Factors determining barrier properties to oxygen transport across model and cell plasma membranes based on EPR spin-label oximetry

Abstract

This review is motivated by the exciting new area of radiation therapy using a phenomenon termed FLASH in which oxygen is thought to have a central role. Well-established principles of radiation biology and physics suggest that if oxygen has a strong role, it should be the level at the DNA. The key aspect discussed is the rate of oxygen diffusion. If oxygen freely diffuses into cells and rapidly equilibrates, then measurements in the extracellular compartment would enable FLASH to be investigated using existing methodologies that can readily measure oxygen in the extracellular compartment. EPR spin-label oximetry allows evaluation of the oxygen permeability coefficient across lipid bilayer membranes. It is established that simple fluid phase lipid bilayers are not barriers to oxygen transport. However, further investigations indicate that many physical and chemical (compositional) factor can significantly decrease this permeation. In biological cell plasma membranes, the lipid bilayer forms the matrix in which integral membrane proteins are immersed, changing organization and properties of the lipid matrix. To evaluate oxygen permeability coefficients across these complex membranes, oxygen permeation across all membrane domains and components must be considered. In this review, we consider many of the factors that affect (decrease) oxygen permeation across cell plasma membranes. Finally, we address the question, can the plasma membrane of the cell form a barrier to the free diffusion of oxygen into the cell interior? If there is a barrier then this must be considered in the investigations of the role of oxygen in FLASH.

1. Introduction

Data on the oxygen concentration and transport in tissues, cells, and subcellular structures are required to understand oxygen-related physiological and pathophysiological phenomena. They are central to comprehending radiation and photodynamic therapy. An exciting new area of therapeutic approaches is the use of very high dose rate irradiation, often termed FLASH. Empirically, there seems to be a decrease in the side effects in normal tissue, while the therapeutic effect on tumors is not diminished. This could greatly improve the efficacy and safety of radiation therapy for cancer. A number of considerations have led to a postulated role of oxygen in these effects, but this has not been shown unambiguously. In order to understand the basis of, and the role of oxygen in, FLASH, it is imperative to better understand the dynamics of oxygen under the conditions of FLASH [1–4]. Considering well-established knowledge of radiation physics and radiobiology, if oxygen does play a role in FLASH, the oxygen of particular pertinence is the intracellular oxygen level at the site of DNA. From experiments with isolated cells in which oxygen was made available at precisely measured time intervals after irradiation, it appears that the reactive intermediate has a lifetime of 1–10 microseconds. A key question then is how to measure the impact of the different types of radiation on the oxygen at that site within the time interval. This requires a high level of understanding of the rate of diffusion of oxygen into and within the cell, especially near the nucleus, in order to ascertain the volume of oxygen that is potentially available for reaction. Having no barriers to the free diffusion of oxygen into and within the cell would greatly simplify many calculations and interpretations of measurements. Conversely, barriers to the free diffusion of oxygen may be key to understanding the FLASH effect, because differences in these barriers between tumors and normal tissues may be involved in the mechanism of FLASH.

Therefore, the purpose of this review is to provide basic data on oxygen diffusion across cellular membranes and within the cell, indicating major factors that can affect this diffusion.

2. EPR approaches to measure oxygen permeation across membranes

The methods available to measure oxygen permeation across membranes are limited and often complex. As far as we are aware, to date, no techniques have been used specifically to investigate oxygen dynamics across membranes under FLASH conditions.

As indicated by the title, although this review will be based on the data obtained with the EPR spin-label oximetry method, we will also add information about other studies that allow measurements of oxygen permeability across model and biological membranes. The quenching of fluorescence probe molecules by oxygen in red blood cell membranes, mitochondrial membranes, or artificial membranes was used to obtain insight into oxygen permeability. However, these methods give only the average oxygen diffusion-concentration product across the membrane because of the large size of the probe and its uncertain localization in the membrane [5–10].Nuclear magnetic resonance techniques provide good spatial distribution of the oxygen diffusion-concentration product across the membrane, but the sensitivity of the method is low [11–13]. EPR measurements have been shown to be a very viable approach and, consequently, a large body of literature discusses the use of EPR for this type of measurement.

Molecular oxygen cannot be monitored directly by electron paramagnetic resonance (EPR) under standard experimental conditions. However, because of its paramagnetic properties, indirect methods exist that allow measurement of the collision rate of paramagnetic oxygen molecules with molecular labels present in the membrane. Based on this, a dual-probe saturation recovery (SR) EPR method that allows observation of the spin-lattice relaxation time (T₁) of lipid spin labels and measurement of bimolecular collision rate between spin labels and molecules of oxygen can indirectly measure oxygen. An important property when using spin labels is that, if the spin labels become associated with cell components, changes in their EPR spectra will be readily detected. Therefore, unlike some other techniques, one can be confident that the spin labels do not have confounding close associations with other molecules. In the early 1980s, at the National Biomedical EPR Center at the Medical College of Wisconsin, the absolute T₁ -sensitive spin-label oximetry method was developed to measure these bimolecular collision rates [14–16]. This method can be used quantitatively because every collision of oxygen with a spin label contributes to a change in the EPR signal in the T₁ and T₂ sensitive methods [15,17–19]. The effects of the collisions of molecular oxygen with spin labels on spin lattice relaxation rates for different spin labels are independent of microwave frequencies [20,21]; therefore, the absolute T₁ -sensitive spin-label oximetry method can be used not only at X-band (9.2 GHz) but also at Q-band (34 GHz) and W-band (94 GHz) [21]. The main benefits of SR EPR spin-label oximetry at Q- and W-bands are the long value of T₁ [20,22]. and small sample volume, 30 nL [23].

The bimolecular collision rate is the product of the local diffusion coefficient and the local concentration of molecular oxygen within the membrane, which is called the local oxygen diffusion-concentration product. In 1982, an oxygen transport parameter (OTP) was introduced by Kusumi et al. [16]. Use of this parameter is a convenient way to measure the collision rate between molecular oxygen and spin labels. OTP is defined as the difference between the spin-lattice relaxation rates of spin labels for samples equilibrated with atmospheric air (T₁⁻¹(x, air)) and those equilibrated with nitrogen (T₁⁻¹(x, N₂)):

$$\text{OTP}={T_1}^{-1}\left(\text{x},\text{air}\right)-{T_1}^{-1}\left(\text{x},{\text{ N}}_2\right)=AD\left(\text{x}\right)C\left(\text{x}\right),A=8\pi p{r}_0$$ (1)

where “x” indicates the position of the nitroxide moiety in the investigated system (i.e., the “depth” in the membrane), r₀ is the interaction distance between oxygen and the nitroxide moiety of the spin label (4.5 A) [24], and p is the probability that an observable event occurs when a collision occurs. It was shown that A is remarkably independent of the hydrophobicity and viscosity of the environment and of the spin-label species [17–19]. Thus, the OTP is proportional to the local oxygen diffusion coefficient D(x) and the local oxygen concentration C(x) in the membrane in the close vicinity of the nitroxide moiety of spin labels. Remember that the OTP is normalized to an oxygen concentration corresponding to the sample equilibrated with air at normal 760 mmHg atmospheric pressure. This method allows profiles of OTPs to be obtained across model and biological membranes (see Fig. 1 for example profiles and additional explanations).

Fig. 1.

Profiles of the OTP across DMPC membranes in the absence of cholesterol obtained at 25°C in fluid-phase membranes (○) and at 8°C in gel-phase membranes (x). Profiles obtained at 25°C in the presence of 50 mol% cholesterol (Chol/DMPC = 1/1) at 25°C (DMPC membrane saturated with cholesterol) (●) and in the presence of 66 mol% cholesterol (Chol/DMPC = 2/1) – across the CBD immersed into the DMPC membrane saturated with cholesterol) (♦). Approximate locations of the nitroxide moieties of spin labels are indicated by arrows.

It is practically impossible to directly measure the oxygen permeability coefficient across the membrane by creating fast-decaying oxygen concentration gradients across the membrane [25–28]. In 1989, Subczynski et al. published a paper [29] in which they describe a way to evaluate membrane oxygen permeability coefficients based on the OTP profiles obtained for membranes in equilibrium with air and nitrogen (such as in Fig. 1), without creating any oxygen gradients. The procedure, developed by Subczynski et al. [29], is based on the theory for permeability of nonelectrolytes across lipid bilayers membranes developed by Diamond and Katz [30]. The membrane permeability coefficient, P_M, connects the oxygen flux across the membrane, J, with the difference in oxygen concentration in the aqueous phase on either side of the membrane (C” - C’):

$$J=-P_{\text{M}}\left(C\text{'}\text{'}-C\text{'}\right)$$ (2)

Comparison of P_M with the permeability coefficient of a water layer of the same thickness as the membrane, P_W, allows one to conclude if the membranes form a barrier to oxygen transport.

3. Model phospholipid membranes

3.1. Effect of temperature (membrane phase)

The T₁ -sensitive spin-label oximetry method was used to calculate the P_M for simple lipid bilayer model membranes [29,31]. The immediate conclusion was that fluid phase phospholipid (PL) membranes do not form barriers to oxygen transport. The P_Ms evaluated for these membranes are significantly greater than the appropriate P_Ws (see Table 1). However, the impact of temperature on membrane organization affects the P_M significantly, and, for gel-phase membranes (for temperatures below the main phase transition temperature) the P_M can be ~5 times smaller than the P_W. Below the pretransition temperature, the P_M can be ~10 times smaller than the P_W (see Table 1). Thus, the physical state of the PL bilayer membrane can significantly (drastically) affect its permeability property (also, see Fig. 1 for appropriate profiles for fluid phase and gel-phase membranes).

Table 1.

Oxygen permeability coefficients P_M for different model membranes.

Membrane	Temperature (°C)	PM (cm/s)	Reference
DMPCa,b	8	5.3	[29]
DMPC	25	105.0	[32]
DMPC	38	185.0	[29]
DMPC - 50% cholesterol	38	38.0	[29]
POPCc	35	157.4	[33]
POPC - 50% cholesterol	35	63.0	[33]
EYPCd	25	119.0	[32]
EYPC	40	201.5	[31]
EYPC - 50% cholesterol	40	73.4	[31]
Water layere	35	60 – 90
CBDf	35	42.5	[34]
Water layer (34 A)	35	85.9

^aGel-phase membrane

^bDMPC - dimyristoylphosphatidylcholine

^cPOPC - 1-palmitoyl-2-oleoylphosphatidylcholine

^dEYPC - egg yolk phosphatidylcholine

^eThe thickness of the water layer is the same as that of the appropriate membrane.

^fThe thickness of the CBD is 34 A.

3.2. Effect of acyl chain unsaturation

Biological membranes contain large amounts of unsaturated acyl chains; thus, the effects of the unsaturated double bonds on PL bilayer P_M were investigated [29,31]. In biological membranes, the double bond is located mainly at the C9–C10 position. It was shown that the presence of a cis or trans double bond at this position decreases P_M as compared with that of saturated membranes (the comparison should be performed at the same temperature and for both membranes in the fluid phase). However, the decrease of the P_M induced by the presence of unsaturation is small, and in pure unsaturated PL membranes, the P_Ms are greater than the appropriate P_Ws (see Table 1).

3.3. Effect of Chol

Cholesterol (Chol) is a significant lipid component of human cell plasma membranes. It plays a major role in the determination of membrane properties such as fluidity and permeability, as well as induction of the formation of coexisting membrane phases and membrane domains (see Fig. 2 and Ref. [35] for more explanation). In this paper, we concentrate on the role of Chol in modifying oxygen permeability across cell plasma membranes. The Chol content in most of the cellular plasma membranes is 10 to 30 mol% (Chol/PL molar ratio between 0.1 and 0.5) [36].However, in certain cells, including the plasma membrane of red blood cells and the myelin membranes of Schwann cells isolating nerve axons, this content is significantly higher, close to 1 (~50 mol% Chol) [37,38]. At this high Chol concentration, Chol saturated the liquid-ordered phase PL bilayer [39]. At greater Chol contents, the PL bilayer cannot accommodate more Chol, and Chol forms pure Chol bilayer domains (CBDs) immersed into saturated with the Chol PL bilayer [40,41]. Only in the plasma membranes of the eye lens fiber cells, especially those of human eye lenses, is the Chol content greater than 1 (>50 mol%), and this high Chol content is physiologically justified [42–44] (see Fig. 2 for a schematic illustration of these situations).

Fig. 2.

Schematic drawings of membrane structures (including phases and domains) formed at different Chol contents in membranes made of Chol/DMPC mixtures. l_d, liquid-disordered phase; l_o, liquid-ordered phase; CBD, Chol bilayer domain (see Ref. [35] for more explanation). As indicated in the beginning of Sect. 3.3, the plasma membranes of different types of cells contain different amounts of Chol. This schematic drawing shows the phases and domains that can be expected in these plasma membranes, thus affecting membrane permeation for oxygen.

In fluid-phase PL membranes (at physiological temperatures), at Chol concentrations up to 50 mol%, Chol decreases oxygen permeation proportionally (practically linearly) to the amount of Chol present in the bilayer. As indicated in Table 1, 50 mol% Chol decreases oxygen permeation by a factor of ~5 in saturated membranes and by factor of ~2.5 in unsaturated membranes [29,31]. Transmembrane profiles of the OTP, which are used to calculate the membrane P_Ms, are bell-shaped with a gradual increase in the OTP toward the membrane center, and all OTP values are greater than in the surrounding water [29,33,45–47]. Chol decreases the OTP near the membrane center and has little or no effect at the membrane center [29,31]. At 50 mol%, the transmembrane OTP profile changes drastically to a rectangular shape with an abrupt increase between the C9 and C10 positions [29,31]. The OTP from the membrane surface to the depth of the C9 becomes ~3 times smaller than in the surrounding water (as low as in gel-phase membranes), and at locations deeper than the C10, it becomes 2–3 times greater than in the surrounding water (as high as in fluid-phase membranes) (see appropriate profiles of the OTP and the explanation in Fig. 1).

Even though EPR spin-labeling is a very powerful technique for oximetry measurements, it has some disadvantages: (a) The EPR spin-labeling approach permits evaluation of the local oxygen diffusion-concentration product but separation of this product on its components is not possible. (b) A nitroxide spin label cannot be placed in certain membrane domains and/or at certain membrane depths. This can lead to gaps in the profiles of the product, making the evaluation of P_Ms less relevant or impossible. A molecular dynamics (MD) simulation can independently provide these components across any bilayer membranes and domains, thus allowing evaluation of the P_M across these membranes and domains. A recent MD simulation of oxygen permeability across membranes in liquid-disordered and liquid-ordered phases membranes performed by Ghysels et al. [48] generated data that can be compared with those obtained with experimental EPR spin-labeling data, e.g., P_Ms. Both methods show that P_Ms across liquid-disordered phase membranes are about three times greater than across liquid-ordered phase membranes. The values obtained with an MD simulation are about five times smaller than those obtained through EPR oximetry (see Table 1). MD simulation-generated results cannot be obtained experimentally, e.g., the values for oxygen diffusion within the membrane and oxygen partitioning into the membrane are separate. Oxygen diffusion was further separated for its normal and parallel components. Combining the experimental and MD simulation results helps better explain the process of oxygen transport across membranes (see [48] and citations therein).

The similarities between the rectangular shapes of the OTP [29,31] and hydrophobicity profiles [49] suggest that hydrophobic molecules of oxygen can be easily transported in the center of the membrane, parallel to the membrane surface [33,34,50]. Skulachev proposed this pathway of transport for small nonpolar molecules within extended membranous systems of mitochondria [51]. This high Chol content creates a barrier to oxygen transport across the membrane.

CBDs, which are formed when Chol content is greater than 50 mol% (see the schematic illustration in Fig. 2), form significant barriers to oxygen transport. As a pure Chol domains, they can be discriminated by EPR using Chol analog spin labels, androstane spin label (ASL) and cholestane spin label (CSL) [41]. The P_M across these domains can be evaluated based on the OTP measured for these spin labels, which significantly decreases the accuracy of the evaluation, especially because the nitroxide moieties of these spin labels are located outside the rigid ring structure of Chol, where the major resistance to oxygen permeation is expected (see appropriate profile in Fig. 1). Nevertheless, EPR evaluations showed that the P_M for the CBD is about two times smaller than that across the water layer with the same thickness as the CBD (see Table 1). Using an MD simulation, which is not limited like EPR spin-label oximetry and can consider the CBD region where the Chol ring structures are located, the evaluated P_M is very low—about 10 times smaller than across the surrounding bulk membrane—supporting the hypothesis that the CBD is a barrier to oxygen transport into the eye lens [52]. Both measurements indicate that CBDs can form a significant barrier to oxygen transport into the cell (see Table 1).

It is commonly accepted that oxidation is a key feature that promotes cataract formation [52–59]. The principal mechanism that was “developed” during evolution to protect the lens against any type of oxidative damage is the maintenance of a very low oxygen concentration (oxygen partial pressure) within the lens through the entire human life [60–64]. The lens fiber cell membranes, with their unique lipid composition including extremally high Chol content, serve this purpose well. With saturating Chol content and the presence of CBDs, they form significant barriers to oxygen transport into the lens interior, helping to maintain low oxygen partial pressure inside the lens and protecting against cataract formation (see Refs. [65–67] for more discussion).

3.4. Effect of transmembrane α-helical peptides

Transmembrane peptides can, to certain degree, model the interaction of integral membrane proteins with the lipid bilayer. Here, we present data on the effects of transmembrane peptides on the oxygen transport within the lipid bilayer membrane and oxygen permeation across this membrane. We compared the effects of two transmembrane α-helical peptides, L₂₄ (Ac-K₂L₂₄K₂-amide) with a smooth hydrophobic surface [45] and (LA)₁₂ (Ac-K₂(LA)₁₂K₂-amide) with the increased roughness of its hydrophobic surface and also the increased motional freedom of its leucine side chains (as compared with L₂₄) [46]. Incorporating 10 mol% of smooth L₂₄ only slightly decreases the OTP close to the POPC bilayer surface, and its effect on the P_M is negligible. Incorporating 10 mol% (LA)₁₂ largely decreases the OTP across the entire POPC membrane and has a stronger effect than L₂₄. However, the resulting P_M is still greater than the P_W. We suggest that these differences are the result of different organizations of the peptide surface and different interactions of peptide side chains with PL acyl chains. We think that this conclusion is valid for integral membrane proteins containing transmembrane α-helical fragments.

3.5. OTP in PL membranes reconstituted with bacteriorhodopsin

The paper by Ashikawa et al. [68] forms a bridge connecting the results obtained for model membranes with those obtained for complex biological membranes. In that paper, Ashikawa et al. present the results of an investigation of the OTP in the lipid bilayer part of reconstituted membranes containing the bacteriorhodopsin (BR) in the DMPC bilayer and in native purple membranes from the Halobacterium halobium JW3. In reconstituted membranes, BR is in the monomeric form when the PL/BR molar ratio is 80/1 (80-rec membranes). When this ratio decreases to 40/1 (40-rec membranes), 25% of BRs are monomers and 75% are trimers and aggregates of trimers [68–70]. OTP measurements with PL-type spin labels indicate that the PL environment for 80-rec membranes is homogeneous, and that the OTP values measured at all membrane depths are smaller by ~1.6 times than those in the pure DMPC bilayer (see appropriate profiles in Fig. 3A). For 40-rec membranes, 2 domains are observed with distinct OTP profiles (see appropriate profiles in Fig. 3A). The OTP profile across one domain is identical to that observed for 80-rec membranes. We assigned it to the bulk-plus-boundary PL region of the membrane where BR is monomeric (as in the 80-rec membranes). OTP in the other domain is ~5 times smaller than that in the bulk-plus-boundary domain and close to that obtained for purple membranes (see appropriate profile in Fig. 3A). In that domain, the OTP (oxygen diffusion-concentration product) is greatly suppressed. We called this domain slow oxygen transport (SLOT) domain, or trapped lipid domain, which is likely formed due to the self-association of BR, and it consists of PLs in contact with two (or more) proteins and/or lipids in contact with protein and boundary lipids (see schematic drawing in Fig. 3B). Based on the OTP profiles across the investigated membranes, P_Ms were obtained for pure DMAP, 80-rec membranes (bulk-plus-boundary domain), 40-rec membranes (bulk-plus-boundary domain), 40-rec membranes (SLOT domain), and purple membranes as follows: 148.1, 88.6, 101.1, 34.7, and 21.3 cm/s, respectively (see Table 2).

Fig. 3.

a) Profiles of the oxygen transport parameter across DMPC membranes in the absence (○) and in the presence of bacteriorhodopsin (BR), with a DMPC/BR molar ratio of 80/1 (Δ), with a DMPC/BR molar ratio of 40/1 (▲,■), and across the purple membrane (□). When BR is in the monomeric form, only one bulk-plus-boundary lipid domain is present (Δ). When BR is aggregated, two lipid domains coexist: bulk-plus-boundary domain (▲) and trapped lipid domain (▲). In the purple membrane, only one trapped lipid domain exists (□). All data were taken at 26°C. Approximate locations of the nitroxide moieties of spin labels are indicated by arrows. b) Schematic drawing of the lateral organization of BR and DMPC molecules at a DMPC/BR ratio of 40/1. Coexisting lipid domains are indicated: bulk (no color), boundary (red), and SLOT (trapped lipid domain) (brown). Lipids in the SLOT domain are trapped between trimers and oligomers of trimers of BR. The shapes of molecules are drawn on the basis of electron microscopy studies [71].

Table 2.

Oxygen permeability coefficients P_M for DMPC membranes reconstituted with BR.

Membrane	Temperature (°C)	PM (cm/s)	Reference
DMPC	35	148,1	[68]
Bulk + boundary (80-rec)	35	88.6	[68]
Bulk + boundary (40-rec)	35	101.1	[68]
SLOT (trapped lipids)	35	34.7	[68]
Purple membrane (trapped lipids)	35	21.3	[68]
Water layer	35	80.8	[68]

Because the lipid exchange rate between the bulk and the boundary domains (10⁷ s⁻¹, [72,73]) is much greater than the inverse of the time window of the T₁-sensitive spin-label oximetry method (T₁⁻¹, 10⁵–10⁶ s⁻¹), these two domains were not discriminated in experiments presented in Ref. [68]. As a result, the OTP values were smaller as compared with those in the pure bulk domain. The lipid exchange rate between the SLOT and BULK domains (4.6×10⁴–7.7×10⁴ s⁻¹ [74]) is much slower than the inverse of T_1, which allows clear discrimination of these domains using our T₁ -sensitive spin-label oximetry approach.

4. Biological (cell plasma) membranes

Cell plasma membranes consist of a variety of lipids that form the lipid bilayer matrix, integral membrane proteins embedded into that lipid matrix, and peripheral proteins forming the cytoskeleton. The different lipid composition induces formation of different lipid environments within the bulk domain. They are raft domains and CBDs [75,76] (see schematic in Fig. 2). The presence of integral membrane proteins, which often form special structures, aggregates, and arrays induces formation of another class of lipid domains, boundary lipids, and trapped lipids [66,77–79]. All of these lipid domains, indicated schematically in Fig. 4, must have different profiles of the trans domain OTP and, thus, different P_Ms. Here, we will extend the analysis of factors that can determine resistance to oxygen permeation in complex biological membranes. Nonetheless, the analysis performed for simple, well-defined model membranes (Sect. 3) will aid our understanding of oxygen transport within and across domains and components of biological membranes.

Fig. 4.

Schematic drawing of the complex biological membranes. In this figure, they represent human eye lens fiber cell plasma membranes at different age of the lens. The composition of the eye lens membranes is depicted from younger membranes at the top to older membranes at the bottom. As membranes age, the Chol/PL ratio increases. In older membranes, the phospholipids are depleted, and protein content is increased. The protein content increases mainly through an increased adherence of peripheral proteins. Some cytoskeletal composition also changes. The different lipid domains expected in these membranes are indicated by various colors (see Footnote 1 in [80] for their detailed description). Cytoskeleton and peripheral proteins, which affect the organization of integral proteins into aggregates and arrays, are also indicated.

4.1. Bulk membrane lipids

It is difficult to obtain information about oxygen transport across the bulk lipids directly in intact membranes. The bulk domain cannot be discriminated by the T₁ -sensitive spin-label oximetry method because of the fast lipid exchange with the boundary domain [72,73]. Because of this lipid exchange, the OTP values measured at all membrane depths are smaller by ~1.6 times than those in pure bulk lipids [68]. This conclusion was based on measurements in simple 1-component lipid bilayer membranes reconstituted with 1 type of monomeric integral protein, BR (see Sect. 3.5). The lipid composition of the bulk domain in intact membranes can be very diverse, including different Chol contents that can induce coexisting phases and domains. The T₁ -sensitive spin-label oximetry method can discriminate the bulk-plus-boundary domain from the trapped lipid domain [68] and, according to the OTP profiles obtained separately with PL-type spin-labels, the P_Ms across these mixed domains. To better understand oxygen permeation across the bulk domain, we then investigated the properties of membranes made from the total lipids extracted from the intact membranes [33,34,50,81–84]. These model membranes made from eye lens lipids are called lens lipid membranes (LLMs) (see [67, 85] for more details). This investigation allowed us to discriminate pure CBDs for membranes with high Chol content and, using Chol analog spin labels ASL and CSL, evaluate the P_M value across the CBD (see Fig. 5 and the explanation in its legend). It should be mentioned that the P_Ms of the lipid bilayers surrounding CBDs, which were obtained with PL analog spin labels, were not contaminated by the presence of the pure CBDs because PL spin labels do not partition into the pure Chol domain. This is also true for the P_Ms of bulk-plus-boundary domains evaluated using the T₁ -sensitive spin-label oximetry method.

Fig. 5.

Profiles of the OTP across eye lens lipid membranes from human donors of different age groups obtained using EPR spin labeling. All profiles were obtained at 37°C. Profiles obtained with the PL-analog spin labels (filled symbols) were not contaminated by the presence of CBDs. Data obtained with Chol-analog spin labels (open symbols) are included in the OTP profiles. As indicated by the OTP profiles (data obtained with ASL and CSL), the CBDs are present in membranes of all age groups, indicating that the surrounding PL bilayer is always saturated with Chol.

Data for all age groups are reproduced from Fig. 4 of “Changes in the Properties and Organization of Human Lens Lipid Membranes Occurring with Age” by L. Mainali, M. Raguz, W. J. O’Brien, and W. K. Subczynski, in Current Eye Research, first published online: 28 Oct 2016, reprinted by permission of the publisher (Taylor &Francis Ltd, http://www.tandfonline.com). Copyright 2016.

A large part of our research has been focused on the oxygen permeation of the eye lens fiber cell plasma membranes, the most unique feature of which is their extremely high Chol content [39,42–44]. In human eye lenses, the Chol content is so high that pure CBDs are formed in lens fiber cell plasma membranes in human of all ages; this ensures that the surrounding PL bilayer is saturated with Chol [78–80]. The PL bilayers saturated with Chol are unique because their bulk properties (including profiles of the OTP and oxygen permeation) become consistent and independent of changes in PL composition [35,66,77,85]. This is illustrated clearly in Fig. 5, where all profiles are practically identical (having typical rectangular shapes) and are independent of the age of the donors and the location within the lens. The P_Ms evaluated for different PL bilayers saturated with Chol are close to each other. The values of P_Ms for the bulk-plus-boundary domain of intact membranes are listed in Table. 3.

Table 3.

Oxygen permeability coefficient P_Ms across the hydrocarbon region of domains in intact membranes in human lens membranes. Age groups are indicated in parentheses. Values were obtained at 37 °C, and data were taken from Ref. [70].

4.2. Boundary lipids

The boundary domain around integral membrane proteins was easily discriminated using the continuous wave EPR method [86,87]; OTPs within this domain and P_Ms across this domain were evaluated only as averaged with the bulk domain [68]. The lipid exchange between bulk and boundary domains is too fast [72,73] to allow separation of OTP and P_Ms for these domains. These values were smaller than those evaluated for membranes without integral proteins (see Table 2). Thus, the P_Ms of the boundary lipids should be smaller than those of the bulk lipids not affected by the membrane protein (evaluated for membranes made of the total lipid extracts). For example, the P_Ms obtained for clear eye lens fiber cell plasma membranes of human donors of different ages were significantly lower than the appropriate P_Ws [79]. Fiber cell plasma membranes are crowded with integral membrane proteins, the amount of which and their organization in aggregates and arrays increases with age (and are greater in nuclear than cortical membranes). These proteins induce formation of boundary lipids, which decreases measured OTP profiles across bulk-plus-boundary domains (see Fig. 6). Based on these profiles, we evaluated the P_Ms across the bulk-plus-boundary domains in membranes from all age groups. We found that the P_Ms are smaller by ~30% in nuclear than in cortical membranes. P_Ms across the bulk-plus-boundary domains in cortical membranes were smaller by ~20% and in nuclear membranes by 47% compared with the P_Ws (see Table 3). As shown in the OTP profiles across the bulk-plus-boundary domain (Fig. 6), the major resistance to oxygen permeation is located near the membrane surface, to the depth of C9. This resistance determines the final P_Ms independent of high OTP values in the membrane center. In principle, to evaluate the total intact membrane permeability for oxygen, such a separation is not needed, because finally oxygen permeation across all membrane components should be added (see Sect. 4.5).

Fig. 6.

Profiles of the OTP across domains in intact cortical and nuclear fiber cell plasma membranes. Profiles were obtained for samples from pools of ~20 clear lenses from donors of two age groups of human donors. All profiles were obtained at 37°C with the PL-analog spin labels, and none were contaminated by the presence of CBDs.

Modified and Reprinted from Fig. 7C of Experimental Eye Research, Vol 132, Raguz M, Mainali L, O’Brien WJ, Subczynski WK, “Lipid domains in intact fiber-cell plasma membranes isolated from cortical and nuclear regions of human eye lenses of donors from different age groups” pages 78–90, Copyright (2015), with permission from Elsevier

4.3. Trapped lipids

The model studies with BR (Sect. 3.5) help to explain the results for oxygen permeation across biological membranes crowded with integral membrane proteins. They showed that in protein-rich membranes, regions of the bilayer matrix can be formed in which the molecular oxygen diffusion is decreased to the level of gel-phase membranes [28,30,37]. In our first publications (which discussed reconstituted membranes), we called these regions SLOT domains [68,74]. For intact biological membranes, we used the term “trapped lipid domain” as it better describes the structure and dynamics of lipids in these membranes [66,78,79] (see Fig. 3B). It is significant to mention that the EPR spin-labeling methods applied to measure oxygen permeability across domains in intact biological membranes (including the trapped lipid domain) allows characterization of these domains in situ without their physical separation from intact membranes.

Over the last 15 years, we have focused on studies of fiber cell plasma membranes of human and animal eye lenses. These membranes are excellent for different membrane studies, including oxygen permeation. For the lens transparency it is significant that during maturation fiber cells los their integral organelles during maturation. Thus, the plasma membrane is the only membrane of the matured fiber cells [88–90]. Thus, the fiber cell plasma membrane, together with the cytoskeleton, form the only supramolecular structure of the mature fiber cell (see schematic in Fig. 4). Additionally, the lipid composition (including Chol content) drastically changes in these membranes with age [43,91–95]. These membranes are crowded with integral membrane proteins; with age comes increases in protein content and protein organization into domains, complexes, and arrays [90,96–100]. Therefore, all domains, the properties of which we investigated in detail (see Sect. 3)—including, bulk, boundary, and trapped lipid domains—as well as pure CBDs, are present in human eye lens fiber cell plasma membranes (see schematic in Fig. 4). The OTP profiles across these domains were obtained with PL analog spin labels (n-SASL), and the data obtained with Chol analog (ASL) for cortical and nuclear porcine [80,101–103] and human [78,79] lenses have been published and the appropriate P_Ms evaluated (see Table 3).

Like the P_Ms across bulk-plus-boundary domains in human lens membranes, the P_Ms across trapped lipid domains were evaluated based on the OTP profiles across these domains. These profiles are presented in Fig. 6 for fiber cell plasma membranes of human lenses from donors of different age groups. The calculated P_Ms values were always smaller (on average by 45%) in nuclear than in cortical membranes [79]. Additionally, the P_Ms evaluated for cortical and nuclear membranes were ~4.7 and ~8.5 times smaller, respectively, than the P_W values. The P_Ms across trapped lipid domain were even smaller than the P_Ms of gel-phase model membranes [29,31]. Interestingly, the values of P_Ms measured for corresponding domains did not change significantly with the age of the donor (see Table 3). We can conclude that the presence of trapped lipid domains formed in membranes crowded with integral membrane proteins is one of factors that potentially can form barriers to the free diffusion of oxygen across cell membranes.

4.4. Membrane proteins, integral and peripheral

The organization of lipids in plasma membranes into domains strongly depends on the organization of membrane integral proteins into arrays and aggregates [90,96–100,104], which, in turn, depends on the interaction with the peripheral proteins forming cytoskeleton [105–108] (see schematic in Fig. 4). As discussed above, the organization of integral membrane proteins into arrays and aggregates strongly affects oxygen permeation across the lipid bilayer portion of the membrane. The integral membrane proteins themselves are nearly impermeable to oxygen [109,110]; thus, the total P_M must be corrected considering the surface occupied by integral membrane proteins (see Sect. 4.5).

The effect of the cytoskeleton on the membrane oxygen permeation has not been fully investigated. Our preliminary results suggest that the intact cytoskeleton (and other urea soluble, peripheral membrane proteins) are responsible for the formation of the rigid lipid membrane environments, and, thus, contributes to the formation of a membrane barrier for oxygen permeation. Removal of the urea-soluble peripheral proteins eliminates SLOT domains in both cortical and nuclear membranes of porcine eye lenses. In nuclear urea treated membranes, the SLOT domains (with an OTP smaller than 0.38 μs⁻¹) that comprised about 18% of SR signal, and thus at least 18% of all spin probes in native membranes, were completely gone.

4.5. Total membrane permeability coefficient

We showed that the T₁-sensitive spin-label oximetry method allows evaluation of P_Ms across different coexisting domains in the lipid bilayer portion of biological membranes. Thus, the total P_M across the entire lipid bilayer portion of a biological membrane should consider all these particular P_Ms as their weighted sum. In the first approximation, the weight for each domain P_M is equal to the surface area occupied by the domain, divided by the total surface area occupied by the lipid bilayer portion of the membrane. We developed EPR spin-labeling methods allowing quantitative evaluation of relative amounts of PLs and Chol in membrane domains, which in turn determines surface area of the domain. The method, which is based on the analysis of continuous wave spectra of PL- and Chol-analog spin labels, allowed evaluation of the amounts of PL (% of total PLs) in bulk and trapped lipid domains and the amount of Chol (% of total Chol) in trapped lipid domains [79,102]. This method was applied for quantitative analysis of lipids in human [79] and porcine [102] cortical and nuclear eye lens fiber cell membranes. Another method that is based on the analysis of SR EPR signals of the PL- and Chol-analog spin labels allowed evaluation of the amounts of PLs and Chol in the trapped lipid domain [103]. This method was applied for quantification of lipids in porcine cortical and nuclear eye lens fiber cell membranes.

As we indicated in Sect. 4.4, integral membrane proteins are practically impermeable to oxygen [109,110]. We did not consider water channels across pores, aquaporins, and ion channels in integral proteins because existing data on how oxygen diffuses across these structures is very limited. The MD simulations suggest that water pores of AQP1 have a very low permeability to oxygen [111,112], and an AQP1 knockout mice study shows that the physiologically relevant conduction of oxygen is minimal for these membrane channels [113]. We want to also add that the surface occupied by these channels is negligible as compared with the total plasma membrane surface. Thus, the total effective oxygen permeability coefficient across the intact membrane should be corrected (decreased) considering the surface occupied by impermeable proteins, by multiplying the final P_M across the entire lipid bilayer portion of the biological membrane by the factor equal to the ratio of the surface area occupied by lipids to the surface area of the entire membrane. In mammalian plasma membranes, the typical protein to lipid ratios (wt/wt) are between 0.4 and 1.5. In protein-poor membranes (like myelin) this ratio is ~0.18, and in protein-rich membranes (like halobacterium purple membrane or mitochondrial inner membrane), this ratio can be as high as 3.0 [114]. In membranes loaded with integral membrane proteins, this correction can significantly affect (decrease) total membrane oxygen permeability.

5. Promising new approach in EPR spin-label oximetry: Measuring oxygen permeability across complex biological membranes

5.1. Stretched oxygen transport parameter

Recently, we applied a stretched exponential function (SEF) to analyze SR EPR signals coming from spin-labeled complex membranes consisting of an undefined number of domains (giving an undefined number of relaxation constants in SR experiments). In this new approach, the SR signals are fitted to the equation:

$$I\left(t\right)=I_o\text{exp}\left(-{\left(t{T}_{1\text{str}}^{-1}\right)}^{\beta }\right),$$ (3)

where I(t) is the signal amplitude at time (t), I_o is a normalized signal amplitude at time zero, T_1str⁻¹ is a characteristic or stretched spin-lattice relaxation rate, and β is the heterogeneity parameter that corresponds to T_1str⁻¹. Fitting parameters T_1str⁻¹ and β are used to construct continuous probability distributions of all possible rates (see schematic explanation in Fig. 1 of reference [109]). This distribution can be used to evaluate the probability of finding a given range of spin-lattice relaxation rates within the signal. The β parameter determines the shape of the distribution; hence, it is termed the heterogeneity parameter. For deoxygenated samples, where T_1str⁻¹ is determined mainly by the rotational diffusion of spin labels [115–119], we can obtain continuous distribution of T₁⁻¹s due to the rotational diffusion (T_1N2⁻¹), which is a measure of membrane fluidity. This application of SEF is described in our first paper [120]. In complex multicomponent biological membranes, the membrane fluidity (sensed by T₁s of lipid spin labels) most likely changes gradually, rather than abruptly, between membrane environments (domains). Therefore, a continuous distribution of relaxation rates associated with the membrane fluidity is beneficial for describing such systems.

In our second paper [121], we further developed the theory for application of SEF to analyze SR signals obtained in the presence of molecular oxygen. When SR signals are recorded in the presence of molecular oxygen, the Heisenberg exchange between the spin label and paramagnetic molecular oxygen, which occurs during bimolecular collisions, increases the observed spin lattice relaxation rates. In that case, two independent processes, namely the rotational diffusion of spin labels and oxygen collisions with spin labels, contribute to the spin lattice relaxation. Because of that, their contributions to spin lattice relaxation rates must be represented by two independent distributions, one determined by the rotational diffusion of spin labels and the other a distribution of relaxations induced strictly by collisions with molecular oxygen. As follows from the theory developed in [121], such SR signals can be fitted to the equation:

$$I\left(t\right)=I_o\text{exp}(-{\left(t{T}_{1\text{strN}2}^{-1}\right)}^{{\beta }_{\text{N}2}})\text{ exp}(-{\left(t{f}_{\text{air}}{W}_{\text{str}}\right)}^{{\beta }_{\text{w}}}),$$ (4)

where T_1strN2⁻¹ and β_N2 are the stretched exponential parameters obtained under nitrogen (for processes determined by the rotational diffusion of spin labels), and W_str and β_W are the stretched exponential parameters for oxygen collision with spin labels in air saturated samples, observed for samples equilibrated with air fraction f_air. It is theoretically justified that f_AirW_str values obtained from Eq. 4 are a linear function of the fraction of air (see Fig. 7 for representative example).

Fig. 7.

Representative data for the fitting parameters f_AirW_str and β_W obtained from air titration of 16-SASL in intact nuclear porcine eye lens membranes plotted versus the air fractions at which the SR signals were obtained. (Top) The black squares are the average f_AirW_str values obtained from fitting at least 3 SR signals using Eq. 4. The vertical bars represent the corresponding standard deviations. The dashed line is the linear fit of these data points with an intercept of −1.5 ×10⁻⁴± 0.022 μs⁻¹ and a slope of 0.78 ± 0.048 μs⁻¹. The values following the ± sign are standard errors of the parameters. The slope obtained from the linear fit of the points represents SOTP W_str. (Bottom) The black squares are the average values of β_Ws obtained from fitting at least 3 SR signals using Eq. 4. The vertical bars represent their standard deviations. The dashed line indicates that the measured values fall in relation to the mean.

These parameters also can be obtained by fitting the SR signals obtained under various air fractions to a general stretched exponential equation (Eq. 3). Plotting the observed parameters versus the air fractions, we found that the observed stretched-spin lattice relaxation rate, T_1strobs⁻¹, increases linearly with air fraction, with the slope close to the W_str obtained in Eq. 4. This allowed us, by analogy with the OTP [16], to introduce the stretched OTP rate parameter (SOTP) [121] as:

$$\text{SOTP}=W_{\text{str}}=T_{1\text{strAir}}^{-1}-T_{1\text{strN}2}^{-1},$$ (5)

where T_1strAir⁻¹ and T_1strN2⁻¹ are stretched spin-lattice relaxation rates obtained for membranes equilibrated with air and nitrogen, respectively. Strictly speaking, Eq. 5 assumes the linearity of the T_1strobs⁻¹ as a function of the fraction of air. This condition is fulfilled when the W_str (which describes the additional relaxation pathway for spin labels that depends on the collisions of oxygen with spin labels) is sufficiently larger than the T_1strN2⁻¹ (the pathway determined by the rotational diffusion of spin labels). This condition is fulfilled in the membrane environment. The heterogeneity parameter β_W that corresponds to W_str can be found by fitting the observed heterogeneity parameters β_obss to a weighted sum of the two heterogeneity parameters β_N2 and β_W (see Eq. 10 in Ref. [121]). Using the fitted parameters W_str and β_W obtained for the SR signal, the local (i.e., around the nitroxide moiety of spin label) probability density distribution of the OTP in membranes equilibrated with air can be constructed as described in [121]. As noted in Ref. [16], the profiles of the oxygen diffusion-concentration product across membranes also describe the profiles of membrane fluidity.

5.2. Global membrane oxygen permeability coefficient

Probability density distributions of the OTP can be constructed based on SR data obtained from different n-SASLs, for the special depths in the membrane at which nitroxide moieties of spin labels are located (see Fig. 8). Such a set of distributions allows the construction of a profile of distributions of OTPs across membranes equilibrated with air. This construction is shown in Fig. 8 for 5-, 9-, 12-, and 16-SASL for intact nuclear fiber cell membranes of porcine eye lenses. It is amazing that use of the SEF approach to analyze SR signals from spin-labeled membranes allows us to obtain profiles of the distributions of OTPs across all membrane domains probed by spin labels.

Fig. 8.

The distributions of OTPs in intact nuclear porcine membranes obtained with 5-, 9-, 12-, and 16-SASL are indicated by thin gray lines. The whiskers (cumulative probability distributions) extend from the 5^th to 95^th percentiles of all possible collision rates in a given sample. The boxes extend from the 25^th to 75^th percentiles, and the dashed line represents the 50^th percentiles of cumulative probability distributions. The OTP value in water is indicated by the dotted line, and the profile of the OTP across the LLMs formed from the total lipids extracted from intact nuclear membranes (●) (points taken from Ref.[101]).

We have published two basic papers [120,121]. A third paper (submitted) introduces a new way to present the probability density distributions of OTPs, indicating where appropriate percentiles of OTPs are located using box-and-whiskers plots. This approach is used in Fig. 8. These results are very promising and indicate a new approach in EPR spin-label oximetry. In Fig. 8, the dashed line indicates the value of the OTP 1.3 μs⁻¹ in water at the same conditions used to obtain the other data. It allows immediate evaluation of the percent of the SR signal (in the first approximation, this is also the percent of spin labels) coming from lipid environments with OTPs slower or faster than the OTP in water. This figure also indicates the profile of the OTP across LLMs made of the total lipids extracted from intact nuclear fiber cell membranes. Using this approach, we can evaluate the contribution of the changes induced in lipid bilayer matrix by integral membrane proteins on the OTP value at a certain depth in the membrane. As shown, the effect on all of the depths is large. We believe that, in the future, we can develop an approach that, based on profiles such as those shown in Fig. 8, will allow evaluation of the total membrane P_M for complex membranes without any assumptions about the presence and number of membrane domains or about their homogeneity. We call this P_M the global membrane P_M.

6. Discussion

As follows from Eq. 2, the oxygen concentration difference across the membrane or through the cell is determined by the P_M and the rate of the oxygen consumption inside the cell or cellular organelle (like mitochondria). While existing investigations have focused on metabolism as the driver of oxygen consumption, in FLASH the mechanism is consumption (often termed “depletion”) of oxygen by reactive intermediates generated by ionizing radiation. The driving force, of course, is the same regardless of the basis of the oxygen consumption. In Ref. [31], this problem is discussed for a suspended single Chinese hamster ovary cell and a single mitochondrion. Data presented in Table V of Ref. [31] indicate the contribution of the membrane to the oxygen concentration gradients of oxygen around the cell and mitochondrion. It follows that, for the Chinese hamster ovary cell (with its typical oxygen consumption rate), the oxygen concentration difference created across cell plasma membranes is of the order of 0.00024 μM (for this membrane the P_M/P_W ratio is 0.5). For the mitochondrion, because of the high oxygen consumption rate, this value is greater: 0.2 μM (assuming realistic P_M/P_W ratio of 0.01). We should remember that for fluid phase PL membranes (with small Chol contents), P_M ≈ P_W. For membranes containing 50 mol% Chol (and at gel phase membranes), P_M ≈ 0.1P_W. In protein rich membranes (purple membranes of Halobacterium halobium), the P_M ≈ 0.01P_W. We can expect that the P_M of mitochondrial inner membranes is as low as that of purple membranes, and, thus, the mitochondrial membranes can be a limiting barrier for oxygen consumption under hypoxic conditions in vivo. We direct readers to the detailed discussion of this problem in [32,122].

7. Concluding remarks

The material covered in this review demonstrates many of the different ways in which cell membranes cannot be considered as freely permeable to oxygen. A principal factor that determines membrane barrier properties is presence of integral proteins, which affects oxygen permeability in two ways: First, integral membrane proteins are nearly impermeable to oxygen, and the total P_M should be appropriately corrected (decreased) (see Sect. 4.5). This correction can significantly decrease the total P_M for membranes dense with integral proteins. Second, these proteins organize lipids around and between them. They induce formation of boundary lipids and trapped lipids, respectively. Even so, we were unable to evaluate the P_M across boundary lipids alone; we evaluated the P_M across the bulk-plus-boundary domain for different membranes and found that it is always smaller than the appropriate P_W. This permeation strongly depends on the type of the membrane and amount of integral proteins. Aggregates and arrays of integral proteins trap lipids between them with very slow exchange with surrounding bulk lipids. This domain, called the SLOT domain, has a P_M as low as that of gel-phase membranes. The presence of Chol, the content of which in plasma membranes is rather high, also decreases the total P_M. The effect of Chol is strong in membranes, where its high content induces formation of CBDs with oxygen permeability about 10 times smaller than that across the appropriate water layer. For example, we showed that, in the Chol- and protein-rich raft domains of the influenza viral membrane, oxygen transport decreased by a factor of 16 as compared with that in bulk lipids [74]. Therefore, it cannot be assumed that measurements of oxygen levels external to the cell cannot be relied upon to indicate the absolute values or even the dynamics of changes of levels of intracellular oxygen. This is a very important conclusion with regard to investigations of the role of oxygen in FLASH. As noted in the Introduction, for particular applications of FLASH, the simplest explanation of differential biological impacts on normal versus tumor tissues due to an oxygen effect should be rooted in differences in the effects of conventional irradiation and FLASH on changes in levels of oxygen availability in the nucleus over time periods relevant to competition between oxygen and other species on chemically unstable radiation-induced changes in DNA. The nucleus is the site of interest because acute cell death from ionizing irradiation occurs via damage to DNA. Therefore, to make direct measurements of depletion of oxygen relevant to FLASH, we must develop methods to measure oxygen at the nucleus. In addition, we need to be able to make these measurements with time resolutions of nanoseconds because the reactive intermediates of DNA that react with oxygen appear to have lifetimes of 10⁻⁵ or 10⁻⁶ seconds. These are very challenging requirements. It may be that we will, least in the short run, need to find less direct ways to determine whether FLASH (or any other process) has a biologically significant effect on local oxygen levels of interest. Changing dose rate and dose per fraction could provide some insights in this regard. However, the data summarized in this review clearly indicate that additional methods are needed to accurately assess the levels of oxygen at the sites of interest for radiation killing of cells.