Type IV P-type ATPases Distinguish Mono- versus Diacyl Phosphatidylserine Using a Cytofacial Exit Gate in the Membrane Domain*

Background: Type IV P-type ATPases (P4-ATPases) Drs2 and Dnf1 are known to recognize different phospholipid headgroups. Results: Dnf1 preferentially recognizes lysophospholipid, but mutations in transmembrane (TM) segments 1, 2, and 3 allow recognition of diacyl phosphatidylserine. Conclusion: Residues within TM1–3 cooperate with the proline + 4 residue in TM4 to define P4-ATPase specificity. Significance: We provide insight into how a P-type ATPase subgroup evolved a new substrate. Type IV P-type ATPases (P4-ATPases) use the energy from ATP to “flip” phospholipid across a lipid bilayer, facilitating membrane trafficking events and maintaining the characteristic plasma membrane phospholipid asymmetry. Preferred translocation substrates for the budding yeast P4-ATPases Dnf1 and Dnf2 include lysophosphatidylcholine, lysophosphatidylethanolamine, derivatives of phosphatidylcholine and phosphatidylethanolamine containing a 7-nitro-2-1,3-benzoxadiazol-4-yl (NBD) group on the sn-2 C6 position, and were presumed to include phosphatidylcholine and phosphatidylethanolamine species with two intact acyl chains. We previously identified several mutations in Dnf1 transmembrane (TM) segments 1 through 4 that greatly enhance recognition and transport of NBD phosphatidylserine (NBD-PS). Here we show that most of these Dnf1 mutants cannot flip diacylated PS to the cytosolic leaflet to establish PS asymmetry. However, mutation of a highly conserved asparagine (Asn-550) in TM3 allowed Dnf1 to restore plasma membrane PS asymmetry in a strain deficient for the P4-ATPase Drs2, the primary PS flippase. Moreover, Dnf1 N550 mutants could replace the Drs2 requirement for growth at low temperature. A screen for additional Dnf1 mutants capable of replacing Drs2 function identified substitutions of TM1 and 2 residues, within a region called the exit gate, that permit recognition of dually acylated PS. These TM1, 2, and 3 residues coordinate with the “proline + 4” residue within TM4 to determine substrate preference at the exit gate. Moreover, residues from Atp8a1, a mammalian ortholog of Drs2, in these positions allow PS recognition by Dnf1. These studies indicate that Dnf1 poorly recognizes diacylated phospholipid and define key substitutions enabling recognition of endogenous PS.

Active transport of substrates across lipid bilayers is an essential process for cell survival. Both prokaryotes and eukaryotes spend considerable energy to actively transport ions, nutrients, waste products, and lipids across the cellular membranes. P-type ATPases include a large family of ion transporters involved in establishing gradients across membranes at the expense of ATP. These ion gradients are required for establishing the resting membrane potential in neurons, signaling at the neural synapse, muscle contraction, and acidification of the stomach lumen. P-type ATPases cycle through conformational states described by the Post-Albers cycle and characteristically form a phosphorylated intermediate when the ␥-phosphate from ATP is transferred temporarily to a conserved aspartate residue on the protein. As the cycle proceeds forward, the phosphate is hydrolyzed from the protein, ensuring a single direction in the reaction cycle. P-type ATPases are grouped into five subclasses on the basis of substrate specificity (1). Type I, II, and III transport heavy metals, cations, and protons, respectively. Interestingly, the type IV (or P4) ATPases are only present in eukaryotic cells and flip phospholipid molecules across the membrane bilayer (flippases). Type V has unknown substrate specificity and is also only present in eukaryotes.
A central conundrum involving P-type ATPases is how they have evolved the ability to transport such a wide range of substrates, from ions to phospholipids. Ion transporters use a highly conserved ion binding site within the center of the transmembrane domain highlighted by the crystal structures of the Ca 2ϩ ATPase (16). We have recently described a series of residues involved in defining the substrate specificity of P4-ATPases that implicated a novel transport mechanism for a P-type ATPase using two gates to select phospholipid on opposing membrane faces (17,18). Substrate-selecting residues near the extracellular or lumenal leaflet surface are proposed to form an "entry gate" where phospholipid is initially selected for transport, and substrate-selecting residues near the cytosolic leaflet surface are proposed to form the "exit gate." Drs2 and its mammalian homologs Atp8a1/Atp8a2 are flippases that catalyze "flip" of phosphatidylserine (PS) and phosphatidylethanolamine (PE) within the Golgi-endosomal system (although PS seems to be the preferred substrate) (19 -23). Even though Drs2 is closely related to Dnf1, they differ in their substrate specificity and localization. Dnf1 and Dnf2 recognize and transport lyso-PC (24), lysophosphatidylethanolamine (25), and C6-NBD derivatives of these lipids (26), but it is unknown whether these pumps recognize intact, dually acylated PC and PE. Our first description of residues involved in phospholipid selection highlighted a residue of particular interest: the "proline ϩ 4" (P ϩ 4) position in TM4. This residue is four amino acids away from a proline that is completely conserved in all P-type ATPases. The presence of a tyrosine at the P ϩ 4 position restricted PS flip, whereas a phenylalanine allowed PS flip. Surprisingly, Atp8a1 and Atp8a2 have a leucine at this position. When we introduced a leucine into either Dnf1 or Drs2, the overall activity of these proteins was reduced, but we did not observe a change in the specificity.
Here we describe new Dnf1 variants that are capable of flipping endogenous, unlabeled PS. Surprisingly, we observed a striking difference among Dnf1 variants capable of flipping NBD-labeled PS in their ability to restore the endogenous PS asymmetry to a drs2⌬ mutant. Identification of new Dnf1 alleles capable of supporting endogenous PS flip revealed substitutions corresponding to the analogous residues in the mammalian PS flippases Atp8a1/Atp8a2. Coordination between the TM4 P ϩ 4 and residues on TMs 1, 2, and 3 define a portion of a phospholipid substrate binding site at the exit gate used by P4-ATPases.

EXPERIMENTAL PROCEDURES
Reagents-All lipids were purchased from Avanti Polar Lipids, Inc. 5-fluoroorotic acid was purchased from Zymo Research. Duramycin was purchased from Sigma-Aldrich. Papuamide B was a gift from Raymond Andersen (University of British Columbia).
Cold Resistance Screen-To generate random mutations, we used procedures reported previously (27). Briefly, we cotransformed a dnf1,2,3⌬drs2⌬ strain (ZHY704) with mutagenized PCR products and pRS313-DNF1 gapped with restriction enzymes AflII and MluI to allow for gap repair through homol-ogous recombination. The transformants were plated onto synthetic dropout plates with 5-fluoroorotic acid, grown for 4 days at 20°C, and then colonies were picked into 96-well plates. We compared growth of the strains expressing dnf1 mutants strains to those expressing WT DNF1 or WT DRS2 and sequenced dnf1 alleles from the strains that grew strongly at 20°C.
Strains and Culture-Escherichia coli strain DH5␣ was used for molecular cloning. All strains and plasmids used in this study are listed in supplemental Tables S1 and S2. Yeast strains were cultured with glucose in standard rich medium (yeast extract, peptone, dextrose (YPD)) or synthetic minimal medium (SD) at the indicated temperatures. Yeast transformation used the lithium acetate method (28).
Growth Assay-To test for drug sensitivity, mid-log phase yeast was distributed into 96-well plates at 0.1 A 600 /ml. Plates were incubated at 30°C for 20 h, and the A 600 /ml was measured using a multimode plate reader Synergy HT (Bio-Tek, Winooski, VT). Relative growth was determined by dividing A 600 values from drug-treated cells by mock-treated cells in the assay.
Lipid Uptake-Lipid uptake was performed essentially as described previously (29). Briefly, overnight cultures were subcultured to 0.15 A 600 /ml and allowed to grow to early mid-log phase. 500 l of cells was harvested, resuspended in ice-cold SD medium containing 2 g/ml NBD lipid (ϳ2.5 M), and incubated on ice for 30 min. Cells were washed twice with SA medium (SD medium, 2% sorbitol (w/v), 20 mM NaN 3 ) and 4% BSA (w/v), washed once with SA medium, and then resuspended in SA medium prior to analysis by flow cytometry.
Flow Cytometry-We used a BD LSRII-3 laser (BD Biosciences) using BD FACSDiva v6.1.3. NBD lipid uptake was measured with the FITC filter set (530/30 bandpass with a 525 longpass). Immediately prior to analysis, propidium iodide was added at a final concentration of 5 M. At least 10,000 events were analyzed using forward/side scatter to identify single cells and propidium iodide fluorescence to exclude dead cells.
Fluorescence Microscopy-Images were collected with an Axioplan microscope (Carl Zeiss, Thornwood, NY) using a Hamamatsu charge coupled device camera and MetaMorph software. To observe GFP-tagged proteins, cells were grown to mid-log phase, pelleted, and resuspended in imaging buffer (10 mM Tris-HCl (pH 7.4), 2% glucose (w/v)). An aliquot of cells was spread onto slides and visualized using a GFP filter set.
Data Analysis-Within each NBD-phospholipid (NBD-PL) uptake experiment, at least three independently isolated transformants containing the same construct were assayed. The data from at least three separate experiments were reported as mean Ϯ S.E. Uptake of NBD-PL for a dnf1,2⌬ strain with an empty vector was subtracted, and NBD-PC uptake by WT Dnf1 in the same strain was used to normalize the data. Each value is reported relative to NDB-PC uptake by WT Dnf1 at 30 min. Substrate preference ratios were determined by the dividing NBD-PE or NBD-PS uptake values by the NBD-PC uptake value for each independent replicate (n Ն 9). The data from each substrate preference ratio is reported as mean Ϯ S.E. For drug sensitivity assays, at least two independently isolated transformants were assayed in at least two separate experiments, and data are reported as mean Ϯ S.E.

RESULTS
Dnf1 N550S Flips Endogenous PS-Deletion of DRS2 disrupts plasma membrane asymmetry for PS and PE. Phospholipid asymmetry can be monitored by sensitivity to pore-forming toxins that specifically interact with either PE (duramycin) or PS (papuamide B, PapB) exposed on the outer leaflet of the plasma membrane. We reported previously that WT Dnf1 and Dnf1[Y618F] (a P ϩ 4 mutation; the brackets are used to indicate the replacement of a Dnf1 residue with a Drs2 residue) were unable to restore the phospholipid asymmetry of a drs2⌬ strain even though Dnf1[Y618F] was capable of NBD-PS transport (17). Importantly, the reciprocal change in Drs2 specifically reduced the ability of this pump to flip endogenous PS, supporting a role for the P ϩ 4 residue in recognizing the phospholipid headgroup (17). We then tested other Dnf1 mutants reported to transport NBD-PS (18), and like Dnf1[Y618F], neither Dnf1[GA3 QQ] (a TM1 entry gate mutation) nor Dnf1[GA3 QQ, Y618F] were able to restore PapB resistance and, therefore, PS asymmetry, to a drs2⌬ strain (Fig. 1A). Incredibly, Dnf1 N550S (a TM3 exit gate mutation) was unique among the Dnf1 mutants that recognize NBD-PS in its ability to fully restore the PS asymmetry of a drs2⌬ strain to near WT levels (Fig. 1A).
The N550S influence on membrane asymmetry was rather specific to PS because Dnf1 N550S conferred only a slight increase in duramycin resistance relative to the other Dnf1 mutants (Fig. 1B). Thus, PE asymmetry was not fully restored in drs2⌬ cells expressing Dnf1 N550S. Even when overexpressed from a 2-m plasmid, neither WT Dnf1 nor Dnf1[Y618F] were able to correct the PS asymmetry defect of drs2⌬ as well as Dnf1 N550S (compare Fig. 1A to C). This suggested that Dnf1[Y618F] was unable to transport endogenous PS, although it was capable of NBD-PS transport. Because Drs2 is required for growth at 20°C, we also tested the ability of the Dnf1 NBD-PS-transporting alleles to suppress the cold-sensitive growth defect. Again, only Dnf1 N550S was capable of suppressing the drs2⌬ cold sensitivity (Fig. 1D). The localization patterns and expression levels of these Dnf1 mutants were similar and, therefore, unlikely to account for the observed phenotypic differences (18). Therefore, the ability to flip endogenous PS and restore membrane asymmetry in vivo correlated with the suppression of drs2⌬ cold sensitivity.
Dnf1 Suppressors of Drs2 Function-We previously used an extensive mutagenesis screen targeted to TM3-4 in Dnf1 and identified several residues within this region involved in NBD-PS selection (18). The observation that Dnf1 N550S was uniquely capable of recognizing endogenous PS and suppressing drs2⌬ cold-sensitive growth suggested that a screen for dnf1 alleles that suppress drs2⌬ cold-sensitive growth would identify additional changes, allowing Dnf1 to flip endogenous PS. Random mutagenesis targeted to TM1-2 produced about 500 dnf1 mutants (of 5000 transformants) that supported growth of a dnf1,2,3⌬drs2⌬ strain at 20°C. We selected 20% of the colonies that were largest, picked these into liquid culture, and grew the strains at 20°C. We recovered 25 strains that grew similarly to dnf1,2,3⌬drs2⌬ harboring WT DRS2 at 20°C. From these strains we sequenced the DNF1 constructs, recovering 17 alleles coding for single amino acid substitutions. There were four positions at which we recovered multiple alleles encoding the same amino acid substitution: F213S, A244V, T254A, and D258E. Random mutagenesis was also targeted to TM3-4 in Dnf1. Screening a similar number of dnf1 mutants yielded only a single new Dnf1 substitution that strongly suppressed drs2⌬ (Dnf1 V553E). Targeting random mutagenesis to TM5-6 failed to produce dnf1 mutants capable of strongly suppressing drs2⌬ cold sensitivity. Except for A244V, these suppressor alleles each arose independently because the codons recovered were different.
We next transformed these DNF1 alleles into a drs2⌬ strain to determine whether the suppression was plasmid-linked or extragenic ( Fig. 2A). In fact, DNF1 A244V failed to suppress drs2⌬ cold sensitivity any better than WT DNF1, and so the original mutation was not plasmid-linked. Each of the other alleles did suppress the cold sensitivity of a drs2⌬ strain, although to varying degrees ( Fig. 2A). We then assayed the NBD-PL uptake activity for each of the new dnf1 alleles in a dnf1,2⌬ background. With the exception of Dnf1 A244V, we observed an increase in the NBD-PS uptake for each Dnf1 mutant. Most of the Dnf1 mutants were able to translocate NBD-PC and NBD-PE at near WT levels, although Dnf1 A244V had a reduced activity for all substrates and Dnf1 V553E dis-zzzz FIGURE 2. Dnf1 suppressors of drs2⌬ cold-sensitive growth transport NBD-PS across the plasma membrane and restore endogenous PS asymmetry. A, growth of drs2⌬ expressing the indicated Drs2 and Dnf1 variants at 30°C or 20°C. B, dnf1⌬dnf2⌬ cells expressing the indicated Dnf1 variants were incubated with NBD-phospholipid for 30 min at 4°C as described under "Experimental Procedures." The amount of lipid transported was normalized to NBD-PC uptake by WT Dnf1. C, the data shown in B was graphed as the ratio of NBD-PE to NBD-PC uptake (open bars) or NBD-PS to NBD-PC uptake (gray bars) to assess relative changes in substrate preference. The dotted lines are the ratios for WT Dnf1, and the shaded region above and below these lines are confidence intervals. Values extending outside of the confidence intervals represent a significant change in substrate specificity. D, the Dnf1 variants indicated were tested for their ability to restore PapB resistance (PS asymmetry) to a drs2⌬ strain. The Dnf1 suppressors of drs2⌬ cold-sensitive growth restore PS asymmetry better than WT Dnf1. played a reduction in NBD-PC uptake activity (Fig. 2B). A measure of substrate preference independent of overall activity can be obtained by plotting the uptake ratios of NBD-PE to NBD-PC and NBD-PS to NBD-PC (Fig. 2C) (18). Ratios outside of the shaded confidence interval are considered significant changes in substrate preference. By these criteria, the N550S, F213S, T254A, D258E, and V553E mutants have a significantly increased ability to recognized NBD-PS relative to WT Dnf1, and V553E also showed an increased preference for NBD-PE, caused by reduced uptake of NBD-PC.
Recognition of Mono-versus Diacylated Phospholipid-Each of these dnf1 mutants were competent for NBD-PS uptake but displayed varying resistance to PapB. We speculated that Dnf1 N550S and Dnf1 T254A were capable or recognizing endogenous, unlabeled PS, whereas Dnf1[Y618F] and Dnf1[GA3 QQ] were unable to do so. To test this possibility, we made use of GFP-LactC2, a probe that binds specifically to PS, and a cho1⌬ strain (30,31). CHO1 (PSS1) encodes the phosphatidylserine synthase in S. cerevisiae, and cho1⌬ mutants completely lack PS (32,33). GFP-LactC2 localizes primarily to the inner leaflet of the plasma membrane of wild-type cells but is diffusely localized to the cytosol in cho1⌬ cells (Fig. 3A) (30). A line scan of fluorescence intensity through the cells clearly shows peaks for the plasma membrane of the wild-type cell that is lacking in the cho1⌬ cells. When cho1⌬ cells are supplemented with lyso-PS applied extracellularly at 4°C, GFP-LactC2 is recruited to the plasma membrane because this substrate is flipped to the cytosolic leaflet (30). The low temperature is sufficient to block endocytosis and reduce the activity of flippases that could pump the lyso-PS back to the outer leaflet (34).
Using a cho1⌬drs2⌬ strain carrying an extra copy of WT DNF1, lyso-PS was rapidly flipped across the plasma membrane at 4°C allowing GFP-LactC2 recruitment to the cytosolic leaflet of the plasma membrane within 5 min of substrate addition (Fig. 3B). In contrast, DLPS (C12 fatty acyl chains in the sn-1 and sn-2 positions) was not appreciably flipped across the membrane, and GFP-LactC2 remained mostly cytosolic up to 150 min of incubation. We then tested whether Dnf1 N550S, Dnf1[Y618F], or Dnf1 T254A were capable of transporting dually acylated DLPS. Consistent with the ability to restore plasma membrane PS asymmetry, DLPS was flipped into the cell by Dnf1 N550S and Dnf1 T254A but not by Dnf1[Y618F]. GFP-LactC2 plasma membrane fluorescence was weak for Dnf1 N550S cells incubated with DLPS, but the Dnf1 T254A cells treated with DLPS displayed a more robust LactC2 plasma membrane fluorescence. Line scans for cells expressing Dnf1 N550S and Dnf1 T254A and incubated 150 min with DLPS showed peaks of enrichment at the plasma membrane that was similar to the same cells treated with lyso-PS. Lyso-PS was quickly taken up across the plasma membrane of cells expressing each Dnf1 variant (Fig. 3B).
Lyso-PS flip was partially mediated by Dnf1 and Dnf2 even though Dnf1 recognizes NBD-PS poorly. After 30 min of incubation with a cho1⌬ dnf1,2⌬ strain at 4°C, lyso-PS crossed the membrane sufficiently to recruit the GFP-LactC2, but it was at least 4-fold slower than with Dnf1 and Dnf2 present ( Fig. 4A  compared with Fig. 3). This is consistent with a previous report suggesting that yeast cells have another undefined transporter at the plasma membrane capable of taking up NBD-PS or lyso-PS that is independent of Dnf1,2,3 and Drs2 (35). For reasons that are unclear, we could not transform the cho1⌬ dnf1,2⌬ strain with plasmids carrying functional alleles of DNF1, whereas empty plasmid transformants were readily obtained. Because Dnf1 appears to be toxic to the cho1⌬ dnf1,2⌬ strain, we could not assay the Dnf1 variants for lipid uptake in this background.
We next performed similar experiments at 30°C, a condition that permits endocytosis, allowing the lyso-PS or DLPS to enter the lumenal leaflet of the Golgi-endosomal system. In cho1⌬dnf1,2⌬, after 1 h at 30°C, both the lyso-PS-and DLPStreated cells displayed GFP-LactC2 localized to the plasma membrane (Fig. 4B). We suspected that this may be due to the lipid being flipped when it reached Drs2 in the Golgi or early endosome. To test this possibility, we examined GFP-LactC2 localization in cho1⌬drs2⌬ cells supplemented with either lyso-PS or DLPS at 30°C. In the absence of Drs2 or in the presence of Drs2[QQ3 GA] (strongly deficient for PS recognition (18)), we observed no recruitment of GFP-LactC2 to the plasma membrane of cells incubated with DLPS (Fig. 4C). GFP-LactC2 was recruited to the plasma membrane of cells incubated with lyso-PS.
In total, these data indicate that Dnf1 and Dnf2 are unable to flip dually acylated PS across the plasma membrane but have a weak capacity to flip lyso-PS or NBD-PS. One class of mutations in Dnf1 (e.g. Y618F) enhances recognition of NBD-PS but does not allow PS recognition. A second class of mutations in Dnf1 (e.g. N550S and T254A) strongly enhances recognition of NBD-PS and endogenous PS. Drs2, in contrast, flips endogenous PS and DLPS efficiently.
P4-ATPase Phospholipid Recognition with P ϩ 4 Leucine-An unresolved question regarding the mechanism of phospholipid recognition involves the "P ϩ 4 position" in TM4. The Drs2 homologs Atp8a1 and Atp8a2 also preferentially flip PS, but have a leucine at the P ϩ 4 position rather than a phenylalanine (19,21). Substitution of leucine for Tyr-618 or Phe-511 (P ϩ 4 residues) in Dnf1 or Drs2, respectively, causes a reduction in overall activity with no apparent influence on PS recognition (17). Several of the drs2⌬ suppressor mutations we identified in Dnf1 through unbiased, random mutagenesis (Fig. 2) were substitutions for the identical residues in Atp8a1 and/or Atp8a2 (Fig. 5A, red boxes). These mutations allowed recognition of endogenous PS and we wondered whether these substitutions combined with a P ϩ 4 leucine could generate a functional Dnf1 capable of PS flip.
To test the potential cooperation of residues in TM1-3 with a P ϩ 4 leucine, we generated combinations of Dnf1 Y618L with substitutions corresponding to Atp8a1 (F213S), Atp8a1/2
To test whether these combinations of Y618L and TM1-3 substitutions were functional as flippases, we tested for the ability to restore PS asymmetry in drs2⌬ cells. In each case, the double mutants restored PS asymmetry better than Dnf1 Y618L (Fig. 5D). Thus, the endogenous TM1,2 residues Phe-213, Thr-254, and Asp-258 couple with Tyr-618 in Dnf1 to prevent PS translocation and cannot function with a Leu at position 618 We next tested whether the function of the Drs2 P ϩ 4 leucine variant (Drs2 F511L) could be restored with these Atp8a1/a2 substitutions. In contrast to Drs2 F511L, each of the Drs2 double mutants tested (F511L combined with Q220S, S261A, or N445S) supported growth of a dnf1,2,3⌬drs2⌬ strain (Fig. 6A). In each case, these double mutants also allowed increased cold-resistant growth of drs2⌬ relative to Drs2 F511L (Fig. 6B). Each of the Drs2 TM1,2 single mutants tested supported growth of the dnf1,2,3⌬drs2⌬ or drs2⌬ strains (Fig. 6, C  and D). Thus, the mammalian TM1,2 residues can function well with the endogenous P ϩ 4 Phe in Drs2. When tested for the ability to restore PS membrane asymmetry, the double mutants all conferred greater resistance to PapB than Drs2 F511L (Fig.  6E). Although the double mutants tested did not complement the mutant phenotypes as well as WT Drs2, they were significantly better that Drs2 F511L. These results indicate that the P ϩ 4 leucine is compatible with PS recognition when it is combined with the appropriate residues on TM1-3.

DISCUSSION
One interesting question preceding this work was why Dnf1 mutants capable of flipping NDB-PS were not able to restore PS asymmetry in drs2⌬ cells (17). The fact that some Dnf1 mutants are fully capable of NBD-PS flip but seemingly unable to flip endogenous PS suggests that these pumps have either the improper localization, abundance, or substrate specificity to restore PS asymmetry in drs2⌬ cells. However, the localization and abundance of Dnf1 N550S was indistinguishable from Dnf1 [Y618F] or Dnf1 [GA3 QQ] (18). Here we show that Dnf1 N550S can replace Drs2 function in establishing endogenous PS asymmetry, whereas Dnf1 Y618F or Dnf1 [GA3 QQ] cannot, even though all three mutant forms of Dnf1 can flip NBD-PS across the plasma membrane. The use of GFP-LactC2 localization in PS-deficient cho1⌬ cells to monitor translocation of unmodified lyso-PS and DLPS provided an explanation as to why Dnf1[Y618F] was unable to establish PS asymmetry; it does not recognize dually acylated PS very well. In fact, WT Dnf1 also appears to flip lyso-PS much better than diacyl PS. Although WT Dnf1 does not flip endogenous PS very well, it seems to have a minor activity to NBD-PS, which can be increased in certain cases (17,26). Dnf1[Y618F] and Dnf1[GA3 QQ] certainly identified residues important in headgroup recognition because reciprocal changes in Drs2 abrogated its ability to recognize endogenous PS, not just NBD-PS. NBD-PL may be capable of mimicking a lyso-PL, a natural substrate for Dnf1, as well as the dually acylated PL substrate of Drs2. However, like drs2⌬, a dnf1,2⌬ strain is also sensitive to PapB (36). This could be because of an increased exposure of lyso-PS on the extracellular leaflet of the dnf1,2⌬ bilayer. Lyso-PS accounts for about 10% of PS in S. cerevisiae, which would likely be enough for PapB to permeabilize the cell if exposed on the outer leaflet of the plasma membrane (30).
A second question is why does yeast maintain two P4-ATPases that appear to function only as lyso-phospholipid transporters? Is this an apparent role for scavenging phospholipid from the environment? Could they be functioning to flip phospholipids that have been damaged by phospholipases in the extracellular environment? Yeast produces lyso-PL when using phospholipids as an acyl donor in triglyceride synthesis (37), and this species can be flipped by Pdr5 and Yor1 to the extracellular leaflet (38). Dnf1,2 may be required to retain the lyso-PL species on the cytosolic leaflet, where they are available to repair enzymes for acylation by Ale1 (39,40). The ecological positioning of true wild-type yeast is in and around plants and animals, scavenging nutrients from the environment. The ability to scavenge nearly whole lipids (lyso-PL) would be a strong selective advantage, saving considerable energy from synthesis in the Kennedy pathway (41). Additionally, it is becoming increasingly apparent that lyso-PL species are involved in signaling events. Recent work has implicated the accumulation of long-chain lyso-PS species in a mammalian neurodegenerative condition (42). Lyso-PC is an early metabolic serum marker for diabetes (43) and also serves as a signaling lipid in plants, enhancing responsiveness for a Na ϩ activated H ϩ channel (44) and activating the plasma membrane proton pump AHA2 (45).
Drs2 may not flip lyso-PL as well as diacyl PL. The fact that Dnf1 N550S rescued drs2⌬ growth defects much better than Dnf1 or Dnf1 [Y618F] implies that dually acylated PS is a crucial substrate for Drs2. In addition, Atp8a1 was found to have a much weaker transport activity toward lyso-PS relative to dioleoyl-PS or DLPS (20). Even with the non-preferred lipid PC, Atp8a1 transported the diacyl variants more quickly than FIGURE 5. Substitutions of specific TM1-3 residues restore function to a Dnf1 P ؉ 4 leucine variant. A, sequence alignment of TM1-2 and TM3-4 from lyso-PC flippases (Dnf1 and Dnf2), PS flippases (Drs2, Atp8a1, Atp8a2), and Ca 2ϩ ATPase. The blue boxes include residues reported previously as contributing to PS specificity. The red boxes indicate newly identified residues contributing to PS specificity in Dnf1 mutants that correspond to equivalent Atp8a1 residues. B, double mutants combining a p ϩ 4 leucine (Y618L) with Atp8a1 residues (F213S, T254A, or D258E) allows Dnf1 complementation of a dnf1,2,3⌬drs2⌬ strain. The 5-fluoroorotic acid kills cells that were unable to lose the URA3-marked plasmid carrying DRS2 in the parental strain. Strains that grow express a Dnf1 variant that can support the viability of dnf1,2,3⌬drs2⌬ in the absence of Drs2. C, the same set of Dnf1 variants were tested for the ability to support growth of drs2⌬ at 20°C. D, similarly, each Y618L double mutant tested partially restored membrane PS asymmetry to drs2⌬ relative to Y618L alone. JULY 5, 2013 • VOLUME 288 • NUMBER 27 lyso-PC (20). This may reflect a difference between the yeast and mammalian proteins, but it seems more likely that Drs2 also prefers intact phospholipid as its substrate. Therefore, we suggest different cellular niches for Drs2 and Dnf1/Dnf2, with Drs2 pumping PS and PE to the cytosolic leaflet and Dnf1/Dnf2 pumping lyso-PC, lysophosphatidylethanolamine, and lyso-PS (in this order of preference).

P4-ATPase Recognition of Lysophospholipid
This work highlights additional residues involved in substrate specificity within TM1-4 in yeast P4-ATPases. The observation that Dnf1 N550S could effectively suppress drs2⌬ mutant phenotypes suggested that a screen for additional DNF1 mutations that suppress drs2⌬ would identify residues important for recognizing dually acylated PS. This prediction was met with the identification of F213S in TM1 and T254A and D258E in TM2 that allowed Dnf1 to flip PS and replace Drs2 function. These residues all cluster near the cytosolic side of the membrane, and we propose they form part of the exit gate where phospholipid is selected prior to release to the cytosolic leaflet. No residues in the entry gate, where phospholipid is initially selected from the exofacial leaflet, were recovered in our screen, implying that distinguishing monoacylated from diacylated substrate is a function of the exit gate.
Our results further support the "boundary lipid to substrate" hypothesis for the evolution of P4-ATPases from a primordial cation pump. The TM1-4 residues in the Ca 2ϩ ATPase analogous to the P4-ATPase exit gate residues form a binding pocket for a phospholipid molecule when the pump is in the E2 conformation ( Fig. 7B and Ref. 46). This binding pocket is inaccessible to phospholipid when the pump is in the E1 conformation, so phospholipid presumably enters and exits this site from the cytosolic leaflet with every ATP consumed. We suggest that this boundary lipid binding site evolved into a substrate binding site in the P4-ATPases. The P4-ATPases also evolved an entry gate to load the phospholipid into this binding site from the exofacial leaflet, thus flipping the phospholipid substrate.