Simulation Study of Breast Cancer Lipid Changes Affecting Membrane Oxygen Permeability: Effects of Chain Length and Cholesterol

Abstract

Tumor radiotherapy relies on intracellular oxygen (O₂) to generate reactive species that trigger cell death, yet hypoxia is common in cancers of the breast. De novo lipid synthesis in tumors supports cell proliferation but also may lead to unusually high levels of the 16:1 palmitoleoyl (Y) phospholipid tail, which is two carbons shorter than the 18:1 oleoyl (O) tail abundant in normal breast tissue. Here, we use atomic resolution molecular dynamics simulations to test two hypotheses: (1) the shorter, 16:1 Y, tail of the de novo lipid biosynthesis product 1-palmitoyl,2-palmitoleoyl-phosphatidylcholine (PYPC) promotes lower membrane permeability relative to the more common lipid 1-palmitoyl,2-oleoylphosphatidylcholine (POPC), by reducing oxygen solubility in the interleaflet region and (2) cholesterol further lessens the permeability of PYPC by reducing overall O₂ solubility and promoting PYPC tail order adjacent to the rigid cholesterol ring system. The simulations indicate that PYPC does have a somewhat lower permeability of 13.6 ± 0.5 cm/s at 37°C, compared with 14.6 ± 0.4 cm/s for POPC. Inclusion of cholesterol in a 1:1 ratio with phospholipid intensifies but also inverts the effect of chain length, giving permeabilities of 9.6 ± 0.3 cm/s for PYPC/cholesterol and 7.9 ± 0.2 cm/s for POPC/cholesterol. These findings indicate that PYPC does influence membrane-level oxygen flux but is unlikely to hinder breast tissue oxygenation.

1. Introduction

Hypoxia is a common attribute of the tumor microenvironment [1]. Yet, tumor radiotherapy relies on intracellular O₂ to generate reactive oxygen radicals that damage DNA and lead to cell death [2, 3]. Thus, factors affecting the intracellular oxygen content of tumor cells are of interest for predicting and enhancing radiotherapy outcomes. This study addresses the possible influence of de novo fatty acids and cholesterol in breast tumors on intracellular oxygenation. Specifically, atomistic molecular dynamics simulations are used to examine membrane-level effects of lipid chain length and cholesterol on oxygen permeability, as well as solubility and diffusion within model membranes. Atomistic simulations enable precise control over lipid composition, along with subnanometer and picosecond timescale resolution. Prior simulation and experimental studies indicate that lipid compositional variations alter membrane oxygen permeability, as well as the pathway of oxygen diffusion [4–8].

The common membrane phospholipid 1-palmitoyl-2-oleoyl-phosphatidylcholine (POPC), or PC(16:0,18:1), has hydrocarbon tails that are well matched in chain length. Though its two tails differ in the number of carbons, the 16:0 saturated tail extends to roughly the same length as the 18:1 monounsaturated tail (Fig. 1A). Here, we use a conventional notation that specifies the number of carbons in the tail, $n$, followed by the number of cis double bonds, $x$, giving $n:x$.

Fig. 1.

Model lipid bilayer images, together with plots of bilayer-depth-dependent properties. (A, B) Simulation snapshot images for PYPC and PYPC/chol. Water molecules shown as cyan dots, lipids as gray bonds with headgroups colored by element, O₂ molecules as red spheres, and cholesterol as white bonds; all H atoms omitted for clarity. (C) Electron density profiles. Legend for all plots: PYPC (blue solid lines), POPC (black solid lines), PYPC/cholesterol (blue dashed lines), POPC/cholesterol (black dashed lines). Vertical lines indicate the approximate bilayer depths of the headgroup electron density peaks and show corresponding depths among the images and plots for PYPC (solid line) or PYPC/chol (dashed line). (D) Relative depth-dependent free energy profiles for O₂, $\mathrm{\Delta }G\left(z\right)$. (E) Depth-dependent O₂ lipid–water partition coefficient, $K\left(z\right)$.

De novo fatty acid biosynthesis is typically upregulated in cancers, and the new fatty acids are used to build membranes for cell proliferation [9]. Products of de novo lipid biosynthesis are unusually abundant in breast cancer cells [10]. In particular, the cell membranes of poor-prognosis breast tumors appear to contain high levels of the 16:1 palmitoleoyl (Y) fatty acyl tail, which can form 1-palmitoyl-2-palmitoleoyl-phosphatidylcholine (PYPC) [9]. The POPC and PYPC lipids are identical in structure, except that the unsaturated tail of PYPC is two carbons shorter than that of POPC (Fig. 1A, B). We anticipated that the shorter tail would introduce chain-length mismatch and, therefore, would promote tail interdigitation at the leaflet-leaflet interface. As such, the first hypothesis tested in this study is that the shorter, 16:1 Y, tail of PYPC promotes lower membrane permeability relative to POPC, by reducing oxygen solubility in the interleaflet region.

The study also examines the effects of cholesterol (chol) in combination with PYPC, as cholesterol modifies the free energy landscape of oxygen translocation in POPC and promotes oxygen solubility in the interleaflet region [6]. Cholesterol accounts for 20–40% of the membrane lipid molecules in most cell types and occurs at higher levels in the eye lens [8, 11] and in red blood cells (RBCs). The RBC membrane cholesterol-to-phospholipid ratio is typically 1:1 (50 mol% cholesterol) [12]. We previously found that POPC bilayer O₂ permeability decreases when cholesterol is enriched [6], and this finding is consistent with experiment [8]. The O₂ lipid-water partition coefficient (or relative solubility) is reduced roughly in proportion to the permeability decrease in POPC/chol [13], as predicted by Overton’s rule [14]. Cholesterol reduces the solubility of oxygen within the bilayer, as a whole, but increases the solubility locally in the interleaflet region [6]. The second hypothesis tested in this study is that cholesterol further lessens the permeability of PYPC by reducing overall O₂ solubility and promoting PYPC tail order adjacent to the rigid cholesterol ring system.

The oxygen permeability and related properties are studied for PYPC alone and with 50% cholesterol. Comparison is made to POPC alone and with 50% cholesterol. The 50% cholesterol level (1:1 mixing ratio) was chosen because permeability changes were previously observed at and above this level [6, 8].

2. Methods

All-atom molecular dynamics (MD) simulations were carried out with graphics processing unit (GPU)-enabled AMBER code, version 14 [15, 16]. The Lipid14 force field, cholesterol extension and TIP3P water model were applied [17–19]. The pressure was fixed at 1 bar using the Monte Carlo barostat with anisotropic scaling, and the temperature was maintained at 310 K using the Langevin thermostat with a collision frequency of 1 ps⁻¹. Bonds to hydrogen were constrained with SHAKE [6]. The simulation timestep was 2 fs, with coordinates written every 1 ps. Periodic boundary conditions were applied, and particle-mesh Ewald was used to calculate nonbonding interactions, with a cutoff distance of 10 Å.

Four simulation systems were constructed: PYPC, POPC, PYPC/chol (1:1 mixing ratio, or 50 mol% cholesterol), and POPC/chol. Compositional details are provided in Table 1. As in other recent work, statistical sampling was enhanced by augmenting the O₂ concentration to reach 110–140 mM averaged over the whole system [5, 6]. Equilibration was conducted over 150 ns for the systems without cholesterol and 400 ns for those with cholesterol (to facilitate lipid mixing). Subsequent production simulations of 300 ns were the basis of all data analysis reported. Each system was simulated in duplicate, to enable assessment of convergence and estimation of error. Calculated values are reported as mean ± standard deviation, averaging over both duplicate trajectories and both bilayer leaflets in each trajectory.

Table 1.

Simulation system composition (number of molecules or lipids per leaflet), bilayer area, and headgroup-to-headgroup thickness from calculated electron density curves, $D_{HH}$

System	POPC	Cholesterol	Water	O2	Area (Å2)	${\mathit{D}}_{\mathit{H}\mathit{H}}$ (Å)
POPC	64 / 64	0 / 0	4,460	20	4,080 ± 90	38.5
PYPC	64 / 64	0 / 0	4,460	20	3,960 ± 80	37.3
POPC/chol	32 / 32	32 / 32	4,460	20	2,690 ± 30	45.5
PYPC/chol	32 / 32	32 / 32	4,460	20	2,640 ± 20	45.8

The membrane O₂ permeability, $P_M$, for each system was computed from unrestrained simulations by tracking O₂ half-permeation events, as in our earlier work [6] but with a factor of 2 correction applied to account for O₂ entry via both bilayer leaflets under periodic boundary conditions. The resulting equation,

$$P_M=\frac{{\mathrm{\Phi }}_{esc}}{4N_w},$$ (1)

generates $P_M$ by normalizing the frequency of O₂ molecule escapes from the bilayer, ${\mathrm{\Phi }}_{esc}$, by the average O₂ population in the water, $N_w$. Overton’s rule predicts $P_M$ to be directly proportional to a solute’s lipid–water partition coefficient, $K$, and its diffusion coefficient within the membrane, $D_M$, but inversely proportional to the membrane thickness, $h$[14]:

$$P_M=\frac{K\cdot D_M}{h}.$$ (2)

Electron density and bilayer-depth-dependent free energy, $\mathrm{\Delta }G\left(z\right)$, were calculated as in our earlier work [6]. Briefly, $\mathrm{\Delta }G\left(z\right)$ was computed as the natural log of the depth-dependent O₂ lipid–water partition coefficient, $K\left(z\right)$:

$$\mathrm{\Delta }G\left(z\right)=-RT\cdot lnK\left(z\right),$$ (3)

where $R$ is the ideal gas constant and $T$ is the simulation temperature.

3. Results and Discussion

Figure 1 provides images of the PYPC and PYPC/chol model systems, along with several plots of bilayer-depth-dependent properties calculated from the simulations. Included are electron density profiles, relative free energy profiles for O₂ (similar to potentials of mean force), and O₂ lipid–water partition coefficient profiles. Electron density is given here as an indicator of physical packing density as a function of the bilayer depth, plotted as distance from the bilayer center. The free energy profile reflects the relative favorability of O₂ occurrence at various depths and, in this study, is equivalent to a normalized log-concentration curve for O₂ (see Eq. 3). The partition coefficient profile is a representation of the bilayer-depth-dependent O₂ concentration, normalized to the O₂ concentration in the water layer. Table 2 provides bilayer permeability values calculated with Eq. 1, along with related physical properties, including a bilayer-average partition coefficient.

Table 2.

Bilayer physical properties: membrane permeability, ${\mathit{P}}_{\mathit{M}}$, permeation thickness, $\mathit{h}$, lipid order parameter plateau, $<{\mathit{S}}_{\mathit{C}\mathit{D}}>$, and average bilayer–water partition coefficient, $<\mathit{K}>$

Bilayer	${\mathit{P}}_{\mathit{M}}$ (cm/s)	$\mathit{h}$ (Å)	${\mathit{P}}_{\mathit{M}}/{\mathit{P}}_{\mathit{w}}$ a	$<{\mathit{S}}_{\mathit{C}\mathit{D}}>$ b	$<\mathit{K}>$
PYPC	13.6 ± 0.5	56	0.30	0.203	5.3 ± 0.4
POPC	14.6 ± 0.4	56	0.32	0.204	5.3 ± 0.5
PYPC/chol	9.6 ± 0.3	62	0.23	0.395	3.5 ± 0.4
POPC/chol	7.9 ± 0.2	62	0.19	0.388	3.8 ± 0.2

^aPermeability of a water layer of the same thickness as the associated bilayer, from $P_w=D_w/h$, where $D_w=2.60\times {10}^{-5}{\text{cm}}^2/s$ is the experimental diffusion coefficient for O₂ in water at 310 K (see ref. [6]). Wherever $h=56\hspace{0.17em}\hspace{0.17em}\AA$, $P_w=46\hspace{0.17em}\hspace{0.17em}\text{cm}/\text{s}$, and wherever $h=62\hspace{0.17em}\hspace{0.17em}\text{Å}$, $P_w=42\hspace{0.17em}\hspace{0.17em}\text{cm}/\text{s}$.

^bCalculated over 16:0 (sn-1) tail carbons 4–8, as in ref. [23]. Standard deviation ~0.006.

Incorporation of cholesterol tightens the packing of the lipids within the bilayer plane and causes the phospholipid tails to lengthen (Fig. 1B vs. 1A). These structural effects are reflected in pronounced changes in the shape of the electron density and free energy profiles (Fig. 1C, D), as well as in the lipid order parameter, $<S_{CD}>$ (Table 2). Earlier studies found similar physical changes upon cholesterol incorporation [6, 20, 21]. The bilayer-depth-dependent electron density curves peak in the headgroup regions and are lowest in the bilayer center, regardless of cholesterol content or phospholipid type (Fig. 1C). Cholesterol incorporation generates an intensification of electron density along the region spanned by the rigid cholesterol ring system (flat zone starting ca. 17 Å from the bilayer center) and a sharp decrease in electron density around the bilayer center (0 Å).

The electron density profiles are nearly identical for the two phospholipid systems (PYPC and POPC, solid lines) and for the two phospholipid-cholesterol systems (PYPC/chol and POPC/chol, dashed lines). The free energy profiles (Fig. 1D) are, likewise, nearly the same for PYPC and POPC, except that PYPC (blue solid line) has a slightly deeper free energy well at the bilayer center. This small difference is reflected in a pronounced peak-height difference in the partition coefficient, $K\left(z\right)$, curves for PYPC and POPC (Fig. 1E). Note that the $K\left(z\right)$ peak is higher for PYPC, counter to our first hypothesis that lower O₂ solubility in the interleaflet region diminishes PYPC permeability. Averaging the partition coefficient over the whole bilayer gives nearly identical values, $<K>$, for POPC and PYPC (Table 2).

As such, the free energy and partition coefficient data suggest that the permeability of PYPC should be the same or somewhat higher than that of POPC. Yet, independent calculation of permeability, $P_M$, using Eq. 1 indicates that PYPC is less permeable than POPC by ~7% (Table 2). Though the error ranges of the two $P_M$ values are close, they do not overlap. Both bilayers have permeabilities about one-third the permeability of an unstirred water layer of the same thickness ($P_M/P_w$ in Table 2).

The free energy profile for PYPC/chol has more pronounced energy barriers than POPC/chol at the headgroups (ca. 25 Å from the bilayer center) and, more prominently, spanning the rigid cholesterol ring system (ca. 10 Å from the center). The increased magnitude of these barriers would tend to favor reduced permeability for PYPC/chol, relative to POPC/chol. However, the permeability of PYPC/chol is found to be greater than that of POPC/chol by ~20%, and the difference lies outside the calculated error ranges (Table 2). The PYPC/chol free energy well is slightly less deep than that of POPC/chol, reflected in a diminished partition coefficient, $K\left(z\right)$, peak for PYPC/chol (Fig. 1E).

Depth-dependent diffusion coefficient profiles estimated from O₂ mean-squared displacements did not reveal any differences (data not shown). The bilayer areas and thicknesses, likewise, are similar (Table 1). Hence, we do not yet have a satisfactory explanation for the higher permeability of PYPC/chol relative to POPC/chol. Both cholesterol-rich bilayers have permeabilities about one-fifth the permeability of a water layer of the same thickness $\left(P_M/P_w\right)$.

Ongoing work will address in more detail the possible roles of local diffusion coefficients, O₂ molecule residence time within the bilayer, and molecular void (“empty” volume) distribution. Limitations of the current work include imperfect O₂ and lipid force fields, which tend to exaggerate O₂ lipid–water partitioning and the POPC bilayer thickness [22], as well as a lack of physical experimental data for PYPC and PYPC/chol bilayers. Additionally, the bilayers studied here are highly simplified and lack the variety of lipids as well as proteins present in real biological membranes. Still, we find these simplified models useful for teasing out molecular physical influences on membrane oxygen permeability.

4. Conclusions

We have studied the effects of lipid chain length and cholesterol on bilayer O₂ permeability, to discern whether de novo fatty acid biosynthesis in breast tumors can modulate cellular hypoxia. The study indicates that the phospholipid PC(16:0/16:1), or PYPC, does somewhat reduce phospholipid bilayer permeability, relative to the more common lipid POPC. The reason for this difference has not emerged from the current work. Cholesterol did reduce PYPC permeability by ~20% (after eliminating thickness effects by dividing by the permeability of a water layer of the same thickness). However, the permeability-reducing effect of cholesterol (at the level of 50 mol%) was less effective for PYPC than for POPC. Assuming that lipid bilayer permeability is a physical determinant of intracellular oxygenation, these findings indicate that increased PYPC abundance in breast tumor cells is unlikely to hinder cellular-level oxygen delivery.