Absence of clustering of phosphatidylinositol-(4,5)-bisphosphate in fluid phosphatidylcholine.

Phosphatidylinositol-(4,5)-bisphosphate [PI(4,5)P(2)] plays a key role in the modulation of actin polymerization and vesicle trafficking. These processes seem to depend on the enrichment of PI(4,5)P(2) in plasma membrane domains. Here, we show that PI(4,5)P(2) does not form domains when in a fluid phosphatidylcholine matrix in the pH range of 4.8-8.4. This finding is at variance with the spontaneous segregation of PI(4,5)P(2) to domains as a mechanism for the compartmentalization of PI(4,5)P(2) in the plasma membrane. Water/bilayer partition of PI(4,5)P(2) is also shown to be dependent on the protonation state of the lipid.

Phosphatidylinositol-(4,5)-bisphosphate [PI(4,5)P 2 ] is found mainly in the plasma membrane, where it is a critical regulator of several cellular functions. It plays a fundamental role, particularly in actin polymerization and vesicle trafficking (1,2). These processes seem to depend on large transient and spatially localized increases of PI(4,5)P 2 concentration, as this phospholipid constitutes only z1% of the lipids in the plasma membrane (3).
Significant evidence indicates that the membrane patches where localized enrichment in PI(4,5)P 2 is observed are cholesterol-rich rafts (4)(5)(6), but partition of these phospholipids to liquid-ordered domains is difficult to explain, as the sn-2 acyl chain of PI(4,5)P 2 is mainly arachidonic acid, a polyunsaturated acyl chain that is not expected to favor partition into rafts. Local enrichment in PI(4,5)P 2 was shown to overlap with enrichment of phosphatidylinositol-4-monophosphate kinases in the same membrane patches, possibly leading to localized synthesis of the inositol (7). However, as argued previously, this effect alone cannot explain the degree of PI(4,5)P 2 compartmentalization observed, as diffusion away from the site of syn-thesis is faster than the synthesis itself (8). However, several proteins [myristoylated alanine-rich C-kinase substrate (MARCKS), growth associated protein 43 (GAP43), cytoskeleton-associated protein 23 (CAP23), and neuronal axonal membrane protein (NAP22)] have been shown to be responsible for the lateral sequestration of PI(4,5)P 2 to specific domains in the membrane in a process that for some cases was dependent on cholesterol (9)(10)(11). This phenomenon, together with localized synthesis, provides a plausible mechanism for PI(4,5)P 2 compartmentalization, as large concentrations of these proteins could prevent free diffusion of the phospholipid.
In a recent study, it was proposed that PI(4,5)P 2 compartmentalization can be achieved simply through hydrogen bonding between PI(4,5)P 2 head groups at or slightly above physiological pH (corresponding to partial or complete deprotonation of the phosphomonoester group) (12), without the contribution of any external agent (cholesterol or proteins). Through fluorescence resonance energy transfer (FRET) experiments in a phosphatidylcholine (PC) matrix in the fluid state, the authors claimed to have detected PI(4,5)P 2 domain formation, and similar behaviors were observed for PI(3,4)P 2 and PI(3,4,5)P 3 as well as for phosphatidylinositol monophosphates in another study (13). Because of the large biological relevance of the problem and the controversy of these conclusions, we propose in this work to investigate this issue using a detailed experimental approach. Our results show clearly that PI(4,5)P 2 does not form domains when in a PC matrix in the pH range of 4. 8-8.4 and that partition of PI(4,5)P 2 to PC bilayers is largely dependent on the pH.

Liposome reconstitution procedure
The desired amounts of phospholipids were mixed in chloroform-methanol (1:2) and dried under an N 2(g) flow. The sample was then kept in a vacuum overnight. Liposomes were prepared with 20 mM HEPES, 100 mM NaCl buffer at pH 8.4 and 7.1 and with 20 mM sodium citrate, 100 mM NaCl buffer at pH 4.8. The hydration step was performed with the addition of buffer followed by freeze-thaw cycles. Anisotropy measurements were performed with large unilamellar vesicles produced by extrusion through polycarbonate filters with a pore size of 100 nm (14).

Fluorescence measurements
Steady-state fluorescence measurements were carried out with an SLM-Aminco 8100 Series 2 spectrofluorimeter described in detail elsewhere (15). Steady-state anisotropies were determined according to Lakowicz (16). NBD-PC and NBD-PI(4,5)P 2 anisotropies were recorded with excitation and emission wavelengths of 460 and 540 nm, respectively, with spectral bandwidths of 4 nm. All measurements were performed at room temperature.
Fluorescence decay measurements of DPH (donor) were carried out with a time-correlated single-photon timing system, which is described elsewhere (15). All measurements were performed at room temperature. Excitation and emission wavelengths were 340 and 430 nm, respectively.
In the FRET experiments, the donor decays in the presence of acceptors [i DA (t)] can be described as where i DA (t) and i D (t) are the fluorescence decays of the donor in the presence and absence of acceptors, respectively. r interplanar is the FRET contribution arising from energy transfer to randomly distributed acceptors in two different planes from the donors (two monolayer leaflets) (17) (Eq: 2) where b 5 (R 0 2 /l) 2 H D 21/3 , R 0 is the Fö rster radius, n 2 is the acceptor density in each leaflet, and l is the distance between the plane of the donor and the two planes of acceptors. The Fö rster radius was calculated as described elsewhere (18).

NBD-PI(4,5)P 2 partition to lipid bilayers
When using a labeled phospholipid as the acceptor in a FRET study, the energy transfer efficiency will depend strongly on its concentration inside these vesicles. As PI(4,5)P 2 is known to generally form micelles in the aqueous medium (19), it is essential in the type of measurements mentioned above to determine whether all NBD-PI(4,5)P 2 incorporates in the liposomes and, if that is not the case, to quantify the effective fraction of NBD-PI(4,5)P 2 incorporated in liposomes. NBD quantum yield decreases with hydration, and clustering of NBD-PI(4,5)P 2 leads to NBD selfquenching (20). Consequently, it was not surprising that NBD-PI(4,5)P 2 in buffer (micellar state) exhibited quantum yields z100 times smaller than when incorporated in PC liposomes (lamellar state) (Fig. 1). This difference allowed us to use fluorescence intensity as a tool to estimate the extent of water/bilayer partition for NBD-PI(4,5)P 2 . From equation 3 (21), the extent of partition can be quantified into partition coefficients (K p ) where I is the fluorescence intensity, I W and I l are the fluorescence intensities in buffer and in liposomes, respectively, g l is the lipid molar volume, and [l] is the lipid concentration. The curves obtained when the NBD-PI(4,5)P 2 fluorescence intensities were plotted versus the lipid concentrations are shown in Fig. 1. Three different pH values (8.4, 7.1, and 4.8) corresponding to three different protonation states of PI(4,5)P 2 (for the micellar state Pk a2 5 6.7 for the 49 position and 7.6 for the 59 position) were studied. The results for the different pH values were clearly different when using POPC bilayers: a larger extent of partition was observed when pH 8.4 was used (corresponding to a total charge of 25), whereas partition for pH 7.1 (24) was almost identical to that for pH 4.8 (23). For POPG, the partition was much less dependent on pH.

Clustering of NBD-PI(4,5)P 2 in POPC
Clustering of NBD-PI(4,5)P 2 inside the POPC matrix would result in a decrease of fluorescence intensity attributable to self-quenching of NBD and in a decrease in fluorescence anisotropy attributable to energy migration (energy homotransfer) inside the clusters (20). The two parameters were studied for increasing NBD-PI(4,5)P 2 concentrations inside a POPC matrix, as shown in Fig. 2. NBD-PI(4,5)P 2 concentrations were corrected for the partition coefficients determined above. As a control, the fluorescence intensity and anisotropies of NBD-PC in POPC and of NBD-PI(4,5)P 2 in DPPC were compared with the data for NBD-PI(4,5)P 2 in POPC. NBD-PC in POPC is homogeneously distributed, whereas in DPPC, clustering is expected for NBD-PI(4,5)P 2 as a result of gel-fluid phase separation.
It is clear that NBD-PI(4,5)P 2 and NBD-PC have identical clustering behavior in POPC. The degree of selfquenching and energy migration of both lipids is almost identical and much smaller than for NBD-PI(4,5)P 2 in DPPC at room temperature, at which clustering is observed as a result of packing restraints in the DPPC gel matrix (22).
We also performed a FRET experiment, using for this purpose the fluorescence decay of the donor. The reason for a time-dependent approach instead of a steady-state approach is that the kinetics of the donor decay gives information on both the distribution (homogeneity/heterogeneity) and concentration of acceptors (15), whereas the steady-state approach is unable to distinguish between the two. Again, we chose POPC as the matrix lipid, the donor was DPH, which is known to have no preference between lipid phases (23), and the acceptor was NBD-PI(4,5)P 2 . In this system, in the case of NBD-PI(4,5)P 2 clustering, two populations of DPH would be detectable, one residing in a NBD-PI(4,5)P 2 -rich site and the other in a NBD-PI(4,5)P 2 -poor area of the bilayer.
We globally fitted the FRET model for a single discrete concentration of acceptors (homogeneous distribution of acceptors) (equations 2, 3) to the data in Fig. 3, and for all pH values used (only results for pH 8.4 are shown), good quality fits were obtained (global Chi-square , 1.5). The acceptor concentrations were the only free parameter during these fits, and the recovered values matched closely (610%) the bilayer concentrations expected from the partition coefficients determined above, again confirming the absence of NBD-PI(4,5)P 2 clustering.

DISCUSSION
Compartmentalization of PI(4,5)P 2 has been reported several times, but its origin had always been attributed to the interaction with the cytoskeleton and/or with specific proteins exhibiting preferences for particular phases inside the membrane. The recent proposal for a spontaneous demixing of completely deprotonated PI(4,5)P 2 when in a PC matrix even at very low concentrations as a result of intermolecular hydrogen bonding (12) suggested an additional mechanism for localized enrichments of this phospholipid. In this work, we tested this hypothesis through a carefully designed approach that considered and tested the existence of possible artifacts that could go undetected in a phenomenological approach (mere observation of increases or decreases in energy transfer as a result of pH variations).
Redfern and Gericke (12) showed by differential scanning calorimetry and Fourier transform infrared spectroscopy that demixing occurred for mixtures of PI(4,5)P 2 and PC lipids in the gel phase. Nevertheless, in vivo, PI(4,5)P 2 contains polyunsaturated acyl chains, and a gel state is highly unlikely for this lipid. With this in mind, these authors used FRET as a tool to detect the immiscibility of PI(4,5)P 2 and PC in the fluid state at low PI(4,5)P 2 concentrations. They used POPC as the lipid matrix and a NBD-labeled PC as the FRET donor to a short-chain (C6) Dipyrrometheneboron difluoride labeled PI(4,5)P 2 . In this system, they observed variations of energy transfer efficiency as the pH was scanned. The authors interpreted the decrease of energy transfer efficiencies at high pH [.Pk a2 of PI(4,5)P 2 ] as a demixing between PI(4,5)P 2 and POPC. However, the use of a very short-chain PI(4,5)P 2 as the acceptor in this FRET ex-periment is likely to have resulted in a significant fraction of acceptors not being incorporated in the membrane. The lipid/water partition coefficients of this short-chainlabeled PI(4,5)P 2 is expected to be significantly smaller than those determined here for NBD-PI(4,5)P 2 , because K p varies dramatically with the number of carbons in the acyl chains of a phospholipid (24). In this case, energy transfer efficiencies would decrease not because of the formation of PI(4,5)P 2 clusters but as a result of changes in the lipid/water partition coefficient of the probe, as the PI(4,5)P 2 group becomes fully deprotonated. Redfern and Gericke (12) acknowledge that the extent of membrane incorporation changed considerably for different labeled phosphatidylinositides.
Analysis of the kinetics of donor decay in a FRET experiment was the method chosen by us to probe the eventual clustering of PI(4,5)P 2 in a zwitterionic bilayer, because it is capable of retrieving information on the homogeneity of the fluorescently labeled PI(4,5)P 2 independently of any quantitative knowledge of the probe's bilayer concentration. The good agreement between the model considering a homogeneous distribution of NBD-PI(4,5)P 2 and the data is solid proof of the absence of PI(4,5)P 2 clustering at low PI(4,5)P 2 concentrations (,5%) even in a completely deprotonated state. The result of this experiment alone cannot exclude the formation of very small PI(4,5)P 2 clusters (zR 0 for this Fö rster pair 5 40 Å ), but, together with the NBD energy migration and self-quenching studies, it is clear that PI(4,5)P 2 has no lateral organization in a fluid POPC matrix. It is