Stimulation of ABCB4/MDR3 ATPase activity requires an intact phosphatidylcholine lipid

Abstract

ABCB4/MDR3 is located in the canalicular membrane of hepatocytes and translocates PC-lipids from the cytoplasmic to the extracellular leaflet. ABCB4 is an ATP-dependent transporter that reduces the harsh detergent effect of the bile salts by counteracting self-digestion. To do so, ABCB4 provides PC lipids for extraction into bile. PC lipids account for 40% of the entire pool of lipids in the canalicular membrane with an unknown distribution over both leaflets. Extracted PC lipids end up in so-called mixed micelles. Mixed micelles are composed of phospholipids, bile salts, and cholesterol. Ninety to ninety-five percent of the phospholipids are members of the PC family, but only a subset of mainly 16.0-18:1 PC and 16:0-18:2 PC variants are present. To elucidate whether ABCB4 is the key discriminator in this enrichment of specific PC lipids, we used in vitro studies to identify crucial determinants in substrate selection. We demonstrate that PC-lipid moieties alone are insufficient for stimulating ABCB4 ATPase activity, and that at least two acyl chains and the backbone itself are required for a productive interaction. The nature of the fatty acids, like length or saturation has a quantitative impact on the ATPase activity. Our data demonstrate a two-step enrichment and protective function of ABCB4 to mitigate the harsh detergent effect of the bile salts, because ABCB4 can translocate more than just the PC-lipid variants found in bile.

Bile salts are synthesized in the liver and translocated across the canalicular (apical) membrane. This membrane forms a meshwork between hepatocytes, which are connected via tight junctions, and several adjacent hepatocytes lead to the formation of the small bile duct (1–3). This area is exposed to relatively high bile salt concentrations, which can reach the millimolar range, and must therefore resist the detergent action of bile salts at these elevated levels (4). Protection is provided by the canalicular membrane itself, which contains medium- to long-chain fatty acids as building blocks of the phospholipids. The canalicular membrane also has a high cholesterol content, which increases the thickness of the bilayer (5). Additionally, cholesterol increases the packing density of fatty acids, through smoothing the kinks of unsaturated fatty acids, and therefore reduces packing defects (6). As a consequence, a more densely packed membrane decreases the probability for detergents to incorporate (7, 8). This obviously requires a strict regulation and continuous supply of the components. The latter function is mainly carried out by three hepatobiliary transporters. One of them, the bile salt export pump (BSEP, ABCB11), is an ABC transporter that translocates conjugated bile salts across the canalicular membrane (1–3). To reduce the detergent action of these bile acids, the other two ABC transporters translocate additional compounds to intercept free bile salts in so-called mixed micelles. They consist mainly of phospholipids, cholesterol, and the bile salts. A balanced ratio of these three components not only reduces the detergent action of bile salts, but also exports cholesterol for excretion without the formation of cholesterol crystals, so-called gallstones. Cholesterol is transported by ABCG5/G8 (9, 10), a heterodimer consisting of two half-size ABC transporters ABCG5 and ABCG8. ABCB4/MDR3 flops phospholipids, more precisely lipids of the PC family, from the cytosolic to the extracellular leaflet of the membrane. ABCB4 is a so-called full-size ABC exporter composed of two transmembrane domains and nucleotide binding domains localized on a single gene (11, 12). Historically, it was termed a multidrug resistance (MDR) efflux pump due to its high amino acid identity (76%) to P-gp/MDR1/ABCB1 (12). However, the physiological role of ABCB4 is clearly different (13–15).

The exact composition of the human canalicular membrane as well as the composition of the two leaflets is currently unknown. However, the overall composition of the rat canalicular membrane has been experimentally determined. It harbors 44% PC lipids, 25% SM, 22% PE lipids, and 8% PS lipids, respectively (16). The phospholipids are also mainly esterified with one saturated and one unsaturated fatty acid (6, 17). Looking at the PCs of the canalicular membrane and the bile in more detail, PC lipids account for 44% of the total lipids of the canalicular membrane (18, 19). In striking contrast, bile, or more precisely the mixed micelles, is composed of approximately 90–95% PC lipids. Especially, two variants, 16:0-18:1 and 16:0-18:2 PC, are drastically enriched (17). This phenomenon has not been sufficiently addressed so far.

A process that is often underestimated but might explain this enrichment is the extraction of phospholipids from the membrane. In this model, extraction of lipids out of the membrane occurs into or through a specific acceptor. In the case of ABCB4, the acceptor seems to be the mixed micelle itself. Specifically, for bile salts, there is evidence that they actively extract phospholipids from the canalicular membrane (20), and it is unlikely that an unknown protein mediates this “lipid extraction” process (21). Considering the detergent action of bile salts and the fact that PC lipids exclusively are solubilized from the extracellular leaflet, it is counterintuitive that bile salts discriminate between lipids of different acyl chain composition (22) and favorably extract lipids independently of their headgroup (16).

In such a scenario, ABCB4 translocates every PC-lipid variant presented to create a diverse pool of PC lipids in the outer leaflet. Bile salts subsequently extract or transfer, dependently or independently of ABCB4, only a subset of PC-lipid variants, namely, those PC-lipid variants that build up bile. Thus, ABCB4 protects the biliary tree from self-destruction by bile salts. Until now, however, it was unknown whether ABCB4 possesses a selectivity for certain PC variants, e.g., different fatty acid composition and/or degree of unsaturation, or whether it flops all PC lipids with similar efficiency.

Therefore, we demonstrated previously that the ATPase activity of ABCB4 is specifically modulated by PC lipids in vitro, reflecting its in vivo functionality (23). Here, we moved one step further and used an in vitro set-up to analyze the physiological function of ABCB4 (24) with respect to the determinants of substrate recognition. The detergent-solubilized and purified ABCB4 permits direct measurements of kinetic parameters, and it is feasible to conclude that ATPase stimulation by a certain substrate might be linked to its transport because most substrates do indeed stimulate the ATPase activity of, for example, ABCB1 (25). Thus, we employed total liver lipid extracts, synthetic lipid mixtures, and pure PC lipids as well as the building blocks of PC lipids to characterize the activity in more detail. This resulted in the understanding that both the lipid chain and the headgroup are recognized by ABCB4 to stimulate ATP hydrolysis as a prerequisite for its transport function.

MATERIAL AND METHODS

Transformation of Pichia pastoris and test expression of ABCB4

All steps were performed as described in (24).

Protein production

X33 Pichia pastoris cells carrying the pSGP18-2μ-MDR3 plasmid were grown on a selection YPD plate containing 200 μg/ml zeocin for 2 days. One liter precultures were grown in 2 liter baffled flasks for 18 h in MGY medium and used for inoculation of a 6 liter benchtop fermenter (I&L) containing minimal glycerol medium (26.7 ml/l phosphoric acid, 0.93 g/l calcium sulfate, 18.2 g/l potassium sulfate, 14.9 g/l magnesium sulfate heptahydrate, 4.13 g/l potassium hydroxide, and 40 g/l glycerol). After batch glycerol was consumed, cells were fed by glycerol for 5 h to increase the biomass. Protein expression was induced by a methanol feed for 24 h. Afterwards, the cells were harvested, washed with 50 mM Tris-HCl (pH 8.0), and repelleted. Usually, a cell mass reaching 1–1.3 kg was obtained, flash-frozen in liquid nitrogen, and stored at −80°C until further use.

Crude membrane vesicle preparation

All of the following purification steps were performed at 4°C as described in Kluth et al. (23). Usually, 100 g of cells were thawed on ice and crude membranes were prepared. Thereafter, cells were resuspended in lysis buffer [0.33 M sucrose, 100 mM 6-aminohexanoic acid,1 mM EGTA, 1 mM EDTA, 50 mM Tris-HCl (pH 8.0), and 75 mM NaCl] in a 1:5 ratio. Cells were ruptured at 2.7 kbar using a cell disruptor system (I&L). Differential centrifugal steps separated cell debris from crude membrane vesicles. Crude membrane vesicles were collected at 138,000 g and resuspended in membrane buffer [50 mM Tris (pH 8.0), 150 mM NaCl, 30% (v/v) glycerin, and protease inhibitor cocktail (Roche)]. Total protein concentration was determined with a Bradford assay following the standard protocol.

Solubilization

Crude membrane vesicles were diluted to 10 mg/ml total protein concentration and supplemented with 1% (w/v) Fos-choline-16 (FC-16; Anatrace). Protein solubilization was performed for 1 h at 18°C. Nonsolubilized parts were removed at 138,000 g for 1 h.

Tandem affinity purification

Imidazole (20 mM) was added to the solution of solubilized membranes to increase binding selectivity. The solution was loaded on a 5 ml HiTrap chelating column (GE Healthcare) loaded with Ni²⁺ ions, washed with washing buffer [50 mM Tris-HCl (pH 8.0), 50 mM NaCl, 30% glycerin (v/v); 0.0011% (w/v) FC-16, and 20 mM imidazole] to baseline level and eluted in one step with an elution buffer containing 50 mM Tris-HCl (pH 8.0), 50 mM NaCl, 30% (v/v) glycerin, 0.0011% (w/v) FC-16, and 200 mM imidazole. Fractions of the immobilized metal ion affinity chromatography with the highest absorbance at 280 nm were pooled and transferred onto a calmodulin affinity resin (GE Healthcare) and incubated overnight with gentle rotation at 4°C. Calmodulin affinity purification was conducted according to GE Healthcare’s standard protocol. Aliquots with the highest protein concentration were pooled and either directly used for ATPase activity measurements or flash-frozen in liquid nitrogen and stored at −80°C until further use.

Liposome preparation

Lipids were purchased from Avanti Polar Lipids and resuspended in chloroform. Chloroform was removed in a vacuum oven at 40°C with stepwise increasing vacuum. After 30 min at the lowest pressure levels, lipids were hydrated in suspension buffer [50 mM Tris-HCl (pH 8.0) and 150 mM NaCl). A sonication step for 5–10 min was needed to decrease the turbidity of the solution. Dilution series for ATPase measurements were prepared with suspension buffer.

Enzyme coupled ATPase assay

ATPase activity of detergent-purified ABCB4 was determined by an enzyme coupled assay. Measurements were performed at 37°C in a 96-well plate. A standard ATPase reaction volume was 200 μl in total with 50 mM Tris-HCl (pH 7.4) at 37°C, 5 mM MgCl₂, 4 mM phosphoenolpyruvate, 0.6 mM NADH (reduced), 13 units pyruvate kinase, 16 units lactate dehydrogenase, and 0.0011% (w/v) Fos-choline 16 and 35 μg/ml (239 nM) detergent-solubilized protein. Substrate stock solutions were prepared, and 10 μl of solution were added to the solution. The reaction was started by addition of 5 mM of ATP. The absorbance of NADH was followed for 60 min. Because NADH oxidation is stoichiometric to ATP hydrolysis, the ATPase activity of ABCB4 was expressed as nanomoles of ATP consumption per milligram of ABCB4 protein and minute. No influence on pyruvate kinase or lactate dehydrogenase performance due to compound or solvent concentrations was observed. A background measurement was subtracted from all data points. Thereafter, separate ATP hydrolysis measurements in which the ABCB4 cysteines in the Walker A motives were labeled with the fluorophore FL-maleimide-Bodipy were conducted. This steric hindrance inhibits ABCB4 specifically (23). Prior to the Bodipy labeling, both samples (labeled and not labeled) were reduced with tris(2-carboxyethyl) phosphine at 10 times the molarity of the protein for 1 min at 22°C. For the labeling reaction, a Bodipy amount 50 times the molarity of the protein was chosen and incubated for 20 min at 22°C. The unlabeled sample was treated in the same way, with the only exception that DMSO was used instead of the maleimide-fluorophore.

To account for batch-to-batch variations in the basal ATPase activity of ABCB4, we decided to present the data as normalized stimulation. To ensure reproducibility and comparability, each batch was analyzed for its basal ATPase activity and in the presence of the highest 1,2-dioleoyl-sn-glycero-3-phosphocholine [18:1-18:1 PC (DOPC)] concentration (250 μM). While we observed variations in the absolute activities, the fold of stimulation was always within the margin of 5.31 ± 0.58.

The physical state of the liposomes

Lipids may adopt different physical states in solution, where the molar ratio between lipids and detergents changes. This solution could contain pure detergent micelles or a mixed species of detergent/lipid micelles up to pure liposomes. To exclude any artifacts evoked by lipid to detergent interactions, the absorption of each ATPase reaction was determined at 492 nm to obtain insights into the physical state of putative liposome formation (26, 27). A molar ratio of the constant detergent concentration against the increasing lipid was calculated. The resulting curve resembles an inverse titration curve that one obtains in liposome reconstitution experiments comparable to (28).

We also performed a dynamic light scattering measurement with a selection of PC lipids to see whether any change in physical state of the PC-lipid variants is correlated to the fold of stimulation. Here, we used 250 μM of lipids diluted with the sample buffer and measured three times for 200 s at 37°C. The measurements were taken with a Malvern Zetasizer Nano ZS (Malvern Panalytical, Kassel, Germany) equipped with a He-Ne-Laser (wavelength of 633 nm) as a light source. The scattered light was detected with a scattering angle of 173° (backscattering).

TLC

We dripped 10 μl of each sample on a silica gel. The phospholipid standards had a concentration of 1 mg/ml. Total buffer as well as buffer-isolated components were dripped on the silica gel next to a sample of purified ABCB4. The running phase consists of 95 parts chloroform, 55 parts methanol, 5.5 parts ammonia, and 5.5 parts double distilled water. For detection, a primuline solution [0.1% (w/v) in 80% acetone and 20% water] was used. For fluorescence imaging, we excited with a short wavelength light and emitted at long wavelength on a ChemiGenius2 Bioimaging System from Syngene.

Data evaluation

For analysis, the background activity of ABCB4-Bodipy was directly subtracted from the data of ABCB4 in the absence of Bodipy (23). ATPase activities were normalized relative to the basal ATPase activity of ABCB4 (57.7 ± 20.3 nmol·min⁻¹·mg⁻¹) Kinetic parameters were analyzed with Prism 7 (GraphPad) and plotted using Michaelis Menten kinetics (equation 1) or a dose response fit (equation 2). Here, v_(c) represents the ATPase activity at a given compound concentration; v₀ is the basal ATPase activity in absence of any compound; v₁ is the maximal ATPase activity in the absence of inhibition; c defines the compound concentration; k₁ is defined as the compound concentration at half-maximal stimulating conditions (at half v₁); k₂ is the compound concentration at half-maximal inhibition of ATPase activity from the value v₁; v_max denotes the ATPase activity in the starting plateau; and v_min denotes the ATPase activity in the maximal reduction plateau.

$$v_{\left(c\right)}=\frac{v_1\cdot c}{k_1+c}+v_0$$ (Eq. 1)

$$v_{\left(c\right)}=v_{\mathit{min}}+\frac{v_{\mathit{max}}-v_{\mathit{min}}}{1+{10}^{\left\{\left[\log \left(k_2\right)-c\right]\right\}·\text{slope}}}$$ (Eq. 2)

RESULTS

Yield and purity of human ABCB4

A purification starting with 100 g wet cell weight pellet of X33 P. pastoris cells expressing wild-type ABCB4 as described in Materials and Methods typically yielded approximately 2–5 mg FC-16 solubilized protein. The homogeneity of the sample was analyzed by SDS-PAGE and Western blot (Fig. 1A). A dominant signal at approximately 135 kDa was clearly visible, which, according to Western blot analysis, corresponded to ABCB4 (calculated theoretical molecular mass including the two affinity tags: 146 kDa). The signals ranging from 45 to 110 kDa correspond to degradation products of ABCB4 as determined by mass spectrometry (23). To determine the specific ATPase activity of ABCB4, labeling with a Bodipy-FL-maleimide was performed. The modification was visualized by UV excitation (Fig. 1B) (24).

Fig. 1.

A: Human ABCB4 was purified by tandem affinity purification in FC-16 micelles. Panel “CBB” displays 2 μg of the elution fraction after TAP stained with Coomassie brilliant blue. Panel “C219” shows the same purification on a Western blot with immunostaining using the monoclonal antibody C219. B: The same purification was loaded onto a different SDS gel without (w/o B) or after covalent modification with the thiol reactive maleimide-Bodipy fluorophore (w/ B) and exposed to UV light. The arrow indicates the position of ABCB4.

ATPase-dependent characterization of PC-lipid variants

Previously, we demonstrated that ABCB4 was purified in a detergent-solubilized state in an active form. Importantly, ATPase activity could be stimulated with DOPC, a member of the PC-lipid family, or crude liver PC lipids, but not by phospholipids bearing other headgroups (23, 24). To analyze the stimulation in more detail, we chose different PC lipids to account for the natural variation of PC lipids with respect to length and degree of unsaturation. Thus, we plotted the ATPase activity expressed as percent of stimulation, where 100% represents the basal activity. Overall, the proteins’ basal activity was 57.7 ± 20.3 nmol·min⁻¹·mg⁻¹. Each activity was defined as 100% and all PC concentrations were displayed relative to their basal activity. For example, the k₁ of 16:0-16:0 PC on ABCB4 ATPase activity was 62.9 ± 11.5 μM and the v_{1_absolute} was 174.0 ± 6.5 nmol·min⁻¹·mg⁻¹. Here, v_{1_absolute} corresponds to the maximal ATPase activity calculated without a reference. While k₁ remains unchanged, the v₁ was calculated to 268.8 ± 10.5% (Fig. 2). Furthermore, we performed an experiment inverse to a typical liposome destabilization experiment. Here, we kept the detergent concentration constant while the lipid concentration excelled. The light scattering of liposomes at 492 nm changes due to variations in the lipid to detergent ratio used in the ATPase assay. This change in the physical state of lipids might have an impact on the stimulation of ABCB4 (28). Therefore, we measured and plotted the change of light scattering at 492 nm against an increasing lipid to detergent ratio (Fig. 3A). Figure 3B shows the slope of the curves. However, no correlation could be shown between the physical property of the PC lipids and the ATPase activity, like liposome formation or intermediate species like mixed detergent-lipid micelles. Additionally, a dynamic light scattering measurement also showed no correlation between the physical state of PC-liposomes and the stimulation capability of the PC lipid (supplemental Fig. S1).

Fig. 2.

Relative ATPase activity measurements of ABCB4 in the presence of the substrate 16:0-16:0 PC. One hundred percent is defined as ATPase activity of ABCB4 in the absence of any lipid (basal activity). Data represent the average of three independent experiments with the errors reported as SD.

Fig. 3.

A: Change in the physical state of lipids with an increasing lipid to detergent molar ratio. E. coli polar lipids treated with Triton X-100 (in gray) were taken from (28) and correspond to a typical titration curve for destabilization of liposomes with detergent. All PC-lipid variants were measured at 492 nm. B: Slopes of the graph in A where a change in the physical state of lipids can be visualized at 492 nm with an increased lipid to detergent molar ratio.

In the next step, the influence of liver total lipid extract (LTLE) from bovine liver on the activity of detergent-solubilized ABCB4 was analyzed. LTLE is composed of 42% PC lipids, 22% PE-lipids, 8% PI-lipids, 1% lyso-PI-lipids, 7% cholesterol, and 20% neutral unspecified lipids (see pie chart in Fig. 4A). The relative ATPase activity of ABCB4 in percent was plotted against increasing concentrations of LTLE in Fig. 4A. This lipid mixture stimulated ABCB4 with a v₁ of 315.2 ± 30.4% and a k₁ of 54.9 ± 26.1 μM. Additionally, we used a synthetic PC-lipid mixture (Fig. 4B), which resembles the composition of the fatty acid distribution of PC lipids in the canalicular membrane (29). Note that we have no direct information about the concrete composition of the human canalicular membrane. We took our information out of a rat canalicular membrane, for which only the overall fatty acid composition is known. Therefore, we concentrated on the nature of the fatty acids even if artificial lipids, such as double unsaturated lipids like 20:4-20:4 PC, had to be used. The composition was visualized as a pie chart and composed of 1.8% 14:0-14:0 PC, 20.6% 16:0-16:0 PC, 0.9% 16:1-16:1 PC, 25.2% 18:1-16:0 PC, 27.5% 18:1-18:1 PC, 19.5% 18:3-18:3 PC, and 4.4% 20:4-20:4 PC, respectively. Here, an even higher stimulation with a v₁ of 375.5 ± 13.6% and a k₁ of 26.9 ± 6.2 μM was evident. This suggested higher flop rates resulting in a higher turnover rate of ATP (Fig. 4B). A detailed analysis of single PC lipids is, of course, necessary to characterize a possible lipid preference of ABCB4. Therefore, we chose a subset of PC lipids covering representatives of this family that differed in their acyl chain length, their degree of unsaturation, and their natural occurrence. The main PC lipids of bile are 16:0-18:1 PC as well as 16:0-18:2 PC, and next to them, 16:0-18:0 PC and 16:0-20:4 PC are also present in bile (Fig. 4C). The chosen bile acids were constant in their sn1 position with 16:0, but varied in their sn2 position, highlighting the importance of the sn2 esterification. Their ATPase activities were in the same range as other PC-lipid variants with a 16-C fatty acid esterification at their sn1 position. The 16:0-18:1 PC stimulated ABCB4 with v₁ of 262.1 ± 12.5% and 16:0-18:2 PC with 236.8 ± 9.1%. Also, their affinity constant k₁s were comparable with 53.3 ± 14.3 μM and 44.7 ± 9.9 μM, respectively. Additionally, 16:0-18:0 PC showed similar ATPase activities of 260.8 ± 25.5% and with a k₁ of 51.2 ± 25.3 μM. The 16:0-20:4 PC variant was an outlier showing more similarities to the 20:4-20:4 PC variant, different from the other variants with a 16:0 fatty acid in the sn1 position. Figure 4D summarizes the influence of the acyl chain length with three different representatives. The 12:0-12:0 PC-variant, which does not occur naturally in the canalicular membrane, was chosen to analyze the biological significance of the observed modulation of ATPase activity. Indeed, this PC lipid was the one that had the least stimulating effect on ABCB4 ATPase activity with a k₁ of 47.2 ± 20.5 μM and a v₁ of 194.4 ± 12.6%. PC lipids with longer acyl chains, such as 14:0-14:0 or 16:0-16:0 PC lipid, activated the protein to a similar degree with a v₁ of 272.1 ± 19.0% for 14:0-14:0 PC and 268.8 ± 10.5% for 16:0-16:0 PC and k₁s of 110.5 ± 3 0.9 μM and 62.9 ± 11.5 μM, respectively (Fig. 4D). Figure 4E addresses the degree of unsaturation for PC lipids with one or four double bonds per acyl chain. The 16:1-16:1 (blue curve) and the 20:4-20:4 PC-variant (purple curve) possessed v₁ values of 272.1 ± 49.0% and 186.1 ± 14.8%, respectively, indicating a preference for a smaller degree of unsaturation. Even if the fatty acid 18:3 is highly abundant (17% of total fatty acids) in the canalicular membrane of hepatocytes [content summarized as a pie chart (Fig. 4B)], we were only able to test the double unsaturated version 18:3-18:3 PC, which is not naturally part of the canalicular membrane. Together with the 18:1 (24% abundance) and the 16:0 (18% abundance) PC lipid, they make up 59% of the fatty acid composition of the PC lipids of the canalicular membrane (29). We determined the highest stimulation for 18:1-18:1 PC (DOPC, v₁ = 564.2 ± 19.7%, k₁ = 37.4 ± 5.3 μM, Fig. 4F). The second highest stimulation of pure PC lipids was observed for the 18:3-18:3 PC-lipid variant with 332.8 ± 31.9% (67.4 ± 32.3 μM). All the kinetic parameters of the ATPase activities of ABCB4 determined for the different PC lipids are summarized in Table 1.

Fig. 4.

Relative ATPase activities of detergent-solubilized and purified ABCB4 in the presence of different PC lipids with increasing concentrations. A: Activity in the presence of LTLE. The composition of LTLE of bovine liver is shown as a pie chart in the inset. B: Activity in the presence of a synthetic PC composition that resembles the natural PC composition of the bile canalicular membrane (shown as pie chart in the inset). C: The influence of the main bile PC lipids 16:0-18:1 PC (POPC, turquoise) and 16:0-18:2 PC (mocha) on the ATPase activity as well as 16:0-18:2 PC in brown and 16:0-20:4 PC in pink. D: ATPase activity in the presence of PC lipids with increasing chain length: 12:0-12:0 PC (orange), 14:0-14:0 PC (green), and 16:0-16:0 PC (red). E: ATPase activity in the presence of PC lipids with an increasing degree of unsaturation: 16:1-16:1 PC (blue) and 20:4-20:4 PC (purple). F: ATPase activity in the presence of symmetric PC lipids: 18:3-18:3 PC (black) and 18:1-18:1 PC (gray). All data points were analyzed using equation 1 and represent the average of three independent experiments with the errors reported as SD (18:1-18:1 PC; n = 6). Each kinetic is the result of approximately one independent protein purification.

TABLE 1.

Summary of the kinetic parameter of the ATPase activity of ABCB4 in the presence of different PC lipids varying in acyl chain length and degree of unsaturation

The transition temperature was taken from (42). The ratio in the canalicular membrane was calculated according to (29). MM, molecular mass; Tt, transition temperature; k₁, concentration at half-maximal stimulation; v₁, maximal stimulation; Mod., nature of modulation on ABCB4, n.d., not determined.

The influence of single PC lipids moieties on the ATPase activity of ABCB4

The ATPase activity of ABCB4 supplies the energy to mediate the flop of PC lipids from the inner to the outer leaflet. Because ABCB4 displays a clear specificity for PC lipids in vivo (15, 30, 31) and in vitro [(23) and this study], it would be plausible to assume that the isolated choline headgroup is capable of stimulating the ATPase activity as well. As a control, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE), an example for a PE-lipid (16:0-18:1 PE) was tested (Fig. 5A). Here, no significant modulation of activity was observed, as was already described (23). In clear contrast, 16:0-18:1 PC stimulated the ATPase activity of ABCB4 with a v₁ of 262.1 ± 12.5% and a k₁ of 53.3 ± 14.3 μM (Fig. 5A). Figure 5B shows the influence of a lyso-PC lipid, 16:0, which lacks one acyl chain. Surprisingly, a large shift in modulation was observed. Now, an inhibition became apparent with a k₂ of 103.9 ± 7.7 μM (note that a log₁₀ representation was used). This value is distinct from the detergent’s critical micelle concentration (cmc) value of 4–8.3 μM and, thus, indicates molecular interactions and not simple solvent effects (32). However, we cannot directly exclude that the presence of lyso-PC caused a removal of annular lipids or leads to a change in the solvation of ABCB4 by changing the detergent micelle. However, we have indications from TLC (supplemental Fig. S2) that no annular lipids are copurified with ABCB4, most likely due to extensive washing with detergent during immobilized metal ion affinity chromatography and calmodulin affinity purification (33). The detection limit of primuline staining was 0.1 mmol. Assuming at least a one lipid to one ABCB4 transporter ratio, we would expect to detect lipids if present because the sample size of ABCB4 on the TLC was 3.4 mmol. Detergents, which share the PC headgroup but lack the core glycerol unit, are Fos-cholines such as FC-16 (Fig. 5E). FC-16 also inhibited the ATPase activity at micromolar ranges with a k₂ of 41.9 ± 11.3 μM, which corresponded to approximately three-times the cmc. It is important to emphasize that FC-16 is the detergent of choice for ABCB4 solubilization and purification. However, all ATPase activity measurements were performed at concentrations of twice the cmc (26 μM). This concentration was present in all other experiments so that the actual concentration of FC-16 was approximately 5-fold above the cmc. An inhibition was determined only under these conditions. We analyzed the interaction of FC-16 in more detail in Fig. 6. Here, we performed FC-16 titration experiments against ascending POPC concentrations. An additional 25 μM (on top of the double cmc of FC-16 which is always present) are tolerable for the transporter. When the concentration is increased near to the k₂ value of approximately 50 μM, the k₁ as well as the v₁ values are dampened. At concentrations of 100 μM, the transporter is inactive until POPC concentration exceeds 100 μM, here POPC is able to “rescue” the transporters’ ATPase activity. But at 500 μM of FC16, the transporter remains inactive due to unspecific detergent effects. To gain further insights into compounds that might directly interact with the transporter, L-α-glycero-phosphocholine (Lαgpc) (lacking both acyl chains) (Fig. 5C) or phosphocholine (ppc) (lacking both acyl chains and the glycerol backbone) (Fig. 5C) and choline itself, which lacks even the phosphate group (Fig. 5C), were analyzed. All three, Lαgpc, ppc, and choline, did not interfere with the ATPase activity of ABCB4 and are likely not recognized by ABCB4. Table 2 summarizes the kinetic parameters of the ATPase activity of ABCB4 in the presence of choline, ppc, Lαgpc, FC-16, lyso-16:0 PC, and 16:0-16:0 PC.

Fig. 5.

Relative ATPase activity of detergent-solubilized and purified ABCB4 in the presence of different building blocks ranging from choline to a PC lipid, POPC (16:0-18:1 PC). 16:0-18:1 PC (black) and 16:0-18:1 PE (POPE, moss green) (A); lyso-16:0 PC (B); Lαgpc (blue), ppc (red), and choline (green) (C). D: Chemical structures of the above-mentioned compounds. Choline moieties are shown in green, phosphate in red, the glycerin backbone in blue, fatty acids in orange and black, and alkyl chains in gray. E: The zwitterionic detergent FC-16 is depicted in gray. Data points represent the average of two independent experiments with the errors reported as SD (16:0-18:1 PC; n = 3). Each kinetic is the result of approximately one independent protein purification.

Fig. 6.

A titration experiment with the concentrations of POPC (16:0-18:1 PC) ascending from 0 to 1,000 μM with four different but constant FC-16 concentrations and an exclusive kinetic without detergent. Each kinetic is plotted as Michaelis-Menten kinetics. The MM fit for orange kinetic of 100 μM FC-16 had to ignore the first three data points in order to be fitted accordingly. Data points that are not used for the MM fit of the 100 μM FC-16 are connected via a dashed line. Each kinetic is the result of approximately one independent protein purification.

TABLE 2.

Summary of the kinetic parameter of the ATPase activity of ABCB4 in the presence of different building blocks ranging from choline to PC lipid

Name	MM (Da)	k1 (%)	k2 (μM)	v1 (%)	Mod.
LPC	495	—	103.9 ± 7.7	—	Inhibition
FC-16	407	—	41.9 ± 11.3	—	Inhibition
16:0-18:1 PC	760	51.3 ± 25.2	—	260.8 ± 25.5	Stimulation

LPC, Lyso-16:0 PC; FC-16, Fos-choline 16; 16:0-18:1 PC, 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine. The PC proportion of the canalicular membrane was calculated according to (29). LPC, Lyso-16:0 PC; FC-16, Fos-choline 16; 16:0-18:1 PC, 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine; MW, molecular mass; k₁, concentration at half-maximal stimulation; k₂, concentration at half-maximal inhibition; v₁, maximal stimulation; Mod., nature of modulation on ABCB4.

DISCUSSION

A two-step enrichment of the PC-lipid variants of the canalicular membrane: ABCB4 protects the biliary tree through lipid preferences

Because ABCB4 is confronted with unknown concentrations of substrate in in vivo or in reconstituted systems, the regulation of the effective concentration might be challenging. Therefore, we chose to use an in vitro environment of a detergent solution, as fine-tuning of composition and concentration of various compounds is possible. ABCB4 was stimulated by a lipid mixture derived from bovine liver (Fig. 4A). This provided us with a starting point for the analysis of specificity of the protein toward the substrate(s) (Fig. 4A). We used the two most abundant phospholipid families, PC and PE, to support this fact. In the case of POPE, no stimulation of the ATPase activity of ABCB4 was observed (Fig. 5A). In clear contrast, 16:0-18:1 PC stimulated the activity of ABCB4. This again highlighted the specificity of ABCB4 for PC lipids. Kluth et al. (23) provided first insights for this stimulation of the ATPase activity with PC lipids in contrast to other abundant phospholipids of the canalicular membrane. Even though an ATPase activity assay does not directly measure substrate translocation, it is now generally accepted that most substrates indeed increase the proteins’ ATPase turnover numbers, because the nucleotide binding domains change conformation and are temporally in close proximity (34). In this study, the phospholipids SM, PE, and PS were analyzed. None of them stimulated ABCB4 above background. The stimulation we observed for LTLE (v₁ 315.2 ± 30.4%) was larger than the stimulation of POPC, suggesting that the nature of the fatty acid had a significant influence on the magnitude of stimulation. Thus, we mimicked the fatty acid distribution of the PC composition of the bile canalicular membrane according to Northfield and colleagues (29). Here, an even higher stimulation was observed (v₁ 375.5 ± 13.6%). The sum of all maximal stimulations of the individual lipids of the mixture multiplied with the amount present in the mixture results in a theoretical stimulation of 357%, strongly suggesting that the effects of PC lipids on the ATPase activity of ABCB4 are additive.

The canalicular membrane contains 40% PC lipids that vary in the nature of their fatty acids. While the canalicular membrane harbors PC-lipid variants of every combination of saturated and unsaturated fatty acids of the categories 16:0 (18%), 18:0 (13%), 18:1 (24%), 18:2 (13%), 18:3 (18%), and 20:4 (4%) (29), bile PC-lipid content differs drastically.

In order to understand the two-step enrichment of specifically 16:0-18:1 and 16:0-18:2 PC lipids specifically, which end up in bile, ABCB4 (35) and bile salts perform a crucial interplay (Fig. 7). First, the PC-lipid content increases from 40% in the canalicular membrane to 95% in bile. Second, the apparent PC-lipid variants change. For example, ABCB4 creates the asymmetry in the canalicular membrane resulting in a canalicular membrane of mainly PC lipids in the outer leaflet and PS-lipids in the inner leaflet (Fig. 7CI). Here, ABCB4 flops PC lipids from the inner to the outer leaflet and ATP8B1 predominantly PS-lipids from the outer to the inner leaflet of the canalicular membrane. Confirmatively, we could show that ABCB4 was indeed stimulated by the analyzed PC-lipid variants, indicating that ABCB4 can translocate a broad spectrum of PC lipids. The high abundance of PC lipids in the outer leaflet may be directly related to the amount found in bile (approximately 90–95%) and answers the first step of PC enrichment. The level of stimulation may go back to a dual recognition of ABCB4 itself or even different accessibilities due to differences in the tightness of packing through fatty acids with a certain level of unsaturation. The second enrichment of 16:0-18:1 PC and 16:0-18:2 PC lipids, specifically, cannot be explained on the basis of the stimulation of ATPase activity of ABCB4, because these PC-lipid variants did not stimulate ABCB4 significantly more strongly than other PC lipids (Table 1, Fig. 4C). Rather, bile salts tend to extract bile-PC lipids out of the canalicular membrane independently of the headgroup but related to the nature of fatty acids (22, 36). Accordingly, bile salts extract mainly phospholipids with 16:0-18:1 and 16:0-18:2 fatty acid chains (22). Because this pool in the outer leaflet of the canalicular membrane consists mainly of PC lipids created by ABCB4, 16:0-18:1 PC and 16:0-18:2 PC are extracted specifically (Fig. 7CII). Consequently, this would protect the biliary tree of lipids from self-digestion (Fig. 7). This selectivity from the site of bile salts is also manifested by the resistance of membranes of SM/cholesterol-rich membranes against the detergent actions of bile salts (37). Following this hypothesis, 16:0-18:1 PC and 16:0-18:2 PC would be completely extracted in ABCB4-deficient systems because no continuous replenishment of the extracted PC-lipid variants could occur. Analysis of the phospholipid content of bile in mdr2^−/− knockout mice showed, indeed, a full absence of PC lipids in bile (30), but the lipid composition of the canalicular membrane is still insufficiently studied to validate this hypothesis further.

Fig. 7.

Two-step enrichment of PC-lipid variants into the mixed micelles. A schematic view on mixed micelles (A) with a ratio of 14.2% other PC-lipid variants, 49.7% 16:0-18:2 PC, 26.4% 16:0-18:1 PC, 4.4% PE, 1.2% PS, 1.4% PI, and 2.7% SM (37) (B). C: Schematic view on the floppase ABCB4 [structure taken from (43)] actively generating a high abundance of PC lipids in the outer leaflet of the canalicular membrane (I). The second step of PC enrichment (II) may be performed by headgroup-independent bile salt selection. D: The canalicular membrane consists of 29.6% other PC-lipid variants, 4.1% 16:0-18:2 PC, 2.9% 16:0-18:1 PC, 24.5% PE, 11.5% PS, 4.5% PI, and 22.8% SM (37).

Fatty acids matter, both sn positions matter

We analyzed the fatty acids with respect to three categories: i) acyl chain length, ii) degree of unsaturation, and iii) natural occurrence. The data summarized in Fig. 4D demonstrate that the stimulation of ATPase activity increased with increasing chain length until a peak was reached for a chain length of 16 carbon atoms. The degree of unsaturation of the acyl chains was investigated in Fig. 4E and F. Here, no clear trend was evident, although the 18:3-18:3 PC lipid was more active than the 16:1-16:1 PC lipid and the 20:4-20:4 PC lipid, but 18:1-18:1 PC had clearly the highest degree of stimulation. Whether unsaturation or chain length had a more pronounced impact cannot finally be answered, because both parameters were changed at the same time. However, a trend manifested, in which the 18 carbon atoms tend to stimulate ABCB4 more than others (Fig. 4F). Table 1 illustrates that the tested PC-lipid variants with 16 carbon atoms at the sn1 position activate ABCB4 to the same extent with the exception of 16:0-20:4 PC, which is highly unsaturated at the sn2 position; therefore, the sn2 position may not be as crucial as the sn1 position but should not be underestimated. C18 variants of the sn1 position especially emerge out of the pool of PC-lipid variants. One plausible explanation for 18:1-18:1 PC activating ABCB4 most efficiently might be increasing packaging defects due to the introduction of the optimal amount of kinks in the PC lipids. DOPC has two double bonds, hence two introduced kinks that might make the headgroup more accessible for the uptake process performed by ABCB4. In category three, we intended to analyze the natural occurrence of PC lipids in bile. Because 16:0-18:1 PC and 16:0-18:2 PC are known as the most prominent bile PCs, ABCB4 might be the discriminating factor and the molecular factor responsible for this enrichment. Unnatural PC lipids as well as PC-lipid variants exclusively found in the canalicular membrane had higher potencies in ATPase stimulation of ABCB4. The 18:1-18:1 PC lipid excelled the bile PC-lipid variants with more than twice the maximal stimulation v₁.

Interestingly as the modulation of the 16:0-16:0 PC lipid was identical to 16:0-18:1 and the 18:1-18:1 PC lipid again was the most potent PC lipid, it might suggest a pivotal role for the sn1 position. Further support of this notion comes from the results of lyso-16:0 PC. Lyso-16:0 PC as well as FC-16 were the only compounds that resulted in inhibition of the ATPase activity of ABCB4. This might be related to either the detergent properties of lyso-16:0 PC and FC-16 or the lack of the fatty acid of the sn1 position (lyso-PC).

The lipid headgroup dictates whether transport occurs, the fatty acids regulate the level of ATPase stimulation

The beneficial effect of lipids on the ATPase activity of ABC transporters such ABCB1 is generally accepted (38). In the case of human ABCB1 lipids, were used for stabilizing the protein and to diminish the detrimental effect of the detergent of choice, n-dodecyl β-D-maltoside (39). ABCB1 purified in detergent micelles often required higher drug concentrations than the reconstituted protein to obtain similar effects on ATPase activity. Although, ABCB4 and ABCB1 are highly homologous, ABCB4 is highly specific for PC lipids (12, 15). Therefore, next to any stabilization effects of lipids, it translocates lipids of the PC family per se. However, a stabilizing effect of phospholipids on detergent-solubilized and purified ABCB4 might also modulate the activity independent of the family of phospholipids. To exclude any changes in the ATPase activity due to the change in the physical state of lipid to detergent ratios, we demonstrated that even if liposomes or liposome-like structures behave differently, with increasing detergent concentrations, no correlation to stimulation was observed (Fig. 3A, B; supplemental Fig. S1). However, we like to emphasize that we cannot distinguish based on the observed stimulation of ATPase activity whether these lipids are indeed substrates of ABCB4 or whether they solely act as activators of ATPase activity, which is a prerequisite of substrate translocation of this primary active membrane transporter. Additionally, effects of annular lipids copurified and still attached may also play a role in the proteins’ activity and even result in a change of the basal activity. However, no copurified lipids could be detected on a TLC (supplemental Fig. S2) and are more likely washed off during the washing steps of the immobilized metal ion chromatography (33). We could previously demonstrate that ABCB4 is specifically stimulated in vitro by lipids of the PC-family (23). This is in line with in vivo studies that only lipids of the PC family are translocated by ABCB4 (15, 30). Both studies imply that the recognition of the substrate occurs within the headgroup, because other phospholipid families share the glycerol core and fatty acids. Using our in vitro system, which displays a PC-lipid-specific stimulation of ATPase activity as read out, we analyzed the ATPase activity of ABCB4 in the presence of the PC headgroup (Fig. 5C). Surprisingly, no stimulation was observed. The same holds true for choline or Lαgpc (Fig. 5C). However, we cannot exclude that the highly water-soluble choline, ppc, and Lαgpc may require greater than sub-millimolar concentrations to show any potential effects on the ATPase activity of ABCB4. Nevertheless, this of course pointed toward a pivotal role of the fatty acids. Interestingly, one fatty acid resulted in inhibition of ATPase activity (Fig. 5B), namely lyso-16:0 PC, a detergent, inhibited the ATPase activity with a k₂ value of 103.9 ± 7.7 μM (Fig. 5B, Table 2).The analysis of the nature of inhibition for FC-16 provided no indication for a competitive inhibition, rather, a negative detergent effect is responsible for the inhibition of ABCB4 (Fig. 6, Table 3). Therefore, only a full choline headgroup, a glycerol core, and at least one fatty acid as found in lyso-16:0-PC need to be present to be recognized by the binding site of ABCB4, and only in the presence of two fatty acids was a stimulation of the ATPase activity of ABCB4 observed. (Fig. 5A). For example, stimulation occurred with 16:0-16:0 PC but was rather low in the context of other ABC transporters, as it was only 268.8 ± 10.5% (174.0 ± 6.5 nmol·min⁻¹·mg⁻¹) in comparison, for example, to chABCB1, which was stimulated up to 3 μmol·min⁻¹·mg⁻¹ (40). The low activity of ABCB4 is however in line with studies from Ishigami et al. (41). The authors pointed out that ABCB4 needs to possess a low activity, as the large pool of surrounding substrates would result in a waste of energy and a loss of membrane integrity if too many substrates are translocated (41).

TABLE 3.

Summary of the kinetic parameters of Fig. 6

Name	k1 (%)	v1 (%)	ATPase Activity Without POPCa
POPC with 0 μM FC-16	36.9 ± 8.3	234.7 ± 7.4	100
POPC with 25 μM FC-16	49.8 ± 9.5	249.9 ±± 9	74.5
POPC with 50 μM FC-16	542 ± 385.5	181.1 ± 36.2	49
POPC with 100 μM FC-16	376.7 ± 209.3	329.9 ± 24	0
POPC with 500 μM FC-16	n.d.	n.d.	0

A FC-16 titration against ascending POPC concentration. k₁, concentration at half-maximal stimulation, v₁, maximal stimulation; n.d., not determined.

^aATPase activity without FC-16 being the Y-intercept of the curves.

With the knowledge that neither the headgroup nor the glycerol backbone was capable of stimulating ABCB4 accordingly, we provided evidence for the requirement that both features of a PC lipid, the head and the tail of the lipid, need to be present to activate ABCB4. Head or tail alone are not sufficient, as PE with the same glycerol backbone core and two acyl chains did not stimulate ABCB4 (Fig. 5A).

CONCLUSIONS

Every PC-lipid variant tested was able to stimulate the ATPase activity of ABCB4. ABCB4 seemed to prefer PC lipids with 18 carbon units at the sn1 position, while the building blocks of PC lipids alone are not sufficient to stimulate ABCB4. The differences in the level of ATPase stimulation by PC lipids regarding their fatty acid chain length and the degree of unsaturation provides an explanation for the protection of the canalicular membrane from the harsh detergent activity of bile salts. However, and equally important, only PC lipids are capable of stimulation of the ATPase activity of ABCB4 originating from the headgroup and the presence of two fatty acid chains.

Data availability

All data are contained within the article.