Report Opsin Is a Phospholipid Flippase

Department of Internal Medicine, Yale University School ofMedicine, New Haven, CT 06520, USASummaryPolar lipids must flip-flop rapidly across biologicalmembranes to sustain cellular life [1, 2], but flipping isenergetically costly [3] and its intrinsic rate is low. To over-come this problem, cells have membrane proteins that func-tion as lipid transporters (flippases) to accelerate flipping toa physiologically relevant rate. Flippases that operate at theplasma membrane of eukaryotes, coupling ATP hydrolysistounidirectionallipidflipping,havebeendefinedatamolec-ular level [2]. On the other hand, ATP-independent bidirec-tional flippases that translocate lipids in biogenic compart-ments, e.g., the endoplasmic reticulum, and specializedmembranes, e.g., photoreceptor discs [4, 5], have not beenidentified even though their activity has been recognizedfor more than 30 years [1]. Here, we demonstrate that opsinis the ATP-independent phospholipid flippase of photore-ceptordiscs.Weshowthatreconstitutionofopsinintolargeunilamellar vesicles promotes rapid (t < 10 s) flipping ofphospholipid probes across the vesicle membrane. This isthe first molecular identification of an ATP-independentphospholipid flippase in any system. It reveals an unex-pected activity for opsin and, in conjunction with recentlyavailable structural information on this G protein-coupledreceptor [6, 7], significantly advances our understanding ofthe mechanism of ATP-independent lipid flip-flop.Results and DiscussionPrevious work showed that phospholipids flip-flop rapidlyacross bovine photoreceptor disc membranes by an ATP-independentmechanism[4,5].Toidentifytheflippaseinques-tion, we used a biochemical reconstitution approach anda fluorescence assay for flippase activity [8](Figure 1A). Largeunilamellar proteoliposomes (Figure 1B) were generatedfrom mixed micelles of Triton X-100-solubilized disc mem-brane proteins (Triton extract [TE]), egg phosphatidylcholine,and a trace amount of nitrobenzoxadiazole-phospholipid(NBD-PL), a fluorescent reporter lipid whose fluorescence iseliminated on reduction with dithionite. Vesicles generatedby this method are sealed because they can retain trappedsmall molecules (Supplemental Information, available online)[9]. The liposomes were prepared with a protein/phospholipidratio in the range w0–2 mg/mmol, corresponding to wzero toten–moleculesofrhodopsin(themajorproteinofTE[Figure1F,inset]) per vesicle on average. Protease protection and colli-sional quenching assays indicated that both rhodopsin andNBD-PLs were reconstituted symmetrically (data not shown).Toestablishassayconditions,wefirstanalyzedprotein-freeliposomes. The fluorescence of these vesicles decreasedrapidly by w50% on adding dithionite (Figure 1C, trace L).Subsequent disruption of the vesicle membrane eliminatedfluorescence entirely (data not shown). Thus NBD-PLs in theouter leaflet of protein-free liposomes are rapidly reduced bydithionite, whereas NBD-PLs in the inner leaflet are protected.When proteoliposomes (protein/phospholipid ratio w1 mg/mmol) were treated with dithionite, fluorescence alsodecreased rapidly (Figure 1C, trace P). However, the extentof reduction was greater (w75%) than for protein-free lipo-somes indicating that in at least a fraction of the vesiclesNBD-PLsarerapidlytranslocatedfromtheinnertothedithion-ite-accessible outer leaflet. We confirmed these results byback-extraction experiments [8, 10] in which we used fattyacid-free serum albumin to extract NBD-PLs from the outerleaflet of the vesicles. Albumin extracted a larger fraction ofNBD-PLs from proteoliposomes than protein-free liposomes,consistentwithtranslocationofNBD-PLsfromtheinnerleafletto the outer leaflet of the protein-containing vesicles (Fig-ure 1D, Supplemental Information, and Figures S1A and S1B).The reconstituted activity was ATP-independent. A varietyof NBD-PLs were flipped, including N-NBD-di-DHA-PE, aphospholipid with polyunsaturated C22:6 acyl chains resem-bling those found in disc phospholipids (Figure 1E). Inclusionof 10 mole percent cholesterol, the average cholesterol con-tent of disc membranes [11], had no detectable effect onNBD-PL flipping (data not shown). Lipid analyses indicatedthat both NBD-PLs and egg phosphatidylcholine (ePC) re-mained intact during reconstitution (not shown). The rate offluorescence loss was comparable in liposomes and proteoli-posomes (Figures 1C and 1D), indicating that flipping occursfasterthantherateatwhichNBD-PLsarereduced.Bysystem-atically varying the dithionite concentration used in the assay,we obtained an upper bound of w10 s for the time constant ofNBD-PL flipping in proteoliposomes.The extent of NBD-PL reduction increased monoexponen-tially as a function of the protein/phospholipid ratio of thesample (Figure 1F), reflecting the probability that a particularvesicleinthesamplehasoneormoreflippases.Themaximumextent of reduction that we observed was w80%–85% ratherthan the expected 100%, indicating that a fraction of thevesicles cannot be accessed by dithionite. It is possible thatsome of these protected vesicles are within aggregates.Assuming that the average molecular weight of proteins beingreconstitutedis40kDaandtheaveragediameteroftherecon-stituted vesicles is 150 nm (Figure 1B), analysis of the data inFigure 1F(Supplemental Information and Figures S1C andS1D)indicatesthatinasamplewithprotein/phospholipidratiow0.33 mg/mmol (corresponding to the monoexponential fitconstant), each vesicle has 1.65 proteins on average, one ofwhichisaphospholipid flippase, i.e.,61% of thereconstitutedprotein molecules are flippases. Similar analysis of an inde-pendent experiment (Figure S1B) indicated that 69% of thereconstituted proteins are flippases. These results suggest