Daily rhythms of glycerophospholipid synthesis in fibroblast cultures involve differential enzyme contributions.

Circadian clocks regulate the temporal organization of several biochemical processes, including lipid metabolism, and their disruption leads to severe metabolic disorders. Immortalized cell lines acting as circadian clocks display daily variations in [(32)P]phospholipid labeling; however, the regulation of glycerophospholipid (GPL) synthesis by internal clocks remains unknown. Here we found that arrested NIH 3T3 cells synchronized with a 2 h-serum shock exhibited temporal oscillations in a) the labeling of total [(3)H] GPLs, with lowest levels around 28 and 56 h, and b) the activity of GPL-synthesizing and GPL-remodeling enzymes, such as phosphatidate phosphohydrolase 1 (PAP-1) and lysophospholipid acyltransferases (LPLAT), respectively, with antiphase profiles. In addition, we investigated the temporal regulation of phosphatidylcholine (PC) biosynthesis. PC is mainly synthesized through the Kennedy pathway with choline kinase (ChoK) and CTP:phosphocholine cytidylyltranferase (CCT) as key regulatory enzymes. We observed that the PC labeling exhibited daily changes, with the lowest levels every ~28 h, that were accompanied by brief increases in CCT activity and the oscillation in ChoK mRNA expression and activity. Results demonstrate that the metabolisms of GPLs and particularly of PC in synchronized fibroblasts are subject to a complex temporal control involving concerted changes in the expression and/or activities of specific synthesizing enzymes.

regulatory step under most metabolic conditions ( 31 ). However, it has been demonstrated that the availability of diacylglycerol (DAG) and regulation of ChoK also infl uence PC biosynthesis ( 27,32,33 ). In most mammals, there are two genes encoding for ChoK: Chka codes for ChoK ␣ 1/2 and Chkb codes for ChoK ␤ ( 34,35 ). Mice lacking ChoK ␣ die early in embryogenesis, whereas mice lacking ChoK ␤ survive to adulthood ( 30 ). ChoK overexpression has been implicated in the development of human carcinogenic processes ( 36,37 ). In mice, there are two genes for CCT: Pcyt1a encodes the CCT ␣ protein from alternative transcripts termed CCT ␣ 1 and CCT ␣ 2, and the Pcyt1b gene encodes the CCT ␤ 2 and CCT ␤ 3 proteins from the differentially alternative spliced mR-NAs CCT ␤ 2 and CCT ␤ 3 ( 38 ).
Several studies have reported the regulation of PC biosynthesis during the cell cycle, which is consistent with a higher PC mass requirement prior to mitosis ( 39 ). In addition, one report showed day/night changes in the content of PC and other phospholipids in the cerebral cortex of rats maintained under a light-dark cycle ( 40 ). Nevertheless, the synthesis of PC has not yet been investigated under constant environmental conditions to reveal whether temporal changes may be generated in an endogenous and self-sustained manner as expected for circadian rhythms. In this regard, we have previously demonstrated that de novo synthesis of whole phospholipids in retinal neurons in vivo or in vitro (41)(42)(43) and in quiescent fi broblasts synchronized by a serum shock is controlled by a circadian clock ( 4,14 ).
Glycerophospholipids (GPL) are essential structural components of all biological membranes and bioactive molecules involved in cellular functions, such as cell signaling, energy balance, vesicular transport, and cell-to-cell communication ( 21,22 ). GPLs are fi rst synthesized from glycerol-3-phosphate via de novo pathway described by Kennedy and Weiss in 1956 ( 22, 23 ) (see Scheme 1 ) and then undergo maturation in the remodeling pathway (Lands cycle) as a result of the concerted action of phospholipase A (PLA) and lysophospholipid acyl transferases (LPLAT) ( 24,25 ).
As a fundamental phospholipid in all eukaryotic membranes, phosphatidylcholine (PC) plays an important role in the structural composition of membranes and in the generation of second messengers involved in key regulatory functions and other processes (26)(27)(28). PC synthesis is crucial for cell growth, proliferation, and survival ( 29 ). In mammals, the disruption of genes encoding phospholipid biosynthetic enzymes has severe physiological consequences or lethality ( 30 ). In most nucleated cells, the biosynthesis of PC occurs via the Kennedy pathway, involving three enzymatic steps catalyzed by choline kinase (ChoK), CTP:phosphocholine cytidylyltranferase (CCT), and CDP-choline:1,2-diacylglycerol cholinephosphotransferase (CPT) ( Scheme 1 ). CCT activity is considered to be the rate-limiting and at different times across several cycles of 28 h, each ranging from 0.5 to 60 h. A 30 min labeling pulse of [ 32 P]Na 2 Orthophosphate (10 µCi/well) or [ 3 H]glycerol (8.5 µCi/well) was given to cultures at different times after the serum shock. Cells were harvested 30 min after addition of the radioactive precursor to the cultures at the different phases assessed and processed for phospholipid labeling.

Determination of radioactivity in phospholipids
The labeling of phospholipids was determined according to Guido and Caputto by the TCA-PTA method ( 46 ). Total phospholipids were extracted with chloroform/methanol (2:1, v/v), and radioactivity was determined in a liquid scintillation counter.

Chromatographic separation of individual phospholipids
Individual phospholipids were separated in silica gel 60 plates (Macherey-Nagel; Duren, Germany) by a one-dimensional, twosolvent system procedure described in Weiss et al. ( 47 ). Standards and individual lipid species were visualized with iodine vapors. Radioactivity incorporated into individual lipids was assessed by autoradiography. The band corresponding to PC at different phases was scraped from the silica plate, and [ 3 H]radioactivity was determined in vials with 2 ml of scintillation liquid in a scintillation counter.

In vitro determination of LPLAT
Confl uent NIH 3T3 fi broblasts from 100 mm dishes were collected at different times from 7 to 56 h after serum shock in PBS, lyophilized, and resuspended in H 2 O containing protease inhibitors. Cell lysates were used as a source of enzyme and endogenous lysophospholipids for determination of total LPLAT activity. The activity of NIH 3T3 fi broblast LPLAT was determined as an "in vitro" labeling by measuring the incorporation of [ 14 C]oleate from [ 14 C]oleoyl-CoA (56 mCi/mmol) into different endogenous lysophospholipid acceptors as described in Castagnet and Giusto ( 48 ) and Garbarino-Pico et al. ( 42 ). Under these experimental conditions, changes in the activity assessed may refl ect both changes in the amount of active enzyme and in the content of endogenous lysophospholipids. The incubation mixture for the assay contained 60 mM Tris-HCl (pH 7.8), 4 M [1-14 C]oleoyl-CoA (10 5 dpm/assay), 10 mM MgCl 2 , 10 mM ATP, 75 M CoA, and 80 g of cellular homogenates in a fi nal volume of 150 l. The reaction was incubated for 10 min with shaking at 37°C and stopped by addition of 5 ml chloroform/methanol (2:1, v/v). The lipids were extracted according to the method of Folch et al. ( 49 ). The lipid extract was dried under N 2 , resuspended in chloroform/methanol (2:1, v/v), and spotted on silica gel H plates. Unlabeled phospholipids were used as standards. The chromatograms were developed by two-dimensional TLC using as system solvents chloroform/methanol/ammonia (65:25:5, v/v/v) fi rst and chloroform/acetone/methanol/acetic acid/water (30:40:10:10:4, v/v/v/v/v) for the second dimension, and visualized with iodine vapors. The spots corresponding to PA, PC, phosphatidylethanolamine (PE), phosphatidylinositol (PI), and phosphatidylserine (PS) were scraped off, and radioactivity was determined by liquid scintillation.

Determination of PAP-1 activity
Confl uent NIH 3T3 fi broblasts from 100 mm dishes were collected at different times from 7 to 56 h after serum shock in PBS and resuspended in H 2 O containing protease inhibitors. PAP-1 activity was determined by monitoring the rate of release of 1,2 diacyl- [2-3 H]glycerol (DAG) from [2-3 H]phosphatidic acid (PA) as previously described by Pasquaré and Giusto ( 42,50,51 ). The In this work, we investigated whether the GPL metabolism is temporally regulated at early steps of de novo biosynthesis and remodeling. In addition, we addressed whether the synthesis of PC changes throughout the day and, if so, how this event may be regulated. To this end, we performed circadian studies in arrested NIH 3T3 cells after serum shock synchronization. We fi rst examined the temporal regulation of total GPLs by metabolic labeling with [ 3 H]glycerol as a precursor and the activity of phosphatidate phosphohydrolase 1 (PAP-1) in desphosphorylating phosphatidic acid (PA) to DAG, a branching point for de novo synthesis of all GPLs. We then assayed LPLAT activities involved in the remodeling of membrane phospholipids. Finally, we evaluated possible changes across time in PC biosynthesis by metabolic labeling with [ 32 P] phosphate and [ 3 H]glycerol, as well as the activity and expression of the two key synthesizing enzymes, ChoK and CCT.

Cell cultures
NIH 3T3 fi broblasts were grown in DMEM (Gibco) supplemented with 10% calf serum (Gibco). Cells reached confl uence after ‫ف‬ 4 days in a CO 2 incubator at 37°C. At time 0, the medium was changed to 50% horse serum (Gibco-BRL)-rich medium. After 2 h, the medium was replaced with serum-free DMEM or DMEM plus 0.5% calf serum and maintained under this condition for several days according to Balsalobre et al. ( 3,4 ).

Phospholipid labeling
The incorporation of [ 32  Each RT-PCR quantifi cation experiment was performed in duplicate (TaqMan) and triplicate (SYBR) for three or four independent experiments.

Immunocytochemistry
The NIH 3T3 were cultured as described previously on 10 mm coverslips to ‫ف‬ 70-80% confl uence. At different times after the serum shock, cells were washed twice with cold PBS and fi xed in 3% paraformaldehyde/4% sucrose in PBS for 15 min at 37°C. Cells were permeabilized in PBS-0.2% Triton × 100 for 5 min at 37°C, blocked for 2 h with 1% BSA-PBS, and then incubated overnight at 4°C with primary antibodies anti-CCT ␣ , anti-CCT ␤ 2, and anti-␣ -tubulin in 1% BSA-PBS. Cells were then washed with PBS and incubated with the corresponding secondary antibody Alexa Fluor 488 or Alexa Fluor 546 (1:1000) for 2 h at room temperature, washed, and mounted with Prolong Antiface kit. The slices were visualized by Axioplan fl uorescence microscopy (Zeiss, Oberkochen, Germany) equipped with a micromax camera (Princeton Instruments, Trenton, NJ). Relative levels of immunofl uorescence associated with CCT ␣ , CCT ␤ 2, and ␣ -tubulin proteins were assessed according to grayscale intensity by using image analysis software Metamorph 6.0 (Universal Imaging Corporation, Downingtown, PA).

Western blot
NIH 3T3 fi broblasts were harvested at different times after the serum shock in radio-immunoprecipitation assay (RIPA) buffer containing protease inhibitor (Sigma). Total protein content in the homogenates was determined by the Bradford method ( 45 ). Homogenates were resuspended in sample buffer and heated at 90°C for 5 min. 50 µg of protein were separated by SDS-gel electrophoresis on 12% polyacrylamide gels according to ( 55 ). Primary antibodies were incubated overnight at 4°C and secondary antibodies were incubated 1 h at room temperature. Finally the membranes were scanned using an Odyssey IR Imager (LI-COR Biosciences). Densitometric quantifi cation of specifi c bands was carried out with ImageJ software (National Institute of Health Bethesda, Maryland).

Propidium-iodide staining and fl ow cytometry
Cells grown to confl uence were collected at different times after synchronization (0, 2, and 24 h), washed in cold PBS, and fi xed with ice-cold 70% ethanol for at least 24 h. Cell pellets were resuspended in 150 µl of staining solution (PBS containing 50 g/ml propidium iodide and 10 g RNase A). Cell-cycle analysis was performed with 60,000 cells on a fl ow cytometer (DB Bioscience). The analysis program used was ModFit software (Verity Software House, Topsham, ME). reaction was stopped at 20 min by addition of chlorophorm/ methanol (2:1, v/v). DAG was separated by TLC and developed with hexane:diethyl ether:acetic acid (35:65:1, v/v/v) ( 52 ). To separate monoacylglycerol (MAG) from PA, the chromatogram was developed with hexane/diethyl ether/acetic acid (20:80:2.3, v/v/v) as developing solvent. PAP activity was expressed as the sum of labeled DAG plus MAG (h × mg of protein) Ϫ 1 .

In vitro assessment of ChoK enzyme activity
Confl uent NIH 3T3 fi broblasts from 100 mm dishes were collected at different times from 7 to 56 h after serum shock in PBS and resuspended in H 2 O plus protease inhibitor. Protein homogenate (100 µg) was assessed with 0.5 µl [methyl-

In vitro assessment of CCT enzyme activity
Confl uent NIH 3T3 fi broblasts from 100 mm dishes were collected at different times from 0.5 to 36 h after serum shock in sterile water and were sonicated three times for 30 s at 4°C. Cell lysates (100 µg of protein) were used as a source of enzyme for determination of total CCT activity. CCT activity was measured by conversion of phosphoryl [methyl-

RNA isolation and reverse transcription
Total RNA was extracted from NIH 3T3 cells using TRIzol® reagent following the manufacturer´s specifi cations (Invitrogen). The yield and purity of RNA were estimated by optical density at 260/280 nm. Total RNA (1 µg) was treated with DNase (Promega) and utilized as a template for the cDNA synthesis reaction using ImPromII reverse transcriptase (Promega) and an equimolar mix of random hexamers and oligo-dT (Biodynamics) in a fi nal volume of 25 µl according to the manufacturer´s indications.

PCR assay (endpoint PCR)
The primers used for RT-PCR are listed in Table 1 . The polymerase chain reaction was performed in a Labnet Multigen Thermal cycler using the GoTaq® DNA Polymerase (Promega). PCR reactions were carried out with an initial denaturation step of 5 min at 94°C, 35 cycles of 30-40 s at 94°C, 30 s at 60°C, 40 s at 72°C, and a fi nal 5 min elongation step at 72°C. Amplifi cation products were separated by 1% agarose gel electrophoresis and visualized by ethidium-bromide staining.

Real-time PCR
Quantitative RT-PCR (qPCR) was performed using SYBR Green or TaqMan Gene Expression assay in a Rotor Q Gene throughout the cell cycle remained constant at all times examined (0, 2 and 24 h after stimulation). Remarkably, these two major features -the nonproliferative condition and serum synchronization -make NIH 3T3 cell cultures a very useful model of peripheral oscillator for circadian studies regardless of both cell division and systemic infl uences from the brain master clock.

Circadian changes in the synthesis of GPLs in synchronized NIH 3T3 fi broblasts
To determine whether the de novo biosynthesis of total phospholipids varies throughout the day, we examined the time course incorporation of [ 3 H]glycerol into GPLs in quiescent NIH 3T3 cells after serum shock. The results showed a signifi cant daily variation in the synthesis of total GPLs ( P < 0.0004 by ANOVA) (

Daily variation in the activity of different NIH 3T3 phospholipid synthesizing enzymes
We explored the possibility that the circadian changes observed in the metabolic labeling of total GPLs in synchronized NIH 3T3 cells were due to variations in the activity of enzymes involved in the de novo synthesis of phospholipids (see Scheme 1 ). For this, we determined the in vitro activities of lysophosphatidic acid acyltranferase (LPAAT) and PAP-1 in homogenates of synchronized NIH

Statistics
Statistical analyses involved a one-way Kruskal-Wallis ANOVA by ranks to test the time effect. Pairwise comparisons were performed by the Mann-Whitney test when appropriate. For further periodic analysis, we performed a linear-circular correlation with the Spearman coeffi cient followed by an aleatorization test with 1,000 iterations to determine the P -value as described by Mardia et al. ( 56 ). The analysis considered a period ( ) of 14, 28, and 35 h and signifi cance at P < 0.05.

Characterization of cell culture conditions
To investigate the circadian regulation of GPL synthesis in NIH 3T3 cells, we fi rst characterized the experimental culture conditions. To this end, cells grown to confl uence and synchronized with a 2 h serum shock were maintained in a basal serum ( р 0.5%) medium for several days.
A synchronization protocol is essential to adjust single cell-autonomous oscillators to the same phase within the culture. Cells synchronized with a brief serum shock displayed a marked circadian rhythmicity in mRNA levels of the clock gene Bmal1 during 56 h after synchronization (supplementary Fig. IA) as previously reported (3)(4)(5). We examined whether cells undergo an important rate of cell division after the serum shock since it may serve as a mitotic stimulus. We found by fl ow cytometry that most cells were arrested at the G 0 /G 1 phases ( ‫ف‬ 90%), while <1% and ‫ف‬ 9% of cells reached the G 2 /M and S phases, respectively (supplementary Fig. IB). The distribution of cell populations   1C and Tables 5 and 6 ).
The statistical analysis showed time to have a major effect ( P < 0.006 by Kruskal-Wallis ANOVA). Pairwise comparisons revealed that levels of activity at 7-21 and 42 h were signifi cantly higher than those at 35 and 56 h. PAP-1 activity is due to the newly defi ned family of lipin proteins. Herein, we assessed mRNA levels for lipin-1 , one of the most abundant lipin isoforms, by using primers that recognize three alternatively spliced transcripts of lipin-1 . The results shown in Fig. 1D exhibit a similar temporal 3T3 fi broblasts collected every 7 h at different times ranging from 7 to 56 h ( Fig. 1B, C ).
De novo synthesis: LPAAT activity. PA, the main precursor of GPLs, is synthesized by the acylation of lysophosphatidic acid (LPA) catalyzed by LPAAT (see Scheme 1 ). In synchronized cells, the acylation of LPA exhibited a significant temporal variation, with the highest levels of PA production at 35 h after serum stimulation. The lowest levels of PA production by acylation were found at 7-21 h and 42-56 h ( Fig. 1B and Table 5 and 6 ). The statistical analysis revealed a signifi cant effect of time on LPAAT activity ( P < 0.06 by Kruskal-Wallis ANOVA).
De novo synthesis: PAP-1 activity. As precursor of all GPLs, PA is dephosphorylated to DAG by PAPs to synthesize PC and PE (see Scheme 1 ) ( 22 ). Although the activities

Circadian changes in the labeling of PC in synchronized NIH 3T3 fi broblasts
When we specifi cally investigated the biosynthesis of PC, the most abundant GPL in eukaryotic cells, a signifi cant temporal variation was observed in the incorporation of  ( Tables 3 and 4 ). These observations are in good agreement with previous fi ndings for the labeling of total [ 32 P] phospholipids ( 4 ).

Temporal control of the fi rst step in de novo synthesis of PC catalyzed by the enzyme ChoK
We investigated whether the daily changes observed in the biosynthesis of PC ( Fig. 3 ) were associated with comparable changes in the activity and/or expression of ChoK, the fi rst enzyme in de novo biosynthesis of PC. To this end, we fi rst studied the activity of ChoK in NIH 3T3 cells synchronized by a 2 h serum shock and harvested at different times during 56 h. We found a signifi cant fl uctuation across time in the in vitro production of [ 14 C]phosphocholine (Kruskal-Wallis ANOVA, P < 0.006) ( Fig. 4A and Tables 5 and 6 ). The results showed an oscillating pattern profi le to that found for PAP-1 activity. Although the ANOVA did not show a signifi cant effect of time across the 56 h examined, the pairwise comparisons revealed that mRNA levels at 7 and 42 h were signifi cantly higher than those at 14 and 49 h.

Temporal contribution of the LPLAT activity to GPL remodeling
To investigate whether the remodeling of GPLs varies across time, we assessed the activity of LPLATs involved in the reacylation of lysophospholipids (Lands cycle) in synchronized NIH 3T3 cells. We found a remarkable temporal variation in the activity of LPLAT for the different GPLs examined ( Fig. 2 and Tables 5 and 6 ). The statistical analysis revealed a signifi cant main effect of time for LPLAT activity irrespective of the lysophospholipid examined ( P < 0.02 by ANOVA) as well as a rhythmic pattern as shown by a linear-circular correlation analysis ( Table 6 ). Posthoc comparisons demonstrated that the maximum lysophosphatidylethanolamine acyltransferase (LPEAT) and lysophosphatidylinositol acyltransferase (LPIAT) activity levels along 21-35 h differ from the minimum values at 14 and 42-49 h after synchronization, whereas the level of lysophosphatidylserine acyltransferase (LPSAT) activity at 28 and 56 h was higher than that from 14 to 21 and from 42 to 49 h ( Fig. 2A ). We found a signifi cant 2-to 3-fold variation in LPSAT, LPEAT, and LPIAT ( Fig. 2 ) activities between maximal and minimal values. Even though we detected markedly distinct profi les for each lysophospholipid, the highest activity was observed during a temporal window centered at 28 h after stimulation and the lowest around 14 and 42-49 h. Remarkably, the activity peak for all LPLAT measured is in antiphase to the rhythm observed in the metabolic labeling of GPLs, which display minimum levels around 28-32 and 56 h after stimulation ( Fig. 1A ).  (29) Statistical analysis was performed with results from three independent experiments (n = 4-6/group) using a one-way Kruskal-Wallis ANOVA to test a time effect. Amplitude (%) was calculated as the ⌬ x/x av ratio, in which ⌬ x is the difference between the media of results at all times examined (x av ) and the lowest values of phospholipid labeling and the media of results (x av ). Maximum and minimum values for each precursor (dpm/mg protein*h) with their corresponding phases of occurrence shown in parentheses are also included to denote the amplitude of each variation along 56 h examined. See text for further detail. Statistical analysis was performed with results from three independent experiments (n = 4-6/group) using a periodic analysis by linear-circular correlation with the Spearman coeffi cient followed by an aleatorization test. Periodic analysis includes r 2 and period ( ) with signifi cance at P < 0.05. See text for further detail.

Temporal control of the second step in de novo synthesis of PC catalyzed by CCT
We examined whether the daily changes observed in the biosynthesis of PC involved comparable changes in the activity and/or expression of the key enzyme CCT.
Temporal profi les of the in vitro CCT activity. To this end, we fi rst assessed total CCT enzyme activity in vitro by measuring the formation of [ 14 C]CDP-choline in homogenates of synchronized cells collected at different times from 0 to 36 h. The results showed a signifi cant increase in CCT activity at 6.5 h, followed by a rapid decrease to basal values between 8 and 34 h ( Fig. 4B and Tables 5 and 6 ). The activity picked up again to maximum levels at 35 h after synchronization ( Fig. 4B ). The Kruskal-Wallis ANOVA of ChoK activity, with higher levels between 14-21 and 49-56 h alternating with lower activity at 7 and 35-42 h. To fi nd out whether the observed rhythmic activity of ChoK was due to similar changes in gene expression at the mRNA level, we studied the temporal profi le of ChoK ␣ and ChoK ␤ transcripts by RT-qPCR during 56 h ( Fig. 5 and Table 6 ). The Kruskal-Wallis ANOVA revealed a significant time effect for ChoK ␣ mRNA ( P < 0.03) but not for ChoK ␤ transcript levels. Moreover, the periodic analysis shown in Table 6 indicated that levels of ChoK ␣ mRNA oscillates with a period ( ) ‫ف‬ 28 h ( P р 0.03) ( Fig. 5A and Table 6 ). Nevertheless, we were unable to detect levels of ChoK ␣ protein by Western blot in the homogenates, likely due to its low expression in nontumor-derived cells ( 36,57 ). Statistical analysis was performed with results from 3-4 independent experiments (n = 3/group) using Kruskal-Wallis one-way analysis to test the time effect. Amplitude (%) was calculated as the ⌬ x/x av ratio in which ⌬ x is the difference between the media of results at all times examined considered as 100% (x av ) and the lowest values of mRNA expression and the media of results at all times examined (x av ). The times at which maximum and minimum values were found for each precursor are included to denote the periodicity of each variation along the 56 h examined. NS, not signifi cant.
a Values for CCT ␣ 1 and PEMT mRNAs at times of peak differed from other times examined by the Mann-Whitney test. See text for further detail.   ( Table 6 ). Nevertheless, CCT ␣ 1 mRNA exhibited an elevated expression at 3 h after synchronization and then returned to basal levels. In contrast, we detected signifi cantly higher levels of CCT ␤ 2 mRNA at 9-12 h after synchronization and minimum expression at 18-36 h. However, the effect of time observed did not represent a signifi cant oscillation as observed by the linear-circular correlation analysis ( Table 6 ). Temporal profi les of CCT ␣ 1 and CCT ␤ 2 mRNA expression. The daily variation found in CCT activity in the synchronized cultures of NIH 3T3 fi broblasts could be the result of equivalent changes in levels of mRNA and protein for the different CCT isoforms. We fi rst examined the presence of the different CCT isoform mRNAs by RT-PCR. Detectable PCR products were observed for CCT ␣ 1 , CCT ␣ 2 , CCT ␤ 2 , and CCT ␤ 3 transcripts compared with positive controls (brain and testis tissues; data not shown). To quantify the most abundant CCT transcripts in cell The LPLAT activity for LPS exhibited temporal changes, with higher activity at 28 and 56 h after synchronization ( P < 0.05) as determined by Mann-Whitney test. The LPLAT activity for LPE (B) and LPI (C) showed similar temporal patterns, exhibiting higher activity between 21 and 35 h after synchronization ( P < 0.01 and P < 0.004, respectively, by Kruskal-Wallis ANOVA). Results are the mean ± SEM of three independent experiments (n = 3 technical replicates/group). in NIH 3T3 cells at different times after a 2 h serum shock. The Western blot for the CCT ␤ protein revealed two bands between 49 and 37 kDa, which may correspond to the isoforms CCT ␤ 2 (43 kDa) and CCT ␤ 3 (39 kDa), respectively. These two bands displayed different temporal patterns of expression: band "a" (CCT ␤ 2) presented a peak at 9 h that differs from those at 12-36 h ( P < 0.05 by Mann-Whitney test), whereas band "b" (CCT ␤ 3) substantially increased by 9-18 h after serum shock ( P < 0.05 by Mann-Whitney test). In contrast, although the ANOVA revealed no signifi cant effect of time for CCT ␣ , Mann-Whitney test indicated that levels at 3 h differ from those at other times after synchronization ( P < 0.05). The relative contribution of the different CCT isoforms clearly changes over time, indicating that although CCT ␣ peaks at 0-3 h, the two CCT ␤ isoforms may act in combination at particular times after serum treatment. In addition, similar changes were observed in the immunoreactivities associated with CCT ␣ and CCT ␤ proteins at the different times tested by ICC (supplementary Fig. IIIE). Moreover, we determined that CCT ␣ protein was mainly confi ned to the cell nucleus, whereas the CCT ␤ 2/3 immunofl uorescence remained outside the nucleus (supplementary Fig. IIIE). Furthermore, the differential subcellular localization of the two proteins remained constant at all times examined. CCT activity in NIH 3T3 fi broblasts grown to confl uence and synchronized with 50% horse serum shock for 2 h. CCT activity was measured in vitro with total homogenates of cells harvested at different times after serum shock between 0 and 36 h. Results are mean ± SEM of nine independent experiments (n = 3-16/group). The daily variation was observed in total CCT activity ( P < 0.001 by ANOVA), with highest levels at 6.5 and 35 h after serum shock. (C) LPCAT activity was determined in homogenates of NIH 3T3 synchronized by 2 h 50% horse serum shock and harvested at different times. The activity was measured by the incorporation of [ 14 C] oleate into LPC. The LPCAT activity exhibited a signifi cant temporal variation ( P < 0.02 by Kruskal-Wallis ANOVA) with the lowest levels around 14 and 49 h after serum shock. circadian clocks results in pathophysiological changes resembling the metabolic syndrome, in which lipid metabolism is strongly altered ( 9,10,14,18,59 ).
The cultured fi broblasts used here were quiescent cells grown to confl uence and then stimulated with a high concentration of serum that synchronizes endogenous clocks located in individual cells ( 3,5 ) (supplementary Fig. I).
Although the 2 h serum shock could trigger cell proliferation, the cell cycle cannot progress in confl uent NIH 3T3 cells because they are inhibited by contact ( 60 ). Moreover, cultures maintained in a basal serum condition after the serum shock present ‫ف‬ 90% of cells arrested at G 0 /G 1 , a percentage that remained constant at the times examined (0, 2, and 24 h after stimulation; supplementary Fig. I).

Daily variations in phospholipid biosynthesis and gene expression in synchronized NIH 3T3 cell cultures
It has been shown that fi broblasts in culture exhibit circadian rhythms in the biosynthesis of [ 32 P]phospholipids in clear antiphase with the rhythm in Per1 expression ( 3,4 ). Moreover, after knocking down Per1 expression, the metabolic rhythm disappeared and cultures of CLOCK mutant fi broblasts -cells with an impaired clock mechanism -displayed a loss of rhythmicity in both PER1 expression and phospholipid labeling. These results clearly show a tight control over phospholipid synthesis by the molecular circadian clock in immortalized cell cultures.
In this article, we characterize the oscillatory behavior of GPL de novo biosynthesis and remodeling in cultured fi broblasts after serum synchronization.  Tables 3 and 4 ). This fi nding is fully consistent with our previous report on total [ 32 P]phospholipid labeling ( 4 ).
To determine whether the observed changes were a consequence of particular time-regulated enzyme activities, we assessed the activity of the two key biosynthetic enzymes, LPAAT and PAP-1. Several enzymes involved in GPL metabolism have been shown to be highly regulated ( 42,43,48,(61)(62)(63). PAP-1 (or lipin), an enzyme that plays an essential role in phospholipid metabolism, dephosphorylates PA to DAG ( 64,65 ). A second type of phosphatidate phosphatase is PAP-2 (or lipid phosphate phosphatase), which is mainly involved in signal transduction mechanisms (66)(67)(68). DAG generated by PAP-1 is specifi cally utilized for PC, PE, and triacylglycerol synthesis, whereas PA is used for PI synthesis through the CDP-DAG pathway ( 22 ) ( Scheme 1 ). Although we detected both PAP activities in the cell preparations, we focused our attention on the temporal regulation of PAP-1 activity in relation to its role in GPL biosynthesis. We found that LPAAT and PAP-1 activities exhibit signifi cant daily variations with totally opposite profi les when assessed in synchronized NIH 3T3 preparations collected at different times ( Fig. 1B, C ). Strikingly, the lowest levels of PAP-1 activity were recorded around 35 and 56 h after synchronization, times at which LPAAT showed the highest activity. Thus, the resulting PA Time-related profi les of other enzymes that contribute to the biosynthesis of PC An alternative PC biosynthetic pathway taking place mainly in the liver has been described in which the enzyme phosphatidylethanolamine methyl transferase (PEMT) converts PE into PC ( 31 ). To address whether this pathway contributes to the temporal control of PC metabolism, we evaluated pemt mRNA expression by RT-qPCR. We found that levels of pemt mRNA displayed a rapid and transient increase in response to the serum shock; however, it did not exhibit a signifi cant circadian rhythmicity (supplementary Fig. IIC and Tables 5 and 6 ).
In addition, the Lands cycle may also participate in maintaining PC content across time. To this end, we investigated the temporal contribution of LPCAT activity in the production of PC in synchronized NIH 3T3 cells. We found a signifi cant temporal variation in the incorporation of [ 14 C]oleate into PC ( P < 0.02 by Kruskal-Wallis ANOVA; Tables 5 and 6 and Fig. 4C ). Posthoc comparisons demonstrated that the maximum LPCAT activity at 7, 21, and 56 h signifi cantly differ from those at 42-49 h. Moreover, the linear-circular correlation analysis showed that the activity exhibited a daily rhythmicity with a period ( ) of ‫ف‬ 14 h ( Table 6 ).

DISCUSSION
In the present work, we described for the fi rst time concerted and sequential changes in specifi c enzyme activities and/or mRNA expression for the temporal control of GPL metabolism and particularly of PC biosynthesis in synchronized fi broblasts. Nevertheless, not all the steps involved in lipid synthesis and tested here are subject to temporal regulation (see Scheme 1 ). However, those steps that were shown to be regulated by internal clocks may refl ect different levels of transcriptional and posttranscriptional control involving changes in mRNA and protein expression and/or activities for the different synthesizing enzymes investigated.

Immortalized cell lines constitute intriguing models of peripheral oscillators for the study of metabolic oscillations
The results presented here constitute the fi rst report demonstrating the temporal organization of the GPL metabolism in cell-autonomous oscillators regardless of the cell cycle and infl uence of the central master clock. Cultures of immortalized fi broblasts synchronized by a serum shock displayed daily oscillations in the expression of clock genes (supplementary Fig. I) ( 4 ), as well as in several enzymatic steps implicated in the de novo biosynthesis and remodeling of GPLs (see Scheme 1 ).
Self-sustained oscillatory behavior at the level of gene expression and metabolic activities has been observed in immortalized cell lines and primary cell cultures, regardless of central control or systemic infl uence ( 2-5, 42, 58 ). Recent studies clearly linked the molecular clock with the regulation of lipid metabolism, and the disruption of axonal transport (41)(42)(43). These metabolic oscillations may represent a general feature of oscillators present either in NIH 3T3 or neuronal cells ( 41,42 ). Moreover, these fi ndings refl ect a similar correlation regarding the circadian-regulated synthesis of GPLs and expression of clock genes, such as Bmal1 , in both cell types.

Time-related mechanisms modulate the biosynthesis of PC in cell cultures
Our observations constitute the fi rst report of significant daily variations in PC biosynthesis in quiescent cultures of NIH 3T3 fi broblasts after synchronization. Indeed, higher levels of precursor incorporation ([ 3 H]glycerol or [ 32 P]phosphate) were found during the fi rst 5-6 h after stimulation along the fi rst cycle and minimum labeling at 28-35 and 53-59 h ( Fig. 3 ). Multiple levels of control may act to regulate the metabolism of PC across time in the cultures. At this point, we can hypothesize that the activation of specifi c key regulatory synthesizing enzymes at the level of expression and/or activity takes place at particular times after serum shock synchronization.
The variations observed in the biosynthesis of PC in synchronized cultures of fi broblasts ( Fig. 3 ) may be due at least in part to 1 ) a higher availability of DAG differentially generated by PAP-1 at certain times ( Fig. 1C ) and 2 ) concerted changes in the activity of the regulatory enzymes ChoK and CCT ( Fig. 4A, B ). In fact, both ChoK activity and expression for the ChoK ␣ isoform mRNA exhibited a signifi cant daily variation in the synchronized NIH 3T3 fi broblasts ( Figs. 4A and 5A and Tables 5 and 6 ). The lowest levels for ChoK ␣ were recorded at 28 h after serum stimulation, and the enzyme activity displayed a delayed profi le across time, likely refl ecting the time required for translation. Strikingly, ChoK has been shown to be strongly expressed in tumor cells ( 36 ), and although this enzyme is not the rate-limiting enzyme in PC synthesis, it has been proposed to function as a key regulatory enzyme ( 27,32,33 ) that seems to be tightly regulated in a circadian manner. Our results suggest that this enzyme may temporally control the biosynthesis of PC in synchronized cells. In addition, the temporal variation in [ 3 H]PC biosynthesis of cell cultures ( Fig. 3 ) is preceded by brief changes in the total activity of CCT ( Fig. 4B ), the ratelimiting enzyme in the Kennedy pathway, indicating that there are other levels of regulation controlling PC content in the cell. Our fi ndings show two sharp peaks of enzyme activity across a 36 h window, appearing at 6.5 and 35 h after serum stimulation with a 28.5 h separation, resembling the period previously described ( 4,5 ). In this respect, a diurnal variation in CCT activity was reported in the rat retina ( 71 ), and changes in CCT activity were associated with those observed in PC synthesis (72)(73)(74) along the cell cycle in cultures of different cell lines. The CCT levels and subcellular distribution of both CCT isoforms did not display circadian rhythmicity (supplementary Fig. III), showing nuclear confi nement for the CCT ␣ isoform and cytoplasmic localization for CCT ␤ . Thus, we may infer that the variation in CCT activity indicates enzyme activation by posttranslational modifi cations content could be transiently utilized for other intracellular functions, such as signaling or PI synthesis. In addition, similar temporal patterns with elevated PAP-1 activity and lipin-1 mRNA levels were found at other times, likely indicating that a higher DAG content can be provided for the de novo synthesis of GPLs during these phases ( Fig. 1C, D and Table 5 ). Strikingly, lipin-1 has also been reported to oscillate in the liver at the mRNA level by microarray assay ( 69 ) and to act as a transcriptional coactivator on a circadian basis ( 70 ).
The generation of lysophospholipids is a consequence not only of the prior sterifi cation of glycerol-3-phosphate but also of phospholipase activity as part of the well-known deacylation-reacylation cycle ( 25 ). Since LPLAT may refl ect the activity of this cycle, we cannot discard the possibility of the differential temporal regulation of PLA or other phospholipases in NIH 3T3 cells. Moreover, the LPLAT activities of the different lysophospholipids examined presented distinct circadian profi les, mostly with highest levels in a time window around 21-35 h, in exact antiphase to [ 3 H]GPL labeling and PAP-1 activity (see Figs. 1A and 2 and Table 5 ). These observations probably refl ect a differential wave of GPLs generated by acylation of LPLAT to keep the membrane homeostasis, likely to give rise to differential phospholipid content and composition in the membrane when de novo lipid synthesis is signifi cantly decreased. This temporal variation may cause signifi cant changes in the fatty acid composition and quality of GPLs, affecting the membrane curvature and fl uidity and ultimately regulating the activity and function of different cellular processes ( 24 ). Although the reported oscillations are driven by an autonomous clock with a period near 24 h, a few LPLAT activities (LPC and LPIAT, Tables 5 and 6 ) displayed a different period from 28 h, likely refl ecting the fact that the total activity came from different acyltransferases having affi nity for the same lysoGPLs, since at least two different LPLAT families with multiple isoenzymes have been described ( 24 ).
We are as yet unable to fi rmly establish whether the changes described in this article are generated by changes in levels of enzyme mRNA and/or protein, by a precise temporal regulation of enzyme activities, by differential accessibility of particular substrates, or by a combination of all these possibilities; and if the latter is the case, whether there is a particular hierarchy in the sequence of such events. Nevertheless, our fi ndings clearly show that the de novo biosynthesis and remodeling of GPLs are subject to endogenous temporal control, likely refl ecting i ) a differential requirement across time of newly synthesized phospholipids for membrane biogenesis and/or generation of second lipid messenger waves, or ii ) the temporal separation of events within the cell, in addition to the spatial organization of reactions in the different cell compartments. The biogenesis of new membrane is required for a number of cellular processes, including cell division, exocytosis, vesicular traffi c, and production of organelles. In fact, we have reported that retinal neurons also display daily rhythms in phospholipid metabolism, which may be necessary for axon and dendrite growth regulation as well as and binding to the endoplasmic reticulum or nuclear membranes (72)(73)(74)(75).
An alternative PC biosynthetic pathway has been described in the liver involving PEMT to catalyze the formation of PC from PE ( 31 ). Although the contribution of PEMT activity to the synthesis of PC in NIH 3T3 cells and whether it is regulated by the endogenous clock is still unknown, we only found a rapid and transient increase in PEMT mRNA levels in response to the serum shock but no detectable circadian oscillation (supplementary Fig. IIC).
In addition to clock regulation, PC metabolism may be regulated by homeostatic mechanisms, such as substrate availability, the state of enzyme activities (posttranslational modifi cations and subcellular localization), and rate-limiting steps, among others ( 22 ). In this respect, cellular levels of CTP, a substrate for CCT, can control PC synthesis as well as choline and phosphocholine availability and the cellular concentration of DAG ( 27,32 ).
Overall, our observations lead us to infer that the biosynthesis of whole GPLs and particularly of PC in NIH 3T3 fi broblast cultures undergoes clear temporal variations somehow sensing the time of day in relation to external cues (food, temperature, hormones, etc.). The circadian regulation of GPL synthesis may be important for a better understanding of the temporal organization of a number of cellular processes, such as the membrane renewal, vesicular traffi cking, exocytosis, membrane protein activity (receptors, channels, etc.) and/or second messenger reservoir changes.