Characterization of human lysophospholipid acyltransferase 3.

Esterifying lysophospholipids may serve a variety of functions, including phospholipid remodeling and limiting the abundance of bioactive lipids. Recently, a yeast enzyme, Lpt1p, that esterifies an array of lysophospholipids was identified. Described here is the characterization of a human homolog of LPT1 that we have called lysophosphatidylcholine acyltransferase 3 (LPCAT3). Expression of LPCAT3 in Sf9 insect cells conferred robust esterification of lysophosphatidylcholine in vitro. Kinetic analysis found apparent cooperativity with a saturated acyl-CoA having the lowest K0.5 (5 microM), a monounsaturated acyl-CoA having the highest apparent Vmax (759 nmol/min/mg), and two polyunsaturated acyl-CoAs showing intermediate values. Lysophosphatidylethanolamine and lysophosphatidylserine were also utilized as substrates. Electrospray ionization mass spectrometric analysis of phospholipids in Sf9 cells expressing LPCAT3 showed a relative increase in phosphatidylcholine containing saturated acyl chains and a decrease in phosphatidylcholine containing unsaturated acyl chains. Targeted reduction of LPCAT3 expression in HEK293 cells had essentially an opposite effect, resulting in decreased abundance of saturated phospholipid species and more unsaturated species. Reduced LPCAT3 expression resulted in more apoptosis and distinctly fewer lamellipodia, suggesting a necessary role for lysophospholipid esterification in maintaining cellular function and structure.

LPCAT assay, radioactive substrate. Lysophosphatidylcholine acyltransferase (LPCAT) activity was measured by the incorporation of [1-14 C]palmitoyl lysoPC into phosphatidylcholine (PC). The reaction contained 100 mM Tris-HCl, pH 7.4, 50 M [1- 14 C] palmitoyl lysoPC (50,000 dpm/nmol), 1-150 M of the respective acyl-CoA, and 1 g of cell lysate protein in a total volume of 100 l. Fixed time assays were performed for 7.5 min at 28°C or 5 min at 37°C. Reactions were stopped by adding chloroform: methanol (2:1) and lipids were extracted, resolved by thin layer chromatography, and quantifi ed as described elsewhere ( 8 ). EZ-Fit software was used for Hill-Plots. To calculate V max /K m , V max was fi rst changed to nM/min/mg to remove volume from the units of the ratio.
LPLAT assays, spectrophotometry. These assays were essentially performed as described previously ( 13 ). The reaction mixture contained 100 mM Tris-HCl, pH 7.4, 50 M of the respective lysoPL, 50 M oleoyl CoA, 1 mM dithionitrobenzoic acid (DTNB), and 2-5 g cell lysate protein in a total volume of 1 ml. Real time assays were performed for 5 min at room temperature or 3 min at 37°C.
Bligh-Dyer lipid extraction. Cellular PL was extracted using chloroform and methanol according to the method of Bligh and Dyer ( 23 ). Following transfection, Sf9 cells were homogenized in 3 ml of methanol: water (2:0.8), transferred to a glass test tube, and then 1.25 ml chloroform was added. Tubes were vortexed for 30 s and allowed to sit for 10 min on ice. After centrifugation at 213 g for 1 min, the chloroform layer was collected. The extraction was repeated and lipid extracts combined, dried under argon, reconstituted with 50 l of methanol:chloroform (2:1), and stored at Ϫ 20°C.
Lipid phosphorus assay. Lipid phosphorus was quantifi ed using malachite green ( 22 ). 10 l of lipid extract was dried under argon. 200 l of perchloric acid was added and heated at 130°C for 2-3 h. To this was added 1 ml of dH 2 O, 1.5 ml of reagent C (4.2 g ammonium molybdate tetrahydrate in 100 ml 5 N HCl and 0.15 g malachite green oxalate in 300 ml dH 2 0), and 200 l of 1.5% (v/v) Tween 20. After 25 min at room temperature, the A 590 was measured.
Electrospray ionization-mass spectrometry (ESI-MS). Lipid extracts (500 pmol/ l) were prepared by reconstituting in chloroform: methanol (2:1). Mass spectrometry was performed as described previously ( 24 ). Samples were analyzed using a Trap XCT ion-trap mass spectrometer (Agilent Technologies, Santa Clara, CA) equipped with an ESI source. 5 l of sample was introduced by means of a fl ow injector into the ESI source at a rate of 0.2 ml/min. The elution solvent was acetonitrile:methanol:water (2:3:1, v/v/v) containing 0.1% (w/v) ammonium formate (pH 6.4). The mass spectrometer was operated in the positive and negative ionization modes with a nitrogen drying gas fl ow-rate of 8.0 L/min at 350°C, nebulizer pressure at 30 psi, and capillary voltage at 3 kV. As previously described ( 24 ), qualitative identification of individual PL molecular species was based on their calculated theoretical monoisotopic mass values and subsequent species utilized is 1-alkyl lysoPC and 1-alkenyl lysoPC, indicating the likely importance of this enzyme in platelet activating factor (PAF) synthesis. A third mouse paralog can also esterify lysoPC with a preference for palmitoyl-CoA ( 11 ).
The S. cerevisiae genome contains no gene with distinct sequence identity to these mouse LPCAT. However, an unrelated gene, LPT1 (also called ALE1 , LPA1, and SLC4 ) encodes for a lysoPL acyltransferase (LPLAT) (12)(13)(14)(15)(16). Lpt1p can esterify all the lyso species of the major phospholipids, has highest capacity for unsaturated acyl-CoA species, and belongs to the membrane bound o-acyltransferase (MBOAT) gene family ( 17 ) which has many representatives in mammalian genomes. Pioneering work by Hishikawa et al. ( 18 ) identifi ed and characterized three mouse MBOAT paralogs, LPCAT3, LPCAT4, and LPEAT1, that can esterify lysoPC, lysoPE and/or lysoPS. Zhou et al. ( 19 ) have shown that human LPCAT3 esterifi ed lysoPC and lysoPS in vitro and that reduction of its expression in Huh7 cells dramatically reduced LPCAT activity.
We have performed similar kinetic analyses of human LPCAT3, with some different results. We have used mass spectrometry to analyze the effect of LPCAT3 overexpression and reduced expression on phospholipid composition. Data are also provided that support a role for LPCAT3 in regulating cell growth and apoptosis.

Methods
Protein sequence analysis. Full-length sequences were aligned using clustalW2 ( 20 ). For percent identity, the conserved amino acids were counted and divided by the length of the shorter of the two proteins compared. Transmembrane domains were predicted by the TMHMM2.0 program ( 21 ).

RESULTS
Identifi cation of human MBOAT5 as a LPT1 homolog Using the yeast Lpt1p amino acid sequence as the query, a blastp search ( 27 ) identifi ed three human proteins with about 25% identity. These likely paralogs were originally identifi ed as MBOAT 1, 2, and 5 with conservation of a putative acyltransferase domain (Pfam PFO3062) ( 17 ). MBOAT5 was chosen for initial analysis because, like Lpt1, it contains a C-terminal ER retention signal, KKXX ( 28 ). Subsequent studies have renamed human MBOAT5 as LPCAT3 ( 19,29,30 ). The predicted protein has 487 amino acids, a molecular weight of 56 kDa, and seven algorithmidentifi ed ( 21 ) transmembrane domains.

Biochemical characterization of LPCAT3
A putative full-length cDNA clone for LPCAT3 was found in the mammalian gene collection ( 31 ) and inserted into baculovirus. This virus was used to drive expression in Sf9 insect cells. Precedents indicated that Sf9 cells harbor low levels of endogenous acyltransferase activity in general ( 22,32 ). Parallel infection with a GST-encoding virus allowed for the measurement of endogenous activity. In vitro LPCAT assays at 28°C found LPCAT3 expression to confer 10-fold higher activity than GST-expression ( Fig. 1 ). This temperature was primarily used for subsequent assays as it yielded 5-fold more activity than parallel assays at the physiological temperature of 37°C ( Fig. 1 ).
To kinetically characterize the substrate specifi city of LPCAT3, acyl-CoA concentration series assays were performed with four qualitatively different species. These represent the main classes of long chain acyl-CoAs: saturated (stearoyl-CoA); monounsaturated (oleoyl-CoA); polyunsaturated, omega-6 (arachidonic acid); and polyunsaturated, omega-3 ( ␣ -linolenoyl-CoA). Since the oleoyl-CoA curve was visibly sigmoidal ( Fig. 2 ), suggesting cooperativity, Hill plots of all kinetic data were generated (see supplementary Figs. I and II). Cooperativity was found in all cases except linolenyl-CoA at 28°C. These same plots were MS n fragmentation to identify the polar head groups. Relative quantifi cation was done by comparison to the most abundant phospholipid in each sample, which corresponded to m/z 760 or 34:1 (16:0, 18:1) PC. MassLynx 4.0 software was used for data analysis. MS n fragmentation was performed on the same instrument under similar conditions except that the ion source and ion optic parameters were optimized with respect to the positive molecular ion of interest. Data from four separate Sf9 infections underwent ANOVA using SAS software. Individual means were compared using Student's t -test.
Transfection of HEK293 cells. HEK293 (human embryonic kidney) cells were seeded in 24 well plates at a concentration of 2.5 × 10 5 cells/well ( ‫ف‬ 30% confl uence) and allowed to grow for 24 h prior to transfection. 1 g of pSM2 either encoding for a short hairpin RNA (shRNA) specifi c for LPCAT3 (Open Biosystems clone #V2HS 200012) or for a nonsilencing (NS) control (with a target sequence not complementary to any known human gene) was gently mixed with 1 l of lipofectamine transfection reagent (InVitrogen Life Sciences) in 100 l of Opti-MEM I reduced-serum medium (Life Sciences). After 20 min at room temperature, the mixtures were added drop-wise onto the cells. Subsequent experimentation was done after culturing in serumfree DMEM at 37°C for 24 h.
Analyses of cell metabolism, abundance, and viability. Activity of Complex I of the electron transport chain was determined by adding 30 l of 5 mg/ml MTT (3-(4, dimethylthiazolyl-2)-2, 5-diphenyltetrazolium bromide) to each well. After 2 h incubation at 37°C, the media was replaced with 800 l of DMSO to dissolve the resulting purple formazan, and the absorbance at 544 nm was measured with a FLUOstar OPTIMA plate reader (BMG Labtechnologies, Inc., Durham, NC). Cell number and viability were assessed using the ViaCount® assay reagent on EasyCyte Mini Guava fl ow cytometer (Guava Technologies, Hayward, CA) according to the manufacturer's instructions. Briefl y, 25 l cell suspensions were combined with 450 l of assay reagent containing 7-Amino-actinomycin D (7-AAD) and propidium iodide (PI). Following 5 min incubation, cell number and viability were assessed by alterations in both 7-AAD and PI staining as well as alterations in forward scatter.
Assessment of apoptosis. Apoptosis and necrosis were assessed using annexin V (apoptotic cell marker) and propidium iodide (PI, necrotic cell marker) staining and fl ow cytometry as previously described ( 25 ) with modifi cations. Cisplatin (50 M) and tert-butylhydroperoxide (500 M) were used as positive controls for apoptosis and necrosis, respectively. Following treatment, media were removed, cells washed twice with PBS and incubated in binding buffer (10 mM HEPES, 140 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , pH = 7.4) containing annexin V-FITC (25 g/ml) and PI (25 g/ml) for 10 min. Cells were washed three times using binding buffer, released from the monolayers using a rubber policeman and staining quantifi ed using a Becton Dickinson FacsCalibur fl ow cytometer. For each measurement, 10,000 events were counted.
Assessment of cell morphology. Cell morphology was determined using phase-contrast microscopy as previously described ( 25,26 ) with modifi cations. Briefl y, cells were washed twice with PBS, fi xed for 20 min using 10% buffered formalin, 4% formaldehyde, washed with PBS, covered with mounting media (Sigma), and cover slips applied. Visualization was performed using a TE300 Eclipse microscope (Nikon, Melville, NY). erence for lysoPC ( Fig. 4 ). However, other species of lysoPE and lysoPS may be utilized to a greater degree. Esterifi cation of lysophosphatidic acid (lysoPA), lysoPG, and lysoPI was assayed for but not detected (data not shown). This was also the case for lysoPS at 37°C (see supplementary Fig. IV). The structural features common to the acylacceptors utilized by LPCAT3 are an amine in the head group and a sn-1 acyl group attached through an ester linkage.

Effect of expressing LPCAT3 on phospholipid species in Sf9 cells
To begin to examine the physiological role of LPCAT3 in PL metabolism, the effect of expressing LPCAT3 on the PL profi le in Sf9 cells was measured. ESI-MS was performed on four independently generated cell lysates expressing LPCAT3 and GST. To account for alteration in ionization effi ciency, the relative abundance of each PL species was expressed relative to the most abundant species (34:1 PC; m/z 760) within the sample. Normalizing to an added standard (28:0 PC, m/z 678) was also performed and showed similar trends (data not shown). After analyzing the abundance of 73 different m/z values in the positive mode, 10 showed a signifi cant difference between the LPCAT3 and GST expressing cells. Assigning specifi c PL species to m/z values was based on their calculated theoretical monoisotopic masses and verifi ed by MS-MS. The fi ve species that increased with LPCAT3 expression were 32:0 PE, 32:0 PC, 36:0 PC, 32:0 (1-alkyl) PC, and 34:1 (1-alkyl) used to determine the K 0.5 and apparent V max values ( Table  1 ). In terms of affi nity (K 0.5 ), the substrate preference was stearoyl-CoA > oleoyl-CoA у arachidonyl-CoA у ␣ -linolenoyl-CoA. Arachidonyl-CoA showed higher affi nity at 28°C than at 37°C. In terms of capacity (V max ), oleoyl-CoA had the highest at 28°C and linolenoyl-CoA at 37°C. Because all of the reactions have the same, although undetermined, enzyme concentration, V max / K 0.5 should be proportional to the catalytic effi ciency ( k cat / K m ). This value was the highest for oleoyl-CoA. Clearly, the degree and placement of double bonds within the acyl-CoA substrate affects utilization by LPCAT3. ␣ -Linolenoyl-CoA was the only substrate for which concentrations above 110 M did not inhibit activity.
The kinetics of LPCAT3 were further analyzed by performing a lysoPC substrate concentration series with oleoyl-CoA as the acyl donor ( Fig. 3 ). Again, cooperativity was observed with a Hill plot showing a Hill number of 1.7, K 0.5 of 13.1 M, and an apparent V max of 486 nmol/min/ mg. Since oleoyl-CoA was kept at 50 M, a value that turned out to be less than saturating, the V max was less than that observed for oleoyl-CoA. Further analysis was performed with a single concentration (50 M) of lysoPC species with varying sn-1 acyl chains ( Fig. 4 ). Neither length nor degree of saturation affected utilization. However, the ether linkage in 1-alkyl lysoPC abrogated activity (data not shown). Varying the head group on the lysoPL also had a distinct effect on activity. Comparing 14:0 lysoPE to 14:0 lysoPC and 18:1 lysoPS to 18:1 lysoPC shows a 7-fold pref-  RT-PCR due to the lack of a commercially available antibody and resulted in signifi cant changes in phospholipid species ( Figs. 6, 7A ) . The trend was greater incorporation of long, polyunsaturated acyl chains into phospholipids when LPCAT3 expression is reduced. There was a concomitant decrease in free arachidonic (20:4) and linoleic (18:2) acid. These changes in phospholipid and fatty acid abundance coincided with a 30% reduction in cell number ( Fig. 7B ). MTT staining, a measure of cellular redox state, was similarly reduced by 30% ( Fig. 7C ), consistent with the reduction in cell number.
To further examine the effect of reduced LPCAT3 expression on cell function, we assessed the apoptotic cell marker, annexin V, and necrotic cell markers, PI and 7-AAD, using fl ow cytometry. As shown in Fig. 7D , shRNA against LPCAT3 resulted in almost a 2-fold increase in annexin V staining, compared with controls, in HEK293 cells. The overall level of apoptosis in these cells was still below 10%. PI and 7-ADD staining were unaffected (data not shown). These data suggest that limiting LPCAT3 expression induced apoptosis-specifi c cell death in a subpopulation of HEK293 cells but did not induce population-wide DNA damage or decrease membrane integrity. However, transfection with LPCAT3 shRNA had a striking effect on cell morphology. There were far fewer lamellipodia and less cell spreading compared with controls ( Fig. 8 ).

DISCUSSION
This study extends our previous characterization of the yeast lysophospholipid acyltransferase, Lpt1, to the human genome. The homologous LPCAT3 was chosen based on its conservation of the ER retention signal and previous studies where LPCAT3-specifi c siRNA reduced LPCAT activity in human Huh7 cells by 90% ( 19 ) and mouse B16 PC. The fi ve species that decreased were 30:0 PC, 34:2 PC, 36:1 PC, 36:2 PC, and 34:2 (1-alkyl) PC ( Fig. 5 ). The overall trend was that the abundance of PC with more saturated chains increased with LPCAT3 expression whereas PC with more unsaturated chains became less abundant. One thing to note is that when normalized to the added standard, 34:1 PC was about 30% more abundant in the LPCAT3 expressing cells. No differences between LPCAT3 and GST expressing cells were detected in the negative mode, suggesting that PI and PS species were unaffected.

Cellular effects of reducing LPCAT3 expression in HEK293 cells
The converse experiment was then performed by reducing LPCAT3 expression in HEK293 cells by transfection with specifi c shRNA. The 21 nucleotide target is in a section of LPCAT3 that shares no homology with LPCAT4 or LPEAT1. Reduced LPCAT3 expression was confi rmed by   the fi ndings of others ( 18,30,33 ). Since 1-alkyl lysoPC and lysoPA were not used as substrates, LPCAT3 does not seem to have a direct role in plasmalogen or de novo phospholipid synthesis. We maintained the use of the LPCAT3 nomenclature based on established precedent, even though the enzyme's substrate specifi city includes lysoPL species with amine head groups.
The observation that human LPCAT3 used unsaturated acyl-CoA species in vitro with higher capacity (i.e., apparent V max ) than a saturated acyl-CoA is similar to that in other studies. However, our studies show generally greater activity with the monounsaturated oleoyl-CoA than the polyunsaturated species. This is in contrast with a recently published mass spectrometry-based assay of LPCAT3 expressed in ale1 ⌬ ( lpt1 ⌬ ) yeast which detected a profound preference for linoleoyl-CoA and arachidonyl-CoA with little utilization of 18:1 acyl-CoA ( 33 ). Expression of human LPCAT3 in HEK293 cells also found a preference for arachidonyl-CoA compared with oleoyl-CoA, along with a fairly robust activity with stearoyl-CoA; a V max 1:5 of that of oleoyl-CoA ( 19 ). The ratio in our study was 1:15.
Consistent with these differences, the effects on phospholipid species as measured using ESI-MS were different compared with other studies. The use of LPCAT3 targeted siRNA in mouse B16 melanoma cells detected a 20% reduction in 16:0, 18:2 and 16:0, 20:4 species of PC ( 18 ). [ 14 C]Arachidonyl-CoA pulse labeling of HeLa treated with LPCAT3 siRNA identifi ed decreased incorporation into PC, PE, and PS ( 30 ). Our experiment of reducing LPCAT3 melanoma cells ( 18 ) by 50%. This preliminary evidence suggests that LPCAT3 is responsible for a large proportion of LPCAT activity in vivo in at least some cell types. Ours is the fi rst report of baculovirus-mediated expression of LPCAT3 in Sf9 insect cells conferring esterifi cation of lysoPC, lysoPE, and lysoPS in vitro. The use of lysoPE as a substrate differs from one earlier report ( 19 ) but confi rms  vestigators, this study is the fi rst to show a morphological phenotype and increased occurrence of apoptosis. expression in HEK293 cells resulted in an increase in polyunsaturated acyl chains in PC. This might be considered an anomaly if not for the concomitant decrease in polyunsatured chains that was observed when LPCAT3 was over expressed in Sf9 cells. Perhaps the contrasting results are due to a combination of different methodologies and cell types. These discrepancies are interesting to compare with LPCAT activity in rat liver microsomes, which preferentially utilize oleoyl-CoA and arachidonyl-CoA ( 34 ).
Particularly unique to these studies is the observed increase in apoptosis with reduction of LPCAT3 expression via shRNA. This is consistent with a recent fi nding that lysoPC is the "death effector" in lipoapoptosis in cultured hepatocytes ( 35 ). While lysoPC levels were below the sensitivity of the ESI-MS methods employed, these will likely increase upon the limitation of LPCAT3 expression. The reduction in lamellipodia is also in line with previous fi ndings that showed RNAi driven reduction of LPCAT3 (C3F) expression in HeLa cells resulting in Golgi fragmentation ( 36 ). Pharmacological inhibition of LPCAT activity resulted in Golgi tubulation in rat clone 9 hepatocytes ( 37 ). Also, disruption of the Golgi by Brefeldin A has been shown to limit lamellipodia in Swiss 3T3 fi broblasts ( 38 ). Taken together, these data suggest that reduced LPCAT3 activity disrupts Golgi structure, limiting plasma membrane extension and cell proliferation.
Also novel was the fi nding that acyl-CoA utilization by LPCAT3 displays cooperativity. One possible explanation is that at low substrate concentrations, diffusion rates into membranes are a limiting factor, and accordingly, activity is disproportionately low. However, there are precedents for two members of the MBOAT gene family showing substrate binding cooperativity. Purifi ed Brassica napus acyl-CoA diacylglycerol acyltransferase 1 (DGAT1) has shown cooperative oleoyl-CoA binding which coincides with protein forming homotetramers ( 39 ). Human acyl-CoA cholesterol acyl-CoA transferase 1 (ACAT1) has been clearly shown to bind cholesterol in a cooperative fashion ( 40 ) and form homotetramers ( 41 ). Whether LPCAT3 forms multimers remains to be determined.
In summary, LPCAT3 expression in tissue culture increased the abundance of phospholipids with relatively more saturated acyl chains, whereas limiting LPCAT3 expression increased the abundance of phospholipids with more unsaturated acyl chains. This is consistent with the kinetic data if cellular acyl-CoA concentrations are 5 M or lower. Kinetic analysis of LPCAT3 also detected a novel sigmoidal relationship between acyl-CoA concentration and activity consistent with cooperativity. While LPCAT3 expression has been reduced in tissue culture by other in -Fig. 8. Effect of LPCAT3-targeted shRNA on cell morphology. HEK293 cells were transfected or not as described in Fig. 6 . A: Nontransfected. B: Nonsilencing shRNA. C: LPCAT3 shRNA. After 24 h., cells were visualized by phase-contrast microscopy (25× magnifi cation). A representative fi eld of view is shown for each. LPCAT, lysophosphatidylcholine acyltransferase.