Solid-State NMR of Highly ¹³C-Enriched Cholesterol in Lipid Bilayers

Abstract

Cholesterol (Chol) is vital for cell function as it is essential to a myriad of biochemical and biophysical processes. The atomistic details of Chol’s interactions with phospholipids and proteins is therefore of fundamental interest, and NMR offers unique opportunities to interrogate these properties at high resolution. Towards this end, here we describe approaches for examining the structure and dynamics of Chol in lipid bilayers using high levels of ¹³C enrichment in combination with magic-angle spinning (MAS) methods. We quantify the incorporation levels and demonstrate high sensitivity and resolution in 2D ¹³C-¹³C and ¹H-¹³C spectra, enabling de novo assignments and site-resolved order parameter measurements obtained in a fraction of the time required for experiments with natural abundance sterols. We envision many potential future applications of these methods to study sterol interactions with drugs, lipids and proteins.

1. Introduction

Sterols are essential in numerous biochemical and biophysical processes inside the cell. Cholesterol (Chol), the major sterol present in mammalian cells, is a key regulator of membrane order, permeability, thickness and lateral organization [1–3] and ultimately membrane protein function [1,4,5]. The regulatory roles of Chol depend directly upon its atomistic interactions with other sterols and phospholipid molecules [6–8]. In fact, Chol has also recently been implicated to have an essential role in HIV-mediated viral fusion [9] and ligand binding leading to apoptosis [10]. The orientation and dynamics of Chol were demonstrated to be essential for its regulatory roles [11,12], although molecular details of the mechanism are still lacking. Additionally, the atomistic specificity of sterol interactions with a variety of proteins is of increased interest, especially because Chol is a major component of mammalian membranes (typically 50 mol%) [13]. For example, Chol interacts with the amyloid precursor protein and may play a key role in amyloidogenesis related to Alzheimer’s disease [14]. These many applications underscore the broad significance of impactful methods to study this ubiquitous molecule in its native bilayer environment.

Much of what is known about the detailed structure and dynamics of Chol in lipid bilayers comes from experimental NMR studies, combined with molecular dynamics (MD). Extensive ²H and ¹³C studies of Chol have reported order parameters and restraints on the orientation in the bilayer from ²H NMR [15–27] and ¹³C NMR [28–30]. However, previous NMR dynamics investigations focused on a limited number of labeled sites [15–25] and/or used bulky analogs of Chol [31,32], which are known to behave differently in membranes than the native sterol [2,33]. Additionally, MD simulations of Chol in bilayers have been utilized to ascertain the orientation and fast timescale dynamics (usually <100 ns) at some sites [11,34]. Unfortunately, the rigorous comparison of MD with NMR data has often been limited to the A ring and tail of Chol, due to the relative ease of labeling these portions of the molecule [29]. For example, the most readily available commercial versions of ²H and ¹³C-Chol are (3,4-¹³C₂), (2,3,4-¹³C₃), (23, 24, 25, 26, 27-¹³C₅), (3-D₁), (6-D1), (7-D1), (2,2,3,4,4,6-D₆) or (25, 26, 26,26, 27, 27, 27-D₇) (from Cambridge Isotope Laboratories, Sigma Aldrich, and Avanti Polar Lipids). Highly enriched Chol would potentially enable more complete global analyses of structure and dynamics, as has been demonstrated to be critical for the high-resolution determination of membrane protein structures.

Here we provide methodological contributions into this active research field including: (1) a cost-effective production of highly ¹³C-enriched Chol on the >10 mg scale; (2) quantitative analysis of labeling patterns for uniform and 2-¹³C-acetate forms; (3) collection and assignment of multidimensional scalar and dipolar magic-angle spinning (MAS) solid-state NMR (SSNMR) spectra; and (4) determination of ¹H-¹³C dipolar order parameters.

2. Materials and Methods

2.1. Cholesterol Biosynthesis

2.1.1. Growth

Several YPD plates were prepared using 10 g/L of yeast extract, 20 g/L of peptone, and 20 g/L of agar. Water was then added to this mixture and autoclaved for 30 minutes. The mixture was cooled for about 20 minutes and 50 mL of 40% dextrose solution was added. The plates were then poured and left to cool and solidify. Once the plates solidified, the Rh6829 strain was streaked onto several plates and left to incubate for 24–48 hours until single colonies were visible. Approximately 1 L of media was prepared for the inoculation of the cell colonies. The media consisted of 950 mL milliQ water, 40 mg uracil, 40 mg leucine, 0.90 g ¹³C-sodium acetate, 7 g yeast nitrogen base with amino acids, and 5 g yeast extract. The mixture was autoclaved for 30 minutes and 50 mL of 40% sterile-filtered dextrose solution was subsequently added to the mixture. 5 colonies of the Rh6829 strain were added to 1 mL of media. 1 mL of the inoculated media was poured back into the 1 L cultures. The 1 L cultures were then incubated at 30 °C for 48–72 hours until the mixture was confluent.

2.1.2. Harvesting

Once confluent, the cells were spun down at 3000g for 30 minutes, 950 mL of the media were poured off and the cells were re-suspended in approximately 50 mL of the remaining media. The re-suspended cell mixtures were transferred to 50 mL conical vials and were then spun down in a centrifuge at 3000g for 5 minutes. The supernatant was discarded and the cells were re-suspended in 50 mL milliQ water (repeated 3 times). After pouring off the water from the final wash, 15 mL of 0.1 M HCl was were added to each vial. The mixtures were vortexed and added to a 1 L round bottom flask. The cells were heated at 90 °C and stirred at 700 rpm for 1 hour. The mixture was then transferred to a 5 L 3-neck round bottom flask. The 1 L round bottom flask was rinsed with 300 mL of 200 proof ethanol and the solution was transferred to the 5 L 3-neck flask. Additionally, the 1 L round bottom flask was rinsed with 1 L of 50% aqueous KOH and transferred to the 5 L flask. The mixture was stirred at 150 rpm and refluxed for 1 hour. After allowing the flask to cool to ambient temperatures, the mixture was transferred to a 4 L separatory funnel. The 3-neck flask was rinsed with 400 mL petroleum ether and poured into the separatory funnel. The organic layer was collected and aqueous layer was rinsed with petroleum ether (4 × 400 mL). The organic layers were combined, dried with sodium sulfate and filtered into a round bottom flask. The solvent was removed under reduced pressure to ca. 5 mL and transferred to a 40 mL IChem vial and the volatiles were removed in vacuo and the sample was dried overnight.

2.1.3. Purification

The crude material after harvesting is light tan (~70% pure). Preparative HPLC yields a >95% pure, colorless solid. The labeled cholesterol was dissolved in 3:1 ethyl acetate:acetone. The sterol was filtered into an HPLC vial. A mixture of isocratic :Ethanol (70:30) was used as the mobile phase to purify the sample using HPLC. The retention time was about 20 minutes. Once the sample has been retained after purification, the solvent was removed under reduced pressure yielding a colorless solid. The sample vial was stored at −20 °C.

2.2. SSNMR Data Collection & Analysis

2.2.1. 13C-¹³C Constant-time uniform-sign cross-peak (CTUC-COSY) [35] NMR Spectra

750 MHz (¹H frequency) spectra were acquired at 20 °C. Spectra were processed using NMRPipe [36] and analyzed using the Sparky program (3.115) for peak-picking [37].

2.2.2. R48₃¹⁸–Symmetry NMR Experiments

600 MHz (¹H frequency) 2D ¹H-¹³C dipolar recoupling spectra were acquired at 20 °C, 13.051 kHz, 36 points in the dephasing dimension, for a maximal dephasing period of 1.38 ms. Spectra were processed using NMRPipe [36] and analyzed in the Sparky program for peak picking. NMRPipe was used to extract out the time-domain trajectories.

2.2.3. R48₃¹⁸–Symmetry Data Analysis

The time-domain dipolar coupling dephasing trajectories were analyzed using an in-house model-free fitting program, Iota_MF (written in Python), which Fourier transforms the data to the frequency domain prior to fitting. The Iota_MF program analyzes the data by Average Liouvillian Theory (ALT). The resulting simulated fits yield a scaling factor and angle for each coupling as well as an overall relaxation value. Line shapes were first fitted using only directly attached protons.

2.3. Lipid Bilayer Reconstitution for Solid-State NMR

SSNMR samples were prepared in approximately 20 mg batches of either 10:3 or 40:1 (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC): U-¹³C Chol. ¹³C-Chol was added to the lipid (Avanti, 850457C, 10 mg/mL in chloroform), swirled until dissolved, and dried under a flow of nitrogen or argon gas. The lipid-sterol film was then placed under high vacuum overnight. Freshly prepared HEPES buffer of pH 7.0 and water were added to the dry film, which was subsequently vortexed (1 min) and sonicated (3 min). The solution then underwent 5 freeze-thaw steps that included freezing the sample vial in liquid nitrogen and thawing under running water. The solution was frozen a sixth time and immediately placed in the lyophilizer overnight to yield a white powder. After lyophilization, the sample container was flushed with argon before removing the powder and recording its final mass, and the powders were stored under argon. The powders were packed into 3.2 mm standard rotors and hydrated to ~33% by mass with deionized water.

2.4. Determining ¹³C Labeling Efficiency Using Quantitative Solution NMR

To ascertain the labeling percentage, quantitative proton and carbon spectra were collected. The quantitative ¹H NMR spectrum was collected without ¹³C decoupling, in order to observe the proton multiplicity due to the ¹³C-¹H scalar coupling. The central peak represents the population of protons attached to ¹²C, while the satellite multiplets depict the population of protons attached to a ¹³C-labeled atom. The ratio of the satellite peaks over all peaks yields the labeling percentage. Integration bounds for the proton spectra were 25 times the half maximal linewidth. This labeling percentage was related to the quantitative carbon spectrum. Integration bounds for the carbon spectra were 25 times the half maximal linewidth, when possible. For example, the [2-¹³C-acetate] Chol spectrum did not have any overlapping carbon resonances so it was possible to use 25 times the half maximal linewidth for all peaks. However, for the U-¹³C-Chol carbon spectrum, due to spectral overlap it is only possible to integrate using that range for 6 out of the 16 resolved resonances.

3. Results and Discussion

3.1. Biosynthesis and Determination of Labeling Efficiency of ¹³C-Chol

We employed a biosynthetic procedure utilizing a genetically modified S. cerevisiae strain capable of producing isotopically labeled Chol [38,39], instead of its native sterol (ergosterol), from ¹³C-glucose. Here we utilized ¹³C-labeled sodium acetate, which enables uniform labeling (with 1,2-¹³C-acetate) or fractional (“skip” or “checkerboard”) labeling with 1-¹³C- or 2-¹³C-acetate, and found the yields to be more cost effective than the fractionally labeled glucose variants. Specifically, the acetate biosynthesis yields approximately 30 mg of Chol from ~0.9 g U-¹³C, 1-¹³C or 2-¹³C-acetate, whereas the reported yield from 1.0 g of U-¹³C-glucose was ~1 mg in 100 mL of culture medium.

In order to determine the percentage of isotopic enrichment site-specifically, we utilized quantitative solution NMR (qNMR) experiments to determine the labeling efficiency. First, we analyzed the relative intensity of the Chol H3 signal (Figure 1A) ¹H{¹³C} doublet to singlet (without ¹³C decoupling), to obtain the percent incorporation for that site. (Due to two- and three-bond ¹H-¹³C J-couplings, each ¹H{¹³C} doublet signal has additional, partially resolved fine structure.) In a similar manner, the C6 labeling percentage was determined by analysis of the H6 doublet to singlet ratio. The relative labeling percentages for other sites in the ¹³C spectrum (Figure 1B) were then determined by comparison of the area of the C3 peak, and validated using the C6 peak. This resulted in an overall labeling of U-¹³C-Chol to be 72 ± 2%, globally (Table 1). Within the measurement error of ~3%, the site-specific determinations were equivalent throughout the molecule. The maximal labeling percentage is determined by the presence of unlabeled carbon sources, especially dextrose (2 g/L), that are necessary for cell growth.

Figure 1. Quantitative Solution NMR of U-¹³C Cholesterol.

A) 600 MHz Quantitative ¹H NMR spectrum showing the C3 region used to interpret labeling incorporation in B) 600 MHz quantitative ¹³C NMR spectrum.

Table 1. ¹³C Incorporation of U-¹³C-labeled Chol.

Asterisks denote an average labeling percentage for overlapped peaks. The error for each measurement is 3%, as determined by integrating a region with no peaks.

Carbon Number	13C % Incorporation
C1	73*
C2	72*
C3	73
C4	73*
C5	75
C6	76
C7	72*
C8	72*
C9	69
C10	73*
C11	72
C12	74*
C13	73*
C14	72*
C15	71*
C16	72*
C17	72*
C18	66
C19	68
C21	68
C22	73*
C23	71*
C24	74*
C25	72*
C26	69*
C27	69*
Average	72
Stdev	2

With U-¹³C Chol on hand, we next proceeded to prepare a sample of Chol in POPC and collect MAS ¹³C-¹³C correlation spectra. We utilized the scalar-based CTUC COSY [35] experiment (Figure 2) to unambiguously identify the one-bond correlations throughout the structure and arrive at complete resonance assignments. In particular, the U-¹³C labeling and correlation spectra enable a de novo assignment that did not rely upon solution NMR chemical shifts and uniquely separated the degenerate signals (C7 v. C8, C10 v. C20 v. C22, C14 v. C17, C15 v. C23).

Figure 2. CTUC COSY of U-¹³C-Chol in POPC Bilayers.

750 MHz, 8.928 kHz MAS SSNMR ¹³C-¹³C 2D CTUC COSY spectrum of 40:1 POPC: U-¹³C. The asterisks in the figure represent POPC peaks, the daggers represent artifacts arising from the subharmonics of the spinning [35].

3.2. Recoupling Experiments to Determine Averaged ¹H-¹³C Dipolar Couplings

Next we proceeded to collect 2D separated local field spectra with the R48₃¹⁸-symmetry based pulse sequence [40,41]. With the high levels of ¹³C enrichment, sensitivity for these experiments was enhanced more than an order of magnitude, relative to samples with natural ¹³C abundance. Experiments were carried out with 3.2 mm rotors and an active volume of ~18 uL, accommodating a sample of 11.6 mg 10:3 POPC:U-¹³C Chol, hydrated with 5.7 mg H₂O. Of this 17.3 mg sample, the signal from the ~2.8 mg of U-¹³C Chol yielded a high-resolution 1D spectrum (Figure 3) with resolution sufficient to observe multiplicity due to J-couplings in only 64 scans (Figure 3).

Figure 3. High Resolution 1D ¹³C Spectrum.

600 MHz MAS SSNMR ¹³C cross polarization spectrum (46 ms, 64 scans) of 10:3 POPC: U-¹³C-Chol with multiplicity due to J-couplings. The additional peaks not labeled are POPC peaks.

The high sensitivity also enabled 2D ^1H-13C dipolar recoupling data sets to be collected in 20 minutes or less. The separated local field line shapes resulting from the ¹H-¹³C dipolar couplings (Figure 4) were fit according to the average Liouvillian theory (ALT) approximations described in Hohwy et al. [42] and Rienstra et al. [43]. Data sets were normalized against the ¹Hα-¹³Cα scaled dipolar coupling of N-acetyl valine (NAV), which is representative of a one-bond rigid lattice coupling (Figure 1S).

Figure 4. Dipolar Spectra.

600 MHz MAS SSNMR R48₃¹⁸ dipolar spectra. Black lines are experimental data and red lines represent the fit.

Remarkably, the Chol data fit with higher quality than small molecule model compounds, as a consequence of the better convergence of the ALT model for fast limit motionally-averaged dipolar couplings that are small compared to the MAS rate and multiple pulse cycle time [43]. A representative set of dipolar coupling line shapes arising from CH, CH₂ and CH₃ groups are depicted in Figure 4.

The order parameters we measured (Table 1S) are consistent with values reported by Ferreira et al. [28] for 34% cholesterol in POPC above the liquid crystal phase transition: for C1 0.39±0.07 (versus ~0.375 for Ferreira), C3 0.38±0.03 (0.4), for C4 0.42±0.01 (0.35), C9 0.47±0.01 (0.45), C11 0.43±0.07 (0.45), and C18 0.21±0.01 (0.15). Additionally, we were able to measure the C6 (0.12±0.01) and C26/C27 (0.03±0.01) order parameters, which had too low a signal-to-noise ratio at natural abundance to be measured accurately.

3.3. Skip-Labeled ¹³C-Chol

Fractional ¹³C-labeling has been demonstrated to be indispensable for structural studies of proteins [44–48], particularly to improve the resolution and sensitivity of 2D ¹³C-¹³C spectra. Thus, we were intrigued by the potential for skip-labeling to yield similar benefits for studies of Chol, to enable future applications where spectral overlap may be problematic for U-¹³C-labeled samples. The biosynthetic approach presented here is compatible with either 1-¹³C or 2-¹³C-acetate, allowing for the production of complementary skip-¹³C-Chol patterns. Here we have evaluated the incorporation percentage for [2-¹³C-acetate]-Chol as described in Methods 2.4 and results summarized in Table 2. As a result of using skip-¹³C-Chol, the multiplet structure in the ¹³C 1D spectrum is considerably sharper due to the absence of most two-bond ¹H-¹³C scalar couplings (Figure 5A). The reduction in linewidth is also observed in the quantitative ¹³C spectrum due to the elimination of most one-bond ¹³C-¹³C scalar couplings (Figure 5B). Additionally, in cases where ¹³C-¹³C couplings are observed due to neighboring ¹³C atoms (Figures 5C and 5D), we were able to leverage this information to analyze the multiplet structures and further validate the percentage of ¹³C incorporation. We also utilized the ¹³C 1D spectrum (Figure 5D) to quantify the ratios of isotopomers. For example, the labeling of C13, C17 and C18 is positively correlated, with all three sites ¹³C-labeled at ~73% (whereas only 60% would be expected as the product of the individual labeling percentages). In addition, for another 24% population, both C13 and either C17 or C18 are labeled. These results are consistent the reported Chol biosynthetic pathway [39], and validated with a dipolar correlation (DARR) ¹³C-¹³C 2D spectrum of [2-¹³C-acetate]-Chol in POPC bilayers (Figure 6) which shows strong cross peaks among the C13, C17 and C18 sites. The C5 olefin peak appears at ~144 ppm and shows strong correlations to C3, C19, C7 and C9 sites. In addition, two-bond correlations are observed for the mostly odd-numbered carbons that are labeled in the sterol rings. Linewidths are approximately 60 Hz for the DARR ¹³C-¹³C 2D spectra. These resolution and sensitivity of the 40:1 POPC: [2-¹³C-acetate]-Chol spectrum demonstrate their potential benefits in future applications where spectral overlap is problematic.

Table 2. ¹³C Incorporation of [2-¹³C-acetate]-Chol.

Error of these measurements is 3% as determined by integrating a region of the spectrum with no peaks.

Carbon Number	13C % incorporation
C1	85
C3	83
C5	80
C7	86
C9	86
C13	90
C15	92
C17	80
C18	82
C19	83
C21	86
C22	89
C24	88
C26	87
C27	87
Average	86
Stdev	4

Figure 5. 600 MHz Solution spectra of Skip ¹³C-Chol.

A) Carbon 3 region of ¹H quantitative NMR spectrum used to obtain a ¹³C labeling ratio applied to B) the quantitative ¹³C Carbon NMR. C) and D) represent labeling percentages of C17 split by is ¹³C labeled neighbor C18 and C13 is split by C17 and C18 neighbors, respectively.

Figure 6. 600 MHz MAS SSNMR of [2-¹³C-acetate]-Chol in POPC.

DARR ¹³C-¹³C 2D (250 ms mixing time) of 40:1 POPC: [2-¹³C-acetate]-Chol, where unlabeled peaks on or near the diagonal are from the lipid, POPC.

4. Conclusions and Outlook

We have obtained a highly incorporated ¹³C-labeled Chol by feeding a genetically modified S. Cerevisiae strain [39] ¹³C-sodium acetate. Using qNMR, we determined the ¹³C-enrichment for U-¹³C Chol and 2-¹³C-acetate Chol to be 72 and 86%, respectively. Using labeled acetate instead of glucose produced a higher degree of incorporation of skip-labeled ¹³C-Chol in comparison to previous studies [39]. We anticipate that the feeding and growth schedule as well ¹³C-acetate to dextrose ratio could be further optimized to yield even higher rates of ¹³C incorporation. Nevertheless, the current incorporation levels already enable highly sensitive measurements. For example, we prepared a sample of U-¹³C-Chol in POPC, collected several MAS ¹³C-¹³C correlation spectra and completed resonance assignments. The high level of enrichment allowed us to measure order parameters in a 20-minute experiment using less than 3 mg of labeled material, whereas prior determinations of equivalent order parameters with natural abundance cholesterol required ~4.8 hrs for a 34 mol% Chol/POPC sample in a 4 mm HR-MAS rotor [28].

We envision that ¹³C-enriched Chol will be useful for a variety of studies by SSNMR in order to examine structure, dynamics and interactions with membrane proteins and drugs. For example, we have previously demonstrated binding of the antifungal drug amphotericin B (AmB) to labeled ergosterol [49], resulting in changes in the chemical shifts, order parameters, paramagnetic relaxation effects and lipid/water correlations. In similar manner, we anticipate that studies with ¹³C-Chol will yield insight into the structural basis of the sterol-specificity of AmB analogs with reduced toxicity [50–52]. A related approach has been demonstrated recently by Gronenborn [53] and colleagues, who used a ¹³C labeled carbohydrate probe with a ¹³C labeled protein to investigate intermolecular distances in a sugar-protein complex in order to elucidate not only the binding site of the sugar to the protein, but also the conformation of the bound sugar. These studies underscore the potential of using ¹³C-labeled small molecules to elucidate their key role in biology.

Highlights.

• Production of cholesterol with three ¹³C labeling patterns

• Reconstitution of ¹³C-cholesterol into lipid bilayer samples

• Magic-angle spinning to yield high resolution spectra

• Resonance assignments of membranous cholesterol

• Rapid order parameter measurements

Abbreviations

Chol

cholesterol

MAS

magic angle spinning

MD

molecular dynamics

SSNMR

solid-state nuclear magnetic resonance

CTUC COSY

constant-time uniform-sign cross-peak correlation spectroscopy

ALT

average Liouvillian theory

POPC

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine

NAV

N-acetyl valine