nonlinear regression and additionally reveals that the variance of these parameter estimates is relatively low, suggesting that the model describes the data well (Table and Figure C–H).

1. Kinetic Parameters of CTL-Ester Transport and Hydrolysis .

Experimental data	Start amount in pool A0	Rate constant and half-time for first step	Rate constant and half-time for second step
Fluorescence intensity of LDL-derived CTL/CTL18:1 in LE/LYSs	96.327 ± 0.102 a.u.	0.0097 ± 0.0011 min–1 (t 1/2 = 71.7 min)	0.0096 ± 0.0012 min–1 (t 1/2 = 72.6 min)
Concentration of LDL-derived CTL in LE/LYSs	0.317 ± 0.034 pmol/μg protein

^aKinetic parameters are shown for the sequential model as determined by hierarchical Bayesian analysis (see Experimental Section). Amplitudes were determined separately, while the rate constant was shared for each data set. Values are given as the mean ± standard deviation of the posterior distribution.

What are the cellular processes underlying these two kinetic steps of LDL processing? Once LDL is bound to its receptor on the cell surface, transport of the LDL-R complex to sorting endosomes (SEs), as well as dissociation of LDL from its receptor inside early endosomes, are fast events with half-times of 2–5 min. ,, The arrival of LDL in late endosomes (LEs) has previously been shown to follow sigmoidal kinetics as the consequence of slow maturation of SEs into LEs. Thus, we hypothesize that the first kinetic step in our model resembles the combined effect of LDL binding to its receptor, transport to SEs, and the onset of endosome maturation into LEs. Uptake of LDL particles might also include other pathways than that mediated by the LDL-R. ,− The second step in our model is likely a combination of continued maturation of SEs into LEs, fusion of LEs with lysosomes, and degradation of the LDL particles upon arrival in LE/LYSs. The latter is supported by our observation that CTL-esters are partially quenched in the LDL particles, as the addition of the detergent Triton X-100 to reconstituted particles resulted in an increase in the CTL signal (Figure S1C). The overall half-time of the sequential hierarchical Bayesian model with shared kinetic parameters is 3 h for the accumulation of CTL fluorescence in LE/LYSs and for the hydrolysis of the CTL-oleate (Figure S6). These results can be compared to values in the literature, which used LDL containing radiolabeled CEs and cell fractionation to show that hydrolysis of CEs takes place after an initial lag-phase. From such studies, the half-time of hydrolysis can be estimated by inspection of the respective graphs with estimated values ranging from about 1.5–2 h in CHO cells ,, and 2–6 h in primary human fibroblasts. ,,, The latter are also in agreement with the sequential transport kinetics described for LDL-R-dependent and independent transport of radiolabeled LDL tagged at the apoprotein in primary human fibroblasts. ,, Thus, a sequential transport model with the first step describing LDL uptake and the second step, CTL-ester hydrolysis/dequenching, is a good description for the trafficking of LDL-derived CTL esters in human fibroblasts. The model predicts that degradation of LDL’s CEs first takes place once the particles have reached LE/LYSs. Only in these organelles is the pH sufficiently acidic for acid lipases to be active in hydrolyzing LDL-derived sterol esters. This conclusion is supported by the experiments with Lalistat (see above).

Lipid-MS also allows us to quantify the re-esterification of CTL, but only as long as we measure the formation of esters other than CTL-oleate. That is, re-esterification to oleate cannot be distinguished from CTL-oleate taken up by cells as part of LDL, so re-esterification must be assessed by the formation of other CTL-esters in the cells. Indeed, we found that cells form CTL-stearate (i.e., CTL-18:0) with monoexponential kinetics and a half-time of 152.7 min (Figure B and Table S3). After prolonged incubation, CTL-esters with polyunsaturated fatty acids, such as CTL-20:4, could also be measured (Figure B, red symbol). These results show that hydrolysis of CTL-oleate and re-esterification of the liberated CTL take place on comparable time scales. We conclude that transport of the liberated CTL to the ER, where ACAT resides, is not rate-limiting and is likely faster than LDL uptake and hydrolysis of LDL-derived CEs.

LDL-Derived Cholesteryl Esters and Triacylglycerols Can Be Tracked by Lipidomics

Fibroblasts primarily accumulated sterols during the uptake of LDL, as expected given the lipid composition of the LDL particles (Figure S1 and S7, violet symbols). This allows us to follow the dynamics of specific CEs, which are enriched in LDL particles but not abundant in cells. In particular, we found that the species CE16:0, CE16:1, CE18:1, and CE18:2 are very abundant in LDL but not in cells, enabling us to trace these CEs specifically in cells (Figure S8). The species CE18:3 is isomeric to CTL18:1 and was, therefore, not included in this analysis. To analyze these data, a hierarchical monoexponential Bayesian model was used, in which the amplitudes for all four species were independently determined, while the rate constant was shared for all four time series. The inferred rate constant was k = 0.00358 ± 0.00003 min^–1, corresponding to a mean half-time for cellular accumulation of all studied LDL-specific CEs of t _1/2 = 193.44 min (Table S4). A monoexponential rise to maximal cellular accumulation with an exponential increase and comparable half-time was also found in previous work using radioactive CE tracers in LDL. The CEs, we determined here, are likely a mixture of LDL-derived internalized CEs and re-esterified cholesterol. Therefore, we cannot further distinguish individual pools in this experiment, in contrast to the studies with the reconstituted CTL18:1 ester (see above).

In addition to CEs, we found that LDL also contains specific TAGs, which are of very low abundance in the nonlabeled cells (Figure S9). These TAGs can also be used to track the uptake of LDL-derived core lipids. Using the hierarchical monoexponential Bayesian model with a shared rate constant, we found that a monoexponential rise to the maximal value also describes the cellular accumulation of LDL-specific TAGs very well, just as we observed for the LDL-associated CEs (compare Figure S8 and S9). The inferred rate constant was k = 0.00168 ± 0.00008 min^–1 lower than that of the accumulation of CEs and corresponds to a half-time of t _1/2 = 411.8 min (Table S5). The slower time course of TAG enrichment in cells compared to that of CEs could be the result of more extensive remodeling, with part of the TAG-derived fatty acids used in β-oxidation and phospholipid synthesis. Together, our results show that lipidomics combined with Bayesian kinetic modeling is very powerful in tracing different lipoprotein-derived lipid species in cells.

Remodeling of Glycerophospho- and Sphingolipids during Uptake and Processing of LDL

Apart from these changes, we observed alterations in the abundance of phospholipids and sphingolipids. Since these lipids are less specific in LDL compared to cells, we could not trace the accumulation of individual species unequivocally but rather assess an overall change in their abundance upon incubating cells with LDL particles. We found a slight decrease in glycerophospholipids during the initial 6 h of incubation, which was followed by an increase after 24 h (Figure S7, green symbols). De novo synthesis of glycerophospholipids to provide membrane lipids for cell division likely contributes to the long-term increase in the level of glycerophospholipids. This was also found for various species of PC (see Table S1 and Figure S10A). A cellular response to the uptake of LDL is a drop in lysoPC, probably due to reacylation (). While the overall abundance of SM increased only slightly, there was a strong formation of ceramides, particularly at later time points (Figure S11). LDL particles contained mainly SM 34:1,2, SM 40:1,2, and SM 42:2,2, and we therefore tracked those SM species over time in cells (Table S2 and Figure S11, dashed lines). The corresponding ceramide species increased over time (Figure S11, straight lines of the same color). This implies the activity of acid sphingomyelinase (e.g., SMPD3), degrading the LDL-associated SM inside LE/LYSs. , The cellular abundance of PE species followed that of PC, with species-dependent remodeling of lyso-PE (Figure S12A and B).

Cell-associated diacyl-PE ethers also showed an initial drop followed by a slight increase for most species, likely due to de novo synthesis in connection with cell division (Figure S12C). Together, these results show the power of lipidomics to uncover the cellular accumulation and metabolic conversion of lipoprotein-associated lipids. Together with the live-cell imaging of LDL-associated CTL-esters, these results demonstrate that Lipid-MS also provides detailed insight into the remodeling of the cellular lipidome upon lipoprotein uptake.

Endolysosomes Containing LDL Align with the Endoplasmic Reticulum for Sterol Transfer

Given that we observed re-esterification of LDL-derived CTL already after 6 h of incubation, we hypothesized that hydrolysis of the CTL-esters inside LE/LYSs is followed by the immediate export of the generated CTL to the ER. Since the mole fraction of cholesterol in the ER is low and the membrane area of this organelle is very large, direct visualization of CTL in the ER is challenging. Nevertheless, already after 6 h of loading cells with LDL containing CTL-ester, we found that LE/LYSs containing the fluorescent sterol and the organelle marker Rh-dextran overlap with ER-tubules labeled with NBD-ceramide (Figure A). NBD-ceramide is a well-known marker for the trans-Golgi network, but it also gives a faint staining in the ER since it is transported between both compartments. − We can confirm this by confocal imaging of fibroblasts double-labeled with NBD-ceramide and ER-tracker (Figure S13). Neither after 6 h nor after 24 h of incubation, CTL was transported to the TGN, which was easily identified by its bright staining with NBD-ceramide next to the nucleus (Figure A, B, yellow arrows). After incubating fibroblasts for 24 h with LDL containing CTL-esters, we found some sterol fluorescence in lipid droplets stained with Nile Red (Figure C). Formation of droplets in LDL-loaded human skin fibroblasts has also been found previously, and droplet-associated CTL fluorescence is in line with our results of the Lipid-MS, demonstrating that the hydrolyzed CTL becomes re-esterified and stored in droplets. After 6 h of incubation with LDL containing CTL-esters, most sterol fluorescence in the PM is homogeneous, suggesting that it stems from hydrolyzed CTL exported to the cell surface.

4.

Fluorescent sterols released from LDL particles colocalize with the ER and LDs. Fibroblasts were incubated in LPDS medium containing 0.1 mg/mL LDL with CTL-oleate ester for either 6 h or 24 h. LE/LYSs were stained with 0.5 mg/mL Rh-dextran overnight, and the ER and Golgi apparatus were labeled with NBD-ceramide added 10 min before imaging. After 6 h (A and A′), endolysosomes containing CTL (ester) and Rh-dextran were aligned with ER tubules which contained NBD-ceramide (white arrows). Faint CTL fluorescence in LE/LYSs overlapped occasionally with the ER (white star, only shown in Rh-dextran panel in A′). Sterol fluorescence in the PM is indicated with blue arrowheads. After 24 h of incubation (B and B′), faint fluorescence of CTL (ester) was additionally found colocalizing with the NBD-ceramide-positive ER structures (white arrows). C, In addition to the costaining with NBD-ceramide, the droplet marker Nile Red was added during the last 10 min of the 24 h incubation period. CTL (ester) fluorescence was found in LDs (inset in part C). Neither after 6 h of incubation nor after 24 h, CTL (ester) fluorescence found in the Golgi apparatus, localized next to the nucleus (yellow arrows in A and B).

We found in independent experiments using a spinning disk confocal microscope that endolysosomes containing the Cathepsin B substrate MagicRed form frequent contacts with the ER, in line with earlier reports (Figure A). We chose MagicRed for these experiments, as its strong red fluorescence is only found in hydrolytically active endolysosomes. CTL fluorescence derived from ingested LDL particles also colocalized with MagicRed in LE/LYSs (not shown). Using our recently developed automated image analysis protocols (see Experimental Section, Figure S14 and), the distance between those endolysosomes and the ER was quantified from 3D confocal stacks. For comparison, the organelle size was measured and determined as the average radius of LE/LYSs (i.e., the distance from the 3D centroid position to the endolysosome surface; see Figure S14 for details of this analysis).

5.

Internalized LDL mediates recruitment of endolysosomes to the ER. Fibroblasts were incubated in LPDS medium containing 0.5 mg/mL MagicRed (red) overnight, followed by incubation with 0.1 mg/mL LDL (right panel in A, “+ LDL”) or without LDL (left panel in A, “–LDL”) for 6 h. During the last 10 min of incubation, 5 μM NBD-ceramide was added to label the ER (green). 3D image stacks were acquired at a spinning disk confocal microscope, deconvolved, and further analyzed to determine the distance between MagicRed-stained LE/LYSs and NBD-ceramide-labeled ER tubules (see Experimental Section). B, C, Histogram of endolysosome size (B) and distance to the ER (C) for cells incubated without LDL (red) or with LDL (blue). The mean distance between the particle centroid and the organelle surface is a measure of endolysosome size, while the shortest distance between endolysosome centroid and the nearest ER tubule quantifies ER-LE/LYSs distance. For this analysis, at least 11 3D stacks of 1–2 cells with 350–620 endolysosomes in each stack were quantified for each condition.

This distribution peaks around 0.3 μm with a long tail of up to about 1.5 μm and a slight shift to larger sizes in LDL-treated cells (Figure B). The closest distance of the centroid of endolysosomes to ER tubules was about 0.7 μm in the absence of LDL but about 0.3 μm in the presence of LDL (Figure B and see Figure S14B and D for the corresponding cumulative frequency plots). Since the latter matches the organelle size, we can conclude that endolysosomes form contacts with the ER upon ingestion of LDL. Moreover, LDL tagged with a fluorescent dye on the apoprotein moiety (Alexa488-LDL) was often found in close proximity to the ER (Figure S15A). Single-particle tracking demonstrates that Alexa488-LDL-containing endolysosomes stayed in contact with ER tubules for several seconds per event, during which their mobility was reduced compared to freely moving vesicles (Figure S15B-J and Supplemental Video 1). From such videos, the distance between the surface of LE/LYSs containing Alexa488-LDL and the skeletonized ER tubules was measured. This distance was often found to be between 0.05 and 0.2 μm, indicating contact formation between the organelles.

Together, these results demonstrate that LE/LYSs become recruited to the ER upon ingestion of LDL, and they suggest that membrane contact sites between endolysosomes and the ER could facilitate the transfer of LDL-derived cholesterol to the ER for esterification. Our approach is not limited to fibroblasts, as the LDL-containing CTL-oleate ester was efficiently internalized in macrophages (Figure S16). CTL fluorescence accumulated in lipid droplets in an ACAT-dependent manner, as demonstrated by using the ACAT inhibitor avasimibe (Figure S16). This establishes our method as a unique tool to study the trafficking of lipoprotein-derived sterol esters in various cell types.

Trafficking of LDL Containing CTL-Ester in Fibroblasts Lacking Niemann-Pick Proteins

NPC disease is a devastating neurological disorder characterized by pronounced accumulation of cholesterol and sphingolipids in LE/LYSs due to the lack of either NPC1 or NPC2 protein. While the exact mechanistic role of NPC1 and NPC2 in trafficking of LDL-derived cholesterol is still debated, it is well established that the absence of either protein prevents esterification of the LDL-derived cholesterol pool in the ER. , As expected, we found that lysosomal accumulation of LDL/CTL-ester was much stronger in cells lacking functional NPC1 or NPC2 compared to control cells (Figure S17A). This phenotype was more pronounced in cells lacking NPC2, in line with earlier observations. We have previously shown that NPC2 protein purified from bovine milk binds CTL and removes excess CTL delivered to cells in complex with BSA from disease fibroblasts. We therefore incubated NPC2-deficient cells in the presence of LDL/CTL-ester with the addition of 100 nM NPC2 protein into the culture medium containing LPDS. We found that purified bovine NPC2 reduced the lysosomal CTL fluorescence (Figure S17B). At the same time, we observed a more dispersed distribution of LE/LYSs containing Rh-dextran and CTL (Figure S17B, lower panel), suggesting that their cholesterol-induced perinuclear clustering becomes reduced in NPC2-deficient cells treated with NPC2. The latter is supported by our previous work in the same cell type, where we found reallocation of endolysosomes toward the PM and increased directed transport of LE/LYSs upon incubating NPC2-deficient fibroblasts with NPC2 protein. , Using a fluorescence-labeled NPC2 protein (Alexa488-NPC2), we observed extensive colocalization between CTL fluorescence and NPC2 in LE/LYSs, showing that the protein acts from inside endolysosomes in rescuing lysosomal sterol accumulation (Figure S17C).

When loading NPC2-deficient cells with LDL/CTL-ester overnight in the presence of 100 nM NPC2, we observed not only the reallocation of sterol-rich LE/LYSs into the cell periphery but also a faint CTL signal colocalizing with the ER marker NBD-ceramide (Figure A and A′). This is more visible after deconvolution, revealing the similarity in sterol distribution with the ER morphology (Figure B and Supplemental Video 2). Also, we found that CTL-positive endolysosomes align with the ER after rescuing lysosomal sterol storage with NPC2 protein (Figure A, inset). ER staining of NBD-ceramide was also found right underneath the PM and often appeared in tubular bundles and sheaths, which matches the pattern observed for the CTL fluorescence (Figure A, D, and F). Endolysosomes containing Rh-dextran and sterol were found again in close proximity to these ER bundles and even attached to their tips (Figure A, inset). We did not observe colocalization of LDL-derived CTL with the bright NBD-ceramide signal in the Golgi apparatus (Figure A, and see Figure S18 for quantification). In contrast, the fluorescence of CTL was found to match the less intense ER staining pattern of NBD-ceramide, spreading throughout the cell (Figure A and enlarged boxes in Figure A′). Co-localization of LDL-derived CTL signal with NBD-ceramide in the ER as well as alignment of Rh-dextran-containing endolysosomes with the ER are also found in individual planes of 3D multicolor stacks (Figure S18 and Supplemental Video 3). These results strongly suggest that LDL-derived cholesterol is directly transferred from LE/LYSs to the ER upon contact formation between both organelles (Figure B). NPC2 inside LE/LYSs likely promotes this transfer by delivering cholesterol to transporters in the lysosomal membrane, such as NPC1 and LIMP2 and eventually other proteins.

6.

LDL-derived CTL reaches the ER in NPC2-deficient cells treated with NPC2 protein. Fibroblasts from a patient lacking NPC2 were incubated with 0.1 mg/mL LDL containing CTL-oleate-ester, 0.5 mg/mL Rh-dextran, and 100 nM NPC2 for 18 h. Cells were washed and labeled with 5 μM of NBD-ceramide. Images were deconvolved and color-coded with CTL in gray (A, A′ and A″), NBD-ceramide in green (B, B′, B″, and F), and Rh-dextran in red (C, C′ and C″). The color overlay is shown in D, D′ and D″. Insets are zoomed versions of the upper small box in A–F or the lower two boxes shown as separate panels in A′–D″, respectively. The arrow points to a contact between LE/LYSs and the ER. G, Summary of our results showing that lysosomal transport and hydrolysis of LDL-derived CEs follow a sequential transport scheme including uptake and transport of LDL to SEs maturing into LEs with a combined half-time of t _1/2 = 71.7 min and accumulation and hydrolysis in LE/LYSs by lysosomal acid lipase (LAL) with a combined half-time of t _1/2= 72.6 min. Our results also indicate that the subsequent export of cholesterol to the ER requires NPC2 (and NPC1) and is followed by re-esterification by ACAT, both steps likely being faster than the transport of LDL to endolysosomes. Some cholesterol might move from the ER to the plasma membrane (PM) mediated by other proteins, as indicated (dashed double arrow, see text for further explanation). Bar, 10 μm.

Conclusion

In this study, we present a novel approach to study the trafficking and metabolism of LDL-derived cholesterol in living cells. The method is based on intrinsically fluorescent sterol esters, which differ from natural CEs by having only two additional double bonds in the steroid ring system. This minimal modification ensures a close resemblance to the natural cholesteryl oleate. We also synthesized a CTL-ether, which cannot be hydrolyzed by lipases, as shown in vitro and in cells. Binding of both LDL containing CTL-ester and -ether could be outcompeted by excess unlabeled LDL to a similar extent as previously described for ¹²⁵I-labeled LDL in human fibroblasts. In line with these earlier studies, we conclude that several binding sites contribute to the uptake of LDL in human fibroblasts, with LDL-R representing the major one. Since our reconstituted LDL particles contain mostly sterol esters but fewer phospholipids compared to native LDL, minor differences in receptor affinity cannot be excluded. For example, phospholipid-depleted LDL has been shown to have higher affinity to proteoglycans compared to native LDL, despite an overall similar apoB structure. ,,, Using Lipid-MS, we demonstrate that LDL-derived CTL-esters become hydrolyzed in endolysosomes of human fibroblasts after an initial delay phase. This is followed by re-esterification of some of the CTL in the ER and its storage in lipid droplets. Transport of the liberated CTL to the PM and ER is likely a fast process, suggesting that trafficking of LDL to LE/LYSs and lysosomal hydrolysis of LDL-associated CEs are the rate-limiting steps in redistributing LDL-derived cholesterol in human fibroblasts (Figure G).

We have developed a hierarchical Bayesian framework, which allows us to infer rate constants for a sequential model, which accounts for both processes and includes uncertainty quantification using imaging and lipidomics data as model input. Since we observed re-esterification of CTL derived from hydrolyzed LDL-CTL esters with a half-time of just 2.5 h, we conclude that cellular accumulation of CEs is accompanied by extensive acyl chain remodeling. Re-esterification of CTL was paralleled by LDL-induced, apposition of endolysosomes with the ER, suggesting that transport of cholesterol derived from hydrolysis of CEs to the ACAT-site in the ER takes place via membrane contact sites forming between LE/LYSs and the ER, as suggested earlier. ,,, Tethering of endolysosomes to the ER is a tightly regulated process, including sorting nexins, such as SNX13 and SNX19, and the sterol transporters StARD3 and Aster B (GRAM1B). ,,, In a recent study, NPC1 has been shown to regulate contacts with the ER to mediate lysosomal cholesterol egress in concert with the ER-resident GRAM1B. Similarly, the oxysterol-binding protein ORP1L has been associated with tether formation, endolysosome positioning, and trafficking of cholesterol from LE/LYSs to the ER. ,, It is likely that one or several of these transporters get activated by ingested LDL, which could explain our findings on the recruitment of LE/LYSs to the ER upon LDL loading. On the other hand, SNX13 has been shown to be a negative regulator of NPC1-mediated cholesterol export, and its absence likely redirects cholesterol transport to the PM independent of NPC1. Other ER-resident proteins, such as the Aster/GRAMD1 family, but also small GTPases, such as rab8, and cytosolic proteins, like StARD4 and FABP7, have been implicated in the export of LDL-derived cholesterol to the ER and PM. ,,, Our current setup does not allow us to quantitatively distinguish between the CTL signal derived from CTL-esters as part of LDL particles bound to the cell surface and the fluorescence of CTL transported to the PM upon lysosomal hydrolysis of the esters. This is particularly relevant at early time points up to 6 h, when a large portion of LDL still resides at the PM before equilibrating with intracellular pools, while lysosomal hydrolysis of CTL-oleate and, as a result, also export of CTL from LE/LYSs has started. , Surface-bound LDL particles containing CTL-oleate esters give a patchy CTL signal in the PM, which resembles that of the LDL particles. After 24 h of incubation with LDL containing CTL-esters but not with LDL containing CTL-ethers, we found predominantly a homogeneous membrane signal of CTL in control cells, indicating that this originates from free CTL exported to the PM. Inhibiting acid lipase with Lalistat abolished this homogeneous membrane staining, indicating that LAL is the major enzyme responsible for the hydrolysis of LDL-derived CEs in fibroblasts, in agreement with other reports. , Re-esterification of hydrolyzed CTL, however, follows immediately the lysosomal degradation of the CTL-esters, such that it is possible that LDL-derived cholesterol destined for the PM traffics via the ER as a transit compartment. Since cholesterol has a much higher affinity for saturated phospholipid species found in the PM compared to the more unsaturated species of the ER, the PM could act as a sink for the lipoprotein-associated cholesterol, even when it passes through the ER network. Supporting that notion are our results obtained in cells lacking either NPC1 or NPC2 protein; in both cases, lysosomal CTL­(ester) fluorescence was elevated, while some sterol intensity was also found in the PM (). Treating NPC2-deficient fibroblasts with purified NPC2 resulted in increased CTL fluorescence in the ER and PM, demonstrating the importance of this protein in lysosomal sterol export. Together, our results strongly suggest that most LDL-derived cholesterol traffics to the ER before being exported to the PM in human fibroblasts. Future studies are warranted to shed light on pathways for trafficking of LDL-derived cholesterol to the cell surface in various cell types.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.analchem.5c05064.

• Experimental procedures (synthetic procedures and analytical data); lipid composition of reconstituted LDL particles; kinetic rate constants of alternative regression procedures; kinetic parameters of Bayesian analysis for CEs and TAGs; reconstitution of LDL with CTL esters and ethers and the associated fluorescence ± Triton X-100; trafficking of LDL-derived CTL-ethers; comparison of PM staining patterns; additional Bayesian model analysis of CTL-esters; simulation of the sequential model; time-course of LDL-associated lipid species; Bayesian analysis of LDL-associated CEs and TAGs; Lipid-MS analysis of PC-, sphingolipid-, and PE-species; colocalization of different ER markers; cumulative frequency plots for distances between ER and LE/LYSs; spinning disk confocal imaging of endolysosomes containing Alexa488-LDL versus ER; transport of LDL containing CTL-ester in fibroblasts from NPC patients; trafficking of LDL containing CTL-esters in murine macrophages (PDF)

• Time-lapse imaging of endolysosomes containing Alexa488-LDL and the ER. (AVI)

• Deconvolution of multicolor images of CTL (ester), ER, and lysosome marker (AVI)

• Three-dimensional multicolor imaging of CTL (ester), ER, and lysosome marker (AVI)

.

All authors contributed in a meaningful manner to the research presented in this manuscript. All authors have given approval to the final version of the manuscript.

This research was funded by the Danish Research Council for Independent Research (2034-00136B to D.W.), the Lundbeck Foundation (R366-2021-2026 to D.W.), the Carlsberg Foundation (CF23-–1086 to D.W.), the Ara Parseghian Medical Research Foundation (to D.W.). and the VILLUM Center for Bioanalytical Sciences (VKR023179 to CSE).

The authors declare no competing financial interest.