Chapter 6: Magnification of Cholesterol-Induced Membrane Resistance on the Tissue Level: Implications for Hypoxia

Abstract

High cellular membrane cholesterol is known to generate membrane resistance and reduce oxygen (O₂) permeability. As such, cholesterol may contribute to the Warburg effect in tumor cells by stimulating intracellular hypoxia that cannot be detected from extracellular oxygen measurements. We probe the tissue- level impact of the phenomenon, asking whether layering of cells can magnify the influence of cholesterol, to modulate hypoxia in relation to capillary proximity. Using molecular dynamics simulations, we affirm that minimally hydrated, adjacent lipid bilayers have independent physical behavior. Combining this insight with published experimental data, we predict linearly increasing impact of membrane cholesterol on oxygen flux across cells layered in tissue.

1 Introduction

Here, we consider intracellular oxygen bioavailability and the potential impact of membrane resistance as a hindrance to oxygen flux. In particular, we examine the kinetics of oxygen transport across a series of several membranes, relevant to cells buried between capillaries, especially within tumors.

Several investigations provide valuable evidence that oxygen flux is reduced by substantial membrane cholesterol incorporation. Swartz and colleagues [1] demonstrated, using electron paramagnetic resonance (EPR) oximetry, that cultured mammalian cell lines with distinct membrane cholesterol levels differ in intracellular oxygenation. Using a technique that differentiates intra- and extracellular oxygen measurements, the group showed that an oxygen gradient could be observed across the plasma membrane. Assays were performed on three Chinese hamster ovary (CHO) cell lines with phenotypic differences in cholesterol metabolism and plasma membrane cholesterol level. The magnitude of the oxygen gradient increased with the plasma membrane cholesterol concentration. Manipulation of the cholesterol level in the cell lines, likewise, influenced the gradient, with increasing cholesterol content tending to decrease intracellular oxygenation relative to extracellular [1].

Such a gradient could arise if the rate of oxygen consumption exceeded the rate of replenishment. Thus, the work suggests that high-cholesterol plasma membranes can present a measureable barrier to oxygen transport. We especially note from the CHO cell study that plasma membrane cholesterol content can influence intra-cellular oxygenation in a manner not directly interpretable from extracellular measurements.

The view of high-cholesterol cell membranes as a barrier is further affirmed in studies of red blood cell (RBC) oxygenation by Buchwald and colleagues [2, 3]. Specifically, the rate of oxygen uptake and release declines with the plasma membrane cholesterol content of RBCs [3]. As a result, RBC oxygen saturation is diminished, and oxygen release upon demand is diminished [2]. The authors propose that angina pectoris in patients with hypercholesterolemia may be caused partly by reduced oxygen release from RBCs with elevated membrane cholesterol, contributing to heart muscle hypoxia and painful lactic acid buildup [3].

These studies, together, present compelling evidence that high-cholesterol membranes present a kinetic barrier to oxygen transport. Biophysical experiments with EPR oximetry reinforce this view. Subczynski and colleagues have found that cholesterol reduces the membrane oxygen permeability coefficient (a descriptor of flux) by several-fold, compared with phospholipid alone [4, 5]. Given this resistance to oxygen transport, Brown and Galea previously suggested that cholesterol may contribute to the Warburg effect in cancer cells [6]. Indeed, the existing evidence strongly implies that intracellular hypoxia is possible, even when extracellular oxygen is abundant.

We investigate the tissue-level impact of reduced oxygen flux through cholesterol- rich membranes. We do so by examining the structural and functional independence of bilayers placed very close together, representing direct juxtaposition of plasma membrane segments in layered cells. We combine the results with experimentally derived flux information to predict how plasma membrane cholesterol impacts oxygen bioavailability within tissue, given the necessity of crossing multiple membranes to reach mitochondria in cells buried between capillaries.

2 Methods

We have used all-atom molecular dynamics simulations of two adjacent bilayers separated by a thin water layer to calculate electron density and oxygen diffusional free energy profiles. All simulations used the GPU/CUDA-accelerated implementation [7] of the Amber 14 or Amber 12 biomolecular simulation software [8, 9], along with the Lipid14 force field [10] and a cholesterol extension by Ross Walker and Benjamin Madej [11]. We developed O₂ parameters in our laboratory, defining the bond length as 1.21 Å from the CRC Handbook [12], with a vibrational force constant of 849.16 kcal/mol • Ų based on Raman spectroscopic measurements [13] and with all other parameters defined the same as the carbonyl oxygen (oC) atom type in Lipid11 [14].

Lipid bilayers were constructed initially using the CHARMM-GUI membrane builder [15, 16]. A bilayer containing 1-palmitoyl-2-oleoylphosphatidylcholine (POPC) and cholesterol in a 1:1 ratio was built with 128 lipids total, including 32 POPC and 32 cholesterol molecules per leaflet and was pre-equilibrated for 500 ns using the GAFFLipid force field [17] with the Lipid11 cholesterol parameters [14]. The Lipid14 force field [10] with a cholesterol extension by Ross Walker and Benjamin Madej [11] was used throughout the remaining simulations. The pre- equilibrated POPC/cholesterol system was further equilibrated for 200 ns with this force-field combination. Its closest 15 water molecules per lipid molecule (“per lipid”) were retained using the AmberTools [8] program CPPTRAJ [18], and this minimally hydrated POPC/cholesterol structure was used as the starting configuration for “double bilayer” simulations.

All simulations used the TIP3P water model [19]. Through trial-and-error, we established that bilayers separated by 15 waters per lipid remained structurally distinct, while bilayers separated by only 10 waters per lipid showed physical fusion behavior early in the simulations.

Bonds to hydrogen were constrained using the SHAKE algorithm [20], allowing a 2-fs timestep. A constant temperature of 310 K (37 °C) was maintained using Langevin dynamics with a collision frequency of 1 ps⁻¹ during the equilibration phases and using the Berendsen thermostat [21] during the production phase. A constant pressure of 1 atm was maintained using the Berendsen barostat [21] during the pre-equilibration and the Monte Carlo barostat (as implemented in Amber 14) thereafter.

A POPC bilayer was constructed with CHARMM-GUI, including 15 waters per lipid. This bilayer was size-matched with the pre-equilibrated POPC/cholesterol structure described above, based on an expected POPC area per lipid calculated from previous simulations. This surface-area matching called for 82 POPC lipids, with 41 in each leaflet. We used PackMol [22] to place the POPC bilayer close to the pre-equilibrated POPC/cholesterol bilayer. This double bilayer system was minimized over 20,000 steps, heated from 100 to 310 K over 100 ps, then equilibrated for 200 ns prior to adding oxygen.

O₂ molecules were introduced by replacing water molecules between the bilayers, and the double bilayer systems with oxygen were minimized and heated in the same manner as above, while applying a Cartesian restraint on the O₂ molecules with a force constant of 5 kcal/mol • Ų. An O₂ concentration of 200 mM was used for the entire molecular system (water and lipid). Though higher than the concentration of air-equilibrated water (approximately 200 μM), this concentration allows the O₂ molecules to diffuse independently while facilitating rapid sampling of the O₂ configurational space [see also 23]. Production simulations were run without restraints for 300 ns.

Free energy and electron density profiles were calculated using CPPTRAJ. The free energy was calculated at 1-Å intervals (bins) along the bilayer normal, z, following the approach of Marrink and Berendsen that applies the relationship G(z) = −RT ln{C(z)/C_max}, where C(z) is the O₂ population of a given z bin and C_max is the population of the most populated bin [24].

3 Results and Discussion

The constituent POPC and POPC/cholesterol bilayers in the minimally hydrated “double bilayer” system remain structurally independent. As such, the electron density and free energy profiles of both bilayers (Fig. 6.1) closely resemble those calculated in our laboratory for single bilayer systems (data not shown). This observation agrees with prior experimental work demonstrating that the structural distinctness of lipid bilayers is maintained by headgroup-associated hydration repulsion [reviewed in 25]. Thus, we treat each bilayer as having independent oxygen transport behavior in the analysis that follows.

Fig. 6.1.

Electron density and free energy profiles are independent for hydrated bilayers in close proximity. The image shows a double bilayer system image overlaid with electron density (white line) and relative free energy (black line ) plotted against relative depth perpendicular to the bilayer plane. The left bilayer contains only POPC, and the right contains POPC and cholesterol in a 1:1 ratio. The bars on the electron density curve indicate the standard deviation. Both curves were calculated from a single trajectory, covering 300 ns of production. The small red points indicate the positions of explicit water molecules, simplified for clarity. Bilayers: POPC carbon atoms—gray, cholesterol carbon atoms—blue, O₂ molecules—green spheres. All hydrogen atoms omitted for clarity

Subczynski and colleagues found in EPR oximetry studies that incorporation of cholesterol in a 1:1 ratio with phospholipid results in a three- to five-fold reduction in the permeability coefficient for oxygen flux [4, 5]. Based on free energy barriers alone and not accounting for the diffusional/collisional behavior of oxygen, we predict a rate reduction of the same order of magnitude as the EPR study (data not shown). We recently have investigated the impact of higher cholesterol concentrations on oxygen transport and have found a consistent decrease in the predicted rate of oxygen transport with modest increases in cholesterol content (publication forthcoming).

Yet, what is the tissue-level impact of membrane resistance? To reach mitochondria, oxygen must diffuse across a minimum of two cholesterol-containing membranes—namely, the plasma membrane of the red blood cell and the plasma membrane of the target cell. Endothelial cell membranes may also intercede, presenting an additional barrier if the cholesterol content is substantial. For cells buried between capillaries, oxygen will then diffuse via the path of least resistance until it reaches its site of consumption. Assuming highly permeable cellular membranes, it seems likely that the most favorable path will normally involve passage from membrane to membrane [26, 27], as the membrane interior provides a low-energy environment for oxygen (observable in Fig. 6.1).

If oxygen were required to pass through several membranes, the rate of diffusion would be reduced in proportion to the number of resistant membranes. Supposing a series of equally resistant membranes, we predict the kinetic impact of passing oxygen across several cell layers. Figure 6.2 projects the cumulative effect on oxygen flux when several cholesterol-rich membranes are traversed, assuming a three- or five-fold reduction in permeability per membrane. The overall rate impact is expected to depend linearly on the number of membranes crossed and on the magnitude of resistance presented by each membrane. In the case that an individual membrane reduces oxygen permeability by three-fold, ten sequential membranes are anticipated to decrease oxygen flux by 30-fold, reaching 1/30 or roughly 3 % of the unhindered flux.

Fig. 6.2.

Predicted magnification of flux reduction as O₂ traverses multiple cell layers

The above analysis describes a “worst-case scenario” for oxygen diffusion across cholesterol-rich tissues. In actual tissue, it is likely that oxygen may find alternative, lower-resistance paths for diffusion that avoid some of the large energy barriers presented by cholesterol-rich membranes. For example, oxygen may pass around cells through the extracellular space. Thus, the degree of overall flux reduction may be lower than predicted with our current model. Even if diffusion through extracellular space were possible, the diffusional path-length required for avoiding passage across resistant cells would increase relative to the membrane-to-membrane diffusion scenario, generating a reduction in flux proportional to the path-length. Further modeling that takes the possibility of diffusion through extracellular space into account may be informative.

It is also important to note that our model greatly simplifies the structure of biological membranes. We have examined membranes incorporating only POPC phospholipid and cholesterol, while actual membranes incorporate many other types of phospholipid, along with substantial amounts of protein. Membrane proteins, like cholesterol, may reduce the permeability of membrane lipids, and the proteins, themselves, may have very low oxygen permeability [4, 28, 29]. Further, we have assumed compositional homogeneity across the leaflets of the bilayer, while plasma membranes are known to differ in the lipid composition of the extra-cellular and intracellular leaflets. We plan to address such complexities of cellular membrane structure in future modeling work. Moreover, we intend to use our models in conjunction with oxygen measurements in living cells, under conditions predicted by the models to influence intracellular oxygenation.

4 Conclusion

The capacity of cholesterol-rich cellular membranes to hinder oxygen diffusion may result in intracellular hypoxia not discernable from extracellular measurements. Even modest reduction of flux across a single membrane could substantially impact tissue oxygenation, due to magnification of the rate effect when oxygen must traverse several membranes or navigate around intervening cells.