2.1.

D38-Cholesterol

Deuterated cholesterol was produced by a yeast strain (RH6829) engineered to produce cholesterol instead of ergosterol.³⁰ An overview of the cholesterol biosynthetic pathway and principles of biosynthetic isotopic labeling has been published.³¹ Yeast growth media for isotopic labeling were 0.7% yeast nitrogen base (US Biological), 0.5% yeast extract (BD), 1.25% glucose, and $30\mathrm{mg}/\mathrm{L}$ uracil and leucine in deuterium oxide (99.8%, ARMAR, Switzerland). Precultures (1 mL) in this medium were used to inoculate 2 L of media, which were grown to a stationary phase with shaking (3 to 4 days) at 30°C. The deuterated cholesterol was purified from harvested cells and analyzed by gas chromatography-mass spectrometry (GC-MS) as described.³⁰ The yield was approximately 10 mg purified deuterated cholesterol per liter of cell culture. Analysis of the GC-MS profile of the deuterated cholesterol showed a GC profile [Fig. 1(b)] identical to commercial cholesterol [Fig. 1(a)], except that the retention time of deuterated cholesterol was slightly earlier than cholesterol. Examination of the high range of intact ions under the peak [Fig. 1(c)] showed an average $m/z$ of 424 denoting an average substitution of 38 hydrogen atoms by deuterium. Analysis of the deuterated cholesterol by NMR confirmed its purity and showed that all positions were substituted between 70% and 90%, consistent with the average of 82% substitution calculated from the MS profile. D7-cholesterol was obtained from Sigma-Aldrich and used without further purification.

Fig. 1

GC-MS analysis of D38-cholesterol produced from yeast. Commercial cholesterol (a) and deuterated cholesterol (b) purified from yeast were analyzed by GC-MS and the GC profiles are shown. The higher masses, representing unfragmented deuterated cholesterol (peak in b), are shown in (c), with an average around 424.6, denoting an average replacement of 38 H by D.

2.2.

Reagents

Cell culture reagents were obtained from Life Technologies or Lonza. Methyl-$\beta$-cyclodextrin, oleic acid, fatty-acid-free bovine serum albumin (BSA), and cholesterol were purchased from Sigma. D38-cholesterol or cholesterol/methyl-$\beta$-cyclodextrin stock complexes were prepared as described previously,³² using a methyl-$\beta$-cyclodextrin to cholesterol (D38-cholesterol) ratio of 6.2:1 and a final concentration of 45 to 50 mM for D38-cholesterol. Lipoprotein-deprived serum (LPDS) was prepared from fetal bovine serum by potassium bromide density ultracentrifugation.³³ Oleic acid was complexed to fatty-acid-free BSA in an 8:1 ratio.³⁴

2.3.

Cell Culture and Lipid Administrations

Y1 cells were cultured in Dulbecco’s modified eagle’s medium (DMEM)/F12 (1:1) with 15% horse serum and 2.5% fetal bovine serum. Culture medium was supplemented with penicillin/streptomycin ($100\mathrm{U}{\mathrm{ml}}^{-1}$ each) and l-glutamine (2 mM). For lipid loadings, Y1 cells were seeded in culture medium for 24 h. Cells were washed three times with phosphate buffered saline (PBS) and incubated with DMEM/F12 medium containing 5% LPDS and $50\text{-}\mu \mathrm{M}$ cholesterol/cyclodextrin, 45- to $50\text{-}\mu \mathrm{M}$ D38-cholesterol/cyclodextrin, $400\text{-}\mu \mathrm{M}$ oleic acid/BSA, or $400\text{-}\mu \mathrm{M}$ oleic acid/BSA plus $50\text{-}\mu \mathrm{M}$ D38-cholesterol/cyclodextrin for 24 h. Control cells were treated with DMEM/F12 medium containing 5% LPDS for 24 h.

2.4.

Lipid Extraction and Quantifications

Lipid extraction and quantifications were performed as described previously.³⁵ Lipids were extracted using chloroform:methanol in a 1:1 ratio. Solvents were evaporated under nitrogen, and dried lipids were dissolved in chloroform:methanol (2:1 ratio). Lipid solutions were spotted on thin layer chromatography (TLC) plates, and cholesteryl ester, triacylglycerols, and free cholesterol were resolved by phase separation using hexane:diethylether:acetic acid (80:20:1 ratio). Cholesteryl ester, triacylglycerols, and free cholesterol were quantified using ImageJ and normalized to protein content, measured by Bio-Rad protein determination.

2.5.

Sample Preparation for Stimulated Raman Scattering Microscopy

Cells grown on No. 1.5 borosilicate coverslips were washed $2\times$ with PBS and fixed with 4% paraformaldehyde for 15 min. Coverslips were rinsed several times with PBS, mounted in PBS, and sealed with epoxy glue to prevent cells from drying.

2.6.

Hyperspectral Imaging with Stimulated Raman Scattering Microscopy

SRS signals were obtained by combining two laser beams: a Stokes beam fixed at 1064 nm ($\sim 9400{\mathrm{cm}}^{-1}$) and a pump beam tuned to the wavelength of interest (Table 1). The system consists of an optical parametric oscillator (OPO; Levante, Emerald OPO, Berlin, Germany) pumped by a 76-MHz mode-locked Nd:Vanadate laser (Picotrain, High-Q, Hohenems, Austria) that delivers a fundamental beam at 1064 nm (Stokes beam) with 7-ps pulses and a second harmonic generated beam at 532 nm. The latter pumps the OPO that generates the pump beam for the SRS process. The tuning of the pump beam was possible by adjusting the crystal temperature, the Lyot filter, and the cavity length of the OPO. The two beams were overlapped both temporally and spatially and sent into a laser scanner (Fluoview 300, Olympus, Center Valley, Pennsylvania), attached to an inverted microscope (IX71, Olympus). The combined beams were then focused through a $20\times$, 0.75 NA objective lens (UplanS Apo, Olympus) onto the sample. SRS images were obtained by detecting the stimulated Raman loss of the pump beam. The Stokes beam was modulated at 10 MHz with an acousto-optic modulator (AOM; Crystal Technology, Palo Alto, California), and the pump intensity modulation was detected by a photodiode (FDS1010; Thorlabs, Newton, New Jersey). The signal at 10 MHz was demodulated with a home-built lock-in amplifier. The average combined power of Stokes and pump beams at the specimen was kept under 50 mW throughout this study to minimize sample photodamage. Additional spontaneous Raman spectra of pure cholesterol were acquired with a commercial Raman microscope (InVia Confocal; Renishaw, Wotton-under-Edge, Gloucestershire, United Kingdom).

Table 1

Pump wavelengths corresponding to the Raman shifts of interest for SRS hyperspectral interrogation at the C─H and C─D stretching ranges.

2.7.

Spectral Analysis with Vertex Component Analysis

Hyperspectral coherent Raman scattering imaging has been previously used to acquire simultaneous chemical and spatial information of biological samples.^13,24,36 In combination with multivariate analysis, we can extract the spectral information from the hyperspectral stacks. In this case, we chose vertex component analysis (VCA) to retrieve the most prominent spectral features in the hyperspectral-SRS stack.^37,38 Briefly, the VCA algorithm identifies the main spectral components in the image that are identified as end members. We use three end members that define the vertices of a polygon (a triangle in this case), and each one is assigned a base color (red, green, and blue—RGB, in this case). Each pixel of the hyperspectral stack (each spectrum in the image) is then defined as a linear combination of the vertex spectra. The result can be visualized in an RGB color map in which the colors denote the spectral class of each pixel.

3.1.

Characterization of D38-Cholesterol

Deuterated lipids increase specificity for Raman, SRS, and CARS imaging due to their specific $\mathrm{C}─\mathrm{D}$ vibrational frequencies from 2000 to $2300{\mathrm{cm}}^{-1}$, situated in a silent region of the Raman spectrum and free of vibrational modes from endogenous compounds. The blue line in Fig. 2 shows the Raman spectrum of cholesterol. The strongest peaks in the spectrum are found at 2700 to $3100{\mathrm{cm}}^{-1}$ ($\mathrm{C}─\mathrm{H}$ stretching modes), with additional peaks at $1670{\mathrm{cm}}^{-1}$ ($\mathrm{C}═\mathrm{C}$ stretching modes) and $1440{\mathrm{cm}}^{-1}$ (${\mathrm{CH}}_2$ scissoring mode). The Raman spectrum of commercially available D7-cholesterol, which has seven hydrogen atoms replaced by deuterium, is shown in red in Fig. 2. The frequency of the CD modes is shifted relative to the CH stretching band, giving rise to a Raman band in an otherwise silent region of the Raman spectrum from 2000 to $2300{\mathrm{cm}}^{-1}$. However, because only 7 out of 46 possible positions are substituted with deuterium, the CD spectral features of D7-cholesterol are relatively weak, while the CH region remains the dominant contribution to the spectrum. The yellow line in Fig. 2 corresponds to the Raman spectrum of D38-cholesterol. By substituting 38 sites with deuterium (82.6% deuteration), the response in the 2000 to $2300{\mathrm{cm}}^{-1}$ range is significantly improved, whereas the magnitude of the CH stretching bands is considerably reduced. D38-cholesterol shows two main peaks in the CD band, centered at 2120 and at $2211{\mathrm{cm}}^{-1}$. This heavily deuterated cholesterol also exhibits a shift of the $\mathrm{C}═\mathrm{C}$ stretching band to $1656{\mathrm{cm}}^{-1}$, relative to natural cholesterol. A similar shift has previously been reported for the deuterated fatty acid D6-arachidonic acid.²⁶ Note that besides a significant reduction of the CH band, the spectral profile of the CH stretches of D38-cholesterol is markedly different from the spectral bandshape of natural cholesterol in this range. The contribution of the symmetric ${\mathrm{CH}}_2$ stretching mode at $2845{\mathrm{cm}}^{-1}$ in D38-cholesterol is relatively weak, whereas the strongest contribution in this range is found at $2925{\mathrm{cm}}^{-1}$. This characteristic bandshape provides an additional handle to identify D38-cholesterol and, as shown in Sec. 3.3, can be utilized to discriminate lipid droplets with predominantly esterified cholesterol from lipid droplets with a high concentration of unesterified cholesterol.

Fig. 2

Spectral properties of cholesterol, D38-cholesterol, and D7-cholesterol. Raman spectra of cholesterol (blue), D38-cholesterol (yellow), and D7-cholesterol (red) are shown in the range from 1400 to $3100{\mathrm{cm}}^{-1}$. Inset: the SRS intensity of D38-cholesterol in cyclohexane relative to the noise floor increases linearly with the concentration of solute. The detection limit was found to be $220\mu \mathrm{M}$.

The sensitivity of the SRS microscope to D38-cholesterol is depicted in the inset of Fig. 2. Here, the SRS signal is measured relative to the noise floor of the experiment, as determined by the photothermal signal of glass. At $2120{\mathrm{cm}}^{-1}$, we find a detection limit of $220\text{-}\mu \mathrm{M}$ D38-cholesterol in cyclohexane at a pixel dwell time of $10\mu \mathrm{s}$, when the SRS signal of D38-cholesterol approaches the noise floor. As expected, there is a linear dependence of the SRS signal on D38-cholesterol concentration.³⁹

3.2.

D38-Cholesterol: Uptake, Metabolic Processing, and Storage in Lipid Droplets

To examine the biocompatibility of D38-cholesterol in mammalian cells, we treated Y1 adrenal cells with $50\text{-}\mu \mathrm{M}$ D38-cholesterol/cyclodextrin for 24 h. Cyclodextrin forms a complex with cholesterol and transfers it to cells via the plasma membrane.⁴⁰ This treatment results in efficient cellular uptake of D38-cholesterol with a 2.9-fold increase in total cholesterol and 3.5-fold for cholesteryl esters [Fig. 3(a)] compared to the LPDS control. Using TLC, we determined that approximately 58% of the sequestered D38-cholesterol was esterified [Fig. 3(a)], confirming enzymatic processing of D38-cholesterol and highlighting its superiority to fluorescent sterol analogs in regard to metabolic processing.^5,20 Importantly, we find that the levels of esterified D38-cholesterol are similar to the levels of esterified natural cholesterol under similar treatment conditions (Fig. 6), indicating that the deuterated probe does not affect acylCoA:cholesterol acyltransferase (ACAT)-mediated esterification. Next, we performed SRS imaging of D38-cholesterol-loaded cells. Figures 3(b)–3(d) show a cell treated with D38-cholesterol imaged at three different frequencies: $2325{\mathrm{cm}}^{-1}$ (off resonance), $2120{\mathrm{cm}}^{-1}$ (${\mathrm{CD}}_2$ stretches), and $2841{\mathrm{cm}}^{-1}$ (${\mathrm{CH}}_2$ symmetric stretches), respectively. The strongest signal at $2120{\mathrm{cm}}^{-1}$ derives from droplet-like structures, as does the more intense signal at $2841{\mathrm{cm}}^{-1}$, indicative of intracellular lipid droplets. This observation provides evidence that D38-cholesterol has been processed by the cell and is stored in lipid droplets.

Fig. 3

Cellular imaging of D38-cholesterol uptake, esterification, and storage inside lipid droplets. Y1 adrenal cells were treated with $50\text{-}\mu \mathrm{M}$ D38-cholesterol/cyclodextrin for 24 h, and cells were used for lipid quantifications (a) or fixed and subjected to SRS imaging (b) $2325{\mathrm{cm}}^{-1}$, off resonance; (c) $2120{\mathrm{cm}}^{-1}$, ${\mathrm{CD}}_2$ stretches; and (d) $2841{\mathrm{cm}}^{-1}$, ${\mathrm{CH}}_2$ symmetric stretches). The result of hyperspectral SRS imaging and multivariate analysis is depicted in yellow for D38-cholesterol in the CD region (e and g), overlaid on the maximum intensity projection of the CH spectral scan in gray scale, and in red for lipid and green for protein in the CH region (f and h). Normalized spectra. $\mathrm{SB}=10\mu \mathrm{m}$.

To further prove the presence of D38-cholesterol in intracellular lipid droplets, we show the result of a hyperspectral SRS multivariate analysis (VCA) in both the CD and CH spectral regions. The CD range spanned from 1965 to $2297{\mathrm{cm}}^{-1}$, with a step size of $8.5{\mathrm{cm}}^{-1}$, whereas the CH region ranges from 2797 to $3055{\mathrm{cm}}^{-1}$, with a step size of $6.6{\mathrm{cm}}^{-1}$. The yellow areas in Fig. 3(e), superimposed on the maximum intensity projection of the spectral scan in the CH stretching region (in gray), exhibit the spectrum depicted in Fig. 3(g). These locations indicate the presence of D38-cholesterol and spatially correlate with the lipid droplet areas as seen in the CH images, confirming that D38-cholesterol is stored in lipid droplets. A corresponding analysis in the CH range shows that the same droplets also contain nondeuterated lipids, whose spectral profile is depicted in red [Figs. 3(f) and 3(h)]. These lipids largely represent fatty acids esterified to D38-cholesterol, some pre-existing nondeuterated cholesterol, and a small fraction of triglycerides (5.1% of total neutral lipids), as confirmed by TLC. The green areas in Figs. 3(f) and 3(h) correspond to the protein-rich content of the cell, which is shown here for context. Hence, these combined results show that D38-cholesterol is processed by the cells, is esterified to a similar degree as natural cholesterol, and is stored in lipid droplets.

3.3.

Hyperspectral Stimulated Raman Scattering Imaging of D38-Cholesterol Identifies Subpopulations of Esterified and Unesterified Cholesterols

Besides identifying cholesterol in lipid droplets, the unique Raman spectrum of D38-cholesterol can be utilized to reveal additional information about the degree of cholesterol esterification. To access this information, we used hyperspectral-SRS imaging combined with VCA. Figure 4 shows a cell that contains two subsets of lipid droplets with different spectral profiles. The CD range reveals droplets that contain D38-cholesterol [Figs. 4(a) and 4(c)]. A subset of these lipid droplets, depicted in blue, exhibit a CH spectrum that resembles the Raman spectrum of unesterified D38-cholesterol. In red, we show the subset of droplets that are represented by a spectrum indicative of both esterified D38-cholesterol and nondeuterated lipids [Figs. 4(b) and 4(d)]. These droplets most likely consist of fatty acids esterified to cholesterol and a small amount of triglycerides and/or pre-existing nondeuterated cholesterol.

Fig. 4

D38-cholesterol allows visualization of a subset of lipid droplets enriched with free cholesterol. Y1 adrenal cells were treated with $50\text{-}\mu \mathrm{M}$ D38-cholesterol/cyclodextrin as in Fig. 3. The result of hyperspectral SRS imaging and multivariate analysis is depicted here in yellow for D38-cholesterol in the CD region (a and c), and in red for esterified cholesterol and blue for free D38-cholesterol in the CH region (b and d). Normalized spectra. $\mathrm{SB}=10\mu \mathrm{m}$.

The CD range of the spectrum contains limited information about the state of cholesterol esterification. However, the spectral information acquired in the CH range helps to discriminate between esterified and free cholesterol accumulations. By combining CD imaging with hyperspectral analysis of the CH region, we thus demonstrate that lipid droplet heterogeneity may not only occur at the level of cholesterol ester and triglyceride partitioning, but also at the level of cholesterol esterification. Some lipid droplets show higher enrichment in free cholesterol than others.

3.4.

Distinct Cholesteryl Ester and Triacylglycerol Containing Lipid Droplets Visualized by D38-Cholesterol and Oleic Acid Administration

Studies based on fluorescent tracers have suggested that steroidogenic cells can store triglyceride and cholesterol esters in distinct lipid droplets resulting in lipid droplet heterogeneity.³² However, this might result from the altered properties of BODIPY-labeled lipid analogs as compared to natural lipids.²⁴ Here, we demonstrate lipid droplet heterogeneity without fluorescent lipid analogs by means of SRS microscopy and D38-cholesterol.

Y1 cells were treated with $50\text{-}\mu \mathrm{M}$ D38-cholesterol and $200\text{-}\mu \mathrm{M}$ oleic acid for 24 h. TLC analysis shows that this treatment results in a 5.8-fold increase in the accumulation of cholesteryl esters ($131\mathrm{ng}/\mu \mathrm{g}$ protein) and a 38-fold increase in the level of triglycerides ($107\mathrm{ng}/\mu \mathrm{g}$ protein) [Fig. 5(a)] with respect to the LPDS-treated cells. Cells subjected to simultaneous D38-cholesterol and oleic acid treatment are depicted in the SRS images shown in Figs. 5(b)–5(f). Lipid droplets containing D38-cholesterol are visualized at the $2120{\mathrm{cm}}^{-1}$ Raman shift in Fig. 5(c). The overall population of lipid droplets is visualized by tuning to the Raman shift of $2841{\mathrm{cm}}^{-1}$, shown in Fig. 5(d), which marks lipid droplets irrespective of the presence of D38-cholesterol. Comparing the 2120 and $2841{\mathrm{cm}}^{-1}$ images makes it clear that the concentration of D38-cholesterol in the available droplets varies significantly. This is made clearer in the hyperspectral image in Fig. 5(e), which highlights the D38-cholesterol containing lipid droplets in yellow [see the corresponding spectra in Fig. 5(g)]. Some lipid droplets, irrespective of size, contain D38-cholesterol at appreciable levels [yellow droplets in Fig. 5(e)], whereas others appear devoid of D38-cholesterol [gray droplets in Fig. 5(e)]. The hyperspectral image obtained in the CH stretching range, shown in Fig. 5(f), reveals two subsets of droplets with different spectral profiles [Fig. 5(h)]. Most of the droplets depicted in red also contain D38-cholesterol, while the droplets that appear in green in the CH range exhibit little to no CD signal. A third subset contains a mixture of both green and red spectra. The hyperspectral information in both Figs. 5(e) (in yellow) and 5(f) (in red and green) is overlaid on the maximum intensity projection of the hyperspectral scan in the CH stretching region (in gray) to demarcate the cells. These results show that intracellular lipid droplets are heterogeneous in terms of chemical composition and that cholesterol partitions inhomogenously among the available lipid reservoirs.

Fig. 5

Visualization of lipid droplet heterogeneity in steroidogenic cells using D38-cholesterol. Y1 adrenal cells were treated with $50\text{-}\mu \mathrm{M}$ D38-cholesterol/cyclodextrin together with $200\text{-}\mu \mathrm{M}$ oleic acid for 24 h. Cells were either used for lipid quantification (a) or subjected to SRS imaging [(b) $2325{\mathrm{cm}}^{-1}$, off resonance; (c) $2120{\mathrm{cm}}^{-1}$, CD2 stretches; and (d) $2841{\mathrm{cm}}^{-1}$, ${\mathrm{CH}}_2$ symmetric stretches]. (Note that LPDS control treatment is the same as in Fig. 3.) The result of hyperspectral SRS imaging and multivariate analysis is depicted in yellow for D38-cholesterol in the CD region (e and g), and in red and green for the lipids in the CH region (f and h). The results are overlaid on the maximum intensity projection of the CH spectral range, shown in gray scale, which outlines the cellular morphology. Normalized spectra. $\mathrm{SB}=20\mu \mathrm{m}$.