Three-dimensional imaging of cholesterol and sphingolipids within a Madin-Darby canine kidney cell Free

Lipids and cholesterol comprise cellular membranes and also play important roles in signal transduction^1,2 and intracellular trafficking.^3,4 Alterations in the distributions of various lipid species within subcellular compartments are linked to a variety of pathologies, the most well-known being Niemann-Pick disease.⁵ Direct imaging of the lipid and cholesterol compositions of various subcellular compartments without the use of fluorophores that may alter organization and intracellular trafficking⁶ would provide a valuable alternative perspective on native cell structure, and ultimately, function. Furthermore, three-dimensional imaging would enable organelles and intracellular processes to be resolved.

We previously used high resolution secondary ion mass spectrometry (SIMS), which we performed with a Cameca NanoSIMS 50, to image the distributions of cholesterol and sphingolipids on the surfaces of fibroblast cells.^7–9 For analyses performed on this instrument, a focused cesium or oxygen primary ion beam is scanned over the sample. Monoatomic and diatomic secondary ions with up to five different mass-to-charge ratios (m/z) are collected in parallel at every pixel, producing images of the elemental or isotopic composition at the surface of the sample with a lateral resolution as good as 50 nm.^10,11 In our approach, distinct stable isotopes that encode for component identity are selectively incorporated into the lipid species of interest with established metabolic labeling techniques.¹² Then, the component-specific isotope enrichment in the membrane cell is imaged with approximately 90 nm lateral resolution with a Cameca NanoSIMS 50 instrument. Use of this approach to visualize metabolically incorporated ¹⁸O-cholesterol and ¹⁵N-sphingolipids on the surfaces of fibroblast cells revealed the plasma membrane contained a fairly uniform cholesterol distribution and micrometer-scale ¹⁵N-sphingolipid patches.⁷ By imaging the effects of drugs on the sphingolipid distribution, we determined the sphingolipids were confined within domains by the cytoskeleton and its associated proteins.^7,8

Having identified the cholesterol and sphingolipid distribution within the plasma membrane, we have now turned our focus to investigating their intracellular distributions, including their abundances within intracellular membranes. The compositions of intracellular membranes have been previously assessed by lysing populations of cells, separating the organelles with centrifugation, and then analyzing the lipid composition in each fraction by biochemical assays or liquid chromatography-mass spectrometry (LC-MS).^13,14 A pitfall of this approach is that the measured cholesterol levels depend on the method used to fractionate the cells, likely due to contamination with membranes from other organelles and spontaneous cholesterol transfer between membranes after cell disruption.¹⁵ SIMS is an attractive alternative approach for probing the distributions of lipids and metabolites within intact cells. A thin layer of material is sputtered from the sample each time the primary ion beam interrogates its surface, allowing a series of images to be acquired at progressively increasing depth below the cell surface by successively reimaging the same location multiple times. Sputtering scans that remove more material from the sample's surface are often inserted between the imaging scans to reduce the analysis time required to depth profile through the cells.^16,17 Both NanoSIMS instruments and new time-of-flight SIMS (ToF-SIMS) instruments that do not need distinct labels for component identification have been used to image the intracellular distributions of nucleotides, lipids, amino acids, nanoparticles, and elements of interest.^18–24 However, the intracellular distributions of cholesterol and sphingolipids have not been imaged in parallel and with submicrometer lateral resolution.

As a first step toward assessing the abundances of sphingolipids and cholesterol within the membranes of distinct organelles, here we use high-resolution SIMS, performed with a Cameca NanoSIMS 50, to directly image the cholesterol and sphingolipid distribution within a portion of a Madin-Darby canine kidney (MDCK) cell.

MDCK cells were obtained from ATCC (CCL-34) and adapted over multiple passages to grow in UltraMDCK serum-free medium (Lonza) supplemented with 104 units/ml penicillin G, and 10 mg/ml streptomycin. Fatty acid-free bovine serum albumin (BSA), ¹³C₁₈-stearic acid (99 at. % carbon-13), and other cell culture materials were obtained from Sigma. Poly-l-lysine and chemical preservation reagents were obtained from Electron Microscopy Sciences. Silicon wafer chips were obtained from Ted Pella, Inc. Uniformly carbon-13 labeled (98 at. %) fatty acids were obtained from Cambridge Isotope Laboratories. Reported methods were used to synthesize the ¹⁵N-sphingolipid precursors, ¹⁵N-sphingosine, and ¹⁵N-sphinganine from ¹⁵N-serine (Cambridge Isotope Laboratories).^25–27 18O-cholesterol was synthesized from i-cholesteryl methyl ether (Sigma) and ¹⁸O-water (Olinax, Inc.) as reported.²⁸ 

MDCK cells were cultured in UltraMDCK serum-free medium supplemented with 3 μM ¹⁵N-sphingolipid precursors at 37 °C in 5% CO₂. Each subsequent day, cell medium was aspirated and fresh applied with added ¹⁵N-sphingolipid precursors. On day 3, cells were passaged in serum-free growth medium supplemented with 3 μM ¹⁵N-sphingolipid precursors, 215 μM ¹³C-fatty acids (4:1:28 mass ratio of UL-¹³C fatty acids/¹³C-stearic acid/fatty acid free bovine BSA), and 50 μM ¹⁸O-cholesterol (2:5 mass ratio of ¹⁸O-cholesterol:fatty acid-free BSA). Each subsequent day, the medium was aspirated and replaced with fresh medium containing ¹⁵N-sphingolipid precursors, ¹³C-fatty acids, and ¹⁸O-cholestrol at the aforementioned concentration. On day 6, cells were passaged into dishes containing poly-l-lysine–coated 5 × 5-mm silicon wafer substrates. The following day, substrates were preserved with glutaraldehyde and osmium tetroxide as described.^7,12 Incorporations of nitrogen-15 and oxygen-18 into the sphingolipids and cholesterol, respectively, were assessed using cells attached to the culture dish as previously reported.^7,12

Prior to SIMS analysis, samples were imaged on a JEOL 7000F analytical scanning electron microscope at 1 keV and a working distance of 10 mm. The samples were then coated with 3 nm of iridium (99.95%) with a Cressington 208HR high-resolution sputter coater equipped with a MTM-20 thickness controller. SIMS was performed on a Cameca NanoSIMS 50 at Lawrence Livermore National Laboratory. The ¹⁶O⁻, ¹⁸O⁻, ¹²C¹⁴N⁻, ¹²C¹⁵N⁻, and ³²S⁻ secondary ions, and secondary electrons were collected simultaneously. Cell membrane imaging was performed with a 0.17-pA, 16-keV ¹³³Cs⁺ primary ion beam, where 12 replicate scans of 512 × 512 pixels with a dwell time of 0.5 ms/pixel were acquired for a 15 × 15-μm analysis region. Depth profiling was performed with a 1.137-pA, 16-keV ¹³³Cs⁺ primary ion beam. Approximately 1620 replicate scans of 512 × 512 pixels with dwell time of 0.5 ms/pixel were acquired of a 10 × 10-μm region.

limage software (L. R. Nittler, Carnegie Institution of Washington) run on the PV-Wave platform (Visual Numerics, Inc.) was used to generate isotope enrichment images. The isotope enrichment is the ¹²C¹⁵N⁻/¹²C¹⁴N⁻ or ¹⁸O⁻/¹⁶O⁻ ratio divided by their terrestrial standard abundance ratio (0.00367 and 0.002005, respectively). Enrichment images showing the ¹⁵N-sphingolipid and ¹⁸O-cholesterol distributions and ¹²C¹⁴N⁻ ion images were constructed using a 5 × 5 pixel moving average smoothing algorithm.

Compared to the depth profiling experiments, cell membrane imaging was performed with a lower dose of primary ions, so fewer secondary ions were detected at each pixel. Because ratioing to small numbers may amplify random signal fluctuations, the pixels where the counts of the naturally abundant ion were below a threshold value were masked so that they appear black in the isotope enrichment images of the plasma membrane. For the ¹⁵N-enrichment images, the masking threshold was set to four counts of ¹²C¹⁴N⁻ ions/pixel, which is 5% of the maximum ¹²C¹⁴N⁻ counts/pixel detected on the surface of the cell. For the ¹⁸O-enrichment images, a masking threshold of 5% of the maximum ¹⁶O⁻ counts/pixel detected on the cell surface, which is one ¹⁶O⁻ count/pixel, was insufficient to remove the pixel-to-pixel variation in ¹⁸O-enrichment that is characteristic of random variations in signal intensity. Therefore, a masking threshold of two ¹⁶O⁻ counts/pixel, which is 10% of the maximum ¹⁶O⁻ counts/pixel detected on the cell, was applied to the ¹⁸O-enrichment images of the plasma membrane. Masking thresholds were not applied to the depth profiling images because significantly more secondary ions were detected at each pixel.

Three-dimensional (3D) representations of the ¹⁵N-sphingolipids and ¹⁸O-cholesterol within the cell were constructed by stacking subsequent images using the built-in 3D viewer imagej plugin run in imagej.²⁹ Each image in the stack was manually aligned in the x- and y-directions to correct for sample and stage drift during the depth profiling experiment. A transparency of 50% was applied to the stack for better visualization.

The approximate depth where each image was acquired was calculated from our raster area, primary ion beam current, and the average sputtering rate measured on other biological samples, as described.¹⁹ Sputtering rate was assumed to be constant at all positions in the sample. Differences between the rate of sputtering on these MDCK cells and the other biological samples, and potential changes in sputter rate caused by differences in chemical composition within the cell would reduce the accuracy of the estimated analysis depths and 3D renderings presented herein.

The distinct stable isotopes, nitrogen-15 and oxygen-18, were metabolically incorporated into MDCK cells as previously described.^7,12 The cells were imaged with SEM under low-kilo-volt to identify well-preserved specimens for analysis with high-resolution SIMS. Figure 1 shows the whole, fixed MDCK cell, where the yellow outline indicates the approximate region that was analyzed with SIMS. No cracks or other blemishes are visible on the cell surface, indicating this cell was well-preserved.

The cholesterol and sphingolipid distribution in the plasma membrane of the same MDCK cell was iridium coated and imaged with high-resolution SIMS using operating conditions that produced a sputtering depth of ∼3 nm, estimated using the average sputtering rate determined for other biological samples.¹⁹ The secondary electron (SE) image that was collected in parallel with the secondary ions [Fig. 2(a)] shows the morphology of the cell. The ¹⁸O-enrichment SIMS image [Fig. 2(b)] shows the ¹⁸O-cholesterol is relatively uniform within the plasma membrane, except for the dark region at the lower left corner of the image. This region appears dark because the ¹⁶O⁻ ion counts were insufficient (≤2 counts/pixel) to calculate the ¹⁸O-enrichment at this tall region on cell (see Sec. II for details). Large portions of this region also appear dark in the ¹⁵N-enrichment image [Fig. 2(c)], which suggests that sample topography interfered with secondary ion collection at this site. Nonetheless, local elevations in ¹⁵N-enrichment that represent ¹⁵N-sphingolipid domains are visible on the cell [Fig. 2(c)]. The ¹⁵N-sphingolipid-enriched features shown in Fig. 2(c) do not correlate with projections or other distinct structures on the cell [Fig. 2(b)], indicating they are not artifacts of cell topography. The finding of sphingolipid-enriched domains and a relatively uniform cholesterol distribution is consistent with the ¹⁸O-cholesterol and ¹⁵N-sphingolipid distributions we previously observed in the plasma membranes of mouse fibroblast cells.^7–9 

The cholesterol and sphingolipid distributions within the MDCK cell were imaged after changing our analysis conditions to more rapidly sputter through the cell. Figure 3 shows the secondary electron, ¹⁸O-enrichment, ¹⁵N-enrichment, and ¹²C¹⁴N⁻ secondary ion images that were acquired during the depth profiling analysis, revealing changes in cell morphology and composition with increasing depth from the cell's surface. Note that the color bars that encode for the isotope enrichment at each pixel in Fig. 3 have lower maxima than those shown in Fig. 2. Each of the images shown in Fig. 3 was constructed using data from ten raster planes, which is estimated to correspond to a thickness of approximately 37 nm. This estimated thickness was calculated from the analysis conditions and the average sputtering rate measured on other biological samples.¹⁹ Although this cell's sputtering rate may differ from that measured on other biological samples or vary with intracellular location, this simple calculation yields a useful approximation of the position of the relatively large intracellular structures that were observed relative to the apical plasma membrane.

Figures 3(a)–3(d) display the secondary electron, ¹⁸O-enrichment, and ¹⁵N-enrichment, and ¹²C¹⁴N⁻ secondary ion images that show sample morphology, ¹⁸O-cholesterol distribution, ¹⁵N-sphingolipid distribution, and ¹²C¹⁴N⁻ counts, respectively, at a calculated depth of approximately 780 nm below the apical plasma membrane. Within this region, the ¹⁸O-distribution is fairly uniform, except for the ¹⁸O-enriched region at the edge of the cell. Comparison to the secondary electron image [Fig. 3(a)] indicates the cell was very thin at this ¹⁸O-enriched region. Therefore, this region is likely ¹⁸O-enriched because ¹⁸O-cholesterol in the plasma membrane at the bottom of the cell was detected. In contrast, ∼0.5–2 μm regions enriched with ¹⁵N-sphingolipids are visible on the thicker regions of the cell [Fig. 3(c)]. These sphingolipid-rich regions do not correlate to the ¹⁵N-sphingolipid domains within the plasma membrane [Fig. 2(c)]. These features may be membrane-bound structures that contain sphingolipids in their lumens, or the sphingolipids may be confined to their membranes but our lateral resolution was insufficient to resolve their lumens.

At an estimated sputtering depth of approximately 4 μm, much of the thin layer of material at the edge of the cell has been eroded away [Fig. 3(e)]. Features enriched with ¹⁸O-cholesterol become visible at this depth [Fig. 3(f)], whereas the ¹⁵N-sphingolipid-rich features are less abundant [Fig. 3(g)]. The average ¹⁸O-enrichment in these features is nearly four times lower than that of the overlaying plasma membrane. This suggests that cholesterol is present at higher abundance in the plasma membrane than in intracellular structures, which is consistent with literature reports.³⁰ The ¹⁸O-enriched features change with increasing depth. At an estimated depth of approximately 6 μm, a significant amount of erosion is visible [Fig. 3(i)]. Figure 3(j) shows that more ¹⁸O-cholesterol-rich features are visible at approximately 6 μm from the cell's upper surface. Very few regions enriched in ¹⁵N-sphingolipids are visible at this depth [Fig. 3(k)], and those that are present are not colocalized with the ¹⁸O-rich regions. As shown in Figs. 3(g) and 3(k), ¹⁵N-sphingolipids appear to be deficient in many regions that are enriched with ¹⁸O-cholesterol. Interestingly, the ¹²C¹⁴N⁻ secondary ion counts are also lower at the sites enriched with ¹⁸O-cholesterol [Figs. 3(h) and 3(l)], which suggests that they contain low levels of nitrogen.

3D visualization of a region that was estimated to span from approximately 3.7–6 μm below the cell surface was performed by stacking individual images acquired at the same sample location. Each layer was manually aligned with the previous image to correct for drift in the location of the analysis region, which produces an image stack that is uneven along its z-axis. Distinct cholesterol-rich features are clearly apparent [Figs. 4(a) and 4(b)]. Some of these features span the entire depth of this analysis, which was calculated to be approximately 2.3 μm, and no decrease in their ¹⁸O-enrichment is visible. In contrast, the ¹⁵N-enriched areas that denote the presence of ¹⁵N-sphingolipids are smaller than the ¹⁸O-enriched regions [Figs. 4(c) and 4(d)]. Little colocalization was observed between the ¹⁸O-rich regions and the ¹⁵N-rich sites. The low counts of ¹²C¹⁴N⁻ detected at the ¹⁸O-enriched features [Figs. 3(h), 3(l)] may suggest that they are lipid droplets.³¹ Lipid droplets are cellular organelles that contain abundant triglycerides and sterol esters,^32,33 which do not contain nitrogen. Lipid droplets range in size from 0.1 to 5 μm,³⁴ which is consistent with the sizes of the ¹⁸O-enriched features we observed. However, further studies are required to confirm the identity of these ¹⁸O-enriched intracellular structures.

The distribution of cholesterol between the plasma membrane and intracellular compartments is tightly controlled.^30,35 The abundance of cholesterol in intracellular organelles has been linked to the membrane's biophysical properties,³⁶ and a variety of cellular processes, including signal transduction^1,2 and membrane trafficking.^3,4 Previous studies have employed fluorescent cholesterol analogs,³⁷ fractionation procedures to separate cellular components of interest coupled with appropriate assays,³⁸ and immunolabeling of cryosections of whole cells³⁹ to analyze the cholesterol distribution within cells.

3D visualization of the cholesterol content of individual organelles within a single cell could improve our current understanding of cholesterol distribution by providing a fluorophore-free in situ method. Others have tracked ¹³C-fatty acid storage in lipid droplets within cell cultures and mouse tissue using high-resolution SIMS performed with a Cameca NanoSIMS.^31,40,41 A similar study in which ¹³C-fatty acid incorporation was simultaneously imaged with ¹⁸O-cholesterol would be required to more definitively assess whether the ¹⁸O-enriched features we observed are lipid droplets. Alternatively, the presence of triacylglyceride and cholesterol esters might be assessed without the need for labels with new ToF-SIMS instruments.

While the relative distribution of cholesterol and sphingolipids within the cell volume can be clearly understood in this study, our estimates of sampling depth may be inaccurate if the sputter rate of this cell differs from that measured on other biological samples, or changes at different locations in the cell. This potential inaccuracy does not undermine the central conclusions of this study, but it would be a problem for a study in which the size of the structures or precise location is a part of the interpretation of the results. Measurements of sample height before and after SIMS analysis would enable more accurate determination of the depths of the observed intracellular structures. Chemical data could be used to account for changes in sputter rate in different structures. Topographic corrections could also help the viewer more readily understand more complex structures. Such corrections warp the image data to follow the surface topography, which can be visualized in the secondary electron images in this study (Fig. 3). Some groups have used AFM to correct the z-axis for sample topography and construct more accurate 3D representations from the 2D data.^42–44 At least one z-corrector toolbox has been developed, though it only works with IONTOF files.²¹ Recently, a Cameca NanoSIMS 50 was combined with high-resolution scanning probe microscopy to allow in situ topographical imaging of the sample surface, which can then be automatically assembled into a 3D representation by an available software program.⁴⁵ This in situ approach has the advantage of being able to correct for any differences in sputter rate across the sample surface, which might occur when a region with a different chemical composition, such as a lipid droplet, comes to the surface. Such effects could be important if attempting to estimate the size of intercellular features with high precision.

Here we have demonstrated that high-resolution SIMS performed with a Cameca NanoSIMS 50 enables the 3D visualization of submicron-sized features within a single cell. Using this approach, we have observed interesting intracellular features with elevated ¹⁸O-cholesterol levels. Future work will focus on confirming the identity of these cholesterol-rich features, and locating specific organelles, such as endosomes, within the cell volume.

This work was partially supported by the U.S. NSF under CHE–1508662 (to M.L.K.), and Lab Directed Research and Development funding (to LLNL). A portion of this work was carried out in the Frederick Seitz Materials Research Laboratory Central Research Facilities, University of Illinois. The authors thank Kaiyan Lou for the synthesis of the ¹⁵N-sphingolipid precursors and ¹⁸O-cholesterol. Work at LLNL was supported by Lab Directed Research and Development funding and performed under the auspices of the U.S. DOE under Contract No. DE-AC52-07NA27344.